Changelog

Everything released on novarchy.com, newest first.

The format follows Keep a Changelog. Every entry says what the player saw before and what they see now. The four oldest releases (beta.1 to beta.4) are in Italian: they were written before the project moved to English.

A release is written up here once it is live, so the newest one appears on this page a little after it reaches the game.

1.450.1 — 2026-10-13

Changed

• Relics are now tuned the way Sector 13's control worlds are (#7321)
  • Before: a relic's automated garrison was priced on the single strongest fleet on the server. A relic held by a guild had no automated garrison at all, so one interceptor could take a relic its guild had left unguarded (#7151).
  • Now: the garrison is priced on the same figure Sector 13 uses: the average fleet of the ten strongest active empires, ignoring anyone more than twice ahead of the rest. A held relic keeps a quarter of that, the same for every attacker, beside the fleets its holder parks there. Passing flotillas bounce off it.
  • It is still not a fortress. The strongest empires can take an unguarded held relic on their own, so keeping a relic still means sending garrisons, as on Sector 13.
  • Rebuilding: a relic's automated garrison is rebuilt when it changes hands, at most once every six hours, and topped up at the six-hour rescale. In between, a relic emptied by a battle keeps a small floor garrison rather than none.
  • Unchanged: attacks on an unclaimed relic still meet a garrison calibrated for your guild.
  • On the day of this update, every relic held today gains its quarter at once. A fleet already flying at a held relic that looked empty will meet it.

1.450.0 — 2026-10-12

New

• Data crates in the Shop (#7310). Until now the Shop sold resources only as Supply crates, Ferrolite : Xenogas : Data at 6 : 3 : 1, so 10M of crate held 1M of Data. A new Data crates section now sits below them, with the same five sizes at the same five prices, all of it Data: Data Crate 10M (€1.99), 100M (€4.99), 1B (€19.99), 10B (€49.99) and 100B (€99.99). Each purchase delivers one sealed crate of Data to your Inventory. You open it into any colony up to that colony's storage, and what does not fit stays sealed until you make room. The Supply crates' rules apply: no purchase limit, no Star Forge needed, no VIP points or supporter tier, and an account linked to Google or Apple. The Terms of sale now name Data crates.

  Known gaps: the confirmation still says "the crates arrive… open each one" for what is a single crate. An app installed before this release labels a Data crate "A Supply crate delivered" in the Activity feed until it updates; the Inventory shows it correctly.

1.449.2 — 2026-10-12

Fixed

The login screen tells an Italian player their account is deleted in Italian (#7288) After deleting their account, a player playing in Italian landed on a login screen in Italian with one English sentence on it: the notice saying the account was gone, with everything it held, and that they can sign up again to start a new empire. That notice is now in Italian too, in wording the owner approved, and it says exactly what the English says. English players see the same sentence as before.

1.449.1 — 2026-10-12

Changed

• The browser no longer keeps its own copy of how a building's sheet is computed (#7196). (PR #7306) Nothing a player reads changes. Until now the web app carried about 3,000 lines that worked out every figure under a building's sheet — production, energy, discounts, defences, the full progression table — duplicating what the game engine already computes for the server. Since the previous releases the sheet has shown the engine's figures; this release deletes the browser's copy. Before deleting it, the old code and the engine were compared over every test colony, building and level: they printed the same text whenever they were given the same colony. What they printed is now frozen in the test suite, so a change to any wording or figure on a building sheet has to be made on purpose. Still to come in the same cleanup: two fallbacks in the sheet that no screen reaches any more, and a few helpers only tests still use.

1.449.0 — 2026-10-12

New

The Season tab gets a Season Chronicle: four milestones on your own season score, each with a title (#7158)

Until now, the only personal goal in a season was the podium of three. While a season runs, the Season tab now shows a Season Chronicle card with four milestones on your own season score. Each one gives a title: Witness, Chronicler, Harbinger and, for the top tenth, Mark of the Eschaton.

The previous season's scorers set the milestones, at roughly their 25th, 50th, 75th and 90th percentile scores. Each milestone is at least double the one before it, and the first is at least one founding. When the previous season had fewer than 20 scorers, fixed figures set them instead. Each milestone you have not reached tells you what closes the gap: founding worlds in the sector (only when you have room left under the two-world limit), or more hours holding a world with your guild while your garrison is above the bar.

Your title shows beside your name on the season's player board. The highest title you reach in a season replaces the lower ones. When the season closes your title is kept, and it stays yours until the next season closes. Titles carry no resources and no Dark Matter.

Titles start being kept with the first season that closes after this release. Until then, the board shows only the titles players reach in the current season, from their score so far.

1.448.2 — 2026-10-12

Changed

• A building's full progression table is computed by the server, like the sheet above it (#7196). (PR #7300) Nothing a player reads changes. Until now the table under every building's sheet — all thirty levels with their price, build time, core fee, the Command Center level each one waits on, and the effects that change from level to level — was worked out in the browser, from formulas the browser kept its own copy of. The game engine now builds it from the same figures as the sheet's rows, and the server answers it when the player opens the table. A test compares the two tables for every building in more than forty colony situations, cell by cell, and they agree. If the server's answer is out of date, slow, or missing, the browser shows the same table built locally by the same engine code. The table is asked for only when the player opens it, and only while the sheet above it also comes from the server. It stays on screen while it is re-read every few minutes. The supply lanes' rule for which lanes still carry goods now lives in one place instead of two. Still open: an open Trade Portal sheet on a hub colony is re-read about every 30 seconds, because the server does not yet say which inbound lane figure it used. That needs its own change.

1.448.1 — 2026-10-12

Changed

• The Trade Portal's sheet figures now come from the game server, and with it every building's (#7298). Part of #7196. The Portal's sheet reads the same sentences as before: open offers across your empire, the value cap, supply lanes out and in, and how much the hub can take in. Its numbers are now computed by the server, which reads your best Portal and the live lanes into the colony itself. Every building in the game now has its sheet figures computed by the server. Still computed in the browser: the progression table under the rows.

1.448.0 — 2026-10-12

New

The catch-up ramp now speeds up construction (#7158)

Before, the catch-up ramp made a newcomer's resource production and research faster, but buildings and ships still took their normal time. Now, while your empire is on the ramp, building and ship orders are faster too, by the square root of the ramp's boost. The speed-up fades as the boost does and ends where the ramp ends. An order keeps the time it was quoted when you placed it. The Leaderboard's ramp line shows your current construction speed next to the boost, and the "Joined late?" note explains the rule. Research works as before, and Anchorage ship bundles keep their normal price.

1.447.0 — 2026-10-12

New

The Leaderboard shows your catch-up ramp and the Gate it fades out at (#7158)

Before, the Leaderboard showed a newcomer how far behind they started. The catch-up ramp that answers that worry was mentioned on this tab only inside the collapsed "Joined late?" note, without its figure. The threshold, half the median Gate of established, active empires, did not appear anywhere on screen.

Now, while your empire is ramped, a line under the Leaderboard header shows your boost and the Gate at which the ramp fades out, in PP. Both figures are the ones the server measured, for example "Catch-up ramp ×8.55: … (315,248 PP)". Power Details now lists the ramp as the last source of Ferrolite, Xenogas and Data production. The Leaderboard line is also translated into Italian; the Power Details row is in English, like the rest of that dialog.

For a few minutes after a server restart, before the ramp is measured again, the line appears without the figure. This is the first part of #7158. The ramp's numbers are unchanged, and the issue's other proposals are not in this release.

1.446.11 — 2026-10-12

Changed

A spy report now says where a parked defence fleet is (#6353) A top-tier spy report used to list a fleet the target had parked on another colony only as "defend (stationed)", with its ships and its return time, and never said where it was. A fleet parked on a friend's colony therefore stayed out of sight for as long as its watch lasted. The row now names the colony it is parked on and gives that colony's coordinates. If that colony is under the Veil, the row says only that the fleet is at a colony under the Veil, with no name and no coordinates. Nothing about the host colony itself is shown: not its hangar, its buildings or the other fleets guarding it. The return time works as before, and reports below the top tier have not changed.

1.446.10 — 2026-10-12

Changed

A guild donation no longer charges Warehouse capacity the weekly war prize does not count (#5927)

Before: every donation cost 0.6% of your home colony's Warehouse capacity, however large it was. Since the war prize stopped counting capacity above its reference, a player with a big Warehouse kept paying for storage the prize ignored.

Now: the donation counts your home Warehouse's capacity only up to a ceiling, the median of the largest home colonies among the empires played in the last week, re-measured every hour. A home Warehouse above that ceiling pays what a Warehouse exactly at it pays, and one below it pays what it paid before. Every resource still costs at least 1. The guild's daily dues are donations, so they pay the same price. The Treasury panel's donation sentence names both figures when the ceiling applies, and the rule under the button says the capacity is counted only up to it.

The iOS and Android apps keep showing the previous, higher price until their next update. Until then, a player above the ceiling who holds the new price but not the old one finds the Donate button dark in the app. The web game shows and charges the new price.

Reported by a player on Discord.

1.446.9 — 2026-10-12

Changed

The account deletion panel is now in Italian (#6561) An Italian player who opened the panel to delete their account, on the phone settings sheet or on the commander card, used to read all of it in English. That covered the warning, the list of what is deleted, what is kept, the guild line, the confirmation prompt, the buttons and both failure messages. The whole panel now reads in Italian, in wording the owner approved, and it says exactly what the English says about what is deleted and what is kept. English players see the same text as before.

Still in English: the one sentence on the login screen after a deletion. It waits on owner-approved Italian (#7288).

1.446.8 — 2026-10-12

Fixed

The privacy page names Anthropic as the AI behind the Wraiths, and says what the AI empires' plans and guilds send (#5234, #3214) Before, the privacy page did not say who commands the Wraiths. It also said the AI empires' planning ran whenever the game used its AI service at all, and it never said that a guild's rank review sends the whole roster.

Now the page has an Anthropic entry: the Wraiths are commanded by Claude, Anthropic's AI model, on a schedule. The entry lists what their briefing carries about players:

• the top five of the leaderboard, with place, commander name, power, colonies and strike history, plus a colony's coordinates when there is a bounty on them;
• the account identifier of at most three proposed targets;
• the names of players whose fleets are heading at a Wraith.

The OpenRouter entry now says the AI empires' plans and guild decisions use the AI service only when it is switched on for them, whether or not they are talking. It also lists what each use sends:

• Plans send spy-report details and, during an alliance war, enemy colonies and guild names.
• Each rank review sends every member's name, rank, points and contribution.
• An application sends the applicant's name, and their power rating when the guild has an entry bar.
• Their chat in a guild channel names who joined, left, changed rank or received from the Fund.

Guild rank titles are now listed among the names that are checked.

Nothing the game sends has changed; only the page has.

1.446.7 — 2026-10-12

Changed

• The rest of the Guild in Italian (#7205). Before, players with Italian on saw these in English:
  • the chat panel's own text: the mute notice, the emoji picker, tags, pins and replies;
  • the one-line descriptions of the guild research branches;
  • relic types and lore;
  • the VIP title on crests and portraits;
  • the busy label on guild confirmations.

  Now they read in Italian. In English, every word and figure reads as before.

  Chat messages stay exactly as written. Names stay as the game gives them, including research node names and relic names. The beta notice no longer names the Guild: it now says only that some screens are still in English.

1.446.6 — 2026-10-12

Changed

• The Shipyard, Nanoforge and Arsenal sheet figures now come from the game server (#7281). Part of #7196. The sheets read the same sentences as before, including «Time per hull» for the reference hull you pick on the card. The numbers are now computed by the server with your pick, your guild's Suspended Yard relic, an Exodus shock or raid disorder, your industrial edict and your VIP level. An open sheet now refreshes when a shock or a disorder begins, not only when it ends, and when a ship order spends your early-build sprint. Still computed in the browser: the Trade Portal, and the progression table under the rows.

1.446.5 — 2026-10-12

Fixed

• Your mother colony's building sheets now come from the game server too (#7279). Part of #7196. Over the last five releases the building sheet moved to figures the server computes, but only your other colonies' sheets read them: the mother colony's never asked the server, and kept the figures your browser computed. They are the same figures, held to the same text by our tests; now every colony's sheet comes from the same place.

1.446.4 — 2026-10-12

Changed

• The Laboratory, Factory, Sensors and Hypergate sheet figures now come from the game server (#7277). Part of #7196. The sheets read the same sentences as before. Their numbers are now computed by the server, including what they read beyond the open colony: the Laboratories on your other colonies, your strongest Sensor Grid, your guild's Science keystone and whether you are in a guild. The server always knows the guild's research, so on a colony other than your mother colony the Laboratory's research-speed row no longer has to wait on the guild call to show its figure. Still computed in the browser: the Shipyard, Nanoforge, Arsenal and Trade Portal, and the progression table under the rows.

1.446.3 — 2026-10-12

Changed

• The defence buildings' sheet figures now come from the game server (#7273). Part of #7196. The Fortress, the Citadel, the Cannon, the Shield and the Batteries read the same sentences as before, with their countdowns ticking as before (a platform regrowing, the Citadel standing back up, the Sanctuary re-arming); the numbers in them are now computed by the server. A battle that breaks the Citadel, arms the Sanctuary or harvests the war value now refreshes an open sheet at the next colony update, and the war value's percentage keeps climbing while the sheet stays open. Still computed in the browser: the Shipyard, Nanoforge, Laboratory, Trade Portal, Factory, Arsenal, Sensors and Hypergate, whose figures read other colonies, the guild or the trade lanes.

1.446.2 — 2026-10-12

Changed

• The building sheet's figures for twelve more buildings now come from the game server (#7268). Part of #7196. The Extractor, Refinery, Compute Node and Stellar Forge (production and co-production), the Zero-Point Reactor (surplus conversion, yield and ceiling), the Accelerator (Dark Matter a day), the Warehouse (storage cap), the Terraformer and the Outpost (malus relief and bonus), the Observatory (warning detail, counter-espionage, reconnaissance travel), and the Reactor and Orbital Panels (their energy) read the same sentences as before; the numbers in them are now computed by the server rather than by the browser, so every client shows the same figure. The sheet re-reads them when the colony's mandate, the empire's guild boost, edict or VIP level, or any building's level changes, and a sheet left open no longer blinks back to the local figures every four minutes. Still computed in the browser: the defence buildings (Fortress, Citadel, Cannon, Shield, Batteries) and the buildings whose figures read other colonies, the guild or the trade lanes (Shipyard, Nanoforge, Laboratory, Trade Portal, Factory, Arsenal, Sensors, Hypergate).

1.446.1 — 2026-10-12

Fixed

Guild research nodes now speak the language on screen (#7252)

Before, the guild research door printed every node's name, and every effect except the nine keystones', in English in every language. The guild log did the same when it named a node someone had started. Now every one of the 40 nodes has its own name and effect in Italian, Open Hands included, both on the cards and on the 3D board, and the guild log names nodes the same way. Every number in an Italian effect is read from the game's own values, so a balance change moves both languages together. One keystone line reads slightly differently in Italian: the reserve-charges node no longer spells out how many charges it adds, and gives the change as 12 → 14. English is unchanged. A node added by a newer server than your app still shows in English instead of as a blank or a code.

1.446.0 — 2026-10-12

New

A System view button on the colony screens (#7195) Getting from a colony to its own system on the Galaxy used to take several steps: leave the colony, open the Galaxy, find the system, then open its view. The Colony tab and the Base now each have a System view button that opens the Galaxy straight onto the system view of the colony you are looking at. On the Colony tab it sits beside the coordinates. On the Base it sits beside the title on desktop. On a phone it is a round button in the top-right corner of the map, next to the planet's traits, and it hides while you place a building. Back returns you to the colony screen. Suggested by a player for the season's system challenges.

1.445.3 — 2026-10-12

Changed

The Nightfall Protocol's counter moved from a Deep Sensor Grid 8 to a Deep Sensor Grid 20 (#7183). Before: one Grid at level 8 anywhere in the defender's empire, about 64k resources, cancelled the 10-billion Spectre capstone completely. Most active empires already had one, and no Nightfall attack had ever launched dark. Now the defender needs a Grid at level 20 anywhere in their empire, or a lower Grid that the Watch lifts to the same reading (for example a Grid 19 with the Watch at level 3, or a Grid 15 with the Watch at level 20). The counter is still empire-wide. The 10-minute floor is unchanged: an attack whose flight is ten minutes or shorter is never hidden, and a hidden attack still shows for at least its last ten minutes, or its last 15% of the flight if that is longer. The Nightfall sheet, the Incoming badge, the late-alarm lines and the Watch card now say level 20.

1.445.2 — 2026-10-12

Changed

• The building sheet's rows start coming from the server (#7196). The "Contribution now" and "Next level" rows of the Command Center, the Singularity Vault, the Recycler, the Silo, the Veil and the Relay, and the Energy row of every building, are now computed by the game server and only worded by your browser, in your language. Nothing moves and nothing is reworded: every row reads exactly as it did. The other buildings' rows follow in the next steps.

1.445.1 — 2026-10-12

Fixed

The phone header no longer runs off a small screen (#7241) On a phone, the top line of the HUD holds your commander card and the controls beside it: Back, the dispatch bell and the colony chip. With several colonies and Back up, those controls could be squeezed below their own size. The colony chip then ran past the right edge of a 320px phone, or under the resource strip on a phone turned sideways. A guest in a guild had it worst, and a medium-length colony name was worse than a long one.

Now the controls keep the room they need, and your commander name gives way first. The header stays inside its space on upright phones and on rotated phones from 568px wide up. On a rotated phone narrower than 620px, the colony name can disappear entirely when there is no room, leaving the chip's two marks. Tap the chip to see the full name in the colony list, and screen readers still read the whole name. On the map tabs, a tap on the GUEST badge now stays on the badge instead of opening the Power details.

Not included: windows that get the rotated layout but are narrower than about 530px, such as split screen, still run out of room.

1.445.0 — 2026-10-12

New

Open Hands: a guild research that lets a guildmate receive a third gift a day (#7188)

A player asked for a way to send guildmates in need more than the small favours the Help board and direct gifts allow. Until now, a commander could receive at most two direct gifts a day, from everyone combined. Logistics in the guild research tree now has a new node, Open Hands: tier 2, next to Open requests, unlocked by Aid capacity 3, for 3,000 Fund and 24 hours of research. Once the guild owns it, each member can receive three direct gifts a day instead of two. The Send dialog's count of gifts a guildmate can still take shows the raised number.

The size of one gift does not change: it is still at most 2% of the smaller of the two warehouses. Each giver can still send the same guildmate at most two gifts a day, so the third has to come from someone else. Recruits still cannot receive gifts. There is no tax on gifts. The node's name and effect appear in English in every language, like the rest of the Logistics branch.

Suggested by a player.

1.444.6 — 2026-10-12

Fixed

Signing out of the iPhone app now clears Apple's and Google's sign-in data from the phone (#7027)

Before, signing out of the game in the iPhone app left behind what Apple's and Google's sign-in kept on the phone. That was Apple's identity token, which carries the Apple account ID and email, the name Apple shared on the first sign-in, and Google's saved session, which the next "Continue with Google" restored without asking. On a shared phone, the next person inherited all of it. Signing out now deletes the Apple token and name from the phone, even if the app was restarted since the sign-in, and signs Google's library out, so the previous account is no longer restored silently.

This reaches players with the next iPhone app build: the app carries its own code, so phones keep the old behaviour until they update. The privacy page still says that signing out does not clear this data. It will be rewritten, naming the app version, once that build is on the App Store (#7248). The fix was checked by tests here, not yet on a device.

The privacy page also now says that signing out removes this device's notification address once our servers confirm the sign-out, rather than as soon as the request reaches them.

1.444.5 — 2026-10-12

Fixed

The Echelon panel speaks Italian (#7230) Before, a player on the Italian locale read the whole Echelon panel on the Leaderboard tab in English: the heading, the countdown to the weekly close, the rules line, the last verdict and its receipt, the board's counts, the purse preview, and the tier mark on the server board. Now every sentence of the panel is translated. Echelon, the tier names, Scrip and the Gate keep their English names, as in the glossary. Figures follow the reader's language, so an Italian reader sees 13.000 and 1.200.000. The purse preview keeps the cycle rule from #7181 in both languages: the cycle that started on 5 October shows the uncapped warehouse share and says the cap starts next cycle, and from 12 October the share is shown capped. In English, a tier with one racer now reads "1 commander" instead of "1 commanders", and counts and ranks above 999 are grouped ("1,234 commanders").

1.444.4 — 2026-10-12

Fixed

On a phone turned sideways, the resource strip now scrolls instead of running off the screen (#6984)

On a phone held sideways, the HUD shows the commander card and the resource strip on one line. That strip did not scroll. It grew to fit every chip and ran past the right edge of the screen, so anything past the edge could not be reached. The commander card beside it was squeezed to make room, and with every chip up it disappeared entirely.

Now the strip scrolls sideways under the thumb, as it does on an upright phone, and stays on the screen. The Build chip stays pinned at the strip's right edge here too. The commander header keeps its width up to half the bar. When it needs more, its name and colony chip shrink as they already do upright. Upright phones, tablets and desktop are unchanged.

Not fixed yet: on a sideways phone narrower than about 660 pixels, a guest at the longest name with several colonies, the Back button showing and a long colony name still has a header that overlaps the start of the strip by a few pixels (#7241).

1.444.3 — 2026-10-12

Changed

• The building sheet's figures now come from the server (#7196). The sheet that opens on a building — what locks it, the Command Center ladder with every level's price and build time, the next level's cap, the rebuild offer after a raid, the core commission and why it would be refused, what the emplacements punish, the accelerator's yield, the power switch — used to be worked out in the browser. Those figures now arrive from the game server, computed by the engine the orders run on, when the sheet opens and again whenever the building changes under it. Nothing moves and nothing is reworded. The "Contribution now" rows and the full progression table still come from the browser for now; they are the next step.

1.444.2 — 2026-10-12

Changed

The Help board now points to sending resources directly (#7188)

The Help board explained how to ask your guild for supplies and how to fill a guildmate's request, but it never mentioned the other way to share: the Send button on the Members roster, which sends resources straight to a guildmate without a request. Under the board's rule about its caps there is now one more line saying so. It adds that a gift uses the same daily hands as the board and pays no Scrip. An Open Members button under it takes you to the roster, on phones and on desktop. The line and the button only appear once your account is past its first two days, because a gift is refused before then. The caps themselves are unchanged.

Suggested by a player.

1.444.1 — 2026-10-11

Changed

• The build menu and the building ledger now come from the server too (#7196). (PR #7244) After the colony screen's numbers moved to the server, two blocks still worked things out in the browser: which rows the build menu offers (and which of them are padlocked, and why: a Tech Tree node, or a Command Center level), and the building ledger on the Overview that lines every colony's buildings up side by side. Both now read what the game server computed: the menu reads the same colony view as the rest of the screen, and the ledger reads a per-colony summary that says, for every building and site, the level, the cap and the Command Center level that would open it. Nothing moves and nothing is reworded. While the server's figures are a step behind what is on screen, the ledger keeps drawing its cells from the colony list it already has, so a build that just finished never shows the old level.

1.444.0 — 2026-10-11

New

The System Survey launches from the colony you choose (#7143)

On the Galaxy map, Survey this system always sent its Spy Probes from your home colony. When the home hangar ran out of probes, the survey refused, even if another colony held plenty, and choosing that colony in the header changed nothing.

Now a Launch from picker sits above the planet list. It lists every colony with at least one Spy Probe, and each entry starts with that colony's probe count. It opens on the colony selected in the header if that colony has a probe; otherwise it opens on your home colony, then on the colony with the most probes. The price, the refusal and each planet's travel time all use the hangar you picked, and the probes leave from there. Your ticked planets stay ticked when you switch colony. The 30-second refresh never changes your choice. If the colony you picked runs out of probes, the panel asks you to pick another.

Suggested by a player.

Send sits next to «All N» in the System Survey (#7144)

The Send button sat under the list of planets, so the most common sweep, ticking «All N» and sending, meant scrolling past every planet. Each group of planets now has its own Send right after its «All N» and «None» buttons, and the Send under the list stays for picking planets one by one. Every Send sends the same sweep. The result appears under the button you pressed. The line explaining why Send cannot be pressed, such as not enough probes or nothing ticked, is now shown above the list as well as below it.

Suggested by a player.

1.443.1 — 2026-10-11

Fixed

• The Muster counts the same people on both sides, so idle alt accounts no longer lift it (#7214) The Muster divided the commanders who struck by the commanders who were active the week before the window, and those were two different groups. A guest account made a week ahead and kept alive only by Marshal strikes was never on the roster but always counted as having turned up. A handful of them could lift the whole server's Muster tier.

  Now every striker is placed when the window opens:

  • On the roster: active in the week before (a build, a ship order, research or a mission). Guest accounts count here as much as anyone.
  • Returning: quiet that week, but signed in with Google or Apple. They still count, and they are added to both sides of the count. The page reads "16 of 46", where it used to read "16 of 40", and it can no longer show more strikers than the total ("12 of 10").
  • Not counted: quiet guest accounts, a guest that links Google during the window, AI-badge empires and accounts under a week old.

  Everyone with a valid strike is still paid the tier the server reaches.

  A small server is now measured against at least five commanders, so a single account can no longer reach the full Muster on an empty roster.

  This makes alt-farming the Muster much more expensive but doesn't make it impossible. Lifting a 40-commander server to the top tier would now take about 35 accounts, each kept active every week or signed in with its own Google or Apple account, where 14 idle guests used to be enough.

1.443.0 — 2026-10-11

New

Supply lanes have one home: Missions › Lanes (#7165) Before, the list of standing lanes appeared under Missions only while a lane was running. The only way to open a lane was a switch inside a Transport or Pickup order, even though a lane flies no ship. The Colony tab's SUPPLY LANES card showed one colony's lanes and pointed nowhere else.

• Missions now always has a Lanes heading. It lists every lane in the empire, each with its Pause, Resume and Delete buttons. Below the list is a lane form of its own: pick FROM and TO among your colonies, and open the lane there, with no fleet order involved. The switch on Transport and Pickup orders still works as before.
• The Colony tab's SUPPLY LANES card links there ("All your lanes: Missions › Lanes"). So does a built Trade Portal's building sheet, the building a lane leaves from.
• Either link lands on the Lanes heading, with the form set to start from the colony you were looking at. On a phone it opens the "In flight" view, where the heading is.
• The card's help line no longer claims to list every running lane. It says it lists this colony's lanes, and that all of them are under Missions › Lanes.

Suggested by a player.

1.442.3 — 2026-10-11

Changed

• The colony screen's numbers now come from the server (#7196). (PR #7232) Until now the browser worked out every figure on the Colony tab and on the Base itself: what the next level of a building costs, how long it takes, whether you can afford it, whether the queue is full, how full the storage is, how the energy balances. Those figures now arrive from the game server, computed by the same engine that runs your orders, so what the screen calls "ready" is what the server will accept. Nothing moves and nothing is reworded: the layout and every sentence are the ones you know. Between two reads the screen keeps counting your resources up at the rate the server gave it, and the moment a build finishes it reads the colony again, so the row never shows the old level's price. The build menu, the placement banner and the hover card read the same figures.

1.442.2 — 2026-10-11

Fixed

The Echelon cycle already running pays under the rule it started with (#7181)

Before: the cap on the Echelon resource purse (a winner is paid on the smaller of their own Warehouse capacity and the median capacity of the largest active empires, v1.434.15) took effect in the middle of the cycle that started on Monday 5 October. Players who had already spent to climb that week would have been paid a smaller purse than the one the panel showed when they started, and the panel's preview already described the cap.

Now: the cycle that started on 5 October is paid on your own Warehouse capacity, as it was when the cycle began, and the panel's preview says so. It adds that the cap starts with the next cycle. The cap applies from the cycle that starts on Monday 12 October. No Echelon purse was ever paid under the cap: no cycle closed between the change going live and this fix.

From now on, a change to what a weekly prize pays takes effect at the start of the next cycle, never during a cycle that is already running.

Suggested by a player.

1.442.1 — 2026-10-11

Fixed

Chat translation no longer publishes a sentence you did not write (#7209, #7150) Before this release, the automatic translator could put an invented reply in the channel under your name, for example a threat or an accusation. It happened when the message was reversed, in leetspeak, keyboard noise, half broken by punctuation, or an English line it took for a foreign one, and the made-up reply was short. Now a translation is shown only when a second check confirms it says what you wrote, and it never adds a coordinate or a link you did not type. Lines that cannot be read as a language, or English lines in disguise, are not sent to the translator at all. Whenever there is any doubt, your message is posted exactly as you typed it. A translated line can now take a moment longer to post, because it is checked before it is stored.

1.442.0 — 2026-10-11

New

• VIP 10 and above now build more ships at once in every colony (#7206). Before, every shipyard ran at most three orders in parallel: one from the Shipyard, one from the Autonomous Factory, one from the Orbital Arsenal. Now supporters get extra lines on top of those, in every colony: one more at VIP 10, two at VIP 12, three at VIP 15, four at VIP 18 and five at VIP 20 — up to eight orders building side by side. The lines still need a Shipyard, and nothing changes below VIP 10. The slot counter in the top bar, the Overview and the Shipyard show the new total, and the Shop's VIP ladder names the perk at each of those levels.

Fixed

• The Factory and Arsenal sheets count the VIP lines too. Their "Shipyard queue slots" row shows the colony's real total, and for a VIP player the advice reads "build it for one more parallel ship order (N in all)" instead of promising a fixed second or third slot. The Arsenal core offer no longer says the yard has "at most three" slots; it says what is true: a core adds no slot.

1.441.1 — 2026-10-11

Changed

The Crucible target lock now has a Fight button above your roster (#7180). Until now the only Fight button sat under the roster panel, which lists one row per ship class. With a full hangar it was below the fold even on a 1440p screen, so fighting again with the same roster meant scrolling past every row first. A compact Fight button now sits at the head of the roster panel, next to its tonnage cap, with the same label and attempt count. The button at the bottom stays for players who edit the roster first. Both buttons are one action: they open and close together, and either press spends the same single attempt. On a desktop the new button is on screen as soon as the sheet opens. On a phone it sits above the roster rows, but you may still need a short scroll to reach it.

Suggested by a player.

1.441.0 — 2026-10-11

New

• The Marshal pays the whole server: richer prizes and the Muster (#7199) An ordinary Marshal night, where the server breaks levels 1 and 2, used to pay 25 Insignia and a 5-minute shard, and nothing else in any bracket. Every level now pays something of value from level 1. Per window, an ordinary night is now worth about:

  | Bracket | Before | Now | | --- | --- | --- | | Patrol | 0 | 100,000 (4 Dark Matter) | | Squadron | 0 | 200,000 | | Fleet | 0 | 425,000 | | Armada, Warhost, Apex | 0 | 2,150,000 |

  On top of that comes 115 Insignia (was 25) and 1 hour of Chronoshards (was 5 minutes), in every bracket.

  The Muster is new, and it pays for turnout, not damage. When 20%, 40% or 60% of the commanders active in the week before the window make a valid strike, every player with a valid strike is paid a Muster tier. The top tier pays 100K crates in Squadron and Fleet, and 1M crates in the big brackets. A small empire moves the Muster exactly as much as the strongest one, so bring everyone. The Marshal page shows the live count ("6 of 20 commanders struck validly · 2 more for Muster 2"), and the result in your diary names the tier the server reached.

  Accounts younger than a week when the window opens, and empires with the AI badge, don't count toward the Muster. They are still paid it when they strike.

Changed

• Bigger podiums. Every podium and guild place pays twice the Insignia. In the big brackets, first and second place now take 1M crates (first was 500K); Fleet's winner takes 500K crates, and Squadron's winner a 100K set. Patrol is paid in Dark Matter, not crates.
• The rename token now costs 1,000 Insignia (was 300). It is there to stop impersonation, priced at about one best week of Insignia, and the Insignia you earn have roughly tripled.
• The rules say what a small strike or a hunt now costs: a strike under a quarter of your military, or a hunt on a striker's colony, loses the level prizes, the Muster and the guild place.

1.440.1 — 2026-10-11

Changed

Calling off a guild raid no longer keeps your fleet away for the whole trip home (#7157) Before: a leader who called off a raid they were still flying alone got their fleet back only after the full return flight, however early they called it off. A cancel a few seconds in kept the ships away for over a minute, with nothing on screen saying why, and it read as a lockout. Now: a called-off raid flies home the way a recalled attack does, for the share of the flight it had covered. Gathering time added to the raid does not shorten that trip: a cancel never comes home faster than an attack recalled after the same time in the air. Once a guildmate has joined, the raid still cannot be called off.

Fixed

The raid composer says when your fleet is back, and why the battle line is empty (#7157) Before: after calling off a raid, the composer showed "No ships" and advised building more, and the receipt said only that the fleet was flying home. The battle-line editor was simply missing until you picked ships. Now: the confirmation quotes the real time home before you tap, the receipt counts down to the landing, and an empty hangar (or one holding no warships) says when your raid fleet lands there. With no ships picked, the composer says the battle line is drawn from the ships you send.

Reported by a player.

1.440.0 — 2026-10-11

Added

• The colony screen is now served as one document: GET /api/v2/colonies and GET /api/v2/colonies/{id} (#7196) (PR #7220). Nothing a player sees changes yet: the web screen switches to it in the next release. Until now the browser worked out every number on the colony screen itself, from the game's rule tables shipped inside the page: what the next level of a building costs, how long it takes, whether the warehouse can pay for it, how full the storage is, how the energy balances. The new route answers all of that at once, computed by the same engine that runs the orders: each resource with its stock, its rate per hour and its storage ceiling; the energy balance and the accelerator; every building with its level, its cap at the current Command Center, whether and why it is locked, and the dry run of its next level: the cost, the seconds, and the same verdict the build order would get (ready, not enough resources, queue full, Command Center too low…), so the screen can never promise a build the server then refuses; the build and ship queues with their limits; the docked hulls; the base layout. A player sees only their own colonies: another player's id answers 404.

1.439.3 — 2026-10-11

Fixed

A failed changelog push now says which command failed and why, and the push is retried once (#7101) When the release script could not commit or push the changelog branch, it printed one generic error for both cases, and git's own message was thrown away, so the cause could not be found afterwards. Now the error says whether the commit or the push failed, shows the last lines git printed, and gives the steps to finish the release by hand. Since the tag and the Release are already public at that point, a failed push is retried once after a short pause before the script gives up.

This release changes only the release tooling. No game code changed, so the deploy step had nothing to ship and ended DEPLOY: SALTATO (production already matched).

1.439.2 — 2026-10-11

Fixed

A relic held by another guild no longer looks unowned in the system view (#7151) The galaxy map's Relics panel already named every relic's holder. One step closer in, it did not. A relic's colony belongs to no guild, so the system view showed a held relic by its name alone. The planet sheet said "Enemy colony" and gave a note that could mean either unclaimed or held. To anyone outside the holder guild, a held relic read as belonging to nobody. Now the relic's caption in the system view carries the holder's tag, like a player's caption carries theirs. The planet sheet adds one line: "Held by guild [TAG]", or "Unclaimed: no guild holds it". Members of the holder guild still see "Relic held by your guild", as before.

Reported by a player.

1.439.1 — 2026-10-11

Fixed

A message spelled out letter by letter no longer comes back as a sentence nobody wrote (#7150)

Before: a line typed with its letters spelled out between dots, such as a question written as W.h.a.t d.o.e.s i.t s.a.y.?, went to the chat translator. The translator could not read it and invented a reply, often a threat, and that reply was posted under the player's name as if they had written it. Each repeat gave a different invented line.

Now: a message whose letters are spelled out one at a time, with dots, spaces, hyphens, underscores or apostrophes between them, or whose words are mostly broken up by punctuation, is never sent to the translator. It is posted exactly as typed. Ordinary chat in other languages is still translated as before. A few short lines (dotted three-letter acronyms, some Russian abbreviations next to one-letter words, and short lines typed with no space after a comma or full stop) now stay as typed instead of being translated.

This closes the spelled-out trick only. The translator can still invent a short reply for other kinds of input, and that is tracked in #7209. Messages already posted with invented text are not changed by this release.

Reported by a player.

1.439.0 — 2026-10-11

Added

• The API now publishes its contract: GET /api/v2/openapi.json, and the first route built on it, GET /api/v2/me (#7196) (PR #7204). Nothing a player sees changes yet. Until now the API's 321 routes had no written contract: a route's shape was whatever its code happened to accept, and the only client that knew it was the web app. The new @novarchy/contract package declares each BFF route once, with its request and response schemas, and derives from that declaration the OpenAPI 3.1 document, the validation the server applies, and the types the clients use. GET /api/v2/me serves the player card exactly as GET /auth/me does, parsed through the contract on its way out so nothing the contract does not name can leave. The document is committed with the code and a test fails when the two disagree. This is step 1 of the plan in DESIGN-BFF.md: game logic server side, clients that render state and send intents, and a phone client rebuilt as a native app in its own repository.

Changed

• The client family a player last played from is now recorded (#7196). The game has received the web / ios / android header on every request for weeks and only used it for the guest throttle; it is now kept per player on the same presence write as the build stamp, so the move from the phone WebView to a native app can be measured without reading web-server logs. A request that does not send the header leaves the value as it was.

1.438.5 — 2026-10-11

Fixed

Opening a battle report in Activity no longer puts its Hide button next to it, and hiding a row leaves a way back where the row was (#7182). Before, every Activity card had its Hide mark, a ✕, pinned to the top-right corner. On a card whose report was opened, that ✕ looked like the button that closes the report. One tap hid the whole row on that device, and it stayed hidden after a reload. The only way to get it back was «Show N hidden» at the bottom of the column, so a player who closed two battle reports this way saw both rows disappear from the feed. Now the ✕ is not drawn while a card's report is open: the card's own Close is the only control there, and the ✕ comes back once the report is closed. Hiding a row in the feed now leaves a one-line «Hidden · Undo» in its place until the page is reloaded, and Undo brings the row back. The exceptions are the Kept chip and the «Show N hidden» view, which work as before. Rows hidden before this change are still behind «Show N hidden», as before. Reported by a player.

1.438.4 — 2026-10-11

Fixed

The attack composer now says whether a spy report read the target's garrisons (#7152) Below the deepest espionage tier a spy report does not read the fleets garrisoned on a colony, and the attack composer's spy line listed only the colony's own hangar, with no word about garrisons at any tier: a garrisoned colony read as lightly defended, and a player lost two fleets that way. The line now ends with one of three notes: that garrisons were not read and a fresh probe at the deepest espionage tier reads them, that nothing else will fight for the colony as of that probe, or how many garrisoned hulls the probe found. The unread note is the same whether or not a garrison is there, so it reveals nothing the tier does not sell, and it is never shown on the Pirate Beacon, which cannot be garrisoned. The fleet matchup card now also says garrisons are not drawn there. In English and Italian.

1.438.3 — 2026-10-11

Changed

The Shipyard's spinning ship can be paused, and it costs far less while it turns (#7162)

Before: the 3D model of the selected hull in the Shipyard spun nonstop, redrawing on every refresh of the display, so 144 times a second on a 144 Hz monitor. It also kept turning the previous hull behind the picture while the next one loaded. The only way to stop it was the operating system's reduce-motion setting.

Now:

• A small pause button sits in the top-left corner of the model, on desktop and on the phone. It pauses or resumes the rotation, and this device remembers the choice, so every hull you open after that stays still until you resume.
• Dragging the ship still stops only the ship you are dragging.
• While it spins, the model redraws about 30 times a second instead of at the display's full rate.
• It no longer redraws at all while hidden during a change of hull.
• With the system's reduce-motion on, the model stays still and the button is not shown.

Reported by a player. The CPU saving on real hardware was not measured; what is measured is the number of frames drawn.

1.438.2 — 2026-10-11

Fixed

Spy probes sent outside the Missions tab now play the spy launch sound (#7187) Since v1.437.0 a spy mission sent from the Missions tab played its own launch sound. The other three ways to send a probe played nothing: the quick-spy dialog behind every Spy button, the Spy action on the Galaxy's Targets board, and the System Survey sweep. All three now play the same sound once the server accepts the launch. A burst counts as one action: a sweep or a Targets-board burst plays the sound once, on the first probe that leaves, not once per probe. A refused launch, or one whose outcome is unknown, stays silent. Only the player who sends the probe hears it.

1.438.1 — 2026-10-11

Fixed

The full progression table prices the core that levels 16 to 30 need (#7156)

The Hypergate's level-by-level table quoted levels 16 to 30 at their construction price alone. Each of those levels also needs a paid core, in resources and Dark Matter, and that price only appeared in the core panel once the gate stood at level 15, so a player planning the climb could not see what it really costs. The table now has a Core fee column with each level's core price beside the unchanged construction cost, and the note above it names the levels the column covers. The Factory and Arsenal tables had the same gap and get the same column.

Reported by a player.

1.438.0 — 2026-10-11

New

Send resources straight to a guildmate (#7185) Before, guildmates could only move resources to each other through the help board. Someone had to post a request first, and the rest of the guild could only answer it. Now every Member's and Officer's row on the guild roster has a Send button; on phones it is the first entry in the ⋯ sheet. It opens a dialog that shows, before you type anything:

• the most you can send of each resource, with a max shortcut;
• how many hands you have left today, in total and toward that commander;
• how many gifts they can still take today;
• anything that would refuse the gift.

The resources leave your home colony and land on theirs. They see "name sent you a gift" in their activity feed, and the guild log records it.

A gift follows the help board's own limits. It uses one of your daily hands, the same allowance that answering requests uses, including the limit of two a day toward the same commander. Both accounts must be at least two days old. One gift is never bigger than what the recipient could ask for on the board, or than what you could ask for yourself. It is also cut to the room left in their warehouse, so nothing is lost to a full store.

Two new limits come with it:

• each commander can receive at most two gifts a day, from everyone combined;
• Recruits cannot receive gifts until they are promoted, just as they cannot ask the board for supplies.

Gifts earn no Scrip and no guild contribution; answering a request still earns both.

Not included: an app that has not been updated since this release shows a received gift as an answered help-board request, until the update arrives. The AI empires can receive gifts and remember them as gifts, but they do not send them.

1.437.1 — 2026-10-11

Changed

• Daily contracts and the guild Store in Italian (#6977). Before, players with Italian on saw these in English:
  • the contracts board, with the Daily Wrap, the Weekly Chest and the Charter;
  • the Season lane switch;
  • the guild Store's item names and descriptions;
  • the contract and Store rows in the Activity tab and the guild log.

  Now they read Italian. In English, every word and figure reads as before. The Store still shows the server's own wording in English.

  The beta notice no longer lists Activity rows: nothing there prints English any more. Part of the Guild is still in English, including most of the guild chat and many guild research effects.

1.437.0 — 2026-10-11

New

Spy probes and attacks leave with a sound of their own (#6905) Before, every launch from the Missions tab played the same engine note, except colonisation. Now a spy mission (sabotage included) leaves with a short, soft probe beep, and an attack (raze and dock raids included) leaves with a low, rising hum. Colonisation keeps its own sound, and every other order keeps the engine note. Only the player who sends the fleet hears it; the defender's incoming-attack alarm is unchanged. Spy missions sent from outside the Missions tab (the quick-spy dialog, the Galaxy list's Spy action, survey probes) still launch without a sound, as before.

1.436.1 — 2026-10-10

Changed

• The AI empires now plan by rules, not with a language model (#7174). The 75 AI empires, the ones with the AI badge, still build, raid, attack, defend, spy and intercept. Until today, each one's daily strategic plan came from a language model: what to invest in, whether to attack or hold, and which colonies to aim at. Now that plan is worked out from what the empire already knows:
  • its temperament sets how aggressive it is;
  • its archetype sets what it builds toward;
  • a capital hit twice at home turns to defence;
  • an empire at war never sits on the defensive;
  • aggressive empires name as targets the colonies of the players who wronged them most, within their attack band and never a guildmate.

  Plans are now revised every 6 hours instead of once a day, and an attack, a spy report or a war gets a fresh plan within the hour. Some empires lean a little more aggressive than under the old plans: 12 of 75 by temperament, against 7 before. Not changed: the empires stay silent in chat, as they have been since September.

1.436.0 — 2026-10-10

New

Synthetic empires will settle the Event Sector, from the first season that opens after December 4 (#4316) Until now the AI empires fought over the Event Sector's six contested worlds but never founded a colony there. On 2026-10-10, 125 human colonies and no AI colonies stood in the sector.

From each AI empire's own switch-over between December 3 and 4, an AI empire that is at its core colony limit may found one world in the Event Sector per season. The rules:

• It starts between one and seven days into the season, at a different moment for each empire, and only during that empire's active hours.
• AI empires together never hold more than half as many Event worlds as human players hold there, counted fresh each time an empire decides.
• They never hold more than four in one system.
• They settle near the contested worlds, preferring the least crowded system.
• An empire that has no colony ship for the purpose may build one, and never more than one per season.

On the current fortnightly calendar that is the season of December 10. Nothing changes in the season running now or in any season that opens before the switch-over. Chat is unchanged.

1.435.6 — 2026-10-10

Also

• Operator runbook: the League podium roll-forward now checks the restore before it swaps (#7171). Nothing changes in the game. After a rollback past the weekly League podium, rolling forward again puts the table of paid weeks back before the new build starts. The step now names the saved file once and stops if it is missing, and it refuses the swap unless the restored rows match the file and are not behind the guild log's latest podium — so a wrong or older file can no longer let a week be paid twice. It also checks the api is answering before it lifts the deploy freeze. The runbook's own test now only accepts a real, complete rebuild before a swap, and the comments describing the League table's window were corrected.

1.435.5 — 2026-10-10

Fixed

The advisor no longer offers a new building you cannot afford (#2141)

Before, a building you had not built yet stayed lit in the Colony tab's advisor even when you were short of resources, and the red price was the only warning. Tapping its button opened the base map. You picked a spot there, and only then was the build refused. Now that row is dimmed like an upgrade you cannot afford, its button is disabled, and the card says "Not enough resources." in plain text, so the reason shows on a phone too. When the build queue is also full, the card says "Queue full." instead, because waiting clears that one first. The Command Center line above the suggestions now names a short purse the same way.

The button for a new building is now a pill that reads "Place", in place of a round plus sign beside the upgrade arrow. It no longer looks like a build button that does nothing. The upgrade arrow is unchanged.

Reported by players.

1.435.4 — 2026-10-10

Changed

A wanted player is told the Veil hides their address, never their place on the Wanted board (#7124)

Before, nothing on screen said what the Eschaton Veil does to a bounty. A wanted player saw their own row on the Wanted board with their coordinates printed, whatever their Veil level, and could reasonably expect a higher Veil to take them off the list. It never does. The board lists every wanted player. The Veil withholds only the address, and only while every one of their colonies is under it, from a hunter whose Sensors cannot resolve it. Your own row always shows your coordinates, because they are yours to see, so it could not show you what a hunter sees.

Now your own row on the Wanted board carries a line under it saying exactly that. The Veil can hide your address from a hunter, depending on their Sensors against it, but only while every one of your colonies is under it. It never takes you off the board, and the coordinates on your row are your view, which a hunter may not share. The bounty chip's tooltip in the header, and the sheet that opens from it on a phone, say the same in one sentence after the decay rate. Only the wanted player sees these lines, in English and Italian. Nothing about the Veil or the bounty itself changed.

Reported by a player.

1.435.3 — 2026-10-10

Also

• Operator runbook: rolling back past the weekly League podium now says how to roll forward safely (#7160). Nothing changes in the game. The rollback instructions for the League podium table (block CP of the deploy ladder) told the operator to restart with a plain up -d, which would have restarted the old build, and said in one place that the table's migration is never applied by hand while another step applied it. The roll-forward is now its own step: rebuild every image, recreate the table and put its saved rows back, then swap to the new images — so a rolled-back week can never be paid twice. The ladder's own test now recognises that build-then-swap shape and still flags a swap that builds nothing.

1.435.2 — 2026-10-10

Changed

A won war pays every member who was there from the start, not only the ones who scored (#1484)

Until now, a guild war's tribute was split by war points alone. A member who scored nothing, including the leader who declared the war, received nothing and was never even listed for a share.

From wars whose scoring week starts on Monday 2026-10-19, 10% of the tribute is split equally among every member who was in the guild when the war week began, scorers included. The other 90% is split by war points as before. Nobody who was there from the start can receive less than a member who scored nothing, and more points always pay more. If nobody scored, the tribute is split equally. A member who joined during the war week gets no part of the 10%; if they scored, they are paid their points share. Being demoted to Recruit does not reset when a member joined. Wars that score in an earlier week, including every war already running when this shipped, keep the old split.

The War Room states the rule for the war in front of it, and the effort tab now lists the members who scored nothing with their share. A winner who scored nothing finds a line on their verdict explaining where their share came from. Each resource is still rounded down per member, and a full Warehouse still refuses what does not fit, as before. A surrendered war is paid by the same split its War Room showed.

Not included: the War Room still does not say which colony the tribute lands in. That part of #1484 stays open.

1.435.1 — 2026-10-10

Fixed

• A news-digest test no longer fails just because no internal-only release shipped recently (#7161). The test required one of the newest 150 changelog versions to have no player-facing section. At v1.435.0 none of them did, so main went red and every queued ship was refused. That case is now tested on a fixed sample changelog. The check on the real changelog keeps only the rule that applies to every version: a version with no player-facing section has no headline. Nothing changes in the game.

1.435.0 — 2026-10-10

New

• The Guild League now pays the guilds at the top of its table, every week (#7146). Until now a League fixture paid each member a small purse and nothing else: where your guild stood on the table won you nothing, and the panel never said how much anything paid. Now, every Monday, the top of each division's table — the same last-eight-fixtures table the panel shows — is paid into the Guild Fund: 150 for 1st, 90 for 2nd, 50 for 3rd (lower divisions a little less). Only guilds that actually played that week collect, and the podium never covers more than half the table. The first podium is paid on Monday 12 October, for this week.
• Winning a League fixture now pays every member who was in the guild when it started: 100M of each resource, 200 Dark Matter and 3 random Star Forge components (#7146). The resources arrive as three sealed 100M crates you open into the colony you choose, so a full warehouse never swallows the prize. Members who joined after the fixture began, and synthetic empires, keep the usual purse but not this prize. Your result in the Activity feed lists exactly what you received.
• The League panel shows its prizes (#7146). A new Prizes block lists what every fixture pays each member, what the weekly podium pays, and the winner's prize — with the podium medals, the crate, the Dark Matter and the components drawn — plus last week's podium, a line when your guild was on it, and a ★ next to those guilds on the table. Your guild log records the podium, and members get a feed row for it.

Changed

• The League table now counts eight finished fixtures, not seven (#7146). The fixture in progress was never on the table, so "last 8 fixtures" showed seven. The table you see on Monday is now exactly the one the podium pays.

1.434.15 — 2026-10-10

Changed

The Echelon weekly purse stops paying more for a bigger Warehouse (#6679)

Before: the resources the Echelon paid at the weekly close were a share of your own total Warehouse capacity, in every tier. So the size of your warehouses decided the payout, and a very large empire could take more from the Echelon than from the War-week podium the Echelon is sized under.

Now: the share is taken on the smaller of your own capacity and the server reference. That is the same reference the War-week prize has used since #5767: the median capacity of the largest empires being played. Below the reference nothing changes. Above it, a bigger Warehouse pays nothing extra. The Echelon panel's purse preview now states the cap next to the share.

1.434.14 — 2026-10-10

Fixed

AI empires that hold the Lance or the Vanguard stop over-sending against a Shield Bubble (#4964)

Before: an AI empire whose own Doctrines slow a Bubble's regeneration (Cascade, the Lance at 15) or skip it in the opening round (Shock, the Vanguard at 10) planned its attack as if it had neither, and sent more ships than the battle needed. Now it counts its own keystones, and the defender's Unyielding (the Mantle at 15), which cancels Cascade. On a bench of 468 matchups under a Bubble, every attack it launches still has enough ships to win, none sends more than before, and 301 send fewer. Strikes it used to call off can now go ahead, and none that it called off before now goes without a reason. The overkill spillover and formations in the same forecast stay open on #4964.

This takes effect gradually from 30 November: each AI empire switches over on its own night, between 30 November and 2 December.

1.434.13 — 2026-10-10

Fixed

AI empires stop sending too few ships against a lone Titan or a few heavy, well-researched hulls (#7132)

Before: against a colony held by a lone Titan, three Battleships or four Reavers with high military research, AI empires sometimes launched a fleet too small to win and lost it. Their battle forecast left out that a ship's shields take part of every salvo, round after round, and it read the defender's research as one flat bonus, below what the battle gives it. Now they price both, so they send enough to win there, and elsewhere a little more than before, never less. When the fleet they gathered still is not enough, they call the strike off and say why, instead of waiting in silence until the plan lapses. On a bench of 1368 matchups, the launches that fell short go from 35 to 6. The 6 left are all a lone Titan with no Bubble up, at worst 94% of the ships the battle needs; #7132 tracks them.

This takes effect gradually from 26 November: each AI empire switches over on its own night, between 26 and 28 November.

1.434.12 — 2026-10-10

Changed

Sixteen more numbers the AI empires plan with are now checked against the game's own rules (#4319)

The synthetic empires keep their own copies of engine numbers: how many empires can join a raid and how strong a raid may be against its target, the relay and stargate travel discounts, the dark matter rate and the warp jump's speed, floor and price, the planetary cannon's damage and cap, the destroyer's per-round hull cap, the price step of each extra copy of a building, and how long a sighted fleet can still be in flight. Until now most copies were checked only against a figure typed again in a test, and four were not checked at all, so a retune of the game left the empires planning on the old number without any test noticing. Each one is now read from the engine and compared, and the check fails if one side changes alone. All sixteen agree today, so nothing in the game changes.

Part of #4319: the other copies the survey lists, and the sweep that would catch a new unchecked copy, are still open.

1.434.11 — 2026-10-10

Fixed

A colony that cannot launch a probe no longer stops the Targets board's whole spy sweep (#7123)

Since the sweep started sending each probe from your nearest colony that has one, a refusal about a single colony stopped the whole sweep. That refusal could be a hangar that was emptier than the board knew, or a colony in the middle of a fold. Probes planned from your other colonies stayed home. Every untried target was cleared from your picks, so you had to pick them all again.

Now that refusal stops only the launches from that colony. Your other colonies keep sending their probes. The targets the refused colony could not launch stay picked, so you can send them again with one tap. The result says how many were "not sent", and says "the sweep stopped there" only when the sweep really stopped. The per-player request limit still stops the whole sweep, as before.

Not included: the board still cannot tell in advance that one of your colonies is folding, so it can still choose a folding colony to launch from. That colony's first launch is refused and its other launches are skipped. Your other colonies are no longer affected.

1.434.10 — 2026-10-10

Fixed

A wanted player is now told how fast their own bounty fades (#2606) A bounty has always shed 15% of itself every day and dropped off the wanted board below grade 1, but only hunters were told. The player carrying it read "it decays on its own if you stop", with no rate and no time. Now the bounty chip's tooltip, its phone sheet and Power details say "It sheds 15% a day on its own: off the wanted board around (day and time) if you stop, sooner if someone collects a share." Your own row on the Wanted board shows "off the board around (day and time)" on a line of its own. The time is in your local clock. When the game cannot name the time, these surfaces state the daily rate alone. All four surfaces are now translated into Italian too. Reported by a player.

1.434.9 — 2026-10-10

Fixed

A won war now tells each winner what they got, where it landed, and what a full Warehouse refused (#1484)

Before, the war verdict card showed every member of both guilds the same thing: who won and the whole tribute pot. A winner's share was paid silently into their mother colony and never recorded anywhere. If that colony's storage was full, the part that did not fit was destroyed without any notice. Winners reasonably concluded that they had received nothing.

Now, on wars that end from this release onwards, each winner who was paid sees one more line on their own verdict card: their share as it actually landed, after the storage cap, and the name and coordinates of the colony that received it. When the cap refused part of the share, a second line says how much was lost and of which resources. Nothing changes for the losing side, for a draw, for a winner whose share was zero, or for any verdict written before this release: those cards read exactly as before. Older wars cannot be backfilled, because the split was never stored.

Not included yet: the overflow is still lost (this release records and shows it but does not change the cap), whether a declarer who scored no war points should be paid at all is still an open design question, and the War Room does not yet say before the verdict where the tribute will land.

Reported by a player.

1.434.8 — 2026-10-10

Changed

On a wide screen, the Campaign tab's boss rematch now sits right under your last engagement (#3834)

Before: when the Campaign tab was wide enough for two columns, the boss rematch was at the bottom of the right column, below the list of your attempts at the level.

Now: the right column runs from the auto battle summary to your last engagement, then the boss rematch, then the attempts list. You hit a wall, and the next thing you see is the one thing you can still do there.

The rematch still uses its quieter button, so it never looks like Engage. Tapping it still opens the engagement sheet first, and your weekly ticket is spent only when you confirm there. Phones, tablets and narrower windows keep the order they had.

1.434.7 — 2026-10-10

Fixed

The Targets board's spy sweep sends each probe from your nearest colony that has one (#2482)

When you picked targets on the Targets board and pressed Spy, every probe left from your mother colony. The board's list spans sectors, so a target two systems from one of your outposts got a probe crossing the galaxy from home.

Now each target takes its probe from the nearest of your colonies that still has a Spy Probe. "Nearest" is measured by flight time, counting that colony's Relay and Observatory. Targets are served in the order you sorted the list. When a colony runs out, the next target goes to the next nearest colony. A target no colony can pay for stays picked, as before.

The confirm dialog shows, for each target, which colony the probe leaves from and the estimated flight. It also lists how many probes leave from each colony. The picking bar counts the probes in all your hangars. The System Survey still launches from your mother colony.

Not included:

• The flight estimate leaves out the own-gate express and guild bonuses. Both can only make the real flight shorter.
• A colony being folded through a hypergate can still be chosen as a launch point. Its refusal stops the sweep.
• If one colony turns out to be short of probes, the sweep stops there, including probes planned from other colonies.

Suggested by a player.

1.434.6 — 2026-10-10

Changed

The AI empires' copy of the production rates is now checked against the game's own (#807) Nothing changes in play: the AI empires and the game already agree on what every Extractor, Refinery, Compute Node and Stellar Forge produces. Before, nothing would have noticed if they stopped agreeing. The empires plan their income from their own copy of those rates, and the only check compared that copy with itself, or compared the game's rates by their total alone. A re-balance that moved output from one resource to another while keeping the total, the shape of the August faucet re-cut, would have left the empires planning on the old figures with every test passing. Now a test reads every producer's rate straight from the game's building catalogue and fails the moment either copy differs, building by building and resource by resource.

Part of #807. Still open there: re-deriving the agents' trade price sheet once stocks converge (an owner decision), and the near-coincidence between the data inflow share and the 0.15 scarcity threshold.

1.434.5 — 2026-10-10

Fixed

Synthetic empires no longer pull a market offer that is still selling (#3991) Since partial fills, a buyer can take a slice of an offer. An agent used to withdraw any offer of its own that was more than a day old, counted from when it was posted, even when buyers were taking slices of it. Players saw an agent pull a listing that was mostly sold. Now an agent counts from the last time it saw the offer sell some, so an offer that is still selling stays on the board. An offer that sold some and then found no buyer for a day is still withdrawn. An offer that never sold is withdrawn exactly as before.

Part of #3991. Still open: an agent does not yet remember a partial sale of its own offer until the offer is sold out, its market lines between two market reads can quote the amount before the last slice, agents do not take slices themselves, and the price precision of very small offers is still with the design desk.

1.434.4 — 2026-10-10

Changed

The rules behind the War Room's figures can now be read on a phone (#3353). The War Room's scoreboard shows what losing costs, what winning pays, your Gate and the war countdown. The rule behind each of those figures used to appear only when a mouse hovered over it. A phone never showed it, and neither did a keyboard. The rules covered the tribute your guild pays from its home colonies if it loses, why the room does not show what the enemy would pay, where the bounty doubles and which strikes are free, and that a surrender can end a war before the countdown runs out. A fold now sits under the scoreboard, "The rules behind these figures". It opens with a tap, a click or the keyboard, and it holds one paragraph for each figure on screen, in the same order. The League's Match room uses the same fold for its purse rule and the Gate. In Diplomacy, the note that a war can end early is printed once under the list. It is no longer a hover on every war row.

This covers the War Room and Diplomacy only. Other screens still keep some decision-critical rules in hovers, and #3353 stays open for them.

1.434.3 — 2026-10-10

Fixed

The Italian resource bar calls spare energy a surplus (#7102)

In the Italian interface, the energy cell of the resource bar used an accounting phrase, the one for a balance in the black, whenever production exceeded consumption. It now reads "in surplus", the same word as in English, and it still fits the narrow row beside the fill bar. Other languages are unchanged.

Reported by a player.

1.434.2 — 2026-10-10

Changed

• Crucible: Last Stand no longer protects only the defender (#7103). Before, in a Crucible duel or spar, the defending snapshot got the Aegis 15 Last Stand cap (no ship group losing more than a quarter of its hull per round). The challenger, held to the same tonnage cap and often holding the same rung, got nothing. At the top of the ladder that kept whoever already held a slot in it, whatever roster the challenger brought. Now a duel and a spar hold no position, so Last Stand is off for both sides there, and so is Interlock, which lends it. At equal tech and the Commodore's 22,000 PP cap, a challenger now needs 0.75×–1.36× of the defence's PP to win against top-of-ladder defences, down from 0.83×–1.59×. Last Stand still works everywhere else, and the rung's description now says it does not apply in the Crucible.

  Not included: the defender still gets its defending-only shield research, and only the challenger can break off a losing fight. Both are tracked in #7112.

1.434.1 — 2026-10-10

Fixed

• Every dropdown's arrow now sits inside the field, clear of its border (#7090). The last release fixed this for the colony picker in the header. Every other dropdown in the game still drew the browser's own arrow pressed against its right edge, with the text running up to it. That covered the night-hour and language settings, the music track, the market and fleet-template pickers, the guild forms, the sector and system pickers on the galaxy screen, the delivery colony in the Anchorage shop, the target board's sector filter, and the gathering picker. They now all draw the same chevron as the header and the building rails. It is set in from the right edge as far as the text is from the left, the text stops just short of it, and it takes the field's colour, so the target board's lit sector filter shows a cyan arrow to match its text.

1.434.0 — 2026-10-10

New

The market's New offer composer shows the ratio of the offer you are posting (#7091)

Before, the composer showed quantities, caps and the fee, but no ratio. Matching or undercutting the board meant working the ratio out by hand, then sorting the board by Ratio to compare.

Now, once both sides of the offer hold something, a line under the fee says the offer's ratio: quantity given divided by quantity asked. It is the same figure the board sorts by and prints, in the same format, and the line says it is not a comparison of worth between resources. When the offer trades one resource for one other resource, a second line gives the highest ratio among other players' open offers for that pair in the same direction, or says that there is none. It appears only after the board has loaded. It leaves out your own offers, the giveaways the board pushes below its divider, and the offers whose accept the board blocks, so one gift or one dead offer cannot set the number.

Suggested by a player.

1.433.5 — 2026-10-10

Fixed

A building row on the Colony tab stays where it is when you queue it (#1743)

Before, queuing a build lifted that row to the top of the building list, and when the build finished the row dropped back to its usual place. Each row moved twice, and the second move came at a moment you did not choose. On a young colony an upgrade can finish seconds after the tap, so a tap aimed at one row could land on the building that had just slid under it. No sort option avoided this.

Now a queued row keeps its slot under every sort option, and it shows its countdown and progress bar in place. The Operations card above the list and the queue pill still list everything that is running. Completing a build moves no row in the default order or A–Z. Under the level and price sorts, a finished building moves to the place its new level gives it, because that is what those sorts order by. A building that reaches its maximum level or its Command Center cap still drops to the bottom group, as before.

The other two parts of #1743 are still open: a focus filter for the Suggested Next advisor, and payback times on the building's own page.

Reported by two players.

1.433.4 — 2026-10-10

Changed

The desktop chat opens and closes from the same button (#1821) Before, the chat button at the bottom-left only opened the chat. To close it you reached up to a small «—» at the top of the panel. Now the button stays in the same spot while the chat is open and closes it, so a quick look at a new message is one click to open and one click to close, without moving the pointer. The «—» in the panel header is gone. The button shows no unread badge while the chat is open, and the field you type in now starts to the right of it, so the button never covers it.

Escape also closes the chat when you are typing in it or have clicked inside it. With a reply or an edit in progress, the first Escape cancels the reply or the edit, as the ✕ on its bar does, and gives back what you were writing. The next Escape closes the chat. Escape does not close the chat over an open dialog, or while a tag's card or an emoji picker is open in the chat.

Suggested by a player.

1.433.3 — 2026-10-10

Fixed

The colony picker in the desktop header no longer draws its arrow against the border (#7090)

The dropdown next to your guild tag drew its arrow flush against the right edge, and a long colony name ran right up to it. It now draws the same arrow as the building-list pickers: set in from the edge by the same distance as the label, with a clear gap before it. A long name also shows slightly more than before, and the whole name is still in the tooltip and in the open list.

Reported by a player.

1.433.2 — 2026-10-10

Fixed

The research queue shows the price the server will charge for a queued Tree level, on Safari and the iOS app too (#7064) When you queued more than one level of the same Tree node, the research queue worked out the later levels' prices in your browser. Safari and the iOS app round that calculation differently from the server, and on one level the queue was off by one: level 5 of the Doctrine of Annihilation showed 521,284 xenogas while the server charges 521,283. Every queued Tree level is now priced from the figures the server sends, so the queue shows what will actually be charged. A level the server has not priced shows no price instead of an estimate. That happens only on an endgame line stacked more than five levels past the next one. Building costs, Discipline costs, core fees and guild research were checked the same way and never differed.

1.433.1 — 2026-10-10

Fixed

The Command Center unlock line in the Colony advisor now says why its button is grey when the build queue is full (#2141)

With the build queue full, the gold line above the advisor's suggestions (the next Command Center level and what it unlocks) showed a greyed upgrade button and nothing else. The reason was only in a hover tooltip, which a phone never opens. The suggestions below it already printed "Queue full." on the card. The Command Center line now prints the same sentence, in the same place and style, under its price. Nothing changes when the queue has room.

This is part of #2141. Still open there: whether first builds the colony cannot afford should be dimmed, and whether the place-on-map button needs a different look.

1.433.0 — 2026-10-09

New

The Targets board can be scoped to one sector (#2328)

Before, the board had bands, a sort, toggles, owner chips and a Gate range, but no way to ask for one sector, so a target known to be in the outer ring had no control that would take the list there. Its count of rows held behind a pact, a shelter or a cordon was also galaxy-wide, so it could not say whether the region you were looking for was the part being held back.

Now a Sector picker sits on the sort row, with All sectors (the default) and sectors 1 to 12. The event sector is not offered, because the board never lists it. Choosing a sector scopes the whole board: the list, the protected, NPC and Sanctioned counts, the owner census and the band chips all describe that sector. If the chosen sector has nothing to show, the empty list says so and points back to All sectors. The choice is kept on the device like the other filters. A board opened for one commander from the Wanted board always starts on All sectors, so none of that commander's colonies is hidden by an older choice. Your saved sector comes back the next time you open the board normally. There is no "jump to sector" button: tapping a row already opens that planet.

1.432.0 — 2026-10-09

New

A building's name on the Colony list opens its full sheet on the Base (#5408)

Before, each row on the Colony list showed only a short, clipped description. There was no way from the list to the full Building Sheet that the Base view opens. Now the building's name on each row is a button. Tap it and the Base opens with that building's sheet showing the full story: traits, the upgrade argument, the notes and the numbers. A second or third Site opens its own sheet, not the first Site's. The name looks the same at rest and turns cyan and underlined on hover or keyboard focus. Rows are no taller on a phone. The More toggle and the upgrade button work as before. Closing the sheet leaves you on the Base: there is no way back to the Colony list from there yet.

Suggested by a player.

1.431.9 — 2026-10-09

Fixed

On desktop, a long colony name cut off in the colony dropdown can now be read in full (#1799)

The colony dropdown at the top of the desktop screen is narrow, and a long colony or outpost name, together with its coordinates and planet type, was cut off with no way to read the rest from the closed dropdown. Hover over it now and its tooltip names the active colony with the whole row, the same text as its entry in the open list, ahead of the usual hotkey hint. The open list itself still shows every row in full. The dropdown keeps its width: a wider one pushed the top bar onto an extra line on common laptop widths. On a touch tablet there is no tooltip, so the open list is where the full name is read. Phones are unchanged; they use the colony chip.

Not included: opening the dropdown on hover, which needs a custom dropdown in place of the browser's own.

Reported by a player.

1.431.8 — 2026-10-09

Fixed

AI guildmates who do not lead their guild now know the guild season (#2586) Only an AI empire that led its guild knew the guild season: the week, the guild's rank and season points, and the Eschaton Project's progress in the final weeks. An AI officer or member asked in the same channel had no season facts at all, while every player sees the week counter on the guild panel. Now every AI guildmate has the same season facts its leader has. As always, it only states facts that are true. Contributing to the Project from the guild Fund is still something only the leader does.

This is part of #2586. Still open: the Calendar does not yet show when the guild season closes.

1.431.7 — 2026-10-09

Fixed

A redirected fleet whose home colony is gone can be intercepted again (#6928)

When a fleet was redirected in flight and the colony it set out from no longer existed, its contact sheet drew the plotted course but kept the Intercept button disabled. The reason line told the player to build Observatory 3, even when they already had it. The server would have accepted the order all along. Now the sheet plots that leg from the point where the fleet turned and lets you launch when one of your colonies can reach it. A contact whose origin is truly hidden is still refused, with the same explanation as before.

1.431.6 — 2026-10-09

Fixed

The guild threat board tells a garrison that is staying from one that is leaving (#785) Before, once a garrison's watch had run out, the board said "watch ending" whether or not it was still standing guard. That covered both a garrison kept in place because a strike past its point of no return was about to land, and a garrison that was about to leave. Now the board names a held garrison as held on station for now by the incoming strike, with the time its watch was due. A garrison that is about to leave still reads "watch ending". The hold has a time limit, so the line says "for now" and does not promise the garrison will be there when the strike lands. The defending total and older app versions are unchanged.

This fixes the threat board part of #785. The two edges about short watches stay open.

1.431.5 — 2026-10-09

Fixed

Fill the hold grows the cargo again after you add freighters (#2104) Before: you picked some freighters, pressed fill the hold, added more freighters and pressed it again, and nothing happened. The cargo stayed sized to the old hold until you pressed clear and filled again. Now: the second press spreads the numbers the fill wrote over the bigger hold. Any amount you typed yourself stays exactly as you typed it, and the fill only grows the cells it wrote. With nothing added, pressing it again changes nothing. Reported by a player.

1.431.4 — 2026-10-09

Changed

From 22 November the AI empires count an enemy's Aegis and an ally's Ward when sizing a strike, and say so when they call one off (#7062, #4984)

Until now, an AI empire sizing a strike counted a defender's Aegis only when an Aegis Bastion stood at the colony, and it never counted the shields that a Ward-holding ally lends a garrison. From 22 November, each AI empire counts both, at its own moment over the following two days, on spy reports it makes from then on:

• the Aegis level a spy report shows on the owner;
• the Ward on garrison rows the report marks as relaying.

In tests against the battle engine, their smallest strikes stop falling short of what those defences take. Some strikes they used to launch too small, and lose, now wait for a bigger fleet.

When an AI empire calls off a strike for this reason, it says why: the last probe showed the owner's tree holding the Aegis, or a garrison whose owner holds the Ward, alone or together with the rest of what the probe showed. Before, such a strike could be put down to a Bastion alone, or dropped with no reason given. It leaves that colony alone for a few days, unless a later probe no longer shows the cause.

1.431.3 — 2026-10-09

Fixed

Guild season points earned after the season's eighth week no longer change its final ranking (#2586)

A guild season ends eight weeks after it starts, but the season only closes the next time someone opens it. Until now, a war won, a Council Directive tier reached or a League match won in that gap still added season points to the season that had already ended, and the close then ranked guilds and paid placement rewards with those late points included. A guild could move up the final standings on points earned after the season was over.

Now a war or League match counts for the season it was fought in. A war or match fought during the season's last week still scores there, even though it is settled after the season ends. A war whose week starts after the end, a war surrendered after the end, and a Directive tier settled after the end add nothing to the ended season. Those points are dropped, not carried into the next season, because that contest had not started when they were earned. Everything else these pay out is unchanged: tribute, Scrip, Fund, guild XP and match rewards. A guild that surrenders a war fought in the last week, in the minutes before the war is settled, can no longer cost the winner its season points that way.

This fixes the second item of #2586. The other two items (season facts for officer agents, and the season close on the Calendar) are still open.

1.431.2 — 2026-10-09

Fixed

BUILD is on screen when the Shipyard opens in a narrower desktop window (#7066)

In a desktop window around 1000 to 1100 pixels wide, the Shipyard switches to its narrow layout, with the hulls in a strip across the top. In that layout the BUILD button opened 8 to 11 pixels below the bottom of the screen, and you had to scroll to reach it. It now opens fully on screen with about 55 pixels to spare, the same as in wider windows.

The build count box also shows a five-digit count in full. At 99,999 it used to cut off the first digit, at every desktop width. The box's small up/down arrows are gone to make room. The minus and plus buttons beside it still change the count.

1.431.1 — 2026-10-09

Fixed

The chat panel's sizing test now catches the resets it used to miss (#7054) A test keeps a site-wide border-box reset off the chat panel, because such a reset would shrink the panel and leave part of the game under it. That test missed resets written with nested or quoted brackets, such as *:not(:is(svg, img)) and *:not([title=")"]). It also failed on the harmless legacy *:before form, which only sizes generated content. It now reads selectors token by token. The CSS reader that the layout tests share also stopped splitting a selector list wrongly when a bracket sits inside quotes. This is test tooling only: nothing changes in the game.

1.431.0 — 2026-10-09

New

The tech card shows the price of every level ahead (#1783) A leveled research card used to price only the next level, so planning a run meant buying your way up to see what came next. A collapsed line under the price now reads "Show the next N levels (to lvl X)". Opened, it lists every later level up to the cap, each in full digits with its Dark Matter leg where there is one. The uncapped endgame lines show the next five. The server computes these prices, so they are the figures the queue will charge, on every device.

Suggested by a player.

1.430.1 — 2026-10-09

Changed

The AI empires' battle forecast can now count an enemy's Aegis and an ally's Ward, but the read is switched off for now (#4984)

Nothing changes in play yet. Until now, an AI empire sizing a strike counted a defender's Aegis only when an Aegis Bastion stood at the colony. It never counted the extra shields that a Ward-holding ally lends a garrison. Both are now part of its forecast:

• the Aegis level a spy report shows on the owner, with Deep Formation from level 10 and Last Stand from level 15;
• the Ward on garrison rows that a spy report marks as relaying.

In tests against the battle engine, the AI's smallest strike stops falling short of what those defences take. In the issue's own case, Bombers against four Reavers and ten Interceptors on the Aegis at 20, it goes from 98 to 142, where the engine needs 131. These reads can only make a strike bigger, never smaller.

The read is shipped switched off. Turning it on would leave some called-off strikes with the wrong reason, or with no reason at all. #7062 gives the Aegis its own call-off first, and only then picks the night the read switches on, one empire at a time over two days.

1.430.0 — 2026-10-09

New

The Base tab now shows the planet you are building on (#2293)

Until now the Base screen, where you actually build, never described your planet. Its bonus and its malus showed up on Colony, Overview, Missions and the Galaxy sheet, but not here. A small badge on the map now shows both sides of the planet's trait. The figures are the ones after your Terraformer and Colonial Outpost have done their work, which is the number that decides whether the next Terraformer level is worth building. On a computer the badge sits in the bottom-right corner of the map. On a phone it sits top-right, just under the header, steps aside while you place a building, and never takes a tap meant for the map.

Suggested by a player.

1.429.34 — 2026-10-09

Fixed

The desktop shipyard shows what an order will cost before you build it (#2264)

Since 1.72.0, the build footer of the shipyard on desktop showed the build time and the Power the order adds, but no price. The order's total was only inside the folded "Order details" panel, and the per-hull price in the console did not change with the count. On a phone the total was always on screen.

The desktop footer now has a "Total ×N" line above the build time. It gives the order's whole price per resource, multiplied by the count, and turns red on any resource you are short of. It updates as you change the count. "Order details" still opens closed and still holds the full breakdown with your stock beside each figure.

Reported by a player.

1.429.33 — 2026-10-09

Fixed

The chat dock's layout test now catches a box-sizing reset hidden inside a media query (#2535) Nothing changes for players. The test that keeps the chat panel's width honest checks that no stylesheet rule applies a border-box reset to every element, because such a reset would shrink the panel and the game band beside it. It read the stylesheet with a pattern that could not see a rule nested inside a media query, so a reset written for mouse-driven screens only would have passed it. It now uses the stylesheet reader that understands nesting, and its failure message names the media query the rule sits in.

1.429.32 — 2026-10-09

Fixed

A cheap upgrade that pays back in hours now reaches its family tab in Suggested next (#7041)

Before, a family tab of Suggested next could show three upgrades that each took two weeks to pay back. Meanwhile a low-level building that paid back in a few hours appeared on no tab at all. Its gain was small next to the colony's whole output, so it was ranked after every bigger row and cut at three.

Now, when a family tab already holds three bigger rows, its last slot goes to the small row with the shortest payback, provided that row pays back sooner than the third row it would replace. If that third row shows no payback, the small row must pay back within a day. The row sits under the "marginal" divider, so you can still tell it is a small gain. A first build of a new Site counts like an upgrade. Small rows that show no payback, such as extra Laboratory levels and defences, still never take the slot.

Not included: the All tab is unchanged. On a colony like the one reported, it still shows the three two-week upgrades, and the fast row appears on its family tab only.

Reported by a player.

1.429.31 — 2026-10-09

Changed

A browser check now keeps the Shipyard's BUILD button on screen when the tab opens (#5521) Players see no change. In September, BUILD sat 7 to 9 pixels below the bottom of the screen on common laptop and desktop sizes until the page was scrolled. That was fixed, but no test checked where the button actually lands. A headless browser now opens the real Shipyard tab at six desktop sizes, from 1280x800 to 1920x1200. It fails if BUILD is not fully visible before any scrolling, and it also fails if the page was scrolled at all. With the old defect put back, it fails at every size.

1.429.30 — 2026-10-09

Fixed

Agents now mention a portal exchange even when its first reply was lost (#7047)

When an agent converted ferrolite at its Trade Portal and the server's first reply never came back, the automatic retry received the stored result. The agent took it for an older conversion, so it never mentioned the exchange it had just made. Now that conversion reaches the agent's facts once, with the amounts the server charged and delivered. A conversion that an earlier turn already reported is still never stated twice.

1.429.29 — 2026-10-09

Changed

AI empires will name a loot preference on their raids (#4338)

When a won raid cannot carry everything, each contingent can ask for one resource to fill its holds first. Until now, AI empires never asked. Their holds always took the even mix, even when the humans in the same raid named a resource, and even right after the empire had asked its guild board for that resource.

From 18 November, each AI empire starts naming one, at its own moment over two days. An empire short of a resource usually asks for that one, and keeps asking while its guild-board request for it is still open. When nothing is short, it follows its plan and its own taste, which can lean a different way from one day to the next. Some empires rarely name anything, as some players never do. Nothing changes before 18 November.

1.429.28 — 2026-10-09

Changed

The layout checks now measure screens with the real bottom navigation bar on them (#5547)

Nothing changes on screen. The automated checks that decide whether a button on the phone layout can actually be tapped used to measure some screens without the bottom navigation bar. Without the bar, a button sitting where the bar would be counted as reachable, even though on a real phone the bar covers it. The checks now place the game's real bottom bar on the mission screen and the quick-buttons picker. On the one test page that cannot carry the bar, they compare every button against the bar's measured position. Rechecked with the bar in place, no button those checks cover turned out to be hidden.

1.429.27 — 2026-10-09

Changed

The AI empires' Trade Portal conversions now have a test that fails if they could ever be charged twice (#5531) Players see no change. An AI empire's Trade Portal conversion is the one spending request its client may re-send after a lost reply, because the server answers a repeated request id with the stored result and converts once. Until now that rested on a comment. The agents' test suite now reads the database schema and the server's replay code, and replays a lost reply against a fake server. Dropping the unique key, or changing how the replay looks it up, now fails the suite before it can ship. The other requests on the retry-safe list are still checked by review only, and the test file now says so.

1.429.26 — 2026-10-09

Changed

The launch form now says why your own Hypergate pair did not speed up a trip (#2105)

Before: the Hypergate card advertised a pair of gates at up to −50%. A transport between two of your own gated worlds could still show only the single gate's discount (for example −22%), and nothing said why.

Now, when the hyperspace row names the single Hypergate or the Relay, a line under it gives the reason the pair lost: the route stays inside one system, which a pair never covers, or the pair's own time on this route (shown) is slower than the discounted time above, because a pair only pays off on longer routes. If you have researched the Relay and the launch colony has none working, the line adds that a Relay there halves any route. The Hypergate card's pair note now says that the hop's discount shortens the hop, not your trip, and that a flat cut on every route comes from the Relay. In Italian, the row now calls the Relay by the building's own Italian name.

Reported by a player.

1.429.25 — 2026-10-09

Changed

The server's background jobs no longer all fire in the same instant every six hours (#2594)

Before: fourteen of the server's recurring jobs, among them the exodus, displacement, campaign and research sweeps, the war-week and Dark Matter ticks, sovereignty, and the faction and relic rescales, all started counting at the moment the server booted. Every six hours they all came due in the same instant and ran back to back, competing with whatever players were doing at that moment. Three more timers sat exactly on another job's tick.

Now: every background timer on the server has its own second, taken from one table, and no two of them are scheduled for the same second. No job runs more or less often than before, and none runs earlier after a restart. A new timer cannot be added without a place in that table: a test reads the source and fails if one skips it.

1.429.24 — 2026-10-09

Changed

Advisor and payback tests now check colonies the game can actually build (#5003) Nothing changes on screen. Several of the Advisor's tests checked cards built from colonies no player can have, such as a Command Center 3 colony holding a level-12 Extractor. Those cards made claims no real player sees. One said upgrading to Command Center 4 releases a level-13 Extractor, and the payback test relied on a price only that impossible colony could produce. Those tests now use colonies at their Command Center cap, and a shared check fails them if they ever go above it again. A Command Center card that names or charges the wrong building now fails the payback test instead of passing it. This is one slice of #5003: the other test fixtures above the cap are still to be checked.

1.429.23 — 2026-10-09

Changed

The AI empires' copies of more engine numbers are now checked against the engine (#4319)

The AI empires keep their own hand-copied versions of a set of game numbers that nothing compared with the server's: the market's offer ceiling curve and settlement fee, the Shield's pool, the stats of a single Defense Battery emplacement and of the Fortress, and the speed-to-flight-time constant. If the server retuned one of them, the agents kept planning with the old value and every test stayed green. Now a test reads each of these numbers from the engine's source and fails as soon as either copy changes alone.

Nothing changes in play: every one of these numbers already agrees with the server. This is one slice of #4319. The raid limits and the other copied numbers that issue lists are still not checked.

1.429.22 — 2026-10-09

Fixed

A new transport no longer opens with a red error about its empty hold, and the phone says it only once (#821)

Before, a transport you had just opened on the launch form showed a red cross with "A transport needs a cargo", even though you had not had a chance to load anything yet. On a phone, the same sentence also appeared twice: once on the sticky Send bar and once in the verdict line right next to it.

Now an empty hold reads as a neutral instruction (grey, with a bullet) until you touch the cargo: typing an amount, a max button, either fill button or clear. If the hold is empty again after that, the red cross comes back. Send stays disabled either way. On a phone, while Send is disabled, its reason is printed only on the bar. The notes and the fix button below it stay where they were.

Reported by a player.

1.429.21 — 2026-10-09

Changed

• A reload brings you back to the screen you were on, not Home (#7007). Reloading the page, from the update banner, the error panel, the browser or a pull-to-refresh, used to throw you back to Home whatever you were doing. You then had to drill back down to the Galaxy system, the colony or the Activity filter you had open. Now a reload reopens the same tab, on the same colony, and screens that remember where you were on them come back there: the Galaxy's sector and system, the Activity chips, Research, the Base, the Colony list, the Shipyard, the Season calendar and the phone's Missions segment. Back after the reload still walks the trail you had before it.

  Some things still start on Home, as before: a fresh visit, a cold start of the app, the Restart button after the game crashes, and another account signing in on the same browser tab. Two cases are deliberate safety exits. If a reloaded screen kills or freezes the page within a few seconds, the next load opens on Home rather than looping on it. Two reloads within about eight seconds also land the second one on Home.

  Not included: a half-filled form is not restored by this. The mission composer already keeps its draft on its own. Other screens reopen on their default view inside the tab.

  Reported by a player in the bugs forum, who separated it from the scroll issue (#6846) they were answering about.

1.429.20 — 2026-10-09

Fixed

• The privacy page now says what the game really keeps on your device (#6385). It used to say the browser holds a single item, a sign-in token, plus a second one if you turned visit counting off. The game keeps much more than that, so the section now lists each kind and why:
  • what keeps you signed in;
  • your settings (language, sound, notification choices, which tabs sit in the menu);
  • work you have not finished (unsent chat, mission, market, raid and defense drafts);
  • what you have already seen, and the snapshot behind the "while you were away" summary.

  It also says what signing out removes and what stays. One thing is new to the page: in the iPhone app, signing in with Apple leaves Apple's identity token and the name Apple shared on the phone, and Google's sign-in keeps its own session there. Signing out of the game does not clear either. This release only discloses that; making sign-out clear it is #7027.

1.429.19 — 2026-10-09

Fixed

The release-notes language check no longer depends on the shell that starts it (#7019) Before: the check that keeps Italian out of release notes inherited the caller's locale. From a shell set to the plain C or POSIX locale it missed the four listed Italian words that end in an accented letter, so notes holding them could pass as English, and the merge check went red on unrelated pull requests whose queue run had started from such a shell. Now: the word search picks its own UTF-8 locale and finds those words from any shell. Where no UTF-8 locale is available, the notes are refused with that reason instead of being read as English.

1.429.18 — 2026-10-09

Fixed

Refreshing stale ship machinery no longer plants a merge conflict (#5503) Before: when a tree ran an old copy of the release scripts, CLAUDE.md, the ship guide and the release, preflight and deploy error messages all said to refresh the whole ship-skill directory. That also copied the ship ledger, a file every ship appends to, so the branch then conflicted with the next row on main even when the code merged cleanly. Now: the advice refreshes every file except the ledger, and the release message says how to put back the merge-base copy of the ledger on a branch where an older refresh already committed it. No player-facing change.

1.429.17 — 2026-10-09

Fixed

Two relic seedings running at once can no longer found a second colony for the same relic (#2552) Relics are placed on the map when the server starts and every six hours after that. If two servers ever did this at the same moment, one relic could end up with two colonies. The map would then show one address while the relic's ownership pointed at the other, and a fleet that beat the garrison at the address on the map would capture nothing. Each relic is now placed by one server at a time, and the second one finds the colony the first one made. Nothing changes in play today: the game runs on one server, so this could not happen yet. The fix makes sure it cannot happen on the day it runs on more than one.

1.429.16 — 2026-10-09

Fixed

AI empires no longer mistake a player for one of their own because the two share a display name (#4340) Before: when a question in galaxy chat named no one, the AI empire that had just spoken took it as meant for itself. It decided who had just spoken by display name, though, and names are not unique. A player who carried an AI empire's name could make that empire answer a question that followed the player's own line. If two AI empires shared a name, the wrong one could pick up the question. The same mix-up could break a tie in favour of the wrong empire when a message named a first name two AI empires share. Now: who spoke is decided by account, not by name. An unnamed question goes to the AI empire that actually spoke last, or to nobody when a person spoke last. With unique names nothing changes. One gap remains: a message that types out a name shared by a player and an AI empire still reads as addressed to the AI empire, because the text carries no account.

1.429.15 — 2026-10-09

Fixed

Research prices show every digit, also on a phone (#2276)

On the Discipline panel and the Tech Tree sheet, a research price of a million or more was shortened to a figure like 12M, and so was the "still needed" amount under it. Every value from 11.5 to 12.49 million printed the same 12M. The exact number was only in a mouse-hover tooltip, which a phone never shows. A player told they were short of 12M Data while holding 12M could not find out by how much.

Both screens now print every price and every missing amount in full digits, as the Building sheet already does. On a phone, Tech Tree prices in the billions wrap onto two lines, with each amount still under its own price. The short form on the Discipline cards and the Bench's "Exact figures" switch are unchanged.

Reported by a player.

1.429.14 — 2026-10-09

Fixed

AI empires can now rescue a stalled young outpost with Orbital Panels (#3743)

Before: an AI empire would rescue a stalled outpost (one with no power and nothing to spend) only once it could afford a Fusion Reactor (80 Ferrolite, 30 Xenogas) plus whatever producers the outpost lacked. A young empire holding 100 Ferrolite and 40 Xenogas never sent the supply run, even though Orbital Panels, at 40/10, would have got the outpost going.

Now: the supply run prices power at Orbital Panels, the cheapest generator, so that empire sends 88 Ferrolite and 34 Xenogas. The outpost spends it on Panels and a Refinery and starts earning Ferrolite and Xenogas. It no longer spends it on a Reactor that left nothing for the Refinery. Outposts and empires that can afford the Reactor and their missing producers still build the Reactor first. The same rule applies to later Reactor levels: an AI colony that is short of power and has no Refinery no longer spends on a Reactor upgrade it cannot pair with a Refinery, and most personas buy the Refinery instead. The dowry a new colony is founded with is unchanged.

1.429.13 — 2026-10-09

Fixed

With the build queue full, the advisor no longer offers a new building it cannot start (#2141). A full queue used to dim the upgrade suggestions, while the new-building suggestions right below them stayed bright with a working ＋ button. That button opened the base map, and the server refused the placement as queue full only after you had picked a cell. Now a full queue dims new-building suggestions as well. Every suggestion it blocks also says "Queue full." on the card, so the reason is visible on a phone and not only in a hover tooltip.

Not included: a new-building suggestion is still not dimmed when you cannot afford it. Its price already shows the missing resource in red. The Command Center unlock button above the list still gives its queue-full reason only on hover.

Reported by two players.

1.429.12 — 2026-10-09

Fixed

The attack advisor's net-loss warning now prices the fleet you are about to launch (#2462)

Before: the forecast row priced the fleet on the form, while the "winning at a loss" warning under it priced the Advisor's recommended fleet. Once you edited the fleet the panel showed two net figures for two different fleets, which read like two calculations of the same raid that disagreed; a green surplus could sit above a red "sets you back", and editing a paying recommendation into a losing fleet gave no warning at all.

Now: the warning, and the "its hold cannot make the raid pay" line, speak for your fleet. An edited fleet that pays carries no warning. When you have edited a fleet that loses and the recommended fleet would do strictly better, a dim line under the warning gives the recommended fleet's figure for comparison; it stays hidden on a tie, and when the recommended fleet's own hold could not bring the loot home. The row and the warning now round figures the same way, so one figure no longer prints one unit apart on the two lines. The row's net still counts the loot your hold brings home, and its "hold full" chip names the gap when the hold caps the haul.

Reported by a player who compared the two figures and spotted that only one of them followed the edited fleet.

1.429.11 — 2026-10-09

Fixed

Four guild actions retry instead of failing when the database briefly blocks them (#4218) Before: if two guild operations hit the same guild at the same moment, the database could cancel one of them to clear the jam. Four actions surfaced that as a server error: handing over leadership, founding a guild, starting or pinning guild research (and opening the guild screen when research had just finished), and opening the guild screen when an absent leader was being replaced. The player had to try again. Now: the cancelled action is retried automatically, up to three attempts and only within the first four seconds, and each retry checks the guild again. A transfer or a succession that another request already completed is not applied twice, and a retried founding does not use up a second name check. If the database stays blocked past those limits, the error is the same as before.

1.429.10 — 2026-10-09

Fixed

A tap on a pirate beacon opens its system, even when the beacon shares its star (#3565)

On the galaxy map, the pirate beacon is the most conspicuous mark on a system, and players aim at it to open that system. A beacon alone on its star already sat on the star's own cell, so that tap worked. But when the beacon shared its star with another sign (your home, one of your colonies with My colonies on, or a named place), it was drawn off to one side, and half of it hung over a neighbouring empty cell. A tap on that half opened nothing. Now a tap anywhere on the beacon opens its system, at the same zooms where tapping a planet opens its card. On a desktop, hovering over it names the system too. Planet taps are unchanged, because the cells the beacon reaches never hold a planet. Further out, a tap on the beacon still zooms in.

1.429.9 — 2026-10-09

Fixed

The Home radar no longer hides strikes past the fourth (#6988) On the desktop Home, the strike list under the threat radar showed four strikes and dropped the rest with no line saying they existed. Now each row is one attacker on one colony: a raid of several fleets reads as one row with ×3, its soonest clock, and every clock in its tooltip. With more than four such groups, the list shows three and a final line counts the strikes left out, for example +3 more strikes · next in 2h; clicking it opens Missions. The radar blips are unchanged.

1.429.8 — 2026-10-09

Fixed

The Home timeline's Threats lane no longer hides a fifth strike (#6988) On the desktop Home timeline, a lane drew at most four points on its axis. With five or more strikes landing inside the four-hour horizon, the Threats lane showed four diamonds, and the later strikes were missing with no counter to show they existed. «Expand all» did not open the lane either. Now the Threats lane gets the same counter as a colony lane, for example +1 · next 3h 30m. The counter's tooltip lists the hidden strikes with their clocks, and opening the lane, by its counter or by «Expand all», draws them all. Strikes that land within the same minute still draw on top of each other. The radar card's strike list still shows four strikes at most; that is being designed separately.

1.429.7 — 2026-10-09

Changed

Agent tests: a check that a true line still goes out now has to prove the whitelist is why (#5556)

The gate that stops agents and Wraiths from pinning an unowned fleet on a real player is tested in both directions. Some of the tests that say "this true line still goes out" would have passed with the gate's list of named owners deleted, because the line never reached that list. Each one now runs the line a second time with the list removed and requires a refusal there. Three tests passed without ever reaching the list and now use facts that do reach it. A fourth stays as it is, because not reaching the list is the point of that test, and it now says so. Nothing changes in the game: this is test coverage only.

1.429.6 — 2026-10-09

Fixed

A number too large to hold no longer stays pasted in a raid coordinate (#4998)

When you pasted something like 1e999, or a long run of digits, into the Sector, System or Planet box of "Open a raid", the pasted text stayed in the box while the target line beside it went blank (—:—:—). The box now puts back what it held before the paste and selects it, so the next paste replaces it, and the target line keeps showing the address. Normal typing, an empty box and the out-of-range warnings work as before.

This is the same fix the other number boxes got in #4226. The two number boxes on the operator-only admin screen are not changed.

1.429.5 — 2026-10-09

Fixed

The Home timeline's «Expand all» tooltip says which lanes it opens (#6987)

On the desktop Home, the «Expand all» button's tooltip said it expanded every lane with more to show. The button opens only lanes that have a +N counter, so the Threats and Guild · season lanes, which have no counter, stayed as they were. The tooltip now reads "Open every lane that shows a +N counter", and «Collapse all» reads "Close every open lane". The Italian tooltips were reworded to match. The button labels are unchanged.

Those two lanes still draw at most four points. That limit is tracked separately in #6988.

1.429.4 — 2026-10-09

Changed

The Speed-up sheet works out its recommended spend in the same tiny time on any order, however long (#5499) Before, the one-tap recommendation on the Speed-up sheet tried every combination of 4h and 1h shards. The longer the order, the more combinations it had to try, and the work grew with the square of the order's length. On the heaviest real board, a 16-day ship batch, that came to about 40,000 tries every second the sheet was open. A year-long order on the biggest hoard in the galaxy would have needed 4.7 million. Now the best combination is worked out directly, so the cost no longer depends on how long the order is or how many shards you hold. The recommendation itself has not changed. On every order and every inventory the sheet suggests exactly the same shards as before: the least time wasted, with the smaller sizes spent first when two options waste the same. Before release, the new calculation was checked against the old search on more than 348,000 boards, and they agreed on every one.

1.429.3 — 2026-10-09

Fixed

On a phone, the Build chip is on screen without a drag (#5498) The resource strip at the top of a phone screen scrolls sideways with no visible scrollbar, and the Build chip (the build queue count, and the tap into the Colony tab) sat at its far end, past the edge of the screen on every upright phone. Nothing showed that the strip continued. Now the Build chip holds the strip's right edge: it is on screen the moment the game opens, the pill partly visible beside it shows that the strip scrolls on, and once the strip is scrolled far enough the chip goes back to its own place. The incoming-attack alarm stays on top of it. On a narrow phone under attack, the chip covers a little of the protection clock until the first drag. Desktop is unchanged. Rotated phones are not covered: their strip does not scroll at all yet, which is tracked in #6984.

1.429.2 — 2026-10-09

Fixed

At 200% text zoom, every ordering arrow stays on screen (#5520) In Settings, under Quick buttons, the Order strip shows each chosen tab as a pill with a left and a right arrow. On a very narrow screen (200% text zoom on a small phone), the pill for a long name such as Leaderboard was wider than the screen, so its right arrow sat past the edge of a sheet that cannot scroll sideways, and the first tab could not be moved right. Now a pill is never wider than its row: when space runs out, the tab's name is shortened with an ellipsis, and both arrows keep their full size. At normal phone widths nothing changes.

1.429.1 — 2026-10-09

Changed

The Home timeline opens every lane in one tap (#6966) On the desktop Home, the next-hours timeline draws three bars per lane and counts the rest behind each lane's own counter. Seeing every clock used to cost one tap per lane on every visit, which on an 8-colony empire meant six or seven taps. A small «Expand all» button now sits above the lane names: one tap opens every lane that has more to show, and it then reads «Collapse all». It appears only while some lane has hidden bars. Opened lanes are still forgotten when you leave Home, as before.

1.429.0 — 2026-10-09

New

• Defend an ally's colony, then harvest the wreckage (#6761). When your garrison fought at a guildmate's or ally's colony, the debris field over it is no longer the owner's alone. The planet sheet now offers Harvest beside Defend, the Missions tab offers the verb, and the launch goes through. Before, the sheet showed only Defend and stand down, and a harvest launched by hand was refused with "only the owner can harvest there". That field holds the attacker's wrecks and your own fallen ships.

  The rule: your garrison must have fought there in the last 72 hours, against the colony's current owner, and you must still be on that owner's side (same guild, alliance or non-aggression pact). Both are checked again when your harvesters arrive. If the pact has lapsed in between, they come home empty and the field is left as it was. The owner's own harvest is unchanged. The attacker, a guildmate who did not garrison, and a former ally are still refused. Relics and contested control worlds stay closed to harvesting by anyone but their owner, as before.

  The field is one pool per planet, so wreckage from a later battle you missed is in it too. You may lift that as well, as long as your own garrison battle is still within the 72 hours. The owner can also lift the same field; whichever fleet arrives first takes it.

  Suggested by two players, one on Discord and one in the suggestions forum: «It's fun to thump an attacker but even more fun to take their carcass and turn it onto weapons. Makes up a little for units lost as well.»

1.428.10 — 2026-10-09

Changed

The starting sectors now say why their bounty follows the place, not the owner (#5713)

In sectors 1-4, a strike far below your Gate costs a doubled bounty, and the bounty starts from a smaller Gate gap. That covers every owner's colony there, including a veteran's oldest and best-developed one. Nothing in the game said why, so it read as beginner protection leaking to veterans. The strike preflight added to the confusion by calling the starting sectors the place "where new empires are born".

The rule is unchanged. The game now states its reason:

• Galaxy tab sector note and the glossary's Starting sectors entry: the price belongs to the place, not to whoever holds it. It keeps the region quiet for the empires starting out there, and it reads the colony's sector, never how old or strong its owner is. A new empire's own shelter is novice protection, which lasts at most 72 hours.
• Targets board header: the tighter band and the doubled price are called a rule of the place, whoever holds the colony.
• Strike preflight bounty line: now reads "double price: a starting sector, whoever holds it". The line is also shorter.

The same changes are in Italian.

Reported by a player on Discord.

1.428.9 — 2026-10-09

Fixed

Last Stand and Interlock no longer promise a cap that a surrender overrides (#6963) Before: the Last Stand and Interlock rows on the Doctrine sheet said the quarter-per-round hull cap held "as long as" something on the defending side still fired. A ship that fired and died in the same round broke that promise: at the end of the round nothing was left able to fire, and every defending ship group, capped or not, was lost in that round. The battle report's "Shielded by the Interlock" line on a covered fleet's card said nothing about that surrender at all. Now: both rows say the cap holds unless a round ends with no ship, Emplacement or Fortress on the defending side able to fire and no Bubble holding. The report line, in English and Italian, carries the same exception.

1.428.8 — 2026-10-09

Fixed

"While you were away" no longer shows your own spending on another device as a loss (#6958)

Before, the sheet measured your absence from the last time you used this device. If you played on your phone and came back to your desktop, everything you did on the phone fell inside the window, and the resources you had spent there showed as a red loss with no attack to explain it.

Now the sheet checks when you last played on any device:

• If you played somewhere else less than 8 hours ago, it does not appear. That was a short break, not a return.
• If it was 8 hours or more, the sheet appears, counting from that last action. Its away time, attacks, espionage and arrivals all start there. The net resource figures are left out, because they would include your own spending; full-warehouse warnings still show.
• If you only use one device, nothing changes.

A switch to another device that lasts only a few minutes can still go unnoticed, and the spending from it can still show in the resource figures.

Reported by a player in the bugs forum.

1.428.7 — 2026-10-09

Fixed

A refused intercept on a flight to its owner's own colony no longer tells you what the flight is doing (#6953) With an Observatory that names a fleet's owner, tapping INTERCEPT on two flights heading for one of that owner's own colonies used to give two different refusals: the housekeeping sentence for a cargo run, the general one for a defence or pickup flight. The refusal told you the mission type, which no Observatory level sells. Both now get the general refusal, at every Observatory level. Which flights can be intercepted is unchanged.

1.428.6 — 2026-10-09

Fixed

The Season tab's lane switch no longer draws its labels over each other on a phone (#6942) On a narrow phone, or with an enlarged system font, the four lanes of the Season tab were squeezed into one row: CONTRACTS was drawn across The Proving Ground and CALENDAR across CONTRACTS, and only the gold outline showed which lane was selected. Now the switch is one row of four only while every label fits whole; otherwise it becomes two rows of two, The Race and The Proving Ground above, Contracts and Calendar below. At the default text size and with system text enlarged by 30%, no label is shortened or cut, the Contracts dot included; only at much larger system text sizes does a label still too wide for its lane end in an ellipsis inside its own lane. At the default text size, phones narrower than about 468px always get two rows; desktop is unchanged.

Reported by a player in the bugs forum, with a screenshot from an Android phone.

1.428.5 — 2026-10-09

Fixed

The Interlock rung no longer promises that covered fleets always hold a quarter a round (#6934)

The Phalanx's Interlock rung on the Doctrine sheet said that no ship group it covers loses more than a quarter of its hull in a round, with no exception. In battle, a defence where nothing fires anymore still surrenders whole: if no ship, Emplacement or Fortress on the defending side has guns left and no Bubble holds, every covered fleet is lost in that round. A guest garrison of Freighters guarding beside unarmed Aegis Bastions lost every ship in round 1 while the rung promised a quarter. The rung now says the cap holds as long as a ship, Emplacement or Fortress on the defending side still fires or the Bubble holds, the same exception the Last Stand rung has carried since #6919. Nothing in combat changed.

1.428.4 — 2026-10-09

Fixed

The battle herald no longer prints the address of a colony under the Veil (#6945)

After a big battle between players, Kepler Observatory posts a line in the galaxy channel. That line used to print the battle's coordinates as plain text, even when the battle was fought at a colony under the Veil, so every reader could see an address the map hides. When the herald posts a new line, it now checks the Veil the same way the Galactic log does: if the colony at that address is under the Veil at that moment, the line says "a world under the Veil" where the coordinates were. Who fought, who won, the losses and the replay chip are unchanged.

Lines posted before this release keep the text they were written with.

1.428.3 — 2026-10-09

Fixed

A refused intercept between two starting sectors no longer hints at what the fleet is doing (#6937)

Before: you could order an intercept on a flight whose origin and destination were both in the starting sectors (an older app version, or a direct API call). The refusal you got back depended on the flight's hidden details. A plain attack or transport was refused as protected space. A spy, explore, colonise or defend flight, a raid contingent, or an owner under night, novice or pact protection got the generic refusal instead. Because a refused order costs nothing, comparing the two answers showed whether the flight carried something your Observatory had not revealed.

Now: every such flight is refused as protected space, whatever it is carrying and whoever owns it. Your own fleets are still refused as your own. Viewers who can see the owner now get the protected-space refusal on these routes, not the night, novice or pact one. Protected space is the reason that still applies after those shelters lift. The set of flights that can be intercepted has not changed.

Not included: one similar case is still open. When a flight goes to one of its owner's own colonies, a viewer who can see the owner can still tell the mission type from the refusal. Fixing it means choosing which hidden detail the refusal keeps, and that decision is tracked in #6953.

1.428.2 — 2026-10-09

Fixed

The Aegis Last Stand rung no longer promises more than a battle delivers (#6919) Before: the Aegis level 15 rung said no group of yours loses more than a quarter of its hull in a round, so it stands at least four rounds, and that the fire it turns aside falls on the rest of its row. Three things were missing. The colony's Emplacements and Fortress were never covered. A defence with nothing left that fires and no Bubble surrenders its ships at the end of that round, cap or no cap. And a row-mate without a cap takes the turned-aside fire in full. Now: the rung says "ship group", and promises the four rounds only as long as a ship, Emplacement or Fortress on your side still fires or the Bubble holds. The Orbital Cannon's shots do not count. The turned-aside fire falls on the rest of the row, up to any caps they have. The rule itself is unchanged. The same gap on the Phalanx Interlock rung is tracked in #6934.

1.428.1 — 2026-10-09

Fixed

An intercept that finds empty space now says the target may have been sheltered (#6940) When a pursuit reached its meeting point and found nothing, the activity feed listed three reasons: the target landed, was recalled, or somebody reached it first. It never mentioned the other reason the game checks at the meeting: the target's owner had become sheltered during the flight, because their night window opened, a pact was signed, or novice protection came back. The row now names that case too, in English and Italian, with the same wording as the note under every intercept solve. The title and the "flying home intact" line are unchanged.

1.428.0 — 2026-10-09

New

The Galaxy hover tip shows the salvage your own battles left (#6830) Before: on desktop, hovering a planet in a system's 3D view showed its name and type, but not the salvage estimate. The system list and the planet sheet had it; the tip had been dropped because the map did not read the feeds the estimate needs while you were just looking around. Now: the tip adds a green line, "salvage ≈ n · from your battle reports", or "· your harvest on its way" while a harvest of yours is flying there. The figure is the same one the list, the sheet and the Missions tab print, and it keeps decaying while the map sits open. The map reads your battle reports, harvests and missions once when you open it and again only when one of your battles or harvests lands, so an idle map still costs no extra request. Not on phones, where the tip does not exist. A recall or a launch made from another device shows up after your next battle or harvest, or when you reopen the Galaxy.

1.427.4 — 2026-10-09

Changed

Battle replays that field the Annihilator, the Praxis Interdictor or the Voidlance Cutter load a lighter model of that hull (#6196) Before, these three ships were each drawn in the replay from their full model: about 2 million triangles and three 4096-pixel textures, roughly 305 MB of graphics memory per hull and an 11–13 MB download. The replay now loads a lighter copy of the same hull, with 1024-pixel textures like every other ship and a reduced mesh, at about 30–41 MB of graphics memory and a 4.4–7.4 MB download. The Shipyard, the galaxy and the planet card still show the full model.

Each hull kept as many triangles as it needed to keep its panels smooth, judged by eye at the replay's closest shot: the Cutter 35%, the Annihilator 50%, the Interdictor 65%. The Interdictor is therefore still well above the ordinary ships' triangle count.

This does not yet claim to fix the iPad replay that would not start (#6188): that needs a test on the device.

1.427.3 — 2026-10-09

Fixed

The public register no longer prints the address of a colony under the Veil (#6521)

Before: the Galactic log, the Calendar's day panel, event chips in chat, and the public replay of a battle printed the exact coordinates of a battle or a new colony, even when that colony was under the Veil. A colony that the galaxy map shows as an empty slot to anyone whose Sensors cannot see through its Veil had its address in plain text for everyone.

Now: wherever those coordinates used to be, the row reads "a world under the Veil". Who fought whom, the outcome, the loot and the Replay button are all still there. Only the address is withheld, and the rule does not depend on the reader's Sensors. The tag picker in chat offers these rows with the same wording. A replay of such a battle shows the fight without naming the place.

What is left out:

• App versions released before this fix do not show these rows at all, until the app is updated. Their Calendar day panel then says the day is only partly shown.
• Chat messages already written keep any address they quoted before the colony was veiled. That includes the big-battle announcement lines. Both are tracked in #6945.

1.427.2 — 2026-10-09

Changed

The Watch Array gives arrival times to the minute, not to the millisecond (#6862)

Before, each fleet a guild's Watch Array listed carried its exact arrival time. A fleet contact also carries its exact arrival time, at every observatory tier, so a guild holding a Watch Array could match even the lowest-tier contact to its Watch row and read the destination planet, which only a tier-3 contact is meant to name.

Now the Watch Array publishes each arrival rounded down to the whole minute, and fleets landing in the same minute are listed by their coordinates, not by their hidden seconds. The panel already showed whole minutes, so the countdown reads the same, except that it can now run up to a minute early, never late. Contacts keep their exact times.

Not closed: when only one Watch row falls in a contact's arrival minute, the minute still matches it. A guild that keeps refreshing the list can also see the first of two same-minute fleets drop off when it lands, which pairs the other one with the contact still in flight, with under a minute to go.

1.427.1 — 2026-10-09

Fixed

The contact sheet no longer promises an intercept the server will refuse (#6924). At Observatory 3 and above, the sheet worked out a meeting point for any flight whose route you had bought, and INTERCEPT stayed live even for flights the server always turns away: probes, exploration and colony ships, raid contingents, a fleet carrying goods between its owner's own worlds, a fleet sheltered by night, novice protection or a pact, and routes inside the starting sectors. Now a flight whose two ends both lie in the starting sectors is refused on the sheet before you tap, using the same sentence the server gives. The note under every solve now says it predicts only the course, that whether the flight can be engaged is decided when you tap and again when the fleets meet, and that a turn or a shelter before then means arriving at empty space. The refusal you get after tapping says nothing left the hangar, explains that the chart shows a course but never a fleet's errand or its shelter, and now covers raid contingents correctly in every interception mode. The sheet still never reveals what a fleet is doing.

1.427.0 — 2026-10-09

New

• Dark Matter packs in the Shop (#6938). Until now Dark Matter could only be bought inside a supporter pack, at 50 a euro, together with VIP points, shards and hulls. The Shop now opens with a Dark Matter section: six packs that pay more Dark Matter per euro and nothing else.

  | Pack | Price | Dark Matter | Bonus | |---|---|---|---| | Dark Matter Vial | €1.99 | 110 | +10% | | Dark Matter Flask | €4.99 | 300 | +20% | | Dark Matter Canister | €9.99 | 650 | +30% | | Dark Matter Reservoir | €19.99 | 1,400 | +40% | | Dark Matter Singularity | €49.99 | 3,750 | +50% | | Event Horizon | €99.99 | 8,000 | +60% |

  Each card shows its bonus over the supporter rate. The packs have no purchase limit and no Star Forge requirement. Like every Shop pack, they need an account linked to Google or Apple, and they give no VIP points, supporter tier or combat bonus.

  Bought Dark Matter does not count as income in the Anchorage's market index, the same as supporter Dark Matter, so a purchase does not move anyone's prices.

  Refunds and disputes:

  • A refund takes the pack's Dark Matter back, as far as your balance holds it.
  • While a payment dispute is open, that Dark Matter is set aside. It comes back if the dispute closes in our favour.
  • The Terms of sale and the refund note in the Shop say so.

  The six card images are cut from the existing Dark Matter medallion: it grows and glows brighter with each pack, and the Event Horizon wears a gold ring. The Shop stays web-only.

1.426.5 — 2026-10-09

Fixed

The Sort and Makes pickers on the building rails now leave a gap between their chevron and the border (#6904) Before, the browser drew the arrow on these pills about 3px inside the right border, and no padding could move it. On the active Sort pill, with its bright cyan border, the arrow looked flush against the edge. The pickers now draw their own chevron, in the pill's colour: cyan when a sort or filter is on, dim otherwise. On a phone it sits 13px in from the border, the same inset the neighbouring chips give their labels, and 12px in on a desktop. Every English and Italian label still fits the pill. Reported in the bugs forum.

1.426.4 — 2026-10-08

Changed

• The AI mark and the Sigil mark in Italian (#6921). Before, players with Italian on saw some marks in English when they hovered over them or used a screen reader. Two marks were affected: the AI mark beside a synthetic empire («Synthetic — played by an AI», «Wraith — hyperintelligent AI») and the Sigil mark. Now they read in Italian everywhere they appear: on names, in chat, in battle results, on the Marshal and on the Market board. In English, every word reads as before.

  The Marshal used to describe an AI player it did not recognise as "synthetic". It now leaves that line plain, as the mark itself does. This only happens on an app version older than the server.

1.426.3 — 2026-10-08

Changed

The number on Overview now counts colonies that went idle since you last looked, and opening Overview clears it (#6925) On the phone, the Overview mark was a hollow gold count of every colony with nothing building, and nothing you did made it go away: measured on production, it was lit for 86–93% of active players at every hour of the day. It also added to the red number on the Empire slot. It now counts only the colonies whose build queue went empty since you last had Overview open, drawn as a red count like the other marks, and opening Overview clears it. A colony you give a build that later goes idle again lights it again. The line under «needs attention» that explained the old number is gone; the Overview's own «build queue idle» chips still list every idle colony. The memory of what you last saw is kept per device: a build queued and finished from another device may not be marked on this one, and on the first run every colony already idle counts as seen.

1.426.2 — 2026-10-08

Fixed

The contact sheet no longer offers to sell you an arrival time you can already see (#6867)

Before, the hint at the bottom of a contact's sheet said Observatory 3 would give "the ETA", and the intercept refusal said Observatory 3 would tell you "when it lands". But every tier already shows the landing countdown at the top of the sheet, so a tier-2 player was offered something already on screen. Both lines, in English and Italian, now say what Observatory 3 actually adds: where the flight came from and where it is going, which is the course an intercept aims at. The refusal also notes that a probe hidden by the Cipher's Blackout never shows its origin, at any level.

A player who already owns Observatory 3 or higher, tapping INTERCEPT on a probe hidden by Blackout, used to be told to build Observatory 3. They are now told that its origin is hidden from every Observatory level and that an intercept needs the whole route.

The Cipher's Blackout card (level 10) now describes the map as it is drawn. On the way out, other players see the probe's mark, a rough heading and its arrival time. An Observatory at 3 or more also sees where it is going. Watched for a while, its track still points roughly back toward the system it left.

Nothing about what can be intercepted, or what each Observatory level reveals, has changed.

1.426.1 — 2026-10-08

Fixed

A pursuit can no longer be aimed at a contact whose course your grid has not plotted (#6863) Before: the game screen already refused a pursuit at a contact seen below Observatory 3 (or a Blackout probe), saying its course has not been plotted. A request sent straight to the server was still accepted, and the pursuit's arrival time was worked out from the target's exact route. Launching and recalling a few pursuits could therefore recover a route the chart had not sold. Now: the server refuses those pursuits too, with the same answer for every contact below that rung. The answer reuses the existing "it cannot be met any more" message, so installed apps need no update. That message's "it has moved" wording does not quite fit this case, which is a known imprecision. Players using the game screen see no change, and so do the AI empires.

1.426.0 — 2026-10-08

New

A guildmate asking for a hand now reaches you outside the Guild screen (#6869)

Before, a request on the guild's help board lit a dot only on the Help board door, inside the Guild screen. A player who never opened that screen never learned that somebody had asked.

Now, while a guildmate's request is waiting and you can still lend a hand today:

• on desktop, the Guild tab in the side rail carries a cyan dot, apart from the gold dot that means a raid forming or a guildmate under attack;
• on a phone, the Guild tab (or the Empire slot, when Guild is not on your bar) carries a dot, apart from the count of guild calls;
• the desktop Home lists "A guildmate has asked for a hand" among its orders and on the Guild card, and the phone Home's Guild pin says "A guildmate needs a hand";
• tapping the order or the pin opens the Help board itself.

Opening the Help board clears every one of these marks; opening the Guild tab on another door does not, because the request was never on screen. A raid forming or an attack on a guildmate always comes first: while one is lit, the tab shows that call instead, the phone pin keeps the guild's name, and the Home routes open the Guild screen as usual so the raid is not hidden behind the board. Nothing lights for a player who has already spent the day's hands, or whose account is still in its first two days.

Suggested by a player in the suggestions forum.

1.425.3 — 2026-10-08

Changed

Last Stand's keystone row reads as three short sentences instead of one that seemed to stop mid-sentence (#6884) The Aegis Doctrine's Last Stand row (level 15) was a single 64-word sentence, four times as long as the other rows on its line. It welded two claims together with a semicolon and ended on a clause that read as if the tooltip had been cut off. It now makes the same claims in three sentences. On defence, no group of yours loses more than a quarter of the hull it brought to the battle in a round, so it stands at least four rounds. The fire the cap turns aside falls on the rest of its row. And each round the cap buys is another round in which part of your guns' fire gets past the attackers' shields, however large their fleet. The rule itself is unchanged. The row is English in every language, as the other keystone rows are. Three edge cases where the row still says more than the engine does, unchanged from before, are tracked in #6919.

Reported in the QoL suggestions thread.

1.425.2 — 2026-10-08

Changed

Your first three buildings go up with one tap (#6795). A new commander's first screen said "Build the Reactor · Build it", but the button did not build: it changed screen and asked you to pick a spot on a map you had never seen, then to confirm. Half of new players left that screen without clicking anything.

Now, for the Reactor, the Extractor and the Refinery, "Build it" builds. The game picks the spot, the step turns to "Reward ready" on the same screen, and the card says what the tap did. You can move the building later from the Base. From the fourth building on, you place buildings yourself, as before.

Also

• Until your first building is queued, the first screen shows that one step. The convoy card and the season line appear right after.
• The new text is in English and Italian.

Nothing changes for anyone past their first building.

1.425.1 — 2026-10-08

Fixed

A forgotten squash subject no longer turns a merge rehearsal into a real merge (#6852) Before: merge.sh <PR> <sha> --dry-run, with the subject left out, took --dry-run as the squash subject and merged the PR for real. That is how main got a commit titled --dry-run on 2026-10-08. Now merge.sh accepts --dry-run in any position. Before any network call, it refuses a subject that is empty, starts with --, is -dry-run, or opens with an en or em dash, and the refusal names the correct usage. A subject that only contains --dry-run still merges. queue.sh add now refuses an argument after the sha and a PR title merge.sh would refuse, and queue.sh drain and withdraw refuse extra arguments, so a stray --dry-run can no longer queue, drain or withdraw for real. Still open: a subject with whitespace before --dry-run, or a blank subject, would still merge. Ships that run the scripts through run.sh pick this up from main. A direct bash merge.sh call from a tree that has not pulled this change, or one run with SHIP_MACHINERY_LOCAL=1, still uses the old parsing.

1.425.0 — 2026-10-08

New

A battle you raided now offers a spy probe on its target (#6864) Before, a spy report had a one-tap button to probe the same target again, but a battle report had no such button. To look at what a raid had left behind, you had to type the target into Missions again or open the planet from the map. Now, on a battle you fought as the attacker of a colony, the Activity feed shows a 🛰 Spy button next to Details, both on the battle's own row and on an archived report card. It opens the quick spy dialog already aimed at that colony, and nothing is sent until you confirm. The button aims only at the coordinates the report already shows. It does not appear on battles you defended, on garrison duty, or on interceptions fought in open space.

1.424.21 — 2026-10-08

Fixed

The advisor's price now colours each resource on its own (#6848)

On a SUGGESTED NEXT card, and on the Command Center line above the cards, the whole price turned red as soon as the colony was short of a single resource. A colony missing only Xenogas read its Ferrolite and its Data in red too, as if it could afford none of the upgrade, while the build list one screen away named the one resource actually missing.

Both advisor prices now use the same per-resource colouring as the Base build menu: each figure the colony can cover is green, and only the short one is red. One visible change follows from that: a price the colony can pay in full now reads green instead of grey.

Reported by a player in the bugs forum, with a screenshot of each surface.

1.424.20 — 2026-10-08

Fixed

«Send from» opens the launch form, at the top of the page (#6846)

Before, «Send from» on an Overview colony card opened the Missions screen (Fleets on a phone) wherever the Overview had been scrolled to. On a tablet that was part-way down the order list. On a phone it was inside the In flight ledger, so you had to scroll back up or tap «+ New mission» before you could compose the order.

Now «Send from» opens Missions at the top. On a phone it also opens on New mission, even when you have saved orders. The colony you tapped is the launch pad, unless you already had an order in progress, which keeps its own pad. An attack heading for you still shows its count on the Alarm pill, and the HUD alarm chip still opens the Alarm.

The other doors on the Overview cards (Base, Colony, Yard, and the rows that need your attention) now open their screen at the top as well. So do the doors on the phone's Empire tab.

Reported by a player on the gameplay forum.

1.424.19 — 2026-10-08

Fixed

The Command Center suggestion no longer vanishes without a word under a category filter (#6847)

Under SUGGESTED NEXT on the Colony tab, the gold Command Center line is shown only under the All and Civic chips, because the Command Center is a Civic building. On Production or Military it disappeared, and since those tabs still listed other upgrades, nothing on screen said a filter was hiding it. One player took it for a fault on their phone.

Now, under Production or Military, a short line stands where the Command Center line was: "Command Center 19 is suggested under Civic.", with a Show Civic button that opens the Civic tab. The Italian interface shows the same line in Italian.

Reported by a player in the gameplay forum, with a screenshot from each device.

1.424.18 — 2026-10-08

Fixed

The Crucible's attempts counter updates the moment a fight ends (#6876)

After a duel or a spar, the Crucible counter in the top bar kept showing the attempts you had before the fight for up to thirty seconds, and only a page reload brought it up to date. The Crucible tab's own Fight and Spar buttons could also stay enabled for a moment after you had spent your last attempt, and pressing one only earned a refusal. Now both take the new count, and the countdown to the next attempt, straight from the fight's result: the top-bar counter and the buttons change as soon as the result appears, and a refresh that was already on its way when you fought can no longer bring back the old number.

Reported by a player in the bugs forum, who also spotted that the Fight button stayed live with no attempts left.

1.424.17 — 2026-10-08

Fixed

The Activity lookup reads a dotted address (#6874) The search field above the Activity feed only understood an address typed with colons: 3.1.1 or 3 1 1 found nothing, with every chip at 0, while 3:1:1 found the rows. It now reads a whole address with dots or spaces the way the mission fields do, and matches it the same way: 1.6.1 finds 1:6:1 and not 1:6:12. A system typed with a dot (3.1) is still not read as an address; type 3:1 for a whole system.

Reported by a player in the bugs forum, with a screenshot.

1.424.16 — 2026-10-08

Fixed

The Guild card on Home no longer tells ordinary members that nobody is active (#6870)

Only a guild's officers and leader can see how recently each guildmate was around. Every other member got a Guild card on the Home tab that read "0 of 28 active" and showed no faces, even while they were chatting with guildmates. The card took "I was not told" to mean "nobody is here".

Now, when the card is not told who is active, it shows how many members the guild has ("28 members") and makes no claim about activity. Officers and the leader see the card exactly as before: the active count and the faces of the members who are around. Who may see guildmates' activity has not changed.

Reported by a player in the bugs forum, with a screenshot.

1.424.15 — 2026-10-08

Fixed

The bottom sheet's close button and the Empire power dialog's close button speak the player's language (#6895). Before, every bottom sheet named its close button "Close" in English, even in Italian, so an Italian screen reader announced it in English; the Empire power dialog's close button showed an English "Close (Esc)" tooltip in every language and had no name besides its glyph. Now both use the same words as the shared dialog close button: Italian players get the Italian name and tooltip, the label changes as soon as the language is switched, and the English is unchanged. The rest of the Empire power dialog is still in English.

1.424.14 — 2026-10-08

Fixed

New commander, colony and raid ids no longer reveal when they were created (#6886)

Before: the ids the game hands out for commanders, colonies and raids began with their creation time, written to the millisecond. Anyone who saw one could decode it. A commander's id, which appears in chat tags and on the boards, gave away the exact moment the account signed up, and the game shows that only to the account's owner. An enemy home colony's id, listed in the War Room, gave away the same moment. A raid's id, which is part of every public battle report, gave away when the raid was opened.

Now: commanders, colonies and raids created from this release on get ids with no time in them. Ids that already exist keep their old form, so accounts, colonies and raids created before this release can still be dated from their id. Guild, battle report and chat message ids are unchanged: the game already shows their creation time to anyone who can see the id.

1.424.13 — 2026-10-08

Fixed

Dialog close buttons are named in your language (#6890) The close button in the corner of the game's shared dialogs had an English name and tooltip in every language. These dialogs include the Campaign refit desk's slot picker, the dispatches box, the building sheet, the profile, the quick spy and the shop's purchase sheets. A screen reader announced the button as "Close" and the tooltip read "Close (Esc)", even in Italian. Both now follow the language you play in, and switch with it while the dialog is open. English is unchanged.

The Italian Shipyard uses the game's word for a refit (#6888) In Italian, the caption under the Shipyard's stat bars used the English loanword for refits. The refit desk's errors, the War Table and the Campaign refit desk all use a native Italian word. The caption now uses that word too. English is unchanged.

1.424.12 — 2026-10-08

Changed

The Campaign refit desk speaks the player's language (#6877) The refit desk, where campaign parts become standing orders, used to be English in every language, including the label a screen reader reads for each class tab's row of filled-bay dots. Now everything on it follows the player's language: the hero and its rules, the class tabs and their dots, the bays, the cooldown chip, the slot picker with its options, refunds and footnotes, and the class Summary. Italian players see the desk in Italian, with decimals spelled the Italian way. Module names, part names and slot names stay as the game names them. One English change rides along: a cooldown that has just run out used to read "refit in ready" for a moment, and now reads "refit in a moment". The Italian refusal for not having enough parts now uses the same word for parts as the desk.

1.424.11 — 2026-10-08

Fixed

A fleet's id no longer tells other players when it launched (#6860) Before, every fleet on the contacts board carried an id. The first characters of that id were its launch time, to the millisecond. Anyone who saw the contact could read when the flight had left, at every sensor tier, Blackout probes included. Now a fleet launched from this release on carries an id that holds no time.

Fleets already in flight when this went live keep their old id until they finish. The two legs of one flight still share a single id. A watcher who sees both legs can still tell when the return leg set off, and can link the two legs at the turnaround. That part of #6860 stays open.

1.424.10 — 2026-10-08

Changed

The chat's Whispers tab, its unread counts and «new messages» now speak the player's language (#6873) In Italian, the chat used to show and say some of its words in English. These were: the tab name «Whispers», the tab's spoken count «Whispers, 3 unread», «new messages» on a room with something unread, and the minimised chat button's «Open chat (3 unread)». They now read «Sussurri», «nuovi messaggi» and «Apri la chat», and the unread counts beside them are said in Italian too. The back button out of a whisper conversation now reads «← Sussurri». Nothing changes in English.

Three icons now carry their screen-reader name in a way every reader announces: the Shipyard's padlock on a locked hull, the Crucible's sparring partner badge, and the dots that count a refit desk tab's filled bays. The bay count is still in English in every language. It will be translated with the rest of the refit desk (#6877).

1.424.9 — 2026-10-08

Changed

A test now guards that a guild page refreshed for new hours always states the leader's hours (#6879)

When an AI leader's hours move, the guild page is refreshed, and the fallback text the desk composes for it adds the leader's hours whenever its own wording left them out. That behaviour was already correct, but no test pinned it: removing the step left every test green. A test now fails if it breaks, so the page cannot go silent about the hours while the leader announces "page updated" in guild chat. Nothing players see changes. Refs #6812.

1.424.8 — 2026-10-08

Fixed

Screen readers now say a whisper conversation's true unread count instead of «99+» (#6868) The gold unread pill on a row of the whisper list, and the one on the chat dock's Whispers tab, show the count capped at «99+». The true count was attached in a way that not every screen reader reads, so some could announce «99+», or nothing at all. Now the pill says its true count («150 unread») as text that only screen readers hear. The cyan dot on a chat room tab says «new messages» the same way. Nothing changes on screen.

On the dock, the count is now spoken in the player's language (in Italian, for example). The tab's own word «Whispers», the room dot's «new messages», and the minimised dock button's «Open chat (3 unread)» are still English. Their translation is tracked in #6873.

1.424.7 — 2026-10-08

Changed

The synthetic empires keep the hours the players keep (#6812)

Before, the AI-badged empires were awake in roughly equal numbers at every hour of the day, and slightly more of them at night (UTC) than in the evening, when most players are online. Now each agent's day is centred on the hours human players are actually around, so more of them are online in the European afternoon and evening and fewer in the small hours. The curve is a dated snapshot of player activity (2026-09-09 to 2026-10-07) and is re-measured each season.

• Gradual roll-out, 2026-10-26 to about 2026-11-15. Each agent starts at its own moment over two weeks and shifts its hours by at most 3 hours a night until it reaches its new schedule. Every day stays one normal day for that agent and every night lasts 5 to 17 hours. No agent's habits jump overnight, and agents never change all at once.
• More than half the agents stay attackable at every hour. An agent's night shelter still opens when it logs off, and its shelter only moves once the game has accepted the new hour, so it is never sheltered while it plays. In the hour with the most agents sheltered, 29 of the 75 piloted agents are (31 before). The rule that no more than half may be sheltered at once holds at the end and throughout the roll-out.
• Guild pages follow their leader. An agent that leads a guild updates its recruitment page once its new hours are in place. The declared hours change only if they were the leader's own, and a description a player edited only has the old hours replaced where they appear word for word. It then mentions the new hours once in guild chat.
• No reason is ever given. While its hours are changing and for a week after, an agent knows they changed and can say when it is usually online now. It is told there is no reason to give and never to invent one (no new job, no move), and its old hours are never put in front of it. The guild-chat line is checked the same way before it is posted.

1.424.6 — 2026-10-08

Fixed

A low-Observatory contact no longer gives away its route by arithmetic (#6856)

Before. A fleet seen at Observatory 1 or 2 carried four fields that let anyone with a script rebuild its route in one or two polls: an exact position re-sampled on every request, the whole route vector as its heading, the exact route length, and the request time as its fix. From those you got its origin, its destination and when it left, which is what Observatory 3 sells. A Blackout probe at Observatory 3 or above gave up its origin with no arithmetic at all.

Now. Below the route rung a contact shows:

• its last fix, on a whole cell, dated at the server's edition (at most 15 s old);
• a bearing on one of 12 directions;
• the distance it still has to fly, rounded up.

On the map the mark stays on that fix while its cone opens 15° either side of the bearing and grows. A mark only glides once you can name its destination. Whether you see a contact, and at what Observatory tier, is also read at the edition, so it changes at most once every 15 s.

Blackout. A Blackout probe's way home is no longer drawn for anyone but its sender. The Cipher card, the contact sheet and the map legend now say what the chart does.

Landed fleets. A fleet that has landed leaves the chart at its arrival.

Agents. A synthetic commander only quotes a route length when the route is sold to it.

Not closed yet.

• #6860: a contact's id still encodes when its mission launched, at every tier, until it gets an opaque per-leg id.
• #6861: a viewer who watches a contact for about 20 minutes can still fit where it is going from its exact arrival time. This is an accepted residual.

Old app versions. Until they update, they still glide a low-tier mark along a bound and draw the old 6° cone.

1.424.5 — 2026-10-08

Fixed

The whisper list speaks your language (#6857) The list of private conversations used to be English in every language: an Italian player whose own last whisper had been deleted read an English "You:" in front of an Italian line. Your own last line, the loading and failure lines, the empty list and the unread count a screen reader announces now all follow the game's language. English is unchanged.

Chat menu messages no longer run off a narrow phone (#6850) The confirmation under a chat message's ⋯ menu and under a player's name was one line centred on the screen. A line wider than the phone was cut off on both sides, and the server's own refusals did this in English too, for example "Only guild messages can be pinned…" on a 320px phone. The message now wraps inside the screen with a margin on each side and grows upward from where it always sat. Messages that already fit look exactly as before. Three Italian pin messages are also shorter, so they stay on one line.

1.424.4 — 2026-10-08

Fixed

Tapping Intercept no longer tells you more than your chart shows (#6537) Before, a refused pursuit could tell you a fleet was a spy probe, or your guildmate's, or sheltered by its owner's night window or novice protection, or flying entirely inside the protected sectors, even when your Observatory had not shown you that fleet's owner or route. A pursuit could also be launched at a fleet that was not on your chart at all. Now a refusal says only what your chart already shows: the owner's side, night and novice protection once the owner is named, and the protected-sector case once the route is shown. Otherwise you get the general sentence. A contact you cannot see is "gone", and a fleet's raid membership is no longer named to bystanders.

1.424.3 — 2026-10-08

Fixed

A deleted chat message reads its line in the player's language (#6843) Until now, a deleted message read "This message was deleted." in English in every language. That was true in the chat, in the quote of any reply to it, and as the last line of a whisper conversation. Since v1.424.2 the delete confirmation has told Italian players that a "deleted" line stays in the message's place, but the line itself was still English. Italian players now read that line in Italian, in all three places. English is unchanged. An app installed before this release still shows the English sentence until it updates.

1.424.2 — 2026-10-08

Changed

The ⋯ menu on a chat message speaks the player's language, report reasons included (#6838) Before, the menu on a chat message was English in every language. That covered every row and note, the delete confirmation, the five report reasons and every confirmation at the bottom of the screen. Now an Italian player reads all of it in Italian: Rispondi, Fissa in cima, Segnala il messaggio…, Insulti, Molestie, Spam, Contenuti illeciti, Altro, «Segnalato. Grazie.». Blocking uses the same words as the menu on a player's name. A report still does exactly what it did and is stored under the same reason. The line a deleted message leaves behind, "This message was deleted.", is still English in every language (#6843).

The merge commit for this change is titled "--dry-run" because of a tooling slip. Its content is the reviewed PR (#6845).

1.424.1 — 2026-10-08

Fixed

The menu on a player's name in chat speaks your language, the AI caption included (#6831)

When you pressed a player's name in chat, the menu that opened was in English whatever language you play in. That covered its rows (Whisper, Invite to guild, Block), the note under Block, the messages after a block or an invite, and the caption under a synthetic player's name. An Italian player who hovered the AI mark read «Sintetico — giocato da un’IA», then opened the menu on the same name and read "Synthetic — played by an AI".

Now the menu follows your language. Whisper uses the same words as the planet sheet. The caption uses the same words as the AI mark beside the name, so the two always agree. English reads exactly as before.

Not included: the ⋯ menu on a single chat message, with its report reasons, is still in English (#6838).

The raid join button counts one ship as one ship (#6837)

When a raid row offered to join with a single ship, the English button read "Join with 1 ships". It now reads "Join with 1 ship".

1.424.0 — 2026-10-08

New

Sector 13 now shows how much Power a garrison needs to earn you points (#6640) Before: the rules said a garrison on a world your guild holds earns you 300 points an hour only at 10% of the full garrison's Power, but nowhere said what that number was. Now: Contested Space, the season guide and the sheet of a world your guild holds all show the current bar, for example "Station at least 190,000 Power here and you earn 300 points an hour yourself". The bar moves with the strength of the field, and the copy says so. When the server does not know the bar yet (the first minutes after a restart), nothing is shown rather than a wrong number.

Changed

AI empires size their Sector 13 watch to that bar, and spend Merit on Chrono Shards (#6640)

• AI empires send a 12-hour watch sized to clear the bar, using at most two thirds of a hangar's fleet. When no hangar within range can do that, they send a third of a hangar, as before. They never send a legendary hull on watch.
• A watch is sized once, when it is sent. If the field grows by more than 15% during the 12 hours, the watch can fall under the bar for the rest of the hold.
• AI empires now spend Crucible Merit on a Chrono Shard when a long wait is worth cutting and no shard they hold fits it. Merit may fill at most half of the weekly shard pool. They never buy a 4h shard with Merit and never more than 2 hours a week, so the other half stays open for shards won in raids and contracts.
• Owner's decision, final: AI empires never get the Annihilator, and they do not buy or open Merit crates.

1.423.0 — 2026-10-08

New

The Galaxy shows the salvage your own battles left (#6762)

Before, Harvest was offered on every free slot, but nothing on the Galaxy said where a debris field was. A trip launched on a guess often came back with nothing to salvage, while the Missions tab already knew from your battle reports where wreckage was.

Now the same estimate reaches the Galaxy:

• System list: a ♻ ≈ n badge in the Planet column wherever your own battle reports put wreckage in orbit, decayed to now and minus what you already lifted. It is explained by a new legend row under the table.
• Planet sheet: one line with the same figure. It says your recovery share is what comes home. Where a colony stands that is not yours or a Beacon's, it says instead that the server refuses a harvest from anyone but the owner. The same applies where your own spy report names a living colony on a slot your scan shows empty.
• Harvest button tooltip: it names the figure when there is one. Otherwise it says salvage lies where a battle was fought, and that your own battle reports are the only sign this screen has.

The figure is the one the Missions tab prints. A fight you were not in leaves no mark: nothing new about other players' battles is published.

Not included:

• The desktop hover tip shows no salvage line. It was cut because the map does not read the battle reports while you hover, so the line would have been blank or out of date (#6830).
• Spy probes do not report wreckage. Whether they should is the owner's call (#6827).
• The list and the sheet have no "still checking" state. While your battle reports have not been read, or the read failed, the map shows no mark at all, which can look like "nothing here".
• The figure comes from your 20 most recent battle reports. A field from an older battle does not show.

Suggested by a player on Discord.

1.422.5 — 2026-10-08

Fixed

The Marshal's launcher speaks Italian (#6818) Until now an Italian player saw the Marshal page in Italian, but the card that launches a strike was still in English. That covered the order buttons, the guild raid rows and why a raid could not be joined, the gathering time, the hint about whether a strike counts for the prizes, the warning that the colony is left undefended, the launch button and the line after a launch. The whole card now reads Italian, in the words the rest of the Marshal page and the guild raid screen already use. The card is headed «Attacco al Marshal». Colony and player names print exactly as typed. The line after a launch now follows a language switch.

In English, three counts of one now read correctly: "Strike with 1 ship" (and the same for opening or joining a raid), "1 seat left" and "in 1 minute". Before, the card said "1 ships", "1 seats left" and "1 minutes". Every other English line is unchanged.

Not included: the guild raid screen's "Join with 1 ships" is still wrong in English, and is left for a follow-up.

1.422.4 — 2026-10-08

Fixed

The AI mark next to a name now speaks your language (#6819) The small mark beside every synthetic empire, the hexagon for an AI player and the triangle for a Wraith, used to give its hover tooltip and its screen-reader name in English in every language: an Italian player heard «Synthetic — played by an AI» beside Italian text. It now uses the words of the language you play in («Sintetico — giocato da un’IA», «Wraith — IA iperintelligente»), the same words the battle alerts already used, and changes with them when you switch language. The mark itself looks the same, and English is unchanged. Still English: the caption under a synthetic player's name in the chat player menu, which is not translated at all yet (#6831).

1.422.3 — 2026-10-08

Changed

AI empires now research Disciplines past level 24, as players do (#6640) Before: once an AI empire finished a Discipline at level 24 it never studied it again. Overdrive was only used by human players. Now: AI empires buy Overdrive levels through the same research queue and on the same terms. A level is bought only after the empire's other research choices, because each level counts for less the higher the line goes. It is bought only when the empire can cover the Dark Matter of every unpaid order already in its queue as well as this level. That rule keeps a parked order from freezing the queue. In chat, an AI empire now gives its real level past 24 (for example Shields level 31, not 24).

Part of #6640: the Sector 13 garrison bar and the Merit Market for AI empires follow in a later release.

1.422.2 — 2026-10-08

Changed

• The Galactic Log and the return sheet in Italian (#6822). Before, players with Italian on saw these in English:
  • the Galactic Log cards in the Activity tab: battles, raids, ambushes, colonies founded, the market's busiest trades, the Eschaton sweep, the season's end and the war week;
  • the "while you were away" sheet.

  Now both read in Italian, with ages like "2 min fa". The galaxy log on the desktop Home follows the language too. In English, every word and figure reads as before.

  These stay English for now:

  • the Calendar's day panel and the event chips in chat, until those screens move;
  • season names, which come from the server;
  • the AI mark's tooltip.

1.422.1 — 2026-10-08

Changed

The rest of the Marshal page speaks Italian (#6791) Before this release, an Italian player who opened the Marshal tab saw an Italian Insignia shop sitting inside an English page. The hero, How it works, the three walls, the prize ladder and podiums, the rules, the standings, the battle feed and the HUD notice for a fallen level were all in English.

Now the whole page is in Italian, in the words the Marshal's other Italian screens already use: the HUD chip, the map board, the refusals and the payout row. A window is called a round, a strike an attack, and a bracket a category. Places read as Italian ordinals. The wall names and bracket names stay in English, as the server gives them. The Italian page calls the Marshal a Wraith commander, not by the word Italian already uses for the top bounty grade, so the page cannot be read as putting a bounty on the Marshal.

The warning about hunting strikers is now one sentence in the language on screen, wherever it appears: on the rules card, on the Marshal launcher and in the Missions attack composer. When a level falls, the screen-reader announcement names an AI empire in Italian too.

English is unchanged, with three small exceptions:

• A one-hour window now reads "1 hour", where it read "1 hours".
• A one-strike window reads "1 strike", where it read "1 strikes".
• A count of 1,000 or more now has a thousands separator. With that many pilots or guilds in a window, the standings boxes would read "1,000" instead of "1000".

Still in English: the body of the Marshal launcher card (#6818), and the AI mark's own tooltip on every board in the game (#6819).

1.422.0 — 2026-10-08

New

Both building lists can be sorted A–Z (#6772)

Before, the Colony tab's building list could be sorted by level or by the next level's price, and the Base build menu had a single «Sort: level ↑» toggle. Neither could put the buildings in alphabetical order.

Now both have a «Sort: A–Z» option («Ordina: A–Z» in Italian). On the Base build menu the toggle has become the same picker the Colony tab uses: off, level ↑ or A–Z. The order follows the name you read, in your language, so the list reorders when you switch between English and Italian («Trade Portal» files under T, «Portale Commerciale» under P). A building's second and later Sites stay right under it. As with the other orders, queued buildings still lead the Colony list and locked buildings still sit at the bottom of the Base menu. Back brings the chosen order with you.

Suggested by a player on Discord.

1.421.10 — 2026-10-08

Changed

When chat translates your message, the game now also keeps the words you typed (#6788)

Before, a translated message in the galaxy or guild channel was stored only as the English translation. If the translation invented a sentence, nobody could see what the player had actually written. Now every newly translated message also keeps the original text beside it. Only its author and the moderators can read it: the API hands it to the author alone, and moderators read it in the database. No screen in the game shows it yet. Untranslated messages, whispers and messages sent before this release are unchanged. If a message is edited, the kept text stays the one behind the first version; a translated edit's own typed words are not kept. The privacy policy now says this in its "Translating chat" entry.

1.421.9 — 2026-10-08

Changed

The cargo stays with the fleet you are choosing: the fleet page carries the load, and a colonisation's endowment comes before its ships (#6760)

On a wide screen, a transport's or a pickup's fleet page used to show «Cargo nothing loaded — set it in the order ›» right under the advisor's «Set the cargo first», so setting the load meant going «‹ Back to the order» and returning. Now the cargo fields themselves are on the fleet page while it is open, between the advisor and the Fleet, and back in the order's column when you close it — the load you typed is kept either way, and so is a supply lane you were drafting.

A colonisation's starting endowment now comes before the fleet on every screen, as a transport's cargo does: the advisor adds the freighters an endowment needs, so the load is set first. The line that used to stand in for the fields above the fleet is gone; what it said — how much the new colony can store of each resource, and how much the hold carries — is now written under the endowment's own heading, and with no Colony Ship or Ark picked it names them and takes you to them. A redeploy is unchanged: its fleet is still chosen first and its supplies after.

Suggested by a player on Discord.

1.421.8 — 2026-10-08

Changed

The Marshal's Insignia shop and the laurel's name on the commander card are now in Italian (#6785)

Before: a player with the game in Italian found the Insignia shop on the Marshal page entirely in English. That covered its title, the balance, each item's name and description, the buy buttons, the lines under each item (how many Insignia are missing, owned, already fitted, nothing to lift), the purchase receipt and the line shown when the shop could not be read. The laurel's tile on the commander card also kept its English name.

Now the shop card speaks Italian, using the words the rest of the Italian game already uses. The shop takes the name the Italian purchase refusals already give it. The rename token takes the Merit Market's name for it, and the refit takes the War Table's. Marshal, Crucible, Wraith, Season Boss and the Insignia currency stay in English, as they do elsewhere. In Italian, figures are grouped with a dot (12.500). The laurel has an Italian name, built like the Crucible's anvil, and the commander card reads it from the same entry as the shop. The two always call the item the same thing.

Players with the game in English see no change. Item ids are unchanged, so every purchase and every equipped laurel stays as it was.

Not included: the rest of the Marshal page (the event header, standings, feed and the notice when a level falls) is still in English in Italian. That is tracked in #6791.

1.421.7 — 2026-10-08

Fixed

The HUD layout check now uses the real width of every chip on the tablet top bar (#6764)

Players see no change: this release only touches tests. The test that keeps the tablet top bar from overlapping used stand-in chips, and two of them were narrower than the real ones: the build/ships/labs counters by 5px and the Crucible bar by 1px. The test could therefore pass a layout that would wrap on a player's screen. The stand-ins are now as wide as the real chips. A new check fails as soon as any stand-in is narrower than the chip it replaces, the night-shelter clock included. The check that keeps 8px between «Go to …» and the # toggle no longer fails a correct layout on sub-pixel rounding, and it still fails when the gap is 1px short.

1.421.6 — 2026-10-08

Fixed

Chat translation no longer turns a short message into a sentence nobody wrote (#6771)

Before: a short message could go to the chat translator and come back as something else entirely. In the reported case, a short English question came back as a one-line threat. That threat was stored under the author's name, shown in the channel, quoted in replies and read by the AI empires.

Now:

• Plainly English messages are not sent to the translator at all. A message whose words are mostly English, with nothing foreign in it, goes out exactly as typed, and it costs nothing from the hourly translation budget.
• Translations are limited by the length of the message. A translation may be at most about twice as long as the original, plus a little room for very short lines, instead of up to 80 characters for any short line. Chinese, Japanese and Korean characters count double, because they carry more per character.

What is still possible, stated plainly:

• A translated line can still contain an invention that fits inside that length limit.
• A line that mixes languages but is mostly English is kept as typed, untranslated.
• The author's original words are not kept beside a translation. That is tracked separately in #6788.

Reported by a player.

1.421.5 — 2026-10-08

Fixed

Ship machinery: the merge queue starts its drain from the machine it runs on (#6803)

The ship sessions run on the queue host itself, and queue.sh always reached the queue host over ssh, which that machine refuses to itself. Every queue.sh add ended ADDED-UNTRIGGERED, every queue.sh status reported the drain as UNKNOWN, and every PR fell back to the manual merge path. Now, on the queue host (its exact full name, as the user QUEUE_HOST names), add and status run the host-side script locally, from that user's home and without the session's git, ship-machinery, queue, gate, merge and GitHub variables, under the same 60-second limit. Every other machine still uses ssh. Nothing about ssh keys or host access changed.

Not included: gate.sh --remote still reaches its gate host over ssh, so from a session on that same host it fails as before. It is opt-in, and the local gate runs on the same machine.

No player-visible change.

1.421.4 — 2026-10-08

Fixed

Ship machinery: two ships building the launcher's cache at the same moment no longer take each other's fresh copy for a tampered one (#6691)

When several ships ran the ship launcher for the first time after a merge that changed the ship scripts, one of them could see a sibling's freshly installed copy, call it "does not match git", and move it aside. That pulled the copy out from under the sibling, which was then refused. The launcher's self-test hit this under load. Now the install itself decides: a run first tries to put its copy in place, and only when that fails does it look at what is there. A correct copy installed by a sibling is kept. Only a copy that is present and wrong is replaced, at most three times, and a run that gives up says so in its own words. The self-test now forces the losing interleaving instead of hoping load produces it.

The ship scripts are tracked per tree. A tree's first build runs that tree's own launcher, before any hand-over to main's, so a tree gets this fix only once it pulls it. Until every tree on a host has pulled, a tree on the old launcher can still knock out a fixed one in the same race. The merge queue's host runs main's launcher directly, and gets the fix at its next start.

No player-visible change.

1.421.3 — 2026-10-08

Changed

• The rest of the Activity feed in Italian (#6793). Before, players with Italian on saw about forty kinds of Activity rows in English. Now they read Italian:
  • exodus offers;
  • chrono shards;
  • faction loot and the Star Forge;
  • bounties;
  • the Eschaton sweep;
  • the season boss and the season's end;
  • destroyers and emplacements;
  • crates and Contested Space;
  • cores, queues, research, contracts and Campaign back-pay.

  In English, every word and figure reads as before. Names stay as they are: players, colonies, guilds, factions, ships, Shop packs and the server's own boss and season names.

  The cards of the Galactic Log inside the Activity tab are still in English. So the beta notice still lists some Activity rows.

1.421.2 — 2026-10-07

Changed

The server's notes on guest accounts now match the hand-off code (#6614)

Nothing changes in the game. Comments in the server code still called a guest's token the only key to the empire, lost for good. Since hand-off codes (#6368), that token can be copied to another browser. The comments now say a guest empire is lost only when every copy of its token is gone, and that its name stays taken until one of the cleanup sweeps deletes the account. The wording players see was already corrected in v1.406.26.

1.421.1 — 2026-10-07

Fixed

Activity can search the whole history of a commander you only know from battle (#6440)

Before: typing the name of someone you had only fought gave you a lookup chip and nothing else. «Search whole history for X» appeared only for names some other row tied to an account, and battle rows carried the name alone. For a combat player that is the most common case.

Now the battle rows written from this release on name each commander together with their account, so the button appears and finds every battle you fought with them. For battles fought before this release, the tab borrows the account from the battle report when that report is among your 20 latest. A battle older than that, against someone no other loaded row names, still shows the chip without the button until newer rows replace it in the feed.

Nothing new is revealed. The battle reports you already receive name the same people the same way.

1.421.0 — 2026-10-07

New

Big battles are announced in the galaxy channel (#6787). When a battle between players wipes out more than 10M in ships, counting both sides, Kepler Observatory posts one line in the galaxy channel. The line says who fought whom and where, who held or broke, and how much went down. A joint raid or a defended world is named by its leader and the number of allies. The coordinates open the map. Only fights with a player leading each side count. Raids on faction worlds, relics and Wraiths stay quiet, and so do small skirmishes. When several big fights land within twenty minutes, only the first is announced and the rest stay in the Galactic log. Picket intercepts are not announced yet.

A battle cited in chat opens its replay (#6787). The battle chip in a chat message used to open only a one-line summary. It now also offers ▶ Replay, the same film the Galactic log plays, and clicking or pressing Escape inside the film no longer closes the card behind it.

1.420.5 — 2026-10-07

Changed

• Faction walls now fire (#6498). The Sector 13 Verge Watch and the roster bastions list a Bubble, a Cannon, Batteries and a Fortress, but none of them had a power plant, so all of those defences fought at zero: no shield to chew through, no opening Cannon salvo, silent emplacements. The attack forecast agreed, so nothing looked wrong. They now fight at full power, in the battle and in the launch forecast alike. At today's garrison sizes the difference is a few percent at break-even, and nearly all of it is the Cannon's shield pierce; a fleet sent with a real margin sees almost no change. Roster bastions recalibrate to each guild as before. The Verge Watch control worlds are slightly harder. Raids already in flight when this shipped land on the live wall.
• Stronger faction garrisons field more kinds of ship. A garrison used to be Battleships, Cruisers and Bombers at every size. It now adds one hull type per tier of power: the Monitor Frigate from 20k, the Orbital Scythe from 200k, the Splinter Lance from 2M, and the Titan Dreadnought from 20M. Below 20k nothing changes. Every garrison is still worth exactly its target power; only the mix varies, and with it how it trades against your fleet. Faction sorties use the same mix, so they now bring these hulls too, and with Titans aboard they arrive later. Existing garrisons switch to the new mix on their next rescale pass (at most six hours).

Fixed

• Garrisons keep tracking their target power. The rescale that resizes bastions and Event garrisons now measures fleet strength, not hull count. With a mix that changes per tier, a hull count could stay put while the target power moved by a third.

Also

• The AI empires price a faction wall at full power too, so they no longer under-send against the Verge Watch.
• Not included: relic wardens have the same unpowered Bubble and Batteries. That is tracked separately in #6776.

1.420.4 — 2026-10-07

Changed

• Last Stand no longer makes a defence almost impossible to break (#6765)
  • Before: a defence under the Aegis's Last Stand lost at most a quarter of what it had left each round, so it never fell. Fights dragged on for dozens of rounds. Splitting the same fleet into garrisons made it last longer still. At equal tech an attacker needed around seven times the defender's fleet to win.
  • Now: no defending group loses more than a quarter of the hull it brought to the battle in a round. It still stands at least four rounds, and falls in the fourth. The fire it turns aside falls on the rest of its row instead of vanishing, so garrisons no longer stack the effect.
  • Measured on equal-tech benches:
    • the full defensive stack drops from about 2.7× to 1.5× the defender's fleet;
    • with four garrisons, from about 6.2× to 1.5×.
  • Overwhelming force still pays a toll: Last Stand's defenders keep firing for those four rounds, however big the attacking fleet.
  • The Doctrine card for Last Stand, and Interlock's, say exactly this.
  • Raised by a top player in galaxy chat, who priced a strike on a Sector 13 base at 2,415 Cutters and 27 Titans. The measurement that followed (#6227) traced 90% of that wall to this one rule.

Also

• Reserve Fleet's second wave no longer turns lost battles around. The Phalanx's level-20 order still holds part of your hangar back until round 3. Its edge came from the old Last Stand, though, and in our benches the held-back wave is no longer better than fielding those ships from the start. A retune of its own is tracked in #6779.

1.420.3 — 2026-10-07

Changed

• A Crucible duel across a tier line is now held at the challenger's own cap (#5652). A Captain at rank 11 can only challenge ranks 8–10, all Commodores, and every one of those duels used to be held at the defender's 22,000 PP cap against a fleet the Captain tier caps at 10,000. That left the seat with no winnable challenge, only a wait for someone above to drop, and the same wall stood at every tier line. Now the duel is held at the lower of the two caps: the Commodore's defence is fitted down to the Captain's 10,000 PP, and promotion is decided by the duel. Climbing further inside the higher tier still meets that tier's full cap. Inside a tier nothing changes, and spars are unchanged.
  • The duel sheet says it: "Held under your Captain cap, not their Commodore one: their defense fields at most 10,000 PP in this fight, fitted down if it was larger." The opponent card shows the defence as you will actually fight it.
  • The bottom seats of a tier are the ones challenged from the tier below, and their defence is fitted down in those duels. The standing panel now shows it on those seats: From the tier below · a 10,000 PP cap.
  • Agents in those seats know it too, and no longer describe their full published roster as the one a challenger from below fought.
  • Installed apps that have not updated still compute the old, higher cap on a cross-tier duel sheet. A roster above your own cap is refused with "over cap", and no attempt is spent. Trim the roster to your own tier's cap, or update the app.

• The Crucible tab says its timers in words. The attempt bar reads "+1 every 2h, up to 5", and the defense editor says "A published defense can be changed every 30 minutes." It had only shown countdowns, and the rule players passed around was "twelve hours".

Not changed, on purpose: the weekly hold points and the top-10 Monday prizes already reset every week. Ranks are not reset weekly: rank is what the ladder pays for, and a weekly wipe would send everyone back into the same tier walls this change removes.

Reported by BryReign, holding rank 11 with a fleet that had no challenge it could win. starman01 agreed and suggested the weekly reset.

1.420.2 — 2026-10-07

Fixed

The Italian launch composer now agrees with itself when a bounty lapses before your fleet commits (#6741)

When a bounty would lapse before your fleet passes the point of no return, the Italian launch composer ended the row by saying the strike is priced as if "they" had never been wanted. That was a masculine plural, but the row names one bounty and one target. The row now says the strike is priced as if there had never been a bounty, which is the form the design desk chose for every Italian surface that says this. The English text has not changed.

1.420.1 — 2026-10-07

Fixed

The Build Advisor's waits now match what your colony actually earns and can store (#6731) The Advisor's "how long until I can pay for this" estimate used a different rate and storage cap from the one the game uses to fill your Warehouse:

• VIP 9 and 10: upgrades priced between the plain Warehouse cap and your VIP cap were treated as never payable and dropped from the suggestions. They now appear, with the real wait.
• Requisition Surge: an upgrade that only fits under the Surge's raised cap is now suggested if you can afford it before the Surge ends.
• A halted colony (after a Ruin, or on declared leave) was still offered upgrades as if they would be paid within the hour. Now it is offered only what its stock already covers. The board says, first, that production is halted.
• Exodus shock, Disorder, Overdrive, power imported over the grid, and the Dynamo brownout grace hour now change the wait the way they change your production, so the order of the suggestions reflects them.

The Colony tab's resource cards and building rows read the same figures as the Advisor, so the two can no longer disagree.

The table shows how much the suggestions moved. It compares 900 generated colonies per case, before and after the fix:

| Case | Waits changed | Order changed | First suggestion changed | |---|---|---|---| | ordinary colony | 0 | 0 | 0 | | VIP 9 | 4 | 4 | 0 | | Requisition Surge (6h) | 0 | 0 | 0 | | exodus shock | 381 | 11 | 2 | | Disorder | 378 | 6 | 2 | | Overdrive | 368 | 5 | 0 | | production halted | 384 | 384 | 305 |

The AI empires use the same suggestions to choose what to build first, so their choices change the same way.

1.420.0 — 2026-10-07

New

Your fleet now tells you when the hunt it is flying on has closed, while you can still turn back (#6689)

You launch a solo attack on a wanted target. Mid-flight, someone else claims their bounty, or it runs out. The strike then lands on a head that is no longer wanted, and it is priced as if they had never been wanted. If they were far weaker than you, that means a bounty on your own head. The rule bills this deliberately: the recall stays open until ten minutes before impact, so pressing on counts as a choice. On a two-hour flight, though, that is up to 1 h 50 m in which nothing told you there was a choice to make.

Now, while your fleet can still be recalled, its card on Missions and Overview shows a gold line right above the Recall button. It says that the hunt is over, whether the bounty was claimed or lapsed, and how long you have left to call the fleet back. The line disappears on the same tick as the button. When a claim closes the hunt, every fleet of yours still able to turn back also gets a «Hunt closed mid-flight» entry in your activity log, under Combat → Wars. That includes your own other waves, so you can read it later if you were not looking.

What has not changed: what a strike costs. A closure after the point of no return is still forgiven once per hunt. A closure before it is still billed. This release only makes sure you can see it coming.

Not included yet:

• A phone notification when the hunt closes. That needs a decision from the game's owner.
• Attacks launched under the Veil. They record no target, so their card stays silent; the strike is still judged the same way when it lands.

Reported on Discord by a player who was billed this way several times.

1.419.3 — 2026-10-07

Changed

• The Wraiths' shipyards are building warships again (#6631). Every Wraith had reached the arm phase, but the strikers' shipyards sat empty and some held hundreds of millions in resources. The autopilot measured hangars already thousands deep against a starting cap of 120 ships per type, and read every one as full. Each striker's fleet ceiling now grows with its bank, as it already does for the other synthetic empires. Every order still fits within a four-hour shipyard lane. The same switch that stops scaled orders for the other synthetic empires now stops them for the Wraiths too, though it has to be set on the operator's machine as well.

  Not included: the scaled ceiling stops at 2,400 ships per type. A striker past that in every offensive hull goes quiet again, and all strikers converge on the same figure. Both are open in #6633.

  This takes effect when the operator's Wraith checkout moves to this release: the server deploy does not run the Wraiths.

1.419.2 — 2026-10-07

Changed

• Every battle replay can be watched the moment the battle ends (#6752). Until now, a battle you had not fought stayed locked for its first 24 hours: tapping ▶ Replay on a fresh clash in the Activity feed showed "replay opens in 4h", and the galaxy's half of the feed padded itself with day-old battles marked "Replay now available", because those were the only ones a stranger could watch. Now any battle in the feed plays as soon as it is fought. The galaxy still sees the public cut of it, without the fighters' home coordinates, research or tree; the people who fought it still see everything.

  The countdown, the "Replay now available" label and the "a day after it is fought" line at the top of Activity are gone, in English and Italian. The galaxy's half of the feed no longer adds older battles below the fresh ones, so what it shows is simply the newest clashes.

  Worth knowing: a fresh replay shows the defender's formation and which attacking ships survived, while those ships may still be flying home. The wait used to hide that; it no longer does.

1.419.1 — 2026-10-07

Fixed

• Your supporter bonus now shows up in your own empire's attack, hull and shields (#6747). Supporter tiers with a combat bonus (Xenogas +1%, Data +3%, Dark Matter +5%, Eschaton +10%) have always fought with it, but the Power details dialog (⚡ on your player card) never listed it, so the Attack, Hull and Shields totals there were lower than the ones that actually fight. An Eschaton Supporter with Energy Weapons 4 read Attack +20% while fighting at +32%. Each of the three stats now ends with a row named after your tier, such as Eschaton Supporter +10%, and when other bonuses sit above it the row says it multiplies them, which is how combat applies it. The totals now match what your fleet fights with in every battle except Crucible duels and spars, which do not apply the supporter bonus today (#6769). Players with no bonus tier see no change, and the bonus itself is unchanged.

  On an app that has not updated yet, the row reads "Eschaton Supporter Lv 1"; the figure is the same.

  Reported in-game by a player who could not find the bonus anywhere in their empire's stats.

1.419.0 — 2026-10-07

New

• The VIP ladder goes to 20 (#6653). It used to stop at VIP 10 (€1,000 lifetime), and past that nothing moved. Ten more levels now run from VIP 11 at €1,500 to VIP 20 at €20,000. Each one keeps stacking the same permanent perks:
  • every level raises production and research, up to +45% production and +40% research at VIP 20;
  • shipyard, fleet speed, storage and Dark Matter production keep rising too;
  • a second extra market slot comes at VIP 15 and a third at VIP 20;
  • the daily Dark Matter stipend grows to 600.

  None of it adds a combat stat, and VIP 1–10 are unchanged. Nobody's level moves at this release: the ladder past 10 starts at €1,500.

• New crests and portrait borders for VIP 11–20. Each level has its own colour: sapphire, emerald, rose gold, solar flame, diamond, void, pulsar, supernova, galaxy, and an eclipsed sun at 20. Halos appear from VIP 16.

Changed

• The Shop's VIP line now says "+2 market slots" when more than one is held.

1.418.2 — 2026-10-07

Changed

The Crucible tab speaks Italian (#6714) Until now the Crucible tab was in English whatever language you had picked. That covered the stage, the target cards, the target lock, the defense editor and its cap meter, the result screen and its class breakdown, your standing, the ladder, the Monday prizes and every explanatory line. All of it, screen-reader labels included, now follows your language. The Italian uses the same words as the rest of the game's Italian screens: la scala, la posizione, il grado, il tetto di tonnellaggio. Your ships in a duel are your «flotta», and a spar is an «allenamento». The Monday payout is named by its day. Lore names stay in English: Crucible, Merit, Wraith and the ranks from Cadet to Admiral. The two battle-line panels on the tab used to stay in English inside the Crucible, and now follow your language too. That includes the saved-formations button, which in Italian now reads «Salva con nome» here as on every other screen; v1.415.7 said the Crucible showed it in English in every language, and that is no longer the case. Five Italian refusals about your roster now call it a «flotta» too, so the one that sends you to the cap meter can no longer be read as pointing at the battle-line panel. In English the tab reads exactly as before. One layout detail changes on phones in every language: a result word longer than eight letters is drawn smaller. In English every result word is seven letters or fewer, so nothing changes there.

1.418.1 — 2026-10-07

Fixed

• The Emergency crate no longer grows with your own Warehouse (#6009). After a real raid, the guild shop's Emergency crate paid 12% of your home colony's Warehouse capacity. For the largest Warehouses that came to about 3.2 billion resources per crate, 6.4 billion a week, which made being attacked profitable. The crate now pays 12% of whichever is smaller: your own Warehouse, or the median home-colony Warehouse among the empires active this week. If your Warehouse is below that median, nothing changes for you: you get exactly what you got before. Above it, a bigger Warehouse no longer makes the crate bigger. Price (30 Scrip), the two-per-week limit and the 48-hour raid window are unchanged.

  Reported by a top-three War-week player in guild chat, who pointed out that a stockpile, a raid and an Emergency crate added up to more than a billion resources.

  Not included: the weekly Echelon reward has the same shape, a share of your own total Warehouse capacity. It is tracked separately in #6679.

1.418.0 — 2026-10-07

New

• The shipping process now measures itself before it changes itself (#6754). Nothing changes in the game. Every update to Novarchy is reviewed by several automated checks, and until now nobody could tell which of them earned their time. A new tool reads the record of every past release, reports what each review cost and caught, and can propose one controlled change at a time. A change is kept only if, over the next releases, it saves more time than chance alone would explain, and no safety signal gets worse. Its verdicts are advice that a person reads. The steps that protect the live game cannot be changed this way at all. Today it proposes nothing: no review step can yet be shown to be safe to cut.

1.417.7 — 2026-10-07

Also

• The layout checks that run before every release now measure the game's own fonts. These checks render screens in a headless browser that cannot reach Google Fonts, so they used to lay text out in whatever substitute font the machine happened to have. Two checks on the HUD compare text widths with fixed sizes. After the build machine was reinstalled they failed on every release, including releases that never touched the interface. The checks now load the same Chakra Petch and IBM Plex Sans the game serves, and they fail loudly if those fonts do not apply. Players see no change. (#6755)

1.417.6 — 2026-10-07

Fixed

• Glossy ships in battle replays no longer glitter with bright pinpoint glints (#6753). After the last update took the white glare off the bright hulls, the Annihilator, the Destroyer and the lit edges of the big ships still sparkled: tiny points of reflected light, each with a glowing halo. Those hulls have an almost mirror-like finish, and the replay's lights turned every edge into a glint. In replays their finish is now satin instead of mirror. The panels still read, the glints and their halos are gone, and engine lights, explosions and weapon fire glow as before. The Shipyard's 3D view is unchanged.

1.417.5 — 2026-10-07

Changed

The desktop settings gear, the Discord link's tooltip and the colony dropdown's hotkey hint now follow your language (#6391)

Three small pieces of English were left for Italian players:

• the gear that opens the Empire menu on desktop, and the menu itself (its screen-reader name, plus a tooltip listing the night window, leave and audio);
• the tooltip on the Discord button (the Empire menu, the phone settings sheet and the login screen);
• the tooltip on the desktop colony dropdown that explains the 1-9, 0 and C / Shift+C colony hotkeys.

All three are translated now. English players see exactly the same text as before.

Still open on #6391: the account-deletion strings, which are held for a separate decision, and the Marshal laurel's name on the commander card, which waits for the Marshal shop to be translated.

1.417.4 — 2026-10-07

Fixed

• Bright ships in battle replays no longer turn into glowing white blobs (#6745). The Bastion, the Cutter and the other near-white hulls used to burn out to a white shape with a halo around it in replays, worst when the ship sat close to the camera. Their lit plating was as bright as an explosion, so the replay's glow effect picked up the whole hull. The brightest panels on those hulls are now toned down in replays only. Plating detail shows again, the hulls still read as white, and engines, explosions and weapon flashes glow as before. The Shipyard's 3D view is unchanged.

1.417.3 — 2026-10-07

Fixed

• Shop: supporter pack cards no longer look cut for a supporter who already holds the tier. Before, once you held a tier, every card dropped its frame/emblem and attack/HP lines and listed only what a new purchase would add. That read as if the packs had been reduced; they had not. Now every card lists the pack's full contents. Lines you already own say so: "already yours", or "your +10% already covers it" when a higher bonus is held. Contents and prices are unchanged.

1.417.2 — 2026-10-07

Fixed

• The merge queue's status line says the drain's state once (#6738). Nothing changes in the game. queue.sh status printed the drain's state twice on one line; it now prints it once. This was the first change to reach main through the new merge queue.

1.417.1 — 2026-10-07

Changed

• The Voidlance Cutter now shows its full 3D model (#6725). The Cutter used to render a decimated model: 113k triangles and 1024 textures, with softened spikes and panel edges. The Shipyard card and the battle replay now load the full model, at nearly 2 million triangles and 4K textures, the same tier as the Annihilator and the Praxis Interdictor. Its card art and engine glow were re-rendered from the new model, so the card opens on the same ship it turns into.

  The model is now about 12.6 MB instead of 1.1 MB. On a slow connection the first replay with a Cutter may show it as a flat illustration while the model finishes downloading; later replays use the cached model.

1.417.0 — 2026-10-07

New

A building you cannot afford yet now says how long until you can (#6707)

On the Colony tab, a building row that cost more than the colony held turned its chips red and showed only the shortfall (3,298 Fe −3,298). You had to divide it by your hourly production yourself, for each resource. The row now adds one line under the chips, for example Affordable in 59m at current production. The figure is the wait for the slowest resource, worked out from the same production and storage cap the resource cards show. It assumes your current production: an upgrade finishing in the queue, a raid or a trade will change it.

During a Requisition Surge, a price that only fits under the surged cap says Affordable in 2h at current production, if spent before the surge ends. Resources above your normal cap are only kept if the game saves them while the Surge is still running, and spending them before the end does exactly that.

When waiting can never cover the price, the line says why, in gold:

• Fe cost exceeds the warehouse cap (12,800): you cannot store that much, so it never accumulates. The cap is the one the resource card shows.
• Fe cost exceeds your normal ceiling (12,800) — the surge ends first: during a Surge, the price fits only under the surged cap, and the row will not be covered before the Surge ends.
• No Xe is being produced here: nothing in this colony is producing that resource right now (none built, powered down, no power, or production halted). A trickle too small for the resource card to show still gets a time.

The line does not appear on rows the Command Center caps. Waiting cannot unlock those, and the note under them already says what will.

Suggested by a player on Discord.

1.416.0 — 2026-10-07

New

• One merge queue for every ship (#6727). Nothing changes in the game. Updates to Novarchy now reach main through a single queue, one at a time and in arrival order, instead of each change re-testing itself against a target that kept moving under it. Each change is still tested together with what it lands on before it merges, and it now gets there without starting over every time another change lands first. Deploying stays with whoever ships the change.

1.415.10 — 2026-10-07

Changed

The Season tab's Race lane speaks Italian (#6704) The Race lane of the Season tab used to be in English whatever language you had picked. That covers the clock and the sector map, a world's sheet, the "if the season ended now" strip, both podiums, the player and guild boards, the "How points are won" tiles and the close. All of it, screen-reader labels included, now follows your language. The Italian uses the same words as the rest of the game's Italian screens for contested worlds, garrisons and Dark Matter. Lore names stay in English: the Verge, the Destroyer, Eschaton. In English nothing on the lane changes except one layout detail on phones: the guilds prize tab now shows «Guilds» with «bigger» on a small line beneath it, instead of «Guilds · bigger» on one line.

1.415.9 — 2026-10-07

Changed

The clock on a Colony building row now says «Build time» (#6706)

Each building row on the Colony tab ends with how long its next upgrade takes: a small clock and a duration. Until now that clock named itself only in a tooltip, which never opens on a touch screen, and in a label only screen readers hear, so on a phone it was a clock and a number. The row now reads Build time 🕓 14m 40s (in Italian, Tempo di costruzione), on phones and on desktop alike. It is the same name and the same figure as the Building Sheet's Build time. Rows that show no clock (an upgrade already running, a building at max level) show no label either. On an unusually narrow screen only the words can shorten; the duration is never cut.

Suggested by a player on Discord, who asked the desk what the timer was and then said it needed a label that says "BUILD TIME".

1.415.8 — 2026-10-07

Fixed

The Shipyard keeps your build count after an order, so repeating it is one tap on BUILD (#6543)

Since the +5 / +50 presets started adding up, the Shipyard put the count back to 1 after every order. Ordering the same 123 hulls again meant tapping ↺ and then BUILD, and on another colony (↺ remembers each colony's own last order) typing the number again.

Now the number you ordered stays in the box. BUILD orders it again, on the same colony or after you switch to the next one. Right after an order the presets read 5 and 50 instead of +5 and +50: a tap sets the count rather than adding to what you just ordered, so +50 after an order of 123 gives 50, not 173. As soon as you touch the count (type, ±, a preset, Max) they go back to adding. ↺ still restores a colony's last order, the same way. An order that fails leaves the count as you typed it.

Reported by a player on Discord.

1.415.7 — 2026-10-07

Changed

The battle line's save button says "Save under this name" (#6713)

Under every battle line, the box of saved formations had a button that said "Save formation". On the Defense Planner it sat under "Defensive line" and above "Save defense plan", and on the Crucible above its Publish button, so it read as saving the line on screen. It never did: it files a named copy in your saved formations, and the line itself is still saved by "Save defense plan" or by publishing your Crucible defense, exactly as before. The button now says "Save under this name", the name you type in its field. On a narrow phone it sits on its own line under that field. The Crucible shows this button in English in every language.

1.415.6 — 2026-10-07

Fixed

A contested world's sheet no longer says the capture award is paid once (#6718)

In the Race tab, a contested world's sheet showed Taking it · +1,000 once. That was wrong: the award is paid on every capture. A guild that takes a world, loses it, and takes it back is paid the 1,000 again. The sheet now reads +1,000 each time. On narrow phones the figure wraps as "+1,000" / "each time" instead of leaving "time" alone on a line.

Retaking a world your guild still holds pays nothing, and neither does a battle that cannot flip the world: one under the one-hour lock, or one won by a guild already at the cap. None of those is a capture.

1.415.5 — 2026-10-07

Fixed

The Crucible roster editor says which hangar it draws on (#6554) The Crucible fields your fleet from your home colony's hangar alone. Before this fix, nothing on the sheet said so. A commander whose heavy hulls were parked on an outpost saw a short list against a much larger cap and read it as the ladder holding them to their own bracket. It does not: a duel is fought under the defender's cap.

Every sheet that lists your ships now carries one line above the list naming the colony: "Drawn from <your home colony>, your home colony. Ships stationed at other colonies cannot be fielded here until they are redeployed home." That covers the duel, the spar, the defense editor, the join form and the Defense segment on a phone. An empty hangar now names your home colony too, where before it said "This colony", which was wrong whenever an outpost was on screen.

The Crucible still fields from the home colony only; this release changes what the sheet says, not which ships you can field. The tab is still English-only; translating it is tracked in #6714.

1.415.4 — 2026-10-07

Fixed

The Italian saved-formations button now says «Salva con nome», in the same verb as its heading (#6551) In Italian, the saved-formations box under the battle line titled its list «Schieramenti salvati» and showed «Salvataggio…» while saving, but its button said «Memorizza», so pressing it changed verb mid-press. The button now reads «Salva con nome», the familiar way to save a copy under a name. It names no object on purpose: it saves a named formation to your list, never the line on screen. On the Defense Planner, under «Schieramento difensivo» and above «Salva il piano», it can no longer be read as saving the plan's line. It is no longer than the English label, so the row fits a phone as before. English is unchanged.

Reported by a player on Discord.

1.415.3 — 2026-10-07

Fixed

A world's sheet no longer makes the capture award look like a second hourly rate (#6609)

When you opened a contested world in the Race tab, its sheet showed «Pays its holder: 300 / hour» right next to «Taking it: +1000». The second figure carried no unit of its own, so it read as a higher rate per hour, and the guild banner's «+900 points an hour» for three worlds looked as if it came from an old rate. The sheet now reads «300 an hour» and «+1,000 once». Nothing about the payouts changed: the guild still earns 300 points an hour for each world it holds, plus 1,000 each time it takes one.

The Race tab is still English-only. Its translation is #6704.

Reported by a player on Discord.

1.415.2 — 2026-10-07

Changed

An idle shipyard is no longer a footnote behind an idle build queue (#6651)

When a colony had both its build queue and its shipyard empty, the Overview showed «build queue idle (+1: shipyard idle)». That happened on most colonies where both queues sat idle, so the shipyard looked like it was never reported at all. Both are now one sentence: «build queue and shipyard idle» («coda edifici e cantiere fermi» in Italian). When something more urgent leads, such as a full warehouse, the pair follows it as one item: «(+1: build queue and shipyard idle)». The Empire tab's «Needs you» list uses the same wording. Tapping the row still opens the colony screen, which lists both queues. A colony whose only problem is an idle shipyard still opens the Shipyard.

Not included: the number on the phone's Overview tab still counts only colonies with nothing building. Whether an idle shipyard should count there is #6701.

Suggested by a player on Discord.

1.415.1 — 2026-10-07

Changed

A Harvest aimed at a field your own last trip found empty now warns before it flies (#6573). The Harvest form used to send the fleet with no remark, even when your previous harvest at the same coordinates had just come home with nothing. It now shows a warning, never a blocker, when your last harvest there came back empty in the last 72 hours, none of your recent battle reports shows wreckage there since, and none of your harvests or attacks is on its way there or has just landed there with its result not yet in. The line speaks only of the reports you hold: a fight you were not in leaves you no report, so what is there now is unknown and the launch stays open. One case can still show the line for a single refresh: an attack of yours whose fleet was wiped out, until its battle report arrives.

1.415.0 — 2026-10-07

New

Synthetic empires notice when their guild loses or takes a contested world (#6669)

Before, when a contested Season Event world changed hands, the AI-badged empires in both guilds carried on as if nothing had happened. They held no grudge, made no attempt to go back for it, and said nothing in guild chat, while their human guildmates read the loss in Activity.

Now:

• They remember who took it. Every member of the losing guild holds it against the player who beat the garrison. They are colder toward that player, but a world changing hands is ordinary Event play: captures alone never make them attack someone far smaller than themselves.
• They want the world back. Once the capture lock lifts, the lost world is the first one they look at, for the next six hours. Whether a fleet actually goes still depends on what their own forecast says they can win.
• They may say something in guild chat. On a loss anyone may speak. On a capture only the boastful ones do: the captor in their own voice, guildmates less often and naming who did it. At most two voices per guild per world in two hours, and only while that empire is awake. No line promises a counter-attack or a retake.

1.414.0 — 2026-10-07

New

The launch form says what your fleet fights with: attack, shield and hull (#6454). Picking ships for an order used to tell you how many hulls you were sending and nothing about what they were worth. The fleet page's totals under the ships counted attack, hull and hold but had no shield, so a fleet with no shields looked the same as a shielded one. The totals now print shield between attack and hull, and on a defence they print the shield your ships actually defend with. On a desktop, the order's Fleet column now shows the same three figures under your ships, in the same words and numbers as the fleet page; clicking them opens it. Hover the line to see what the figures include (your research, tech tree, refit, mastery, supporter bonus, Wraith refit and stance) and what they leave out (shield penetration, guild war bonuses, counters, formation rows, the Ward and the Phalanx on a defence, and the target's defences). They describe what you send, not how the battle will go, and there is no comparison with the target. On a phone and a narrow desktop the fleet page's totals are already on the form, so nothing new appears there.

Fixed

The fleet page no longer undercounts a supporter's fleet or a refitted corvette's shield (#6454). The figures under your ships, and on each ship card, left out the supporter bonus (up to +10% on attack, hull and shield) and the Wraith refit (+12% shield on up to 40 corvettes), although battles, the battle forecast and the Shipyard already counted both. They are counted now, and the line under the totals names each of them, so the bonus is no longer credited to research. The Marshal's strike screen gains the shield total but not these two terms, because the strike's score counts neither.

1.413.2 — 2026-10-07

Fixed

The galaxy map keeps your Layers, Places search and Signals sort while you play (#6602). These used to reset every time you left the map: a trip to another tab, or any system view, including opening a Signals row and pressing «Political map». Every layer came back on, the Signals board went back to sorting by tier, and the Places board lost its search and the band you had opened. Now the map comes back the way you left it for the rest of the session. A page reload or a logout still starts from the defaults, so the next player on the same device never inherits your switches. A Places band is only remembered if you opened or closed it yourself, so a relic you win later still opens its own band.

1.413.1 — 2026-10-07

Fixed

The guild log no longer prints an engine id for an entry your app does not recognise yet (#6675) Every new kind of guild-log entry reaches every app, including installed ones built before that kind existed. Until now such an entry showed its internal id, for example "Vex: event_control_lost". From this version, an entry the app does not recognise reads "Vex: activity logged." ("Vex: attività registrata." in Italian), the same wording the Activity feed already uses for its own unrecognised entries. This protects apps built from this version onward. An app installed before it keeps the old line for new kinds until it is updated.

1.413.0 — 2026-10-07

New

A guild that loses a contested world is told on its phones (#6688)

Before this, a lost contested world showed up as an Activity row and a guild-log line. Nothing reached the phone, so members who were not in the app found out late, sometimes after the hour in which the world could not change hands.

Now every member of the guild that held the world gets a push in two cases: a rival takes the world ("Ember Crown lost"), or a guild at the cap wins there and knocks it neutral ("Ember Crown torn away"). The push names the world and who took it, along with their guild's tag and name. If the attacker is synthetic, the push says so in words.

• Who gets it: each member of the guild that lost the world. The capturing guild gets no push.
• How often: at most one of these pushes per player per hour. A second world lost inside the hour is still listed in Activity, but it does not ring the phone again.
• Tapping it: opens Activity, where the row shows what the world was paying and when it can be taken again.
• Same rules as every other push: it counts toward the daily limit, never rings while you are in the app or during your night hours, and is written in your language (English or Italian).

1.412.1 — 2026-10-07

Changed

A world freed by a guild disbanding can't be taken for an hour, as after any capture (#6676)

When a guild that held contested worlds disbanded (its last member left, the reaper closed an empty guild, an account was deleted, or a move emptied it), each world went back to neutral with its capture lock lifted. A world taken at 12:00 and freed by a disband at 12:10 could then change hands again at 12:15, inside the hour the losing guild had been told it could not. Now a disband restarts the lock: the world is neutral and sealed for a full hour from the moment the guild dissolved, and the board shows it as locked until then, the same way it shows a world a full guild tore off its holder. Nobody scores on the freed world during that hour, and nothing about points, awards or the per-guild cap changed.

1.412.0 — 2026-10-07

New

Orders over the map on phones and tablets (#6604)

Before, on a phone, a tablet, or any window narrower than 1180px, the planet sheet's Attack, Defend, Redeploy, Harvest, Colonise and Explore left the Galaxy for the Missions tab. So did the Signals board's «Send with escort…».

Now, on every window under 1180px, phones included, they open the real Missions composer in a full-size sheet over the map. The v1.409.0 note that phones were unchanged no longer holds.

• The sheet: a «‹ Map» button and the order's title, such as «Attack · 3:142:7». On a phone the form is the Missions tab's own, with Send stuck to the bottom of the sheet; a tablet gets the three-column layout.
• Launch from: the colony picker is on the form, so the launching colony is never picked silently.
• After a launch: the form stays. A receipt line appears above Send, with «Show it», and Send becomes «Send another fleet».
• Closing: «‹ Map», Back or Escape closes the sheet and reopens the planet's own sheet. A tap on the dark border around it does not, so a half-written order is not lost to a stray tap.
• Rotating or resizing: the order stays open and its fleet is kept.

Windows 1180px and wider keep the drawer from v1.409.0. Installed iOS and Android apps get this with their next build.

1.411.8 — 2026-10-07

Fixed

The phone Settings sheet shows each device setting once (#6611)

On a phone, the Settings sheet could fill up with the same row repeated over and over: Replay diagnostics (and, in recent builds, Camera shake too), one copy under another, with more appearing every time the sheet was opened and closed. Each row is now on the sheet exactly once, however often you open it, and it still shows the value your device holds each time you open the sheet.

Reported by a player on Discord, from the iOS app.

1.411.7 — 2026-10-07

Fixed

A contested world changing hands is now written down for both guilds (#6608)

Before, when a contested world changed hands, no one got a line. A guild that lost a world earning it 300 points an hour found out from the map; on the day this was reported, one guild lost all six worlds that way. The guild that took a world got no line in Activity either.

Now:

• The captor reads "World taken": the +1,000 award, the rate the guild now earns, the hour in which the world cannot change hands, and the rule that a garrison of their own there earns them extra points.
• Every other member of the capturing guild reads "World taken by your guild", with the captor named. That hour is their window to station a garrison.
• Every member of the guild that held it reads "World lost": who took it, for which guild, what it was paying, and that after the hour any guild under the cap can take it.
• A world torn away by a guild at the cap reads "World torn away" for the guild that held it: the world is neutral and held by nobody.
• Both guild logs record the change.

The rows sit under Combat, next to the battle that caused them. Who wins, and what a capture pays, are unchanged.

Reported on Discord by a player.

1.411.6 — 2026-10-07

Changed

• Notifications in Italian (#6659). Before, players with Italian on saw these in English:
  • the dispatches panel behind the bell;
  • the notification switch;
  • three messages the game writes to a player: the colony ruin notice, the removal from a guild for inactivity, and a move to another guild by the team.

  Now the panel reads Italian, and those three messages arrive in the player's language. Pushes now also follow the device's language for players who left the language on Automatic. Before, they arrived in English unless Italian was picked by hand.

  In English, every word reads as before. Dispatches the team writes by hand and the daily "What changed" digest stay in English. The beta notice now says so instead of listing notifications.

1.411.5 — 2026-10-07

Changed

• The record of every ship now uses one name per review, and says where its review depth came from (#6673). Nothing changes in the game. Each ship records which reviews ran and what they found. Until now a review could be named any way: the 4,279 recorded review runs used about 140 spellings for 8 reviews, so no review's track record could be read. Reviews are now named from a fixed list; old spellings are grouped under the right name when the record is read, without rewriting history. Every ship also records whether its review depth was printed by the tooling, worked out by hand or guessed.

1.411.4 — 2026-10-07

Fixed

Spy reports on the contested worlds now name the world, not the garrison (#6557)

Until now every spy report on one of the six contested Event Sector worlds read "The Verge Watch", the neutral garrison that holds a world nobody has captured. Two reports on two different worlds looked identical in your Reports list, and the only way to tell them apart was to re-read the coordinates.

They now carry the world's own name, the same one the map shows: [13:9:12] · Hollow Meridian, [13:15:15] · The Long Silence. Five more places follow the same rule: the launch finder's Spied rows, the recent targets on the launch pad, the Scouted Systems list in Missions, the colony end of a supply lane on the report, and the Activity search. Searching "Hollow Meridian" now finds that world's reports. "The Verge Watch" still finds them too, in the Activity search and in the launch finder. Spy reports on every other world still name their commander.

Battle reports still say you fought The Verge Watch, because that is who you fought.

Reported by a player on Discord.

1.411.3 — 2026-10-07

Changed

• Every ship now runs the same release machinery, whichever branch it ships from (#6667). Nothing changes in the game. The scripts that review, merge, deploy and tag each release used to run from whatever copy the shipping branch held, so ships running in parallel from branches of different ages ran different versions of them. They now always run the current copy from main, which is checked against git before each run, so a fix to the shipping process applies to the next ship from any branch.

1.411.2 — 2026-10-07

Changed

The commander card has one name: Commander profile (#6655)

The card that opens when you tap your portrait used to call itself "Commander identity", and so did the hints on your portrait, your name and the Guest chip. Every other screen that sends you there (logging out, moving to another browser, the DM Market and Merit Market receipts) already said "commander profile". Now the card's title and all of those hints say "Commander profile" too.

In Italian the card used to have three names: «Identità del comandante», «profilo comandante» and «profilo da comandante». Now it has one, «Profilo comandante», on the card's title, the header hints, the DM Market and Merit Market receipts, and the "you already own this" message.

1.411.1 — 2026-10-07

Changed

• A guild at the Sector 13 cap can no longer attack the contested worlds it does not hold (#6618). Before, a guild already holding the most contested worlds it may hold could still send fleets at the others. A win tore a rival's world away and left it neutral, sealed for an hour against everyone, and the attacker kept the battle's points. One guild could keep its rivals at zero worlds without holding any more itself. Now the launch is refused, whether a rival holds the world or nobody does, and the mission composer says so before you spend a fleet: "Your guild is at the cap of 2 contested worlds: it cannot attack a world it does not hold until it holds fewer than 2." The Contested Space rules, your guild's cap note and each world's sheet say the same. A fleet already in flight when your guild reaches the cap still lands and fights; if your guild is at the cap when it wins, the win does not take the world, as before. Leaving your guild to launch and rejoining before the fleet lands does not get around it.

  If you play on an app version from before this release, a refused launch reads only "Invalid destination." until you update.

  Suggested by astroeddie88 on Discord, who worked out that the cap stopped a guild from owning the sector but not from denying it to everyone else.

1.411.0 — 2026-10-07

New

• A fourth background track: Atmos 01 (#6648). The music dropdown in the 🔊 panel, and in the More sheet on a phone, now offers Atmos 01 next to Ambient Light, Deep Space and Eschaton Theme, and shuffle can pick it too. It plays at the same loudness as Ambient Light, and fades in and out so a loop or a shuffle hand-off doesn't cut mid-phrase.

Fixed

• Shuffle no longer gives up on music with tracks still untried (#6648). When a track failed to load in shuffle (offline, a flaky connection), the fallback could bounce between two tracks and then stop the music, without ever trying the rest. It now tries each other track once before it stops.

1.410.1 — 2026-10-07

Also

• A flaky test of hand-off links no longer turns other releases red (#6660). The suite that checks a hand-off link pasted into an open tab waited a fixed 10 ms after each Back, Forward and paste; on a busy machine the browser events came later, and the test failed about one run in three on its own. It now waits for the events themselves, and a navigation that fires nothing fails outright. Nothing changes in the game.

1.410.0 — 2026-10-07

New

• Shop: Sovereign Seal packs. Before, Seals came only from fighting bastions, at most four a week. Now the Shop sells them: 1 Seal for €3.99, 5 for €16.99, 15 (a hull's whole track) for €44.99, 40 for €99.99. They go straight into your Star Forge and open once you have built it (any Tier-9 capstone).
• Shop: Supply crates. Ferrolite, Xenogas and Data at 6 : 3 : 1:

  | Crate | Price | Contents | |---|---|---| | 10M | €1.99 | 6M Ferrolite, 3M Xenogas, 1M Data | | 100M | €4.99 | 60M Ferrolite, 30M Xenogas, 10M Data | | 1B | €19.99 | 600M Ferrolite, 300M Xenogas, 100M Data | | 10B | €49.99 | 6B Ferrolite, 3B Xenogas, 1B Data | | 100B | €99.99 | 60B Ferrolite, 30B Xenogas, 10B Data |

  They arrive sealed in your Inventory. You open each one into a colony of your choice, up to its storage; what does not fit stays sealed until you make room.

Changed

• No purchase limits on Shop packs. Before, Forge Caches had a weekly limit: 45 components per family shared with bastion drops, and two purchases a week. Now every Shop pack can be bought as many times as you like. What you buy no longer counts toward the weekly limits on what battle pays, so bastion drops are always paid in full.
• Forge Cache 45 now costs €44.99 (was €49.99), so a bigger pack is never dearer per component.
• Payment disputes: a Supply crate stays sealed while a dispute on its payment is open. The Terms of sale describe all of this, including refunds: what you still hold is taken back, and resources already taken out of a crate stay with you.

1.409.0 — 2026-10-07

New

Plan a mission without leaving the galaxy map (#6604) Pressing Attack, Defend, Redeploy, Harvest, Colonise or Explore on a planet in the Galaxy (or «Send with escort…» on the Signals board) used to take you to the Missions tab, and you had to find your way back to the map. In windows 1180px and wider (desktop, and large tablets in landscape), these buttons now open the same Missions order form in a panel on the right of the map, titled with the order, e.g. «Attack · 3:142:7». The map stays visible and movable on the left, and centres on the target in the space beside the panel. «Launch from» is at the top of the panel, so you always see which colony the fleet leaves from.

Close the panel with «Back to the map», ✕, Escape or Back. You return to the planet's card, and the half-built order is dropped. The order you keep saved on the Missions tab is never touched. After a launch, «Show it» opens the mission in the Missions tab. Spy keeps its quick dialog.

Phones unchanged: on a phone, and in windows narrower than 1180px, these buttons still take you to the Missions tab. The phone version of the panel is not part of this release.

1.408.1 — 2026-10-07

Fixed

The commander card in the header speaks Italian: the GUEST badge, the portrait hint and the guild hint (#6644) An Italian player used to see the header's commander card in English: the GUEST badge next to the name, its tooltip and its screen-reader label, the "Edit commander identity" hint on the portrait and the name, and the "Guild: …" hint on the guild tag. They now read «OSPITE», «Modifica l’identità del comandante» and «Gilda: …», and the badge's explanation is in Italian too. English is unchanged, with one exception: if a guild's name has not loaded, the guild hint now reads "Guild: " rather than "Guild: null". «OSPITE» is one letter wider than «GUEST», so on narrow screens the commander name gives up about 3.5px. The badge is still shown whole.

1.408.0 — 2026-10-07

New

The phone alarm names who is coming, when your Observatory already knows (#6528): the incoming-attack push used to name only the colony and the arrival time. When the Observatory identified the attacker at the moment the alarm was written (the diary's rule since v1.405.0), the banner now names them too, and an AI empire carries its marker in words ("Synthetic — played by an AI", "Wraith — hyperintelligent AI"). An unidentified fleet reads exactly as before. Also fixed: when a fleet you had been warned about was recalled while another colony's alarm still led, the next banner for that colony kept leading with the recalled fleet ("arrives in about 13 minutes"); it now announces the fleet that is actually coming. A player who chose Italian gets the same banner in Italian («Una flotta di Vex è diretta su Lantea…»), with the marker in Italian too.

1.407.0 — 2026-10-07

New

You can now support the game without wearing it in public (#6616). Your commander card has a new checkbox: Hide my VIP level and supporter status from other players. Tick it and save, and chat, the leaderboard and guild rosters show you exactly like a commander who never gave — no VIP crest, no VIP border, no supporter ring — your own rows included. Every perk you paid for stays exactly as it was, and the portrait preview on the card shows you what everyone else now sees.

It is there before any purchase too, so you can give anonymously from your very first one. One thing it does not hide: a reserved supporter emblem you pick as your portrait stays visible, since it is a portrait you chose and can change. Chat lines already open on someone's screen keep the old badge until they reload.

Requested by many players.

1.406.26 — 2026-10-07

Fixed

The login screen, the log-out warning and the guest badge stop promising «only on this device» and «for good» (#6614) Since a one-time code can carry a guest empire to another browser, several guest sentences said more than was true.

• Log-out confirmation: a guest used to read "This empire lives only on this device: logging out loses it for good". It now reads "A guest empire has no account to sign back into, so logging out here loses it for good, and your commander name stays taken", the same words the commander card already uses. The variant for a profile that has not loaded, and the notice shown when a hand-off link opens in a browser that is already signed in, also say "logging out here".
• Login screen, first visit: "sign in with Apple or Google to keep it for good" is now "…to keep it on every device". A linked empire is also retired after a long silence, only later than a guest's, so "for good" was not true.
• Login screen, returning device: "A guest empire's key stays in the browser or app it was made in" is now "A guest empire opens only in a browser or app that holds its key". Logging out here removes the key from this browser, and a hand-off copies it to another browser.
• Guest badge in the header: its screen-reader label no longer says "this empire lives only on this device". Its tooltip now promises "keep your empire on every device".

Italian follows the same wording. Some code comments that still repeat the old claims are left for a later change.

1.406.25 — 2026-10-07

Fixed

Closing the map drawer no longer lets you send a second explorer to the same signal (#6623)

On the Signals board, Explore sends one explorer at a time: while a send is in the air, every other send on the board waits. That guard lived inside the board, so closing the map drawer, opening a system or switching tab threw it away. Reopen Signals before the first send had landed, and the same signal offered its Explore button again. A second tap launched a second hull and used a second of the day's explorations, and that explorer came home with nothing.

Now the board keeps the send in progress, and the lines it has already shown ("Spy Probe away" and the rest), across the drawer closing, the system view and a tab switch. A reopened board shows "Sending…" with the sends disabled until the first one and its refresh have landed. Logging out clears all of it, so the next player on the same device starts with an empty board.

Changed

On iOS 15, a request that never answers now gives up after 25 seconds, as it already did on newer devices (#6623)

On iOS 15 the app had no time limit on a request. One that never answered could keep a screen waiting until the app was restarted. It now stops after 25 seconds, the same limit iOS 16 and later have always had, and says the connection was lost and the action may still have gone through. This applies across the whole app on iOS 15, not only to Signals: for example, a colony that takes longer than 25 seconds to load at startup now shows the error screen instead of waiting.

1.406.24 — 2026-10-06

Changed

• The whole Market in Italian (#6621). Before, players with Italian on saw the Dark Matter market, the Merit market and the Portal exchange in English. Now they read Italian:
  • the Dark Matter market: convoy charters, shard purchases, boosts, ship bundles, Star Forge components and cosmetics, with their confirmations, refusals and receipts;
  • the Merit market, with its shelf, prices and claims;
  • the Portal exchange;
  • the market receipts in the Activity tab.

  The cosmetic names you buy are now Italian in the commander profile too. In English, every word and figure reads as before.

  The beta notice no longer lists the Market. Engine and setting names stay English: the Anchorage, hull and component names, boosts and Crucible ranks. So does the tooltip on the AI mark next to a synthetic seller.

1.406.23 — 2026-10-06

Changed

• Sector 13: a guild can now take at most 2 of the six control worlds, down from 3 (#6641). Before, one guild could sit on three of the contested worlds. Now a guild that already holds two cannot take a third: winning on a rival's world tears it off and leaves it unclaimed (sealed for an hour, at full garrison), and winning on an unclaimed world changes nothing. A guild holding one still captures as before. With six worlds, filling the sector now takes at least three guilds. The attack composer, the Contested Space panel and the Race world sheet all say 2.

  Not retroactive: no world was taken from anyone at the update. A guild that held three keeps them until it loses one, and cannot take another while it holds two or more. Activity entries for capped wins from before the update still name the cap of 3 that applied then.

  Players on an older app build may still see "cap of 3" in the attack warning until they update; the panel and the world sheet already show 2.

1.406.22 — 2026-10-06

Fixed

The galaxy map no longer opens where the previous player left it (#6624) The galaxy map remembers where you left the camera, so coming back from a system view puts you on the same spot. Logging out did not clear that memory, because a logout does not reload the page. The next account to sign in on the same tab then opened the galaxy map centred and zoomed on wherever the previous player had been looking, such as their home or a target they were watching. That happened even when they logged out with the map still on screen. Now a logout forgets the camera, and the next account opens the map on the whole galaxy. Within one session nothing changes: leaving for a system view and coming back still returns you to the framing you left.

1.406.21 — 2026-10-06

Changed

Replay diagnostics now describe the battle, not just its last few seconds (#6191) Before: with Replay diagnostics on (in Settings, or ?replaydiag=1), the copied block gave frame times only for the last 240 frames drawn, a few seconds at most. The replay keeps drawing while it is paused or showing its finale. So a block copied at the end described the calm finale, not the battle that stuttered. Now the block also summarises every frame drawn while the battle was actually playing. Frames from loading, pauses, the finale and a lost canvas are not counted. The summary gives the median, 95th and 99th percentile, the longest frame, and how many frames took over 25 ms and over 50 ms. It lists runs of three or more slow frames in a row, with roughly where in the battle each began. It says whether the stutter continued after the replay had already dropped to its lowest quality setting. When frame times look capped at 30 fps, as iOS Low Power Mode does, it says so. Nothing is sent anywhere: the block still leaves the device only when you paste it. This does not fix the iPad stutter; it is the measurement that will show what causes it.

1.406.20 — 2026-10-06

Changed

• AI empires plan their day on a newer model (#6634). The once-a-day strategy each AI-badged empire writes (posture, targets, what to build) now comes from a newer model. It still runs once a day per empire, and you will not see any of it as chat.
• AI empires no longer strike or raid right before they log off (#6634). An attack spends the attacker's own night shelter for two hours. Before, an AI empire could strike half an hour before bedtime and sleep exposed. Now it stops starting fights two to three hours before its night, and the exact cut-off differs from night to night. Beacon strikes are unaffected.
• AI empires go after rival-held Sector 13 worlds first (#6634). A held world keeps only a quarter of the garrison of a neutral one, so AI guilds now look at those first. They never go after a world held by a guild they have a pact or an alliance with.

Fixed

• A bot-led guild no longer hands out Officer rank by a donation count (#6634). Without its leader's model, a bot guild promoted whoever topped the contribution chart, every twelve hours, for good. Donating at the cap was enough to get the kick, the research pin and a weekly share of the Fund. Bot leaders now leave ranks as they are. Officers already appointed keep theirs.
• AI guilds stop pledging short of the seasonal project they are waiting for (#6634). Pledges stopped at the price of the cheapest projects, while the leader waited for the pinned one, so the Endowment could be wiped at the season close unspent. AI guilds also keep the Fund for the research project they are saving toward, instead of spending it on the finale or an anchorage first. They now move their research pin when its branch is complete, while the leader is awake.
• AI empires read their income correctly after a raid (#6634). After an unopposed raid, or an exodus, a colony produces less for a while, and AI empires now plan with the reduced figure.

1.406.19 — 2026-10-06

Changed

• Shop: Star Forge Caches are built, not yet on sale. The Shop can now sell Forge Caches: 4, 18 or 45 Star Forge components of one family you choose, for €4.99, €19.99 or €49.99. They fill the same weekly allowance as bastion drops and the DM and Merit Markets (45 per family in any seven days, at most two Cache purchases per family). They open with a Tier-9 capstone and need an account linked to Google or Apple. They are switched off for now, so the Shop looks the same as before.
• Terms of sale. The Terms now describe Star Forge Caches: what they deliver (no VIP points or supporter tier), their limits, card-only payment, and the automatic full refund if the family's room is gone when a payment is confirmed late. They also cover how refunds and payment disputes treat a Cache. The Refunds and dispute rules for supporter packs are unchanged; those sentences now say "supporter pack".

1.406.18 — 2026-10-06

Changed

Push notifications arrive in Italian for players who chose Italiano (#6613) Every notification the game composed itself used to be in English, even for a player who had picked Italiano in Settings. Those players now get the incoming fleet alarm (including the line about a second fleet behind it), the battle report, the missed interception, the Eschaton Finale, the First Convoy and the Marshal window in Italian. The time left is rounded the same way in both languages, so an Italian alarm never promises more time than the English one would. Players on the Automatic language setting still get English for now, because the game cannot yet see which language their device uses (#6629). Dispatches still arrive exactly as their author wrote them.

1.406.17 — 2026-10-06

Fixed

A strike your guild launched no longer fights your own garrison on a world or relic your guild has since taken (#6600) Before: when a guild sent several strikes at a Sector 13 world or a relic and took it with the first, the others kept flying. When they landed, they fought the site's defenders, and those now included the guild's own stationed garrisons. Your fleet shot at your guildmates' ships for a capture you already held. While those strikes were in the air, the "Guild holdings under attack" board listed them as an unknown attacker, and the Guild door and the Home line lit red. Now a strike that arrives at a relic or contested world its own guild already holds turns back whole, before any battle. This is the rule the launch already applied ("Your guild already holds that world"), now checked again on arrival. The pilot's activity line says "your guild already held that site when the fleet arrived". Those strikes are no longer listed on the holder's board and no longer light the door. A strike by any other guild is still listed and still lands.

The same issue asked for three other checks of the holdings board, and none needed a change. Another guild's reliefs were already left off a holding's row. A Sector 13 row already offered no "Find it on the map" button and said that only a garrison can help. The board heading, the Guild door and the Home line already name a guild holding rather than an ally when no guildmate is under attack.

1.406.16 — 2026-10-06

Changed

The commander card's account and sign-out sections are translated, and the guest warning no longer says linking prevents retirement (#6391) Before: on the commander card, the Account section (linking Google or Apple) and the "Sign out other devices" control were in English in every language. The guest warning there also said that linking an account prevents your empire from being retired after a long absence. That was not true: a linked empire is retired too, only later. Now both sections are in Italian, every refusal of a link attempt included. In English they read as before, except four guest sentences:

• The guest warning used to say "about three weeks" and that linking "prevents" retirement. It now says a guest empire is retired after 20 days away, and that linking an account delays retirement to 30 days of silence instead of preventing it: later, not never. It also no longer says this device holds the only key, because a one-time code can copy the empire to another browser.
• The two refusals that tell a guest not to log out (when the chosen Google or Apple account already has an empire) no longer say the empire lives on this device only. They now say a guest empire has no account to sign back into, so logging out here loses it.
• For a guest, the "Sign out other devices" confirmation no longer says those devices are locked out for good. It now says they cannot get back in on their own, and can only be let back in from this device with a one-time code (or by linking an account, where the game offers a sign-in button).

The same "only on this device" wording on the login screen, in the log-out confirmation and on the guest badge is #6614. Refs #6391.

1.406.15 — 2026-10-06

Changed

• Shop: card disputes. Before, disputing a Shop payment with your bank changed nothing in the game. Now purchases from your account are suspended while a dispute is open. For purchases whose payment page opened on or after 7 October 2026 (00:00 UTC), what the payment delivered is also held while the dispute is open: the VIP points and supporter tier it gave stop counting, and Dark Matter up to what the pack delivered is set aside. If the dispute closes in our favour, all of that comes back. If it closes in yours, it stays removed, the pack's Chrono shards and hulls are taken back as far as you still hold them, and purchases stay suspended until support has reviewed the account. The Terms of sale have a new "Payment disputes" section with the full rules.
• Shop: payment pages close after 30 minutes. Before, a Stripe payment page you opened stayed valid for 24 hours. Now it closes after 30 minutes; open a new one from the Shop if it timed out.

1.406.14 — 2026-10-06

Changed

The galaxy map keeps the board you were working through (#6427) Before: tapping a row in Targets, Signals, Places (a bookmark) or Contacts closed that board as it opened the planet's card, the contact's sheet or the system view. Closing the card left you on a bare map, so you reopened the board for every row. Ordering an Attack, Defend, Redeploy, Harvest, Colonize or Explore from the card, or opening the Marshal from its board, sent you to another tab, and coming back also showed the map with every board shut. Now a row leaves its board open under the card. Closing a planet's card shows the board where you left it, scrolled to the same place. Closing a contact's sheet brings the Contacts board back. After a trip to Missions (or to the Marshal), going back reopens the board that was open. From a system you reached through Signals, "Political map" brings Signals back. One consequence: after you come back from Missions with Back, the board is open again, so leaving the Galaxy tab with Back takes one more press (the first closes the board). The map does not remember its layer filters, its contacts setting or the Signals and Places filters yet; that is #6602. Composing an order without leaving the map is #6604. Refs #6427.

1.406.13 — 2026-10-06

Changed

The starting sectors say new empires are placed there while there is room (#6508) Before: the Displacement panel, the Galaxy sector notes and the glossary said the starting sectors are "where new empires are placed first", and a refused fold said those planets are "kept for new empires". To a player who started in the Frontier, that read as out of date. Now every one of these says the condition: new empires are placed in the starting sectors while there is room, and in the Frontier once those sectors are full. The refused fold no longer claims the planets are kept for anyone (any empire can colonise there; only a fold back in from the Frontier is refused). Same change in Italian. No rule changed: where a new empire is placed, and the one-way Displacement rule, are exactly as before.

1.406.12 — 2026-10-06

Changed

The commander card speaks Italian (#6391) — the card you open from your own portrait was English in every language, apart from the name rule and the log-out button. In Italian it now reads in Italian: the title and subtitle, the private «Commander since» line, every section heading, the names of the portraits and of the sixteen free emblems, the tint swatches, the Shop door, Cancel and Save, and the hints under the name field as you type. English is unchanged.

• The portraits are named after the face painted on them, so a woman's portrait reads «Veterana» or «Ingegnera».
• Names of things the Shop and the markets sell stay as they are sold, in English: market, Merit, supporter, Marshal and Crucible emblems, frames, name colours and the supporter tiers. A bought colour reads «Nome in Gold».
• The DM Market is «Mercato MO» on the card, as on its tab. The six Italian refusals that still said «Mercato DM» say «Mercato MO» too: the emblem, frame and name-colour refusals, and the three notices that ask you to reload the market (a bundle no longer on the board, an item or a component no longer on sale).
• Renaming yourself onto a name another commander already holds used to tell you a guild name was already in use. It now says the name is already taken by another commander, in English and in Italian.

Still in English on the card: the account-linking section and «Sign out other devices», which are the next part of #6391, and the account-deletion panel. Refs #6391, which stays open.

1.406.11 — 2026-10-06

Changed

Speed research now says it cannot hurry a fleet already in flight (#6448) — a flight's arrival is fixed when it launches, so finishing or rushing Propulsion, Solar Sails, Warp Drives or Beyond the Horizon never changed when a fleet already in the air would land. Nothing said so. The rule itself is unchanged, and the screens that sell speed now say it:

• the Speed up sheet, on a speed research order, above the shards, the Dark Matter rush and «Ask the guild»: finishing it sooner will not hurry fleets already in the air, only those launched or put on a new course after it finishes;
• the research cards for those four lines, under their effect;
• the Veil Doctrine's sheet: a level bought now will not hurry fleets already in the air, only those launched or put on a new course after it.

"Put on a new course" is said on purpose: Change course does price the new leg at the speed you hold at that moment.

Suggested by a player.

1.406.10 — 2026-10-06

Fixed

A half-built mission you aimed and then left no longer comes back with a mission type you never picked (#6576)

Before: a mission holding ships or cargo was saved with whatever mission type the form showed, picked or not. A draft left on its own launch colony reopens without a type (since v1.406.5). But if you then picked a target and left the tab before choosing a type, the next visit brought the old type back as if you had chosen it. On a phone the order form opened showing it. On a desktop, the type a destination suggests stopped following your next target, and the target list opened on «Your colonies» for a type nobody chose.

Now: a type you never picked is saved as not picked. On a phone the mission reopens on the target step, with the target still set, no type selected, and a line saying your fleet or cargo is kept. On a desktop the suggested type keeps following each new target. A type you did pick reopens exactly as before, and missions saved before this update reopen as they always did.

1.406.9 — 2026-10-06

Fixed

The layout test that flagged the Italian sort picker on the Colony tab as too wide was over-counting it, and nothing was ever clipped (#6567)

Before: a phone layout test held one named exception. It said the Italian building sort picker on the Colony tab, at its widest option «Ordina: tot. ↑», needed 153px against a 150px cap, so that face might clip. The test was reading the browser's natural width for the picker, which adds a four-space indent in front of every option inside a group. The closed picker never draws that indent: the face itself needs 136px, well inside the cap.

Now: the test measures what the closed face needs, with the grouping flattened, and the exception is gone. A new check makes sure the measurement still sees the indent it leaves out. This is a test-only change: no text, size or style in the game changed, and players see nothing different. It was checked in headless Chrome; an iPhone check of the Italian picker is still owed.

1.406.8 — 2026-10-06

Fixed

The browser layout checks no longer lose seconds to a frame stall under virtual time (#6570) Test infrastructure only: nothing a player sees changed. Before, every layout check that gives the headless browser a virtual-time budget could wait 0.5 to 1.9 s of real time for a single animation frame. Meanwhile the page clock jumped seconds ahead, which ate into the time those checks are allowed. Now those dumps ask Chrome to draw only after every compositor stage. The refusal check's sequence fell from 4.7–6.6 s of page clock to 1.0–1.1 s, and from 5.2–5.9 s to 3.0–3.6 s of wall time per dump, with the same readings. The derived dump ceiling from #6553 stays in place as the safety net.

1.406.7 — 2026-10-06

Fixed

The notifications switch says which way it failed (#6564). When switching attack notifications on or off did not work, the switch on the iPhone settings sheet and the iPad's Empire menu always said it "couldn't reach the server" — even when the failure was on Apple's side and no request reached the server at all. It now says what actually happened: "couldn't switch them on — try again" beside Off, or "couldn't switch them off — try again" beside On, in English and in Italian. What the switch does is unchanged.

1.406.6 — 2026-10-06

Changed

AI empires now remember who feinted at them whenever a human defender's Activity log would name the attacker (#6527)

Before: when an attack aimed at an AI empire was called off before it landed, the empire could only put a name to it if it had happened to see the attacker on its Incoming panel while the fleet was still in the air. A short flight, a missed check, or an attacker under a Veil left it with a scare and no one to remember it by.

Now: the empire reads the attacker straight off the withdrawal entry, the same entry a human defender sees in their Activity log. It names the attacker only when that entry does, which happens when the defender's Observatory (with Cold Trail, Cold Read and Omniscience) had identified them when the alarm went off. Below that tier the attack stays nameless, exactly as it does for a human in the same seat. A named feint is remembered as a minor grievance that colours how the empire talks about that player, and it still never counts as grounds for a retaliatory strike.

1.406.5 — 2026-10-06

Fixed

A mission draft saved on its own launch colony no longer shows a verb you never chose (#6555) Before, on the phone, a draft whose target was its own launch colony reopened on the order form showing a verb such as «Pickup ✎». The composer treated that verb as never picked: the verb list lit nothing, and one tap on a target's quick order replaced it. Now such a draft reopens on the «Where does the fleet go?» step with its ships and cargo kept and no verb chosen, and you pick the destination and the order again. The step says what was kept («What you had prepared is kept: 4 ships and the cargo…»), so it no longer looks like a blank composer. A Harvest over the salvage above your own launch colony is a real order, and it now comes back aimed with Harvest chosen on the phone and the desktop, even if the salvage has gone in the meantime. A Harvest picked that way before any ships were chosen now survives switching tabs too. The desktop is otherwise unchanged. Two related questions are filed separately and are not fixed here: a Harvest over an empty field is not warned about (#6573), and a parked draft that you aim and then leave can come back with its old verb shown as chosen (#6576).

1.406.4 — 2026-10-06

Changed

The colony strip and the colony step buttons name themselves in Italian (#6562) These are screen-reader names and tooltips; nothing visible changes. Before, in every language, the row of colony chips on the Home and Empire tabs was read aloud as «Colonies», and the mother colony's chip as «(mother colony)». The previous / next colony buttons in the desktop header and the phone colony sheet were «Step through colonies» and «Previous colony» / «Next colony». An Italian player now hears «Colonie», «(colonia madre)», «Scorri le colonie», «Colonia precedente» and «Colonia successiva», tooltips included. The keyboard hint stays «Shift+C» / «C». English is unchanged. The rest of #6391, including the commander card, is still English.

1.406.3 — 2026-10-06

Fixed

A test that checks the phone's refused-launch layout no longer fails on a busy test machine (#6553) This is a test-infrastructure change only: nothing changes for players. The automated check that a refused launch on a phone does not cover the cargo controls used to give its headless browser a fixed 40 seconds. On a heavily loaded test machine that was sometimes not enough, and two full test runs failed although the game was fine. The time limit now comes from the time the test is allowed to run anyway. When the check does fail, it says which step the page reached and how long it had.

1.406.2 — 2026-10-06

Changed

• New Shop prices, VAT included. The six packs now cost €1.99, €5.99, €11.99, €29.99, €59.99 and €99.99 (they were €1, €5, €10, €25, €50 and €100). They deliver more in return: 100, 300, 600, 1,500, 3,000 and 5,000 Dark Matter, and 200 to 10,000 VIP points. One Data Pack now reaches VIP 3 and one Dark Matter Pack VIP 4. Shards, hulls and bonuses are unchanged.
• Nobody loses a tier. The totals that unlock each supporter tier stay where they were: €1, €5, €10, €25 and €50, with Eschaton now reached at €99.99. Everything you have already earned stays yours. The profile's emblem list now shows those totals instead of pack prices.

1.406.1 — 2026-10-06

Fixed

Source comments stop describing a daily recall cap that no longer exists (#6529) Comments and design docs only: there is no player-visible change. Recall has been free since #6247, but a few code comments and one design document still named a three-a-day recall cap as what holds feints back. They now describe the game as it plays: recalling a fleet costs nothing, and a defender who saw the attack coming is told it was called off.

1.406.0 — 2026-10-06

New

• Holding a Sector 13 world now pays: five times the guild points, points for the players who garrison it, and Dark Matter on the guild podium (#6563)

  Before: a world your guild held paid it 60 points an hour and 250 for the capture. The players whose fleets kept it earned nothing for it. At the close, placed guilds paid their members crates only.

  Now:

  • a held world pays its guild 300 points an hour, and a capture pays 1,000;
  • every member of the holding guild with a fleet stationed on a held world earns 300 points an hour on their own season score, provided your fleets there have at least 10% of the Power of the full garrison an unheld world carries (not of the reduced one left on a world your guild holds). A token fleet earns nothing. Several garrisons of yours on one world count together and are paid once, and each world you guard pays separately;
  • at the close, the top three guilds also pay Dark Matter to every member who is paid crates, by the guild's place and your own rank on the event board: 1st place 1,000 / 500 / 200, 2nd 500 / 250 / 100, 3rd 250 / 125 / 50 (top 3 / top 10 / everyone else). The crates are unchanged.

  The Contested Space rules, the Race panel, the guild prize table, the activity receipt and the wiki show the new figures.

  This season is live, and nothing is recalculated. The new rates apply from this release on. Points already scored stay as they were, and the Dark Matter is paid at this season's close. Two small limits: a garrison's points are settled every five minutes, so the last few minutes before a world changes hands, and before the season ends, pay the guild but not the garrison. An app that has not updated yet still shows the old rates, but the server pays the new ones.

1.405.1 — 2026-10-06

Changed

The attack-notifications switch speaks your language (#6365). On the iPhone settings sheet and in the iPad's Empire menu, the switch for attack notifications was English in every language: its name, its On and Off states, the line shown once iOS has refused notifications, and the line shown when switching them on or off did not work. The menu section around it, «This device», was English too. All of it now follows the game's language; what the switch does is unchanged. This completes the settings screen's move to the game's language, except for the account-deletion panel, whose wording waits on the owner (#6561).

1.405.0 — 2026-10-06

New

Incoming-attack alarms name the commander your Observatory identified (#6439)

Before, an "Incoming attack" line in the Activity diary never said who sent the fleet. So a whole-history search for a commander never found their attacks, and the line that closed the alarm ("turned around") was anonymous too. The name was only ever on the live Incoming panel, from Observatory 4.

Now the alarm names the commander when your Observatory (or Omniscience) identified them at the moment the alarm was written. That is the same step that names them on the Incoming panel: "a hostile fleet from Vex arrives in 12 min". An AI empire carries its badge beside the name. The line that closes the alarm names them the same way, copied from the alarm. A named alarm is found when you search that commander's history, and it offers them as a suggestion.

Below that step, the alarm reads exactly as before. A later Observatory upgrade does not add a name to an alarm already written. The raid summary row still names nobody, and alarms written before this release stay as they were.

1.404.1 — 2026-10-06

Changed

The battle line's row names stay the same in every language (#6070)

Vanguard, Center and Rearguard, and the short Front, Center and Rear on the row buttons, are names, like a ship's: they read the same in English and in Italian, so a guildmate writing "Battleships in Rearguard" matches the button on every member's screen. Italian already printed them this way («la Vanguard», «il Center», «la Rearguard»), so Italian players see no change beyond the Crucible kicker below.

The one visible change, in every language: in the Crucible, the formation stack used to label its first row VANGUARD while the rest of the tab called it Front. It now reads FRONT, like the row buttons and the War Table. Center and Rear are unchanged.

1.404.0 — 2026-10-06

New

The Colony tab and the Base build menu can now filter buildings by what they make, and fade what you cannot afford yet (#6426) Before, the building lists could be sorted by cost and filtered by family (Production, Military, Civic), but nothing picked out the buildings that produce a given thing, and nothing showed at a glance what could be bought right now. Both lists now carry two new controls beside the sort:

• «Makes: …» keeps only the buildings that produce Ferrolite, Xenogas, Data, Energy (power for this colony's grid) or Dark Matter (the empire's currency). On the Colony tab the family chips re-count under it, so an empty family shows 0 before you tap it. On the Base menu, picking Dark Matter also shows the Particle Accelerator while it is still locked, dimmed, so you can see what to unlock.
• «Affordable now» dims every building whose next level you cannot buy right now: not enough resources, already upgrading, at its maximum level, capped by the Command Center, or still locked. Nothing is hidden and nothing moves, because your resources change every second. A full build queue does not dim a building you can pay for.

Back brings both controls back as you left them. On a phone the filter row takes one more line than before.

1.403.2 — 2026-10-06

Changed

On the phone, a verb you chose survives changing the target, and your own colony no longer carries a Transport button (#6428) Before: in New mission on a phone, each of your colonies in the target list had a Transport button beside it, the only button on the row. Tapping it chose Transport and jumped to the fleet. The button also stayed there when you went back from the order list to change the target, so a Redeploy or a Pickup you had already chosen could become a Transport in one tap. Now: tapping one of your colonies sets the destination and asks what the fleet should do there: Transport, Pickup, Redeploy or Defend, none chosen for you. Once you have chosen an order, changing the target never replaces it. A Transport costs one more tap than before. Salvage above your launch colony still offers Harvest in one tap, and a colony you spied on still offers Attack or Espionage, because there the row already knows which order fits. Not changed: the desktop, where picking your own colony still lights Transport in the column beside the target.

1.403.1 — 2026-10-06

Changed

Declaring a leave speaks your language (#6365). On the phone settings sheet and in the desktop Empire menu, everything about a declared leave was English in every language: the prompt and the picker, the notice while a leave runs (cordon kept or lost, days left), the warning that ending a leave drops the cordon and any hours bought with it, the end-leave button, a guild leader's warning about their seat, and the rule about what happens without a leave. All of it now follows the game's language, with the thresholds still read from the game's own rules. One English line changed too: the desktop prompt said a leave "shields" the empire, the old word for what the game now calls the Cordon everywhere else; it now says a leave puts the empire under a cordon. Refs #6365: the notification switch on the same screens is still English and follows in the next update.

1.403.0 — 2026-10-06

New

• A guild can hold at most three of Sector 13's six control worlds (#6545). Since the worlds became takeable again they changed hands all six at once: whoever was online with the biggest fleet held everything, and every world's hourly points went to one guild. Now a guild takes a world only while it holds fewer than three:
  • Under three, a won battle captures the world as before.
  • At three or more, on a world another guild holds, the win tears the world off its holder and leaves it unclaimed. The holder is paid for the hours it held, the world is sealed for an hour, and its garrison is rebuilt at full strength.
  • At three or more, on an unclaimed world, nothing changes hands; the garrison you beat is rebuilt within minutes. Either way the battle, its losses and its season points stand, and your Activity feed says which case it was.
• You are told before you fly. The Contested Space panel states the rule and, for a guild at the cap, where it stands; the Season Race world sheet says what a win there would do; the attack composer warns (it does not block) when your guild is at the cap.
• A freed world reads "sealed" with the minutes left, on the Contested Space board and the world sheet, instead of looking like an ordinary garrisoned world. The refusal for a world inside its hour now says it "changed hands" rather than that another guild took it.

Changed

• Nothing was taken away at deploy. The cap applies to the next capture; a guild holding more than three keeps them until it loses them, and every player is told why such a guild is above the cap. When this shipped, four guilds shared the sector (two, two, one and one worlds), so no guild was above it.
• Synthetic guilds at the cap stop setting out to capture worlds they could not keep; they still garrison the ones they hold.

1.402.3 — 2026-10-06

Fixed

On the phone, a target picked on New mission is no longer lost when you leave the Missions tab (#6494)

Before: if you picked a target on New mission but had not yet chosen ships or cargo, leaving the Missions tab lost it. That happened on a trip through the tab bar, a tap on the alarm chip, or Back, and New mission reopened on an empty target step.

Now: the target is still there when you come back, already filled in on the target step. If you had also picked a mission type, you come back to the form with that type picked. A target saved without ships or cargo is kept for 24 hours from the last time you aimed it, and it never replaces a mission you had saved with ships or cargo.

1.402.2 — 2026-10-06

Fixed

AI empires no longer send a picket after a raid fleet they could never engage (#6518)

Before: when a joint raid was aimed at an AI empire's guildmate, or at a relic its guild holds, the AI empire could fly a picket out to meet one of the raid's fleets. Only the raid's own target can intercept a raid fleet. So the picket stood its watch, engaged nothing and came home, and the empire had spent its intercept cooldown on it.

Now: an AI empire leaves those raid fleets alone. It still goes out to meet an ordinary attack on a guildmate, and a raid aimed at one of its own colonies.

Not included: a raid fleet aimed at an empire outside the AI empire's guild, or a raid fleet on its way home, can still draw such a picket. That needs the server to say which fleets a viewer may engage (#6536).

1.402.1 — 2026-10-06

Fixed

• A line of Emplacements now dies in a Cruiser-sized explosion (#6513). In the battle replay, when the last Emplacement fell, the explosion was smaller than a Corsair's, because it took the size of the replay's default dart model. It now explodes at a Cruiser's size, still smaller than a Fortress's or a Citadel's, so a defence line going down reads as a ladder of growing explosions. The Emplacements look exactly as before, and every other explosion is unchanged.

1.402.0 — 2026-10-06

New

• A Camera shake setting for the battle replay (#6430). Until now, the only way to calm the replay's camera was your device's reduced-motion setting, and that turns the shake off entirely. Settings now has a Camera shake row with three steps: Full (the default), Half and Off. It is under "This device" in the desktop menu and on the phone's Settings sheet, just above Replay diagnostics. The choice is saved on this device only and applies from the next replay you open. If your device asks for reduced motion, that still wins: the row shows Off, its steps cannot be changed, and it says why. Your choice comes back once the system setting is off.

1.401.1 — 2026-10-06

Changed

• The night window speaks your language (#6365). On the phone settings sheet and in the desktop Empire menu, the night window was English in every language: its label, the line saying which clock the hours are in, the 24-hour rule (and when it frees up again), and the ten-minute undo button. The commander card's name field stated its own 24-hour rule and wait in English too. All of it now follows the game's language, and in Italian the refusal you get while the window is still locked now uses the same words as the row above it. English is unchanged. Refs #6365: the leave row and the notification switch on the same screens are still English, and follow in the next updates.

1.401.0 — 2026-10-06

New

• AI guilds now defend the worlds and relics they hold (#6480). Since #6474, guild members can see attacks on their guild's contested worlds and relics on the "Allies under attack" board. The AI empires (marked with the AI badge) could not see those rows, so an AI guild that held a world or a relic left it unguarded. AI members now read those attacks too, and send a garrison when one is coming, on the same terms they defend a guildmate: only while they are awake, and only when they can spare the ships.

  They commit about a third of their hangar at most, the same share they leave on a relic in peacetime, because the alarm on a held site never says how big the incoming fleet is. When one sends ships, it may say so in guild chat ("someone is going for our relic at 9:2:1, sent 20 ships"). It never names the attacker, which the board does not show.

  AI members also count a world or relic their guild holds as their own ground when deciding whether to intercept a fleet aimed at it.

  No AI guild holds a world or a relic right now, so this cannot be seen in play until one takes a site. The deploy confirms only that the AI service restarted cleanly with all 75 empires active. A failed board read would not show in production logs either: the API does not log requests, and the agents log a refused read only at debug level.

1.400.0 — 2026-10-06

New

• The Shipyard's «For you» strip now shows the hulls you actually build (#6420). Before, the three tiles were the first hull of each class in progression order, the same for every empire at the same tech level, and they never changed once the early classes were open. Now the first tile is your last order from that colony (or, on a device that has none, your empire's most recent order), and the next tiles are the hulls your empire ordered most in the last 30 days. Any slot left over is filled the old way, from a class the strip does not already show. A new empire, with no orders yet, sees the same starting hulls as before, and the strip is labelled «Start here» until something on it is yours. Hover a tile on desktop to see why it is there. A hull you order most is shown even when you cannot afford it right now; the build bar names what you are short. The strip is still settled once per visit: it does not reorder while you are choosing, and an order you place shows up on it the next time you open the Shipyard or switch colony. Favourites you pick yourself are not part of this release.

  Suggested by ancalego on Discord.

1.399.4 — 2026-10-06

Fixed

• «Check your missions» stays on «In flight» while you are under attack (#6493). On a phone, the quick spy dialog's «Check your missions» opens Missions on «In flight», so you can see whether your probe left. Before, if an attack was inbound, the tab switched to the Alarm a moment later, as soon as the first inbound check answered, and you were moved away from what you came to see without tapping anything. Now it stays on «In flight». The threat is still shown: the ⚠ Alarm pill keeps its count, the board's alarm banner stays up, and the under-attack chip at the top still takes you to the Alarm.

1.399.3 — 2026-10-06

Fixed

• The battle replay's camera no longer piles shake on shake (#6430). Before, every hit, kill, lance volley, Bubble break and finale added its knock on top of the last one. A round full of kills pushed the camera up to its ceiling and held it there, so on a big battle, or at 2x and 4x, the screen read as bouncing the whole time. Now each knock lifts the shake only to its own height. A single hit or kill shakes exactly as much as before, but a busy round can never shake harder than its biggest single knock. On a measured fourteen-group battle, the share of fighting time spent visibly shaking drops from 29% to 14% at 1x and from 77% to 35% at 4x. Players who turned on reduced motion in their operating system already got no shake from most of these. They now get none from the Bubble breaking either, which was the one source that ignored the setting. An in-game control for the shake is not in this release, and #6430 stays open for it (Refs #6430).

1.399.2 — 2026-10-06

Changed

• More of the settings screen speaks your language (refs #6365). Several settings strings were English in every language: the names a screen reader gives the phone sheet's round music and mute buttons, the desktop Empire menu's music chip and mute button, the Changelog link and its tooltip on both, and the title of the phone's colony picker. They now follow the game's language. In Italian they read «Spegni/Accendi la musica», «Silenzia tutto», «Riattiva l’audio», «♪ accesa/spenta», «Novità» and «Colonie». English is unchanged. The night window, the leave row and the notification switch on the same screens are still English, and follow in the next updates.

1.399.1 — 2026-10-06

Fixed

• Hand-off links pasted into a tab that is already open (#6387). Three cases that used to go wrong now behave as expected:
  • A newer link wins. Before, in a signed-out tab still holding an older code from the "This browser already holds an empire" notice, a newly pasted link was ignored and the expired code stayed in the field. Now the link you just pasted is the code offered.
  • A paste is not a Back. Before, pasting a link while a dialog was open closed the dialog as if Back had been pressed. If that dialog was your commander profile, the pasted code was lost, and the notice turned up later on an unrelated Back. Now the dialog stays open and the notice opens over it right away, also over a confirmation nested inside another dialog and on a phone. Closing the notice returns you to the dialog.
  • Old codes stay gone. Before, if you pasted two links before signing in, pressing Back past the start of the game later reached the first code and announced it as "a code for an empire started in another browser". Now a code the sign-in page already offered does not come back through Back or Forward. Signed in, it simply leaves the address with no notice. Signed out, the page keeps offering the code it already showed.

1.399.0 — 2026-10-06

New

• The Shipyard's count chips now add: three taps on +50 make 150 (#6417, #6505). The two quantity chips next to the order count used to read ×5 and ×50, and each tap replaced the number in the box. Any order between 50 and Max meant typing it, which is awkward on a phone keyboard. The chips now read +5 and +50, and each tap adds to the count. This works the same on the desktop console, on the phone's count sheet and in the ship's specs sheet.
  • On a fresh order, the first tap sets the number. When the box holds 1 or less, +50 makes it 50, not 51, so three taps from a fresh hull read 50, 100, 150. The trade-off: from 1, +5 gives 5, not 6. For 6, type it or tap + once.
  • Past what you can afford, the chips keep adding. The red «Short …» line says what is missing and BUILD stays off, as it does for a typed number. Max sets the count back to the most you can afford.
  • The chips no longer light up. After three taps nothing equals 50, so the count box is now the only place the number is shown. Max still lights up when the count is exactly the maximum.
  • After a successful order, that hull's count goes back to 1, so the next +50 starts from 50 and not from the order you just placed. The ↺ Repeat chip beside BUILD still brings back the order you just placed. If the order fails, the count stays for the retry. On desktop, a number you change while the order is still being sent is kept.
  • Screen readers announce the chips as «Add 5 to the count» and «Add 50 to the count».

  The War Table's count buttons are unchanged and still set the number you tap. The chip's symbol tells the two apart: a «+» adds, while a bare number, Max or All sets.

  Suggested by mark_of_the_sherminator on Discord. His in-game name is not yet established, so the credit is held back until it is, not withheld.

1.398.2 — 2026-10-06

Fixed

• The Home timeline's «+2 · next …» counter now shows a chevron, so you can tell it opens (#6409, #6502). On the desktop and tablet Home, a lane with more than three bars prints how many it hides and when the next one lands. Tapping that line opens the lane and shows every bar, but it looked like a plain dim label, so many players never found out. Some concluded their missions were missing from Home altogether. The counter now starts with the same small down chevron as the Operations rail's «+N». The chevron flips while the lane is open and flips back when you close it. Screen readers already announced the counter as an expandable button, and nothing changes there. On a long lane name, the clock is cut off a little sooner to make room for the chevron. The chevron itself is never cut off.

  Suggested by mark_of_the_sherminator on Discord. His in-game name is not yet established, so the credit is held back until it is, not withheld.

1.398.1 — 2026-10-06

Fixed

• Hovering a planetary mandate chip no longer brings scrollbars into the Colony tab's side column (#6463). On a desktop, moving the mouse over the «Transit» mandate made scrollbars appear in the left column and the panel jump a little; the other mandates did not do it. Transit's description is the longest, so it wraps and is the bottom row, and the invisible strip that keeps a hovered button clickable at its lower edge reached one pixel past the column's end. The chip row now leaves room for that strip, so hovering any mandate, stance or edict chip leaves the layout exactly where it was. The Transit wording is unchanged.

  Reported on Discord by antiproduct5001, with a screenshot that showed exactly which chip did it.

1.398.0 — 2026-10-06

New

• On a phone, Back now returns you to the Missions screen you were on (#5898). Before, Back onto Missions picked the screen afresh each time. A player reading «In flight» could come back on «New mission» because a half-built mission was saved. With an attack inbound, every Back landed on the Alarm, whatever you had been reading.

  Now, Back (the browser's, Android's, or the on-screen ←) onto Missions reopens the Alarm, «In flight» or «New mission», whichever you left:

  • With an attack inbound, «In flight» and «New mission» stay where you left them. The attack is still shown by the under-attack chip at the top of the screen, the red count on the ⚠ Alarm button, the banner on «In flight» and the red edge when impact is close. One tap on the chip still takes you to the Alarm.
  • If you left on the Alarm, Back opens the Alarm. If the attack has gone by the time the game checks, it opens the screen a fresh visit would show instead, never an empty Alarm. Screen readers announce the landing only once the attack is confirmed.
  • Links into Missions still open where they always did: «Check your missions» opens «In flight», and an Attack from the Galaxy, a guild or a report opens «New mission».

  What has not changed

  • Opening Missions from the tab bar or after a reload chooses the screen as before.
  • The step of a half-built mission (target or objective) is not restored.
  • Nothing changes on a computer, where the three are shown together.

  Part of #5898. Still open there: the native items listed on the issue.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.397.0 — 2026-10-06

New

• Your guild is warned when a world or relic it holds is under attack (#6474). Until now an attack on a Sector 13 contested world or a relic your guild held reached nobody: the "Allies under attack" board in the Guild tab listed only guildmates' colonies, so a holding could be taken while its holders were busy elsewhere. Now each such attack gets its own row on the board: the structure's name and coordinates, a line saying "Contested world held by your guild" or "Relic held by your guild", the impact countdown, and your guild's garrisons already there or on the way. Send defence works from the row.

  The warning says where and when, never who or with what: your guild has no Observatory of its own on a site it holds, and the row says so. On a contested world the row says that no interception is possible, because the Event Sector is off the galaxy chart; on a relic, "Find it on the map" works as it does for a guildmate.

  The rest of the game reacts too. The board's heading names holdings when one is on it. The Guild door turns red ("a guild holding is under attack"), the Guild tab's dot counts it as one more call, and Home lists "A relic or world your guild holds is under attack" among its urgent orders. On a phone, the board's heading now wraps onto a second line instead of running off the screen, in English and in Italian.

1.396.4 — 2026-10-06

Fixed

• Every log-out button now warns a guest that logging out loses the empire for good, and says what keeps it (#6380). A guest empire's only key is kept in the browser it was started in, and logging out throws that key away. Before, only the "Log out" in the commander profile said so. The "log out" in the desktop settings menu ended the session on one press. The one in the phone settings sheet asked "Confirm log out" without a word about what a guest loses.
  • All three now ask first, with "Stay" as the first button.
  • A guest reads that the empire lives only on this device, and how to keep it: «Continue in another browser» in the commander profile gives a one-time code that carries the empire to another browser or device. Where this build offers a sign-in to link, the settings menu and sheet also suggest linking an account.
  • A player whose empire is linked to an account sees no warning. They are told that signing back in brings them to exactly where they left off.
  • If the profile has not loaded yet, the warning is given as a condition ("if this is a guest empire…"), so nobody is told something that might be false.

1.396.3 — 2026-10-06

Fixed

• The Base and the Colony tab now offer the Reload their "could not be loaded" notices ask for (#6458). After an update, a Base whose 3D view could not be fetched said "Restarting the app, or reloading the page, fetches it again", but its only button was "Open the Colony tab". On the Colony tab, the planet picture's line said the same thing in the red banner at the top, which clears itself after eight seconds on a phone. An app added to a home screen has no address bar and no pull-to-refresh, so there was no way to do what either sentence said. Now the Base card offers Reload next to "Open the Colony tab". The planet's line stays in the colony card, where the planet picture would be, with a Reload beside it. It no longer appears in the banner. Switching colony clears it, and the next planet is tried again. Reload only appears when the sentence asks for one: on a device that cannot draw 3D, the line explains that and offers no Reload, as before. This is the same Reload the Galaxy tab got in 1.394.5.

1.396.2 — 2026-10-06

Fixed

• On a phone, a refused launch no longer covers the cargo fields you need to fix it (#6429). Before, a refusal such as «Not enough resources.» or night protection appeared as a floating red notice that stood over the middle of the mission form. On smaller phones it sat on «Fill every», on the dati field or on «Fill hold»: the very controls that answer «not enough resources». It could already be closed (its «ok», a tap on it, or waiting eight seconds), but it was in the way until then. Now the refusal appears on the launch bar itself, right above Send, with a Dismiss button. Send stays where it was, and the form scrolls up by the height of that line, so the fields that were just above the bar stay in view. Retrying keeps the line in place instead of pushing the form further up each time, and dismissing it puts the form back where it was. If you leave the form or change target while the launch is still waiting for an answer, the refusal still reaches you in the floating notice, as before. Other errors and other screens are unchanged.

  Reported by a player on Discord, composing a redeploy and a pickup on an iPhone.

1.396.1 — 2026-10-06

Also

• Ship tooling: the Market's translation test no longer turns unrelated changes red on a busy build machine (#6489). One test read every screen of the player-trade board twice, in English and in the test locale that checks every string goes through the translation catalogue. On a loaded machine it ran out of its 60-second limit. The read it gave up on kept running and collided with the next test. That test then failed too, and the end-of-run check reported dozens of Market strings as never checked. Changes that never touched the Market were blocked three ways at once. The board is now read in nine parts, each in its own tests. The slowest of them took 12 seconds where the old one took 36, with the machine under heavier load. A test that times out now stops its own read and cannot disturb the next one. Every string checked before is still checked. Nothing in the game changes.

1.396.0 — 2026-10-06

Added

• The server can now warn a guild when a world or relic it holds is under attack (#6474). This is the first half of that issue. Before this, an attack on a Sector 13 contested world or a relic your guild held reached nobody in the guild: the "Allies under attack" board listed only guildmates' colonies. The server now answers attacks on the guild's held structures and raises a guild-door flag for them. Nothing changes on screen yet: the board, the Guild door and the Home line that show these warnings arrive with the next release. Apps that predate it get exactly the board they had.

  The warning says where and when, never who or with what: a guild has no Observatory of its own on a site it holds. Anchorages are not on the list: an attack that can bring one down is already an attack on a guildmate's colony in that system, and anchorages have their own siege alarm.

1.395.2 — 2026-10-06

Fixed

• The attack forecast now says whose ships each list holds (#6464).
  • Before: the forecast row listed the ships your fleet was expected to lose with no side word, next to the defender's "survivors on target". A player attacking the held Sector 13 control worlds saw Titans and Lances in that list and took them for a defence the spy report did not show. They were the player's own hulls, priced as losses. The forecast itself was fighting exactly the garrison the spy report read.
  • Now: the row reads "your losses N% (…)" and "their survivors: …". In Italian it reads "le tue perdite …" and "i loro superstiti: …". The Italian quick-attack confirm also says whose losses they are ("Perdi circa N% …"), as the English one already did.

1.395.1 — 2026-10-06

Also

• Ship tooling: the gate's Python leg no longer races the engine build (#6479). Some Python tests run the game engine through node. When a change was small enough that only a few of those tests ran, they started before the engine had been built and failed with "Cannot find module". That falsely blocked a two-line text fix for #6464. The Python leg now waits until the engine build has run. Nothing in the game changes.

1.395.0 — 2026-10-06

New

• Back now returns you to the Base with the map where you left it (#5898). Before, Back onto the Base always opened the map at the default zoom, centred. A player who had zoomed in on one corner of the colony, opened Research from a building and pressed Back had to find that corner again.

  Now, Back (the browser's, Android's, or the on-screen ←) onto the Base opens the map at the zoom you left, centred on the same spot of the ground, straight away, with no fly-in. It does so:

  • on any colony, as switching colony inside the Base already keeps the map where it is;
  • after a rotation or a resize, because the spot is remembered on the ground, not on the screen, so a phone turned sideways in between still shows the same spot.

  Once you move the map, rotating the phone or switching colony keeps that same spot at the centre, and Back returns to it too.

  What has not changed

  • Opening the Base from another screen, or after a reload, frames the map as before.
  • The ⤢ button still frames the whole base, and Back then opens the default framing too.

  Part of #5898. Still open there: the galaxy's political-map camera, and the Missions segment on a phone.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.394.7 — 2026-10-06

Changed

• The Shop says who sells, and links its terms. The Shop and the confirmation before payment now name the seller: AIgen Solutions S.r.l., Via dei Maffei 42, Roma, VAT IT18165291008. Both link the new Terms of sale page, which covers what a pack is, final prices with VAT included, delivery, the right of withdrawal and refunds.
• An invoice after every purchase. After you buy a pack, an invoice is e-mailed to you. It shows the VAT and confirms the choice you made before paying: immediate delivery, with the 14-day right of withdrawal ending once the items arrive.
• The privacy page names who runs the game. It now names AIgen Solutions S.r.l. as the data controller, instead of calling Novarchy a hobby project.

1.394.6 — 2026-10-06

Fixed

• Merge commits on main no longer carry attribution trailers (#6465). Until now each squash commit body was assembled by GitHub from the branch's commit messages, so the Claude-Session: and Co-authored-by: lines of those commits ended up in main's history. The ship's merge step now writes the body itself: the PR's commit messages, read whole, with every line that starts with either key removed and everything else kept. If the messages cannot be read, the merge still goes ahead with the PR title and a "body unavailable" note, and says so on a MERGE-BODY: line. It never falls back to GitHub's default body. Nothing changes in the game.

1.394.5 — 2026-10-06

Fixed

• A Galaxy tab that could not load after an update says so once, with a Reload button under it (#6401). When a tab left open across an update could not load its 3D views, the Galaxy tab showed two errors in two wordings: the map's notice ("Restarting the app, or reloading the page, fetches it again") and a second line under it ("Reload the page to fetch it again"). The only button was Open the system view, which goes around the problem rather than fixing it, and on a game installed to the home screen there was no address bar to reload with. Now each screen says it once. The map's notice has Reload first and Open the system view beside it. The system view's notice uses the same wording as the map's, with the same Reload under it, and its list keeps working. Reload is offered only when the sentence asks for a reload: a device that cannot draw 3D is still told so, with no Reload, because reloading does not fix that. Not included: the base view and the Colony tab's planet still name a reload with no button for it. That is #6458.

  Reported by mark_of_the_sherminator on Discord, whose in-game name is not yet established: the credit is suspended rather than withheld.

1.394.4 — 2026-10-06

Fixed

• Number boxes no longer start an accidental drag that can lock the page (#6416). When you click into a number box, such as a cargo amount, the whole value gets selected. If the mouse moved a few pixels during that click, the browser could start dragging the selected digits away instead of placing the cursor. On Windows, a drag holds the mouse and the keyboard until it ends. Number boxes now refuse to start that drag. Typing, selecting and dropping a number into a box work as before. This removes one likely cause of the page that keeps updating and scrolling but ignores every click and key until you reload. That freeze has not been reproduced yet, so if it happens to you again, tell us whether pressing Escape brings the page back.

  Reported by deadbeef1337 on Discord. They noted the three details that ruled most suspects out: the counters kept updating, the wheel kept scrolling, and the mouse and keyboard died together. Their in-game name is not yet established, so the credit is suspended rather than withheld.

1.394.3 — 2026-10-06

Fixed

• On a phone, typing in chat pulls the screen back into place (#6447). Before this fix, tapping the chat box could push the whole screen up when the keyboard opened. The messages and the reply box went off the top, and you had to drag the page back down before every message. The game now moves the screen back for the chat box and the Go-to sheet, including when the page was a few pixels too tall or when the phone's scroll arrived late. This is a partial fix, not yet confirmed on a real iPhone: if iOS shifts the view in a way the page cannot undo, the problem can still appear, and #6447 stays open until it has been checked on a device. The iOS app gets this with its next update. The website gets it now.

  Reported by antiproduct5001 on Discord, with a screenshot. Their in-game name is not yet established, so the credit is suspended rather than withheld.

1.394.2 — 2026-10-06

Fixed

The research board's lights are calmer. The 3D research board flashed like a strobe: every node you could start pulsed with a growing ring, the level being researched blinked on and off, the core and the capstone beams beat, sparks raced along every link, and a bright light at the core flared on the crystal nearest it. Now the lights hold steady. Only a node you can start still breathes, slowly — about once every six seconds — so it still stands out without flashing. The level being researched is a steady half-lit dot, the crystals are dimmer, the glow is softer, each link carries a single faint spark, and the floor no longer mirrors the core.

Reported by the owner after playing the board on the live game.

1.394.1 — 2026-10-06

Changed

• The Market's trade board in Italian (#6455). Before, players with Italian on saw the trade board in English. Now these parts read Italian:
  • the board itself, with its filters, sorting and offer rows;
  • posting an offer, with the fee and the Trade Portal's daily intake;
  • buying a slice of an offer, and your own offers;
  • the refusals and notices on these screens.

  In English, every word and figure reads as before.

  Still in English: the Dark Matter market, the Merit market and the Portal exchange panel, plus the market receipts in the Activity feed. They come in the next update, so the beta notice still lists the Market.

1.394.0 — 2026-10-06

New

• Back now returns you to the Calendar on the month and day you were reading (#5898). Before, Back onto the Season tab's Calendar always opened this month, on today. A player who paged back to July, opened the 14th, followed an event's coordinates to the Galaxy and pressed Back had to page back to July and find the 14th again.

  Now, Back (the browser's, Android's, or the on-screen ←) onto the Calendar:

  • opens the month you were on;
  • selects the day you had open in it, and loads that day's events as a tap would. A day in the future is never selected this way.

  What has not changed

  • Opening the Calendar from another screen, after a reload or from a notification opens it as before, on this month and today.
  • Switching to another lane of the Season tab and back still resets the Calendar.
  • With no day open, this month still falls back to today.
  • A month more than two years away opens on this month.
  • A day you tapped in the faded edge of the grid, one that belongs to the month before or after, is not remembered.

  Part of #5898. Still open there: the Missions segment on a phone, and the map cameras.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.393.0 — 2026-10-06

New

• Search your whole history for a place or a commander (#6395). Until now the Activity lookup searched only the events your feed had already loaded, which for most players goes back about a day. A commander who had changed their name also showed up as two different spellings. Now a «Search whole history for X» button appears under the lookup field, under All or Mine. It shows up when you type an address, or a name your feed already links to one commander. One tap loads the first page of your own events at that place, or involving that commander, from your whole history. «Older in this history» then walks further back a page at a time. Rows found under an older spelling, or that name nobody (a bounty on you, for example), carry one line saying why they were found. The suggestion chips now show each commander once, under their current name.
• This searches among your own events only. Some things it does not cover yet:
  • incoming-attack alarms record no attacker, so they cannot be found by who sent them (#6439);
  • a commander you only know from battles gets no whole-history button yet, because battle rows carry no commander id (#6440);
  • the Galaxy register is still searched only among the rows loaded (#6441).

1.392.1 — 2026-10-06

Changed

• The game stalls less while someone opens the leaderboard (#6457). Building the leaderboard used to re-price every level of every building in every colony, twice, each time anyone opened it, and the server could serve nobody else while it worked. That total is now remembered per building and level. Measured on the live server, the work that holds everyone else up dropped by more than half, from about 540 ms per request to about 235 ms. The board itself loads only a little sooner, because most of its time goes to reading the database. Scores and Power do not change: the new sum is checked against the old one for every building, up to well past the level cap.

1.392.0 — 2026-10-05

New

• The Shop opens under Markets. Supporting the game used to be a section at the very bottom of the commander's profile. It is now a tab of its own, the fourth door under Markets. It has a painted hero, your standing and VIP progress, and six packs from the Energy Pack (€1) to the Eschaton Pack (€100). Each card lists exactly what one purchase delivers to you: Dark Matter, Chrono shards, VIP points, hulls from €25 up, and the frame, emblem and bonus your running total reaches. The profile keeps one button, Open the Shop. The Shop is on novarchy.com only; the iOS and Android apps do not show it.

Changed

• Packs are purchases, and the words say so. Packs deliver items in the game, so they are sold as goods, not taken as donations. Prices on the cards are final and include VAT. Badges read "Data Supporter" instead of "Data Donator". The diary, the Inventory and the Chandlery now speak of shards "in a Shop pack" instead of "a donation".
• One question before payment. Buying a pack opens a short confirmation with the total. It asks you to confirm that you want the items delivered now, and that the 14-day right of withdrawal ends once they are delivered. Payment on Stripe's page opens only after you tick it.

1.391.4 — 2026-10-05

Fixed

The research board turns around its core. On the 3D research board, opening a branch aimed the camera at the middle of that branch: the guild core sat off to one side, and dragging turned the board around that point instead of its centre. Now the whole-tree view and every branch view aim at the core and turn around it. A branch is seen from behind the core looking out along it, with the core low in the middle and the branch rising above it to its capstone; on a phone the camera stands further back so the capstone still fits. Only a tapped crystal's close-up turns around the crystal, and going back brings it to the core again. The open branch's name on the rim no longer sits on top of its capstone's label.

Reported by the owner from a screenshot of the live board.

1.391.3 — 2026-10-05

Fixed

• The ship ledger no longer counts review runs with no recorded duration as zero minutes (#6370). The ledger in the repository records, for each review run, how many problems it found and how long it took. 196 runs had no duration recorded: 176 had none at all and 20 had placeholder text. The report counted each as zero minutes but kept its findings, which inflated the findings-per-hour rate. The report listed only the 20 placeholder runs as unreadable, and said nothing about the other 176. Now:
  • each review figure is computed only over the runs that recorded what it needs. The totals and each reviewer's rate say how many runs that is;
  • the overall rate goes from 12.8 to 12.6 findings per hour. Of the 126 reviewers with any recorded time, 14 change rate (the largest goes from 26.9 to 18.6), and 23 now show ? instead of 0.0;
  • a run can record an unknown duration or count as ?, stored as empty rather than as a number. Values such as unknown, an empty value, or a reviewer name containing spaces are refused when the row is written. The space rule is aimed at 46 existing rows whose fields ended up inside the reviewer's name. Those rows stay as they are, and the report lists them separately.
• The ledger records where each release's review lane came from (#6370). A ship row can now say whether its lane (express or full review) was printed by the pre-check, worked out by hand, or guessed. The report counts each. Every existing row reads as "unrecorded", because none of them say.
• The ship instructions quote unknown values (#6370). They used to show minutes=?, which zsh, the default shell here, treats as a filename pattern. When no file matches, zsh stops the command and no row is written. The instructions and the tool's own error messages now show 'minutes=?', which works in zsh and bash alike.
• The partial row a release writes for itself when no ship row reaches it ran for the first time on v1.391.1. It printed the expected summary line, added exactly one row of the expected shape, and the report counts that row as partial.

1.391.2 — 2026-10-05

Fixed

• The Activity lookup now finds Event Sector addresses (#6436). Typing an address in sector 13, the season's Event Sector (for example 13:1:3), used to hide every row and say nothing loaded mentioned it, even when the column held battles fought there. The lookup now accepts the same addresses as the rest of the game, so 13:1:3 and 13:1 find their rows. Addresses past the Event Sector still match nothing.

1.391.1 — 2026-10-05

Changed

• On a wide desktop, a transport's cargo is set right in the order, without opening the fleet page (#6408). Before, on a screen 1180px or wider, the WITH column of a Transport or a Pickup showed only a line, «Cargo nothing loaded set it on the fleet page ›». The cargo fields, the hold warning and its «Add N Freighters» button were on the fleet page. Typing the load is what makes the Advisor pick your Freighters, so that step always cost an extra tap.

  Now, on those two orders, the whole cargo section sits in the WITH column, under the fleet tiles and above «Choose ships — open the fleet page ›»:

  • the three fields and their «max» buttons, «fill the hold» and «fill every carrier»;
  • what the destination already holds, and how much room is left there;
  • the hold warning and its «Add N Freighters» button;
  • the supply-lane switch.

  Type a load and the Advisor's Freighters appear in the tiles just above. When the column is too short for everything, it scrolls, and the fleet-page button stays pinned at its bottom. On the fleet page, where the fields used to be, a line now says what is loaded, «Cargo 10,000 Fe set it in the order ›». Tap it to go back to the Ferrolite field. If your ships cannot lift the load, the line also repeats the hold warning.

  This changes the approved three-column layout of the order: the WITH column used to only describe a transport's cargo, and now it is where you set it.

  What has not changed

  • A Redeploy keeps its cargo on the fleet page, after the fleet, with the line in WITH. A Colonize order keeps its starting-endowment line.
  • Every other order keeps the fleet-page button exactly as it was.
  • On a phone and on a narrower desktop, the cargo still comes right after the Advisor and before the fleet (#6324).

1.391.0 — 2026-10-05

New

• Back now returns you to the Shipyard on the hull you were reading (#5898). Before, Back onto the Shipyard always opened it on the first hull, the Interceptor. A player who was comparing a Cruiser, stepped out to Research for the tech it needs and pressed Back had to find the Cruiser again. On a phone, the All-hulls list was back on its class sort too.

  Now, Back (the browser's, Android's, or the on-screen ←) onto the Shipyard:

  • reopens it on the hull you had selected, on any colony;
  • on a phone, also keeps the All-hulls list on the sort you left it on: by class, by owned or by cost. Closing and reopening that list inside the Shipyard already kept the sort; now leaving the screen and coming back does too.

  What has not changed

  • Opening the Shipyard from another screen, or after a reload, opens it as before: on the first hull, with the list sorted by class.
  • Back never opens a sheet: the hull's details and the All-hulls list stay closed.
  • The number of ships you had set to build is not restored.

  Part of #5898. Still open there: the Missions segment on a phone, the Calendar's month and day, and the map cameras.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.390.3 — 2026-10-05

Changed

• Battles in the Event Sector no longer put a bounty, a grudge or guild tension on anyone (#6384). Sector 13 is the season's arena, but until now a fight there was billed like one anywhere else: the attacker took a bounty for hitting a smaller empire, a wanted player's pot could be claimed off a temporary event colony, the victim earned a bounty-free reply, and the two guilds' tension climbed toward a war. Now none of that happens in Sector 13, whether you strike alone or in a raid. The Marshal's predator bounty is not billed there either. Hitting a pilot's colony while their fleet is away still costs you the rest of your Marshal strikes for the window and your share of its rewards, but it no longer puts a bounty on you. Points on the weekly war board still count exactly as before.

  Before you launch, the composer and the planet sheet now say so in one line: "Event Sector: battles here are the season's sport. No bounty on you, no claim on theirs, no reply owed, no tension between guilds. War board points still count." A target far stronger than you still carries its red warning, and the raid door simply stops quoting a pack price there.

  The game's AI empires treat an arena hit the same way: they remember it and can talk about it, but it never turns into a grudge or a revenge plan outside the arena.

1.390.2 — 2026-10-05

Fixed

• A ship-pipeline safety check now refuses every copy of the deploy script that still uses the live lock or worktree paths (#6411). Before, a new use of those paths, beyond the constants the check's harness rewrites, would pass it whenever the match sat early in the file. The deploy script's server-lock probe can no longer mistake an answer followed by more output for "the server did not answer" (today its marker is the probe's last line, so this was latent). The remaining ship scripts check their own output the same race-free way, and their self-tests refuse the racy shape if it comes back. Nothing a player sees changes.

1.390.1 — 2026-10-05

Changed

• The fleet template picker and the battle-line editor in Italian (#6393). Before, players with Italian on saw them in English, inside the Missions composer and the guild raid forms. Now they read Italian:
  • the template picker, with saving, loading and the note when a template is trimmed to the ships you have;
  • the battle-line editor, with the row hints, the formation presets and the summary of where each class stands;
  • the launch line that tells you how your fleet will deploy.

  In English, every word and figure reads as before. The Crucible and the campaign sheet stay English for now, as they were. The beta notice still says part of the Guild is in English, because the Guild Store's item names and descriptions come from the server in English.

1.390.0 — 2026-10-05

New

• «Allies under attack» can take you straight to an inbound fleet on the Galaxy map (#6398). A live row already said that only an interception from the fleet's contact on the Galaxy map can stop a fleet in the air, but you had to switch tabs and find the right moving triangle by hand. The row now has a Find it on the map button (Find them on the map when several fleets are coming). It opens the Galaxy map, even if you had left the tab on a system view. If your own sensors hold the fleet, its contact sheet opens on the soonest one, ready to plan the interception. If they don't, the Contacts board opens and says so, naming your ally, or tells you the strike landed while the chart was loading. The button never appears for fleets nobody but the raid's target may intercept (an ally's joint-raid contingents), or for routes with both ends in the starting sectors.

1.389.1 — 2026-10-05

Changed

• Sector 13 can be contested again: one runaway empire no longer prices the garrison, and a held world keeps only a quarter of it (#6407). Before, every control world sat under a neutral garrison of about 32.11M PP. It was priced on the ten strongest empires active in the last week, and one of them, at 139.01M, almost four times the next, pulled the price far above the typical top player. The wall was also rebuilt at full strength on a world a guild had just captured. One guild held all six worlds without stationing a single ship on any of them, and the rest of the server had stopped trying.

  Now the garrison leaves out any empire more than twice as strong as the median of those ten, and the next empire down takes its place. On the 5 October field that drops only the 139.01M empire, and the neutral wall falls to about 19.26M.

  A world a guild holds has its neutral garrison rebuilt at only a quarter of that, about 4.81M today, and never below the minimum garrison. Taking a neutral world still costs the full wall. Keeping a held one is now up to the holder's own fleets. A world whose guild disbands goes back to the full wall, normally within five minutes. The one-hour capture lock is unchanged.

  This lands mid-season: the restart re-prices every world at once. Held worlds drop to the quarter wall and neutral worlds to the new price within minutes of this release. The Contested Space panel and the wiki's Event page describe both rules.

1.389.0 — 2026-10-05

New

• Back now returns you to the Colony tab's building list as you filtered and sorted it (#5898). Before, Back onto the Colony tab opened the building list fresh: on All, unsorted. A player who picked Military, sorted by total cost and stepped out to the Base came back to the whole list in the shipped order.

  Now, Back (the browser's, Android's, or the on-screen ←) onto the Colony tab:

  • on the same colony, reopens the list on the family filter and the sort you left;
  • on another colony, keeps the sort and opens the filter on All. Switching colony on the tab has always reset the filter this way, so that a filter picked on one colony cannot leave the next one showing an empty list.

  What has not changed

  • Opening the Colony tab from another screen, or after a reload, opens it as before: All, unsorted.
  • The building descriptions you had opened and the Core quotes are not restored.

  Part of #5898. Still open there: the Missions segment on a phone, the Shipyard's selected hull, the Calendar's month and day, and the map cameras.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.388.3 — 2026-10-05

Changed

• Transport and pickup now show the cargo before the fleet (#6324). On these two missions the Advisor picks the Freighters for whatever load you type, as long as you have not chosen the ships yourself. But the cargo fields sat below the ship picker, and an empty cargo hold showed two prompts telling you to pick ships by hand («No carrier picked yet — pick one under Fleet…» and «Pick at least one ship.»). Following them took the fleet away from the Advisor, which then stopped sizing it.
  • The order is now Advisor, then cargo, then fleet, on phones, narrow screens and the desktop fleet page.
  • The cargo section moves as a whole, so the hold-gap warning, its «Add N Freighters» button and the supply-lane switch come with it.
  • Redeploy and colonize keep their current order.
• While the Advisor is still choosing your ships and the cargo is empty:
  • The line reads «No carrier picked yet — set the load and the advisor picks the carriers for it, or use fill every carrier.»
  • «Pick at least one ship.» is gone from the launch verdict. Launch stays blocked until you set a cargo.
  • The launch bar reads «Set the cargo for an estimate.»
• Once you pick or clear ships yourself, the prompts are the same as before, and an empty fleet still reads «Pick at least one ship.».
• On a redeploy with a load typed and no ships picked, the form no longer shows «pick one under Fleet» next to the hold warning's «Add N Freighters» button. Only the warning and its button remain.

There is no second Send button on the cargo line: Send is already on screen in every layout. Cargo fields in the desktop's main WITH column, so a transport can be set up without opening the fleet page, are tracked in #6408.

1.388.2 — 2026-10-05

Fixed

• The ship pipeline's self-tests no longer fail, or pass, by chance (#6405). Several checks in the release machinery fed a command's output to a matcher that stops reading at its first hit. When the machine was busy, that early stop could make the command report failure even though its output matched. One ledger check could turn any pull request's gate red at random. Three of the release script's checks for forbidden code patterns could pass with the pattern present: in a planted test, on 9 of 12 runs. Every check now captures the output first and matches it afterwards. Both scripts' self-tests now refuse the racy shape if it comes back. Nothing a player sees changes. The same shape in the other ship scripts is tracked in #6411.

1.388.1 — 2026-10-05

Fixed

• «Allies under attack» now says what can still stop an inbound fleet (#6284). A guildmate's row used to offer a garrison and nothing else, while it printed where the inbound fleet came from. A player raided that origin, spending a raid and their shield, and the fleet, already in the air, flew on. Each live row now says that a raid on the origin cannot stop a fleet in flight, nor catch it before it has flown home, and names what can:
  • an interception, from the fleet's contact on the Galaxy map, if your own sensors pick it up; the map says whether it can be met;
  • when the fleet is one contingent of a joint raid on a guildmate, only the raid's target may intercept it, so the way to help is a garrison;
  • when both ends of its route are inside the starting sectors, nobody may intercept it, so the way to help is a garrison.

  The open-raid form also adds a line when the address you type is the origin of a fleet still in the air: a raid there cannot stop it, nor catch it before it has flown home.

  Not included: a button that opens the fleet's contact on the map straight from the row. That is #6398.

  Reported on the bugs forum by a player who destroyed the 107 ships standing at the source colony and found 82 still on their way.

1.388.0 — 2026-10-05

New

Find a commander or an address in your Activity feed (#6288). The feed could be filtered by whose rows they were and by kind, never by who was in them or where they happened. The only way to keep hold of a row was to Keep it. Now a field under the chips, «Commander or address…», filters the feed among the events loaded:

• A name finds the rows that name that commander: the other side of a battle, a spy target, who spied on you, a buyer or seller, a register entry. Case and accents do not matter, so «zorg» finds «Zörg». It matches the name as each row spells it, so a commander who renamed shows up under both spellings.
• An address (1:6:1, or 1:6 for a whole system) finds the rows that happened there or came from there. 1:6:1 does not find 1:6:12.
• While the field is empty and focused, up to eight chips offer the names and places your loaded events mention most. Your own name is not one of them, but you can still type it.
• The search works with every chip, Kept included, and the chip numbers count what the search leaves.
• A line under the field says how far back the search reached and how to reach further. Older loads more under a type chip. When nothing matches under the chips you picked, the feed says which chip has the matches.

Suggested by a player on the suggestions forum.

Not included: rows older than what the feed has loaded are not searched, and alarms of an incoming attack cannot be found by the attacker's name, because they do not store it. Both are tracked in #6395.

1.387.0 — 2026-10-05

New

• Whisper a planet's owner straight from the planet panel (#6280). The planet panel printed the owner's name, but you could not write to them from there: you had to find them in the chat first. Now the panel's action row ends with a Whisper <name> button on every world held by another empire, whether enemy, guildmate or ally, and whether the owner is a person or carries the AI badge. Tapping it closes the panel and opens your private thread with them: on a computer the chat dock opens on that thread, on a phone the Chat tab does. Pirate Beacon, faction and relic worlds have no button, since nobody there can answer. If you blocked the owner, the panel says so instead and points you to Unblock in the chat.

  Suggested by a player on the suggestions forum. Not included: whispering from the system list or the map's hover tip, and blocking or inviting from the map.

1.386.0 — 2026-10-05

New

The hold warning can add the freighters it asks for (#6262). When the cargo you set outran the ships you had picked, the cargo panel said so in red and named the fix ("Add 8 Freighters, or the launch is refused"). To act on it you had to go back to the fleet, open the Freighter row and type the new count yourself. Now an Add 8 Freighters button sits under that sentence. One tap adds exactly those ships on top of the ones you picked, the warning clears, and Send becomes available. It works on desktop and phone, and on a redeploy too, where the game picks no ships for you.

• The button only adds ships. It never replaces the ships you picked and never swaps in the advisor's fleet. After the tap the fleet is yours, as after any edit by hand.
• It appears only when your spare freighters can carry the load on their own. If the rest could only travel in warship bays, or nothing left in the hangar covers the load, the sentence stays as it was and there is no button.
• The "fill the hold" and "fill every carrier" buttons still never change your fleet.

Suggested in the suggestions forum. The issue does not name the reporter, so the credit is suspended rather than withheld.

1.385.0 — 2026-10-05

New

Guild research is now a 3D research board. Opening the research door used to show a flat list of rows: the Fund and Endowment in a line of grey text, plain yard boxes, and every node as a text row whose Research button looked the same whether it could start or not. Now the door opens on the whole Guild Tech Tree in 3D: seven branches around the guild core, tiers 1–3 on the inner ring and the seasonal tiers 4–5 on a raised gold ring past the dashed season line.

• Every node is a model made for it: a crystal pylon for a node you level up, a sealed module for a one-time node, a gold spire for a capstone, and a scaffold while a node is still in development.
• Crystals glow cyan on permanent tiers and gold on seasonal ones, go dark while locked, and pulse when they can start. The links between nodes carry sparks once a prerequisite is met.
• The camera flies to the branch you open. Tap a crystal to see its name, level and effect on the board; Open its card takes you to its card below.
• On a phone, one finger still scrolls the page and two fingers turn and zoom the board. On a computer, click into the board to zoom with the wheel.
• Close up, the models are drawn at full detail; they load as you get near, so a first look is quick.

Below the board the panel is redesigned too: the Fund and Endowment as two meters, a branch list with each branch's progress, the two yards with progress bars, and each branch as a spine of tiers split into Permanent and Seasonal. Every card shows its level ring, its price and time (in red when the guild is short), and the reason it cannot start yet.

1.384.0 — 2026-10-05

New

A guest can continue an empire in another browser or on another device (#6368). A guest empire used to live in the one browser it was started in. Opened anywhere else, the game did not know you: Play or a sign-in there started a second, empty empire, and the first stayed out of reach. It caught players who started inside another app's built-in browser, such as Reddit's, and then opened their own.

Now your commander profile (tap your portrait) has "Continue in another browser". It gives you a one-time code. On the first page of the other browser, choose "Started in another browser? Enter your code", right under Play, and type it. The same empire opens there. You can also copy a link that carries the code.

• The code lasts ten minutes and works once. Asking for a new one cancels the previous one, and so does "Sign out other devices".
• Both browsers stay signed in.
• It is for guest empires. An empire linked to Google or Apple already opens wherever you sign in.

Also

• Inside another app's built-in browser, the first page no longer suggests signing in with Google there. It says to press Play, then move the empire to your own browser from the commander profile.
• If you open a code link in a browser that is already signed in to an empire, a notice says so and nothing is changed. Log out there and the code is offered on the first page.
• The new text is in English and Italian.

Not verified: none of this was tried on a real phone inside the Reddit app.

Known: after a code from a link has expired, a newer link pasted into that same tab still shows the old code. Type the new code over it, or open the link in a new tab (#6387).

1.383.10 — 2026-10-05

Fixed

Battle replays keep their glow when loading runs past the first second (#6264). The replay's quality ladder ignores the opening frames, where shaders compile and textures upload, and used to stop ignoring them after one second. On a device that loads for longer than that, the rest of the loading was judged like a scene that cannot keep up, and bloom was switched off in the opening seconds and stayed off for the whole viewing. The ladder now keeps waiting while loading goes on: it starts judging once the device draws three smooth frames in a row, and after 3 seconds at most.

What it costs, said plainly. On a device that is slow on every frame, the first quality step now comes at about 4 seconds instead of about 2. A stutter that starts inside the first 3 seconds looks the same as loading, so it is only judged if it shows again afterwards. Loading that lasts longer than about 4 seconds still costs a step. The 3-second limit is a choice, not a device measurement: everything here was measured on synthetic frame traces through the game's own renderer setup, not on a phone or tablet. The replay diagnostics (?replaydiag=1) now print how each warm-up ended, which is what a real device's report would need to tune it (#6191).

1.383.9 — 2026-10-05

Fixed

• In Italian, the Marshal chip's tooltip and screen-reader name are now in Italian (#6360). While a Marshal window is live, the HUD shows a chip with its countdown. Hovering it, or reaching it with a screen reader, gives one sentence: the level, the phase, how much of the level is left, when the round closes, and how many strikes you have left. Before, that sentence was in English for every player, with English time units. Now an Italian player gets it in Italian, with Italian time units, and the strikes part is the same sentence the Marshal board on the map already shows. In English nothing changes.
• Only the tooltip and the screen-reader name changed. The countdown printed on the chip and the Marshal tab itself are still in English.

1.383.8 — 2026-10-05

Fixed

• The Contracts dot no longer lights on a board with nothing left to do (#6328). The dot on the Season tab, and its badge in the phone menu, used to come on at every contract-day roll at 05:00 UTC, whatever the board held. It came on over a board whose cards were all claimed. If the browser had lost its record of the board being opened, as an iPad can when it clears site storage, it also came back on after a reload over a board already read. Now it comes on only while today's daily board still has something to do: the new hand has not been dealt yet (opening the board deals it), or a card is still unclaimed. A fully claimed board keeps it off, even after that record is lost. At the roll it also waits for the server's new day, so it no longer lights while the board still shows yesterday's hand.
• Two cases keep the old behaviour. If the record is lost while cards are still unclaimed, the dot still comes back on every reload until the board is opened. A day whose hand comes out empty, because no contract fits the empire, still gets one dot that day.
• Contracts are still dealt every day up to the end of a season: nothing pauses them before a reset.

Reported by astroeddie88 on Discord, whose in-game name is not yet established: the credit is suspended rather than withheld.

1.383.7 — 2026-10-05

Changed

• Every release now leaves a row in the ship ledger, including the ships whose record is incomplete (#5770). The ledger in the repository records what each release cost: review rounds, minutes, gate runs and reds. Before, a ship left no row at all in these cases:
  • it lost its worker partway through;
  • it was handed over to a second worker;
  • its ledger flag was never passed;
  • its row was due after the release step.

  Every summary computed from the ledger therefore covered only the ships that went smoothly, and nothing in the file showed that. Now:

  • a ship row can mark a value as unknown (?), which is stored as empty and never as a guessed number;
  • the row says whether it is complete, and if not, which fields are missing;
  • the row can say that the ship was handed over;
  • the row can count gate runs that were stopped before they reached a verdict, separately from the gate runs;
  • when no ship row reaches the release, the release writes a partial row itself, recording only the pull request, the version and that it shipped.
• The ledger report counts the missing parts instead of hiding them. The report now opens with the share of partial rows. Every figure is computed only over the ships that know that value, and the report says how many that is. On the existing ledger, 18 of 1,135 ships (1.6%) now show as partial: they never recorded their minutes. Before, all 18 entered the median ship time as zero minutes. The median is still 95 minutes, now over the 1,117 ships that recorded it.
• A ship that never reaches the release step still leaves no row, so the partial share is a lower bound, and the report says so. Recording where each ship's lane came from, and how lens rows with unknown costs are counted, is #6370.

1.383.6 — 2026-10-05

Fixed

• The phone's settings sheet is titled Settings, Impostazioni in Italian (#6343). It was titled Command, the old Menu's name, in every language, so the Settings door on the Empire tab opened a sheet with a different name on it.

1.383.5 — 2026-10-05

Fixed

• In Italian, a strike on the Marshal now has one name: «attacco» (#6359). Before, the Marshal launcher counted your strikes as «attacchi», but three of its launch refusals (too late for this round, all strikes used, strikes forfeited) called the same thing a «colpo», so one screen used two words for one thing. Now those three refusals, the Marshal entry in the glossary and the Marshal report's "no threshold prize" row all say «attacco», as the launcher does. The English text is unchanged.

1.383.4 — 2026-10-05

Fixed

• The Marshal's dot in the Events menu goes out when no strike can go out anymore (#6346). On desktop the dot used to stay lit for the last 11 minutes of every Marshal window, even though in those minutes every launch is refused («Too late for this round»), so it led you to a tab with nothing to do. It now goes out at the same moment the launcher stops accepting strikes. The window's clock stays in the menu, in gold, without the dot.
• The Marshal's line in the Events menu is translated (#6347). «Live · closes in 1h 12m» and «Opens in 1d 3h» were always English. In Italian they now read «In corso · chiude tra 1h 12m» and «Apre tra 1g 3h», with the same words the map's Marshal board uses for the same clock. English is unchanged.

1.383.3 — 2026-10-05

Fixed

• A line that wraps no longer starts with a lone «·» (#6350). With an Overdrive level already queued, the Overdrive track's cost line reads «Cost lvl 28 · after lvl 27 completes». On a narrow panel, the line could break right before the dot, so the next line started with «·» on its own. The space before the dot can no longer break, so the dot stays at the end of the text it follows. The same fix applies everywhere else a «·» clause is appended in the same way:
  • the incoming-attack card's intel line (origin, attacker, guild, fleet size, ships, classes, composition, and the unreadable and partial notes);
  • the launch summary's travel and return times;
  • the Target panel's spy line (hangar, defenders, stock);
  • the Raids panel's ally and raid rows;
  • the Last Convoy buttons;
  • the Bubble line on the planet sheet.

  The words are unchanged. On an incoming-attack card with nothing before the attacker (a lone fleet whose launch colony was razed mid-flight), the line still opens with the attacker's «·», with no indent, as before.

  Not measured: the panel widths at which each of these lines actually wraps.

1.383.2 — 2026-10-05

Fixed

• The Empire tab reads Italian (#6331). Behind the phone's fifth slot, an Italian player used to see an English screen with Italian sentences inside the rows: «EMPIRE», the colony and hull count, «Needs you», the BUILD · YARD · STORAGE · DEFEND buttons, «Show all» / «Show fewer», the Events / Markets / Doors headings, the Settings door, the stock tiles' «at cap» and «in transit», and the loading, failed and fleets-in-flight notes. All of it is now Italian, in the Overview's words: «Da controllare», «scafi: N», «magazzino pieno», and the build button reads «EDIFICI», like the base's own build button. English is unchanged. Not yet translated: the sheet the Settings door opens is still titled «Command» in English (#6343).
• The dots on the Empire tab's doors are red (#6335). The Season door's dot (today's contracts unopened) and the Activity door's dot (a probe detected) were gold, the colour that since #5578 means a condition no tap clears, like the hollow idle-colony count on Overview. These dots go away when you open the door, so they are now red, the same as the dots on the bottom bar.

1.383.1 — 2026-10-05

Changed

• The test badge on the repository's front page keeps up with the code (#5820). Before, the badge moved only when a release had measured both test suites. Most releases run the Python suite partly or not at all, so the count fell behind with every test added in between. It reached 134 tests behind on 5 October. Now the badge reads "last recorded full-suite count of each suite". Each count moves when its own suite was measured in full on the code that actually shipped. A count that was not measured is never republished as if it had been.
• The badge is checked against what shipped. A count is published only if the code it measured matches the release, apart from the changelog, the README and the ship ledger. A count taken on an older version of the change, or on code that moved before the merge, leaves the badge as it is.
• This changes the release tooling only, and it reaches only the working copies that pull it. Copies still on the old tooling keep the old, stricter behaviour: their badge does not move.

1.383.0 — 2026-10-05

New

• You can bookmark planets on the galaxy map, with a note (#5579). Before, the galaxy had no saved places: to come back to a coordinate you had to remember it, or dig it out of a report again. Now a ☆ beside the ✕ on any planet sheet bookmarks that planet to your account, so the list follows you to every device. Once bookmarked, a one-line note field opens under the planet's name ("Why this planet?", up to 60 characters, private to you).

  Your bookmarks are listed in the map's Places drawer, in a BOOKMARKS section above the landmarks:

  • Each row shows the coordinate and who was there when you saved it (with the AI badge where it applies), your note, and how long ago you last scouted it.
  • If the planet has changed hands since, the row says so: "now held by …", or "no longer on your scan". It never says "empty": the scan cannot see a veiled colony, so a missing colony is not proof the planet is free.
  • Tapping a row takes the camera there and opens the planet's sheet. The ★ on a row removes the bookmark.
  • An Event Sector bookmark opens the system view while a season runs.

  You can keep twenty bookmarks. You add them from the planet sheet only; the Targets board has no star of its own, but every row there already opens the planet sheet.

  Suggested by a player in the suggestions forum. It also answers the other half of #2841: TartarusMkII asked whether there was a way to record the targets he had scouted, or whether he should keep a note outside the game. Now there is a way inside it.

1.382.3 — 2026-10-05

Fixed

• League byes now fall on AI guilds before human ones (#6355). When the number of guilds is odd, one guild sits out the weekly League round. Until now the choice ignored who was playing, so a guild full of people could be benched while two all-AI guilds played each other. This week it happened to The Chilling Arc Ship. Now a guild with no human member takes the bye first. A human guild sits out only when no AI guild can, and one human member makes a guild human. One exception: if benching the AI guild would leave other guilds with no legal opponent (for example two allied guilds that can each only play it), a human guild rests instead so the match still happens. This week's round was corrected by hand: The Chilling Arc Ship now plays Scourge of the Stars, and Perimeter Holding rests. The Chilling Arc Ship starts the week at zero, so its Monday construction before the correction (00:00–10:00 UTC) did not score.

1.382.2 — 2026-10-05

Fixed

• A cost caption no longer starts a line with a lone «·» (#6320). The line above each price (for example «Cost lvl 21 · after lvl 20 completes · still needed below each price») wraps on a narrow panel, most often in Italian, where it is longer. Before, it could break just before a separator dot, and the next line then opened with «·». Now each dot stays at the end of the segment before it, and the line can break only after it. This applies to the Discipline panel, the Tree node sheet and the Building sheet. The words are unchanged. The Overdrive track's cost line has the same issue and is tracked in #6350.

1.382.1 — 2026-10-05

Fixed

• A screen reader now hears the marks on the desktop tab rail (#6332). Before, the rail's marks said nothing useful. A screen reader read Activity's unread battle count as a bare number after the tab name, and read nothing at all for the probe dot, the Guild dot or the dots on the Events menu. Now each button says what its mark means, with the same sentences the phone's bar uses since #5578:
  • «Activity, 2 unread battle reports» or «Activity, probe detected»;
  • «Guild, 1 call from your guild»;
  • «Events, today’s contracts are waiting and you have a Marshal strike left» on the closed Events button, naming every dot that is lit, in menu order;
  • «Season, today’s contracts are waiting» and «Marshal, you have a Marshal strike left» on the items inside, where the Marshal’s clock line is still read as its description.

  The marks themselves are silent, so nothing is said twice. In Italian, for example, «Attività: 1 rapporto di battaglia da leggere» and «Eventi: i contratti di oggi ti aspettano e ti resta un attacco al Marshal». Nothing changes on screen.

  Not included: the Overview idle-colony count, Missions’ inbound count and Home’s reward dot are still not drawn on the desktop rail, so they are not named there either.

1.382.0 — 2026-10-05

New

• Colony names on the desktop Home now take you to that colony (#5613). Before, only the planet labels on the 3D table did, and on a machine without 3D support there was no way at all from a colony's name to the colony. Now the name in each storage row, in each colony's lane on the timeline, and the colony half of a threat on the radar opens that colony's Base. At rest the name looks as it did; hover or keyboard focus turns it cyan and underlined. A threat whose target the game could not resolve still prints «?» and is not a link. The alert card's «Defend …» title and the order rows are unchanged (they already have their own buttons).
  • On a timeline lane with a «+N» counter, clicking the colony's name used to open the lane instead. The name and the counter are now two separate targets, one above the other, and the lane keeps its height.
  • On the radar, the AI mark now sits beside the attacker it describes, not after your colony's name.

Changed

• The 3D panel on the desktop Home is now called «Theatre of operations» (#5613). It was «Empire theatre», which reads as the name of a building. Only the title changes, and the Italian title changes to match.

Suggested by a player on the suggestions forum, who also pointed out the wording of the panel's name.

1.381.0 — 2026-10-05

New

• The Shipyard, Nanoforge and Orbital Arsenal cards quote build time for the hull you pick (#5702). Until now, «Time per hull» on these three cards was always for a Battleship. A player building fighters read a number about fifteen times larger than their real timer and had to convert it before every upgrade. A Reference hull menu now sits at the top of «Contribution now» on those cards. It lists the hulls your empire can order, in the Shipyard's order, with the Battleship as the default. The time, the next-level time, the saving, the first-fleet allowance, the note and the full level table all follow your pick. One choice covers all three cards, so they never disagree. It is remembered on this device, for your account only.

  When the arrow in «Next level» does not move, a line under it says why:

  • the Arsenal does not shorten civilian hulls;
  • the hull is already at the ten-second order minimum;
  • or, while the first-fleet allowance holds it there, the next level will shorten it once the allowance is spent.

  The Nanoforge's heavy and light percentage rows, and the Arsenal's percentage row, stay as they were: they describe the mechanic, not your fleet.

Suggested by a player on Discord.

1.380.1 — 2026-10-05

Fixed

• AI empires no longer waste Marshal strikes that cannot count toward the prizes (#5627). A strike only counts toward the window's threshold prizes when it carries at least a quarter of the striker's whole military. Before, an AI empire sent its usual share of the warships at home, so one whose fleet was mostly at outposts or in flight launched strikes that scored but could never qualify: a careful empire with 40% of its fleet at home committed about 18%.

  Now an AI empire's strike carries enough of its home fleet to qualify, up to a limit set by its temperament: a careful empire never sends more than about two thirds of what it has at home. When even that is not enough, it holds a strike it would otherwise have flown alone, and tries again on a later look once its fleets are back. In chat it says that it held, and why, rather than inventing a reason. A guild raid it has been called to still gets its usual contribution.

1.380.0 — 2026-10-05

New

• The guild raid and defence doors now have the War Table's per-hull controls (#5709). Composing a fleet there used to mean typing a number into a bare field for every hull. Now:
  • every hull card has a Max button;
  • tapping a ship's picture opens the War Table's own editor under it: None, the count ladder, Half, All, and the − / + steppers you can hold;
  • a new Everything useful · N button fills every hull the door shows bright, at the launch colony's full stock, in one tap.
• Everything useful only fills the form. It never sends: Send, Join and Launch are still the only buttons that launch. A count you typed on a dimmed ship is kept.

Fixed

• A garrison no longer takes your founding ships into battle (#6317). The Muster tap and the quick garrison in Missions sent the Colonial Ark along with the defenders, while their own advice said colony ships stay home. Now:
  • both founding hulls (Colony Ship and Colonial Ark) and the Spy Probe stay home on every one-tap fill or send;
  • on a defence they are dimmed with their reason («keep it home», «no weapons, dies first»), the same as on an attack;
  • you can still send one by typing its count on its own card.

1.379.3 — 2026-10-05

Changed

• The number on Overview no longer looks like a message you can clear (#5578). On a phone, Overview counts your colonies that have nothing building. It was drawn as the same filled red badge as Chat's unread count, so tapping around looked like the way to clear it, and nothing did. The only thing that lowers it is giving those colonies a build. It is now a hollow gold badge: a gold ring with gold digits. Every badge that does clear when you look at it (battle reports, chat, guild calls, inbound attacks) stays filled red. The Empire button in the bottom bar still adds up everything behind it. It turns red if anything behind it is news, and hollow gold when the only thing behind it is idle colonies.
• Overview and the Empire tab now say it in words. Under NEEDS ATTENTION on the phone's Overview, and in the Empire tab's «Needs you», a line reads «2 colonies have nothing building — that is the number on Overview, and it drops as each one starts building». The header above it still counts every colony that needs you for any reason, so the two numbers can differ, and the line says which one the badge is. The desktop is unchanged: its tab bar does not show this count.
• Screen readers hear the badges. Every badge on the phone's bottom bar and on the Empire tab's doors is now read out with its tab, as in «Chat, 3 unread messages» or «Overview, 2 colonies with nothing building». Before, the bar's badges were silent and the doors read «OVERVIEW 2» with no meaning.

  Reported by a player in the bugs forum, who tapped the idle notices on Overview and found the red 2 would not go away.

1.379.2 — 2026-10-05

Fixed

• The Orbital Arsenal's hull time now warns you when the first-fleet allowance is in it (#6315). The Arsenal's «Time per military hull» figure has always included the first-fleet allowance while it is open, the same as the Shipyard's and Nanoforge's per-hull times. Its note, though, read only "Reference hull: Battleship; includes this colony’s current shipyard, Nanoforge and empire modifiers." A new player could read a hull time that rises on its own as they order, with nothing on the sheet saying so. While the allowance is open, the note now reads "reference hull: Battleship · first-fleet allowance 40% left, hull times rise to full length as you order · includes this colony’s current Shipyard, Nanoforge and empire modifiers". Once the allowance is spent, it drops back to the reference hull and the includes clause.

Changed

• The Shipyard's «Time per hull» note says what else its figure includes (#6315). The Shipyard's hull time also counts this colony's Nanoforge and Orbital Arsenal, but its note never said so. Every form of that note now ends with "includes this colony’s current Nanoforge, Orbital Arsenal and empire modifiers", including the one shown before the Shipyard is built. The Shipyard, Nanoforge and Orbital Arsenal sheets now word their per-hull notes the same way. The Nanoforge's note is unchanged.

1.379.1 — 2026-10-05

Fixed

• Attack forecasts now count your Wraith-pattern refit (#5625). Before, if you owned the refit, the attack forecast and the Advisor fought your corvettes without it, while the real battle refits up to 40 of them with +12% shield. On a close fight that was the wrong verdict: twelve corvettes against twelve read as a loss in every simulation, but the battle wins. Now the forecast row and the Advisor's recommendation fight with your refit, just as the battle does. The defender's refit is still unknown to the forecast, because no spy report shows it, so against another refit owner the forecast is a little optimistic.
• The Hulls sheet shows your refitted corvette shield (#5625). If you own the refit, the corvette row now carries a shield refitted figure next to its shield. If you hold the shield-defence nodes, it also shows a shield refitted defending figure. A note explains that up to 40 corvettes per battle carry the full refit, and that a bigger group shares the extra, so each corvette lands at or between the two figures. The refit is not carried against the Marshal, the Season Boss, the Crucible or the Campaign. The sheet's list of what it leaves out now also names the shield the Ward gives your garrisons and the hull the Phalanx gives when you defend a colony. Not included: the Shipyard's stat bars still show the unrefitted corvette shield.

1.379.0 — 2026-10-05

New

• The fleet ledger shows what is in production (#5686). The ledger on the Shipyard, the Overview and the Fleets tab counted only hulls that already exist: in a hangar, in flight, or on garrison. A batch on the slips was invisible in the table you read before ordering more. Now, while any shipyard has an order running, the ledger gains an In production column after Total. It shows how many of each hull are being built across the empire. Total still counts only the ships you own, so it keeps answering what can fly today. A hull whose first batch is still being built now gets its own row, with a dash under Total. On a phone, the Overview's fleet line adds «· N in production», and the hull chips still count only ships you own. Dry-dock repairs are not counted as production.

  Suggested by a player on the suggestions forum, who pointed out that with only a few slips per colony, what is already queued is what decides the next order.

1.378.23 — 2026-10-05

Fixed

• The ship gate no longer prints a shell error when a test module is missing (#5802). Before, if the merge check was handed a Python test module that was not on disk, it printed one No such file or directory line per missing module into its log, even though the check itself went on correctly. Now it prints nothing for them. The missing module still gets a test shard, as before, and two self-test rows pin both of those behaviours. No path does this today, because every list of modules is taken from the files on disk, so players see no change. This is ship tooling, and it reaches only the trees that pull the new gate.sh.
• A test-count figure in the ship instructions now matches the commit it cites (#5802). The figure was 45,316, read from a different commit. It is now 45,336, the count in the README at the cited commit.

The other two items left open on #5802 are covered. The ledger-path check it named had already been fixed on main: a directory or a FIFO is now refused, and the refusal names the real cause.

1.378.22 — 2026-10-05

Changed

• The Nanoforge sheet now shows how much build time it saves per hull, in minutes and seconds (#5701). Its Heavy and Light hull rows gave only a percentage of a build time the sheet never stated. The Shipyard and Orbital Arsenal sheets show a duration per hull, so the three buildings that speed up ship construction could not be compared. A new Time per hull row, between the two percentage rows, shows a Battleship's build time at this Nanoforge level and at the next, and how much the upgrade takes off. On the same colony it is the same figure the Shipyard sheet shows: it includes this colony's Shipyard, Orbital Arsenal and empire modifiers, and while the first-fleet allowance is still open it says so, exactly as the Shipyard does. The row appears only on colonies that have a Shipyard. The percentage rows are unchanged.

  Suggested by a player on Discord.

1.378.21 — 2026-10-05

Fixed

• A wrapped price no longer loses its «Still needed» label (#6274). The red figures under each price (what you still need to bank) used to be labelled «Still needed» in a column beside the prices. That column took width from the prices. On the Discipline panel and the Tree node sheet on desktop, and on a narrow phone, the prices then wrapped onto a second line, and that line had no label. In Italian this happened at every panel width. Now the caption above the prices names them: «Cost lvl 20 · still needed below each price», worded the same way in Italian. The prices get the full width, and in every case we measured they fit on one line. The Building sheet uses the same caption. A very long price still wraps: a Singularity Vault at level 30, or a Dark Matter node's four prices on the desktop Tree sheet. When it does, the caption sits above both lines. Screen readers read each figure the same way as before.

1.378.20 — 2026-10-05

Fixed

• The Marshal's Discord close post now says which wall the window closed on (#6281). A window that closed at level 1 with the Shield Line down was announced on Discord as "0 levels of the Marshal fell", which read as false to a server that had just watched a wall fall. The post now adds where the window closed, in the same words as the battle display and the Activity result: "…and 0 levels of the Marshal fell — the window closed on the Wraith Escort of level 1." When levels were cleared, it names the wall standing on the next level, for example "2 levels of the Marshal fell — the window closed on the Shield Line of level 3."

  The level count and the number of commanders are unchanged, and the post still names no one. This starts with windows that close after this release. Windows that closed before the Marshal's results began recording the wall (v1.378.13) keep the old sentence.

1.378.19 — 2026-10-05

Also

• A maintainer note on the two shields that share one stored name (#5906). Nothing changes in the game. The Planetary Shield building and the guild treasury's 8-hour Peace shield both use the same saved id, scudo, in two unrelated places. Each declaration in the code now points to the other and says why neither can be renamed: the building's id also names its artwork files and lives in every colony's saved state, and the shield's id is recorded on every guild shop purchase, where it feeds the weekly price escalation. A future rename that changed both by search-and-replace would have broken purchase history and colonies' saved buildings.

1.378.18 — 2026-10-05

Fixed

• The deploy step no longer reports a deploy for code that is already live (#5839). Players see no change. This is ship tooling. Before this release, deploy.sh --pr <N> let the PR's own runtime files decide whether to deploy. When a sibling ship had already deployed main, including this PR's merge, the step rebuilt the tree production was already running and reported it as a deploy. Run without --pr, the same tree correctly ended SALTATO. Now the decision always compares what production runs against main, so both forms give the same verdict. --pr still records the PR number in the deploy journal and the run state. The script's own self-test reproduces the sibling case: red before the fix, green after. This reaches only the working trees that pull the new deploy.sh and SKILL.md. A tree that has not pulled them keeps the old behaviour, and there the bare form, without --pr, gives the right verdict.

1.378.17 — 2026-10-05

Fixed

• Five more API test suites stop reloading the whole module graph to re-read one setting (#6293). Nothing changes in the game: this is the test suite only. It follows #6267. The Google sign-in, first-convoy, first-steps, PvE calibration and event-garrison tests each re-read an environment variable by reloading 170 to 342 modules, one round trip to the test runner at a time. They now reload only the module that reads the setting, which is one to four modules per test. Every assertion is the same as before. When each module fetch was delayed by 250 ms, five of these tests timed out before this change. After it, all of them passed in about a second or less.

1.378.16 — 2026-10-05

Fixed

• The Building sheet puts each «Still needed» figure under its own price (#6268). The sheet a tap on the map opens printed the price as one row (every resource the level costs) and «Still needed» as another row listing only the short resources. With one resource covered, every deficit after it slid under the wrong price: with Ferrolite covered, «−3,120 Xe» sat under the Ferrolite price. Now each resource is its own column, as on the research Tree sheet and the Discipline panel since #5696: the price on top, and under it either the deficit in red or a green ✓ when you hold enough. Figures stay exact. The «Cost lvl N» caption now sits on its own line above the prices. With an empty treasury, each column shows its own full deficit instead of the phrase «the full price». Screen readers hear «Still needed» before each missing amount, and the ✓ reads as «Ferrolite: covered».
• «the full price» is translated. On the map's hover card and in the Build-menu row, an empty treasury collapses the deficit to «the full price». That phrase was always English, so Italian players read «Mancano the full price». It now reads «Mancano tutte le risorse».

Not included: when the label and three exact prices are wider than the sheet, the prices wrap onto a second line that has no visible «Still needed» label. Each red figure still sits under its own price. Tracked in #6274.

1.378.15 — 2026-10-05

Fixed

• An anchorage hit at the moment the season closes now counts under the new season's rules (#6162). A battle that resolved in the same instant as the close could be scored against the old season's breach limit. The worst case was a bar shown full, with its modules dark, yet the anchorage still standing until somebody attacked it again. In another case the attacker's notch was simply lost. Now the battle is re-scored against the bar as it stands after the close. If that hit fills the bar, the anchorage falls then and there, and the notch is never lost. The battle report shows what actually happened to the anchorage, not what was expected before the close.
• The same re-scoring also covers a battle that lands at the same moment as a repair or another battle on the same anchorage. Before, that notch was dropped too.

1.378.14 — 2026-10-05

Changed

• Espionage past level 24 now pays off against defended targets (#6295). Each Espionage Overdrive level used to add +1% to the chance a probe goes unnoticed. Against a well-defended target that chance is small, so a level was worth a third of a point or less. Against a target that already spots almost every probe, it was worth less than a tenth of a point. Each level now takes half a point off the chance the target notices your probe, and it keeps working even against a defence already at 95% detection or backed by a guild running Counter-Intelligence. If you hold the Cipher, it multiplies what is left, so each level is worth a little more. A probe is still caught at least 5% of the time, so against a target your probe already slips past 95% of the time a level adds nothing, as before. Against lightly defended targets, where a probe already goes unnoticed about 70% of the time or more, a level is worth slightly less than the old +1%. Levels you already own keep their count. The research panel and the wiki now describe the effect in points instead of "+1%".

1.378.13 — 2026-10-05

Fixed

• The Marshal's result now names the walls that fell before the window closed (#6261). When a window closed with a wall down but no full level cleared, the result in your Activity feed still said "No level was cleared, so no threshold paid". That was true, but it read as false to anyone who had just watched the Shield Line fall. From the next window on, the result names the same walls the battle display shows. It says which walls of the current level fell and which one the window closed on. When a wall fell but the level did not, it explains that a threshold pays only on a whole level. For example: "Level 1 lost the Shield Line; the window closed on the Wraith Escort. A threshold pays on a whole level, so none paid."

  The level count, the prizes and the title are unchanged, and so is what each window pays. Results from windows that closed before this release keep their old wording, including the 4 October window that prompted this fix. The Discord close announcement still counts only levels; that is tracked in #6281.

  Reported by a player on the bugs forum.

1.378.12 — 2026-10-04

Fixed

• The protection rules now say the Marshal is free too (#6225). Striking the Marshal, alone or in a guild raid, has never cost you anything: your empire and colony cordons, your novice shelter and your night window all stay up. The game didn't say so. The cordon banner, the protection tooltip and sheet, the War board's hints, the spy report, the launch warnings, the attack confirmation and the Store's Cordon rows all named the Pirate Beacon as the only target you could attack for free. A protected player who believed them could skip a Marshal window to keep hours they would never have lost. Every one of those sentences now names the Marshal beside the Beacon, in English and Italian. The night-window rule says the same, and the cordon banner and Store rows gain a comma so "the Marshal" and "the cordon" no longer read as one phrase. Nothing changes in how protection works.

1.378.11 — 2026-10-04

Also

• The Missions tab's translations are now all checked by tests in every language (#5749). Nothing on screen changes in this release. The Missions tab has 149 places where it looks up text in the translation catalogue. We swapped each one in turn for the English it would print if someone had typed it in by hand, and checked whether any Italian or test-language run noticed. Twelve slipped through every time: a label left in English would have stayed in English for Italian players, and no test would have failed. Those twelve are:
  • the «Fleet» headings (three of them);
  • «clear» on the cargo row;
  • the wording of the Send button;
  • «Not now» on the garrison and bounty confirmations, and the line under the bounty one;
  • the raze, sabotage and no-watch entries in the "Also" line;
  • the target named in a quick attack's confirmation.

  Each of the twelve is now checked in English, in Italian and in the test language. 137 of the 149 were already checked.

1.378.10 — 2026-10-04

Fixed

• The rate-limit tests no longer time out on a busy test machine (#6267). Nothing changes in the game: this is the test suite only. Three API tests checked that the per-player rate limit reads its settings from the environment. To read them again, each one reloaded about 160 unrelated shared modules, one round trip at a time to the test runner. Run alone, a test took half a second. In a full parallel run one of them was seen to cross its 60-second limit. Now each test reloads only the module it checks, two modules per test, and every assertion is the same as before. When each module fetch was delayed by 250 ms, all three tests timed out before this change and passed in under a second after it.

1.378.9 — 2026-10-04

Changed

• On a phone, disorder is now a chip, not a paragraph (#6275). A follow-up to the disorder rule: a colony in disorder after an undefended raid used to show a full paragraph at the top of every phone tab, pushing the tab's own content down for the whole day. It now sits in the same row as the shield and siege chips, with the time left always visible ("Disorder · 5h 12m"). A tap opens the full explanation underneath, and another tap folds it away. Desktop is unchanged.

1.378.8 — 2026-10-04

Fixed

• Research: each «Still needed» figure now sits under its own price (#5696). On the Discipline panel and on the Tech Tree node sheet, the «Still needed» line listed only the resources you were short of. Once one resource was covered, every deficit after it slid one column over. Cybernetics at level 19 showed 8.37M Fe · 8.37M Xe · 7.17M Dt, with −1.66M Fe · −7.93M Xe underneath, so the Fe deficit sat under the Xe price and the Xe deficit under the Data price.

  Now each resource is its own column: the price on top, and under it either the amount still needed in red or a green ✓ when you already have enough. A screen reader says the ✓ as «Data: covered». The «Cost lvl N» line has a line of its own above the prices. On the Tree sheet, the Dark Matter price of the tier-9 nodes gets the same column. Screen readers now hear «Still needed» right before each red figure.

  One thing is left over. When the prices don't fit on one line, they wrap, and the second line has no «Still needed» label beside it. Each figure is still under its own price. This was measured in a test browser with a substitute font, so the widths may be a few pixels off. In English it happens on the Discipline panel when the panel is narrower than about 268px. In Italian it happens across the panel's whole desktop width. With four prices it also happens on a 320px phone. It is tracked in #6274. The building sheet has the same column problem and is tracked in #6268.

  Reported by a player in the bugs forum, with a screenshot of the Cybernetics panel, who also proposed the check mark.

1.378.7 — 2026-10-04

Changed

• Fleeing has a price: a raid on an emptied colony leaves it in disorder for 24 hours (#6275). Flying your fleet away before an attack used to cost only the resources the Shielded Silo did not protect: ships came home intact, defences regrew, and production never noticed. Now a raid that wins with nothing in the field (no ship and no defence) puts that colony in disorder for a day: production −20%, build and ship times +20%, and research times +20% too if it is your home colony. Buildings are never touched. Keep anything in the field, such as a ship, a Battery, a Fortezza or the Bubble, and there is no disorder. Being raided again during the day restarts it from that raid; it never stacks deeper. Your Activity feed names it, a banner counts it down, and every production and time figure already includes it. Guild joint raids apply it the same way. A colony razed while in disorder does not carry the disorder to its new home.

1.378.6 — 2026-10-04

Fixed

• A launch-screen test no longer fails at random on a busy build machine (#5703). Nothing changes in the game. The test checks that the phone and desktop launch screens show the same guild travel-time note. It sometimes failed because the travel-time request settled halfway through, so one screen was read while still waiting for the server and the other after it had answered. The test now holds the request open and reads both screens only after it has been sent, so a busy machine can no longer change the result.

1.378.5 — 2026-10-04

Fixed

• A refused trade now shows its red refusal right above the Accept button (#6253). On a phone, the Accept button sits pinned at the bottom of the trade sheet, but the server's refusal was added at the very end of the scrolling content, below the breakdown, the delivery picker and the copy-to-composer button. Nothing scrolled to it, so tapping Accept seemed to do nothing until you scrolled down by hand. Now a refused accept, or a refused cancel on your own offer, scrolls the red line into view just above the button you tapped. It does this again on a second refusal with the same message, and screen readers announce the line. The desktop dialog brings its refusal into view in the same way. The wording and position of the message are unchanged.

  This was checked with automated tests, not on an Android device. The installed Android and iOS apps get the fix only with their next app build. The web game at novarchy.com has it now.

  Reported by a player on the bugs forum, with two screenshots.

1.378.4 — 2026-10-04

Changed

• Change course now works in sector 13 (#6254). Turning a fleet in flight onto a new target used to be refused for any order touching the Event Sector («No course can be plotted from where this fleet is to there»). That covered a fleet heading for sector 13 and a fleet turned toward a sector-13 world. Now you can change course into, out of and inside sector 13 until the point of no return, like anywhere else, so a feint at one Event world can turn onto another. Sector 13 is off the hex map, so the new arrival is never earlier than a direct launch from home at your original departure would land. A recall after such a turn brings the fleet home on a route it could really fly, and a warped launch does not get its warp back for free. A fleet that turns off the map cannot be seen on it or intercepted by pickets, just like a fleet launched from a sector-13 colony. The defender still gets the inbound alarm and countdown.

1.378.3 — 2026-10-04

Fixed

• The event-loop monitor's tests stop failing on a busy build machine (#5688). Nothing changes in the game. The tests for the API's stall gauge, the one behind the admin runtime panel, used to fail now and then during full test runs. They read 22, 34 and 42 ms where a block of at least 80-100 ms was expected. The tests slept for fixed times, and the gauge needs one tick of its own before a stall in order to measure it. When the build machine paused the test at the wrong moment, the stall went unseen. The test of the real gauge now waits on the gauge's own sample count. The tests of its arithmetic write their readings into a real histogram. A new test pins how finely the gauge samples. The gauge itself behaves exactly as before in production. The second flaky test named in the same report, a 60-second timeout in the rate-limit suite, is not fixed here: it moved to #6267.

1.378.2 — 2026-10-04

Changed

• A stuttering battle replay now gives up bloom and resolution within seconds, including when the stutter comes in bursts (#6236)
  • Before: the replay judged its smoothness 45 frames at a time and stepped down only when more than half of those frames were slow.
    • Burst stutter never counted. 22 slow frames in every 45 is more than two seconds of hitching in every three, and the replay stayed at full quality.
    • One extra slow frame per 45 could flip the replay from keeping everything to losing every step.
    • A slow device took longest to react: about 19 s at 10 fps, and minutes slower still.
  • Now: the replay judges each second of play on how much of that second was spent in slow frames.
    • At 10 fps it steps down once per second, about 2 to 5 s after the replay opens, where it used to take 5.6 to 19 s.
    • Burst stutter counts.
    • Milder stutter takes longer to trigger a step than severe stutter. A single long hitch (a pause or a load) does not trigger one at all.
    • The first second of every opening, and of every return to the page, is not measured.
  • No step is ever given back while you watch: once the replay has lowered its quality, it stays lowered until you open a replay again. The home screen's planet scene uses the same rule.
  • The tuning has not been verified on an iPad. The numbers were set from synthetic frame traces, not from the iPad whose report started this (#6191 stays open for that device's trace).
  • Loading that runs past the first second can still cost a step. That follow-up is #6264.
  • The replay diagnostics (?replaydiag=1) now report the worst second as frames over budget and the share of that second's time they took.

1.378.1 — 2026-10-04

Fixed

• Two of the game's own checks could pass without checking anything (#6238). Nothing in the game itself changed. This is about the tests that guard it.
  • The checks on the HUD band's commander line (the GUEST pill, the guild tag, the Crucible chip) took their reading when the page finished loading. They assumed the player's profile had arrived by then, and under load it sometimes had not. They now wait until the pill, the name and the chip are actually on screen. A new check delivers the profile late on purpose and fails if the reading is taken without it. The measured layout figures are identical to before.
  • The deploy self-test that proves the server lock's time limit ends a live hold could pass on a run where the hold never started. It now passes only if the hold read at least one heartbeat before the limit ended it. How production takes and releases the lock is unchanged.
  • Not included: the back-button layout check (backControlPage) still takes its reading on page load. When the profile is late it fails loudly, with an "unsettled" marker, so it cannot pass by mistake.

1.378.0 — 2026-10-04

New

• Settings has a Replay diagnostics switch, so the replay's frame diagnostics can be turned on without a URL (#6249). Until now the battle replay's frame diagnostics panel (#6191) came on only when novarchy.com was opened with ?replaydiag=1. The iOS and Android apps have no address bar to type that into, and the app is exactly where the stutter it measures was reported. Settings now has a Replay diagnostics row with an On/Off switch and one line saying what it does.
  • Computer or tablet: the Settings menu, last row of This device.
  • Phone: the Command sheet, under the attack-notifications row.

  It sets the same switch the address does, on this device only, never on the account. It takes effect the next time a replay opens. The row always shows the current state: after the panel's own Turn off, the row reads Off the next time Settings opens. If the browser blocks site storage, the switch stays Off and says the setting could not be saved. The row is in English and Italian.

  Not in this release: the installed iOS and Android apps get the row only with a new app build, and shipping that is the owner's call. Until then the row is on novarchy.com in a browser. The player who reported #6191 has not yet been told where the switch is. That message is the owner's to send, and #6191 stays open until a trace from a real device exists.

1.377.1 — 2026-10-04

Changed

• Recall is free: call a fleet back as often as you like (#6247). Calling off an attack used to be capped at three per rolling 24 hours: the fourth Recall of the day was refused, the button greyed out with «No recalls left today (3/3)», the Missions tab counted down «N of 3 recalls left today», and turning an attack away with Change course spent from the same three. That limit is gone. An attack can be recalled, turned away or (for a raid leader) called off as many times as you want, right up to its point of no return in the last stretch of the flight. The countdown lines and the greyed button are gone with it. The target that saw the attack is still told it turned away.

Fixed

• AI empires remember a raze even after a string of feints (#6247). Fleets launched at an AI empire and recalled before arriving count as feints, and they could fill its grudge to the brim first, so a real raze that followed was shrugged off: once its first strike back was spent, the empire held nothing against the player who had razed its colony, and its alliance and guild council ignored it too. Real harm now always outweighs feints, whichever came first.
• Installed apps stop showing the old recall allowance (#6247). An app left open across this update would have kept its last «N of 3 recalls left today» line on the Missions tab until it was restarted. It now clears on the next refresh.

1.377.0 — 2026-10-04

New

• The battle replay can show what its frames cost, on a device that asks for it (#6191). Before, a replay that "kind of stutters" left nothing to go on: no frame timings, and no way to tell whether the game had already lowered its picture quality to keep up. Now, opening novarchy.com with ?replaydiag=1 turns on a small panel at the bottom-left of every replay, on that device only. The panel shows:
  • how far the replay has stepped its quality down (bloom, then resolution);
  • the typical, slow and slowest frame times over the last 240 frames;
  • how many of those frames took longer than 25 ms.

  Copy diagnostics puts the same figures on the clipboard, with the device, the history of each quality step and a frame-time histogram, ready to paste into a forum post. If the clipboard refuses, the text appears on screen to copy by hand.

  The panel is off unless asked for, and nothing is sent anywhere. Hide folds it into one chip, so it does not cover the finale. Turn off (two taps) or ?replaydiag=0 switches it off again. The labels are in English and Italian. The native apps have no address bar, so on a phone or tablet open novarchy.com in the browser instead: it is the same replay on the same hardware.

  The stutter itself is not fixed by this release. #6191 stays open until a trace from a real device exists. Nobody has asked the player who reported it to run the tool yet: that request is the owner's to make.

  Reported by a player on the bugs forum, who mentioned the stutter after the replay started opening again on their iPad (#6188).

1.376.1 — 2026-10-04

Fixed

• Three test checks that sometimes failed on a busy build server now wait for the page they measure (#6220). Nothing changes in the game. Two headless-browser layout checks, one on the launch pads and one on the floating action bars, sometimes read a screen before it had been drawn, about 3 times in 200 under heavy load. That made them report "no controls at all" or a button that "found nothing". They now wait until the page has finished its own work. If a page is still busy when time runs out, they say so directly instead of measuring an empty screen.
• The deploy script's lock self-test no longer fails when the server is busy (#6220). When the server-lock helper got a lock and lost it at the same moment, it reported that it "could not take" the lock. It now records the lock as taken, and refuses the next step because the holder is gone. A deploy that skips work, and the read-only check run, now also confirm they still hold the lock before giving their verdict. Before, they could give it without checking.

Two related weaknesses in the test machinery are tracked in #6238 and are not fixed here.

1.376.0 — 2026-10-04

New

• Change course in flight (#6235). A fleet on its way out used to have one order, Recall: its destination was fixed at launch. Attacks, spies, transports, redeployments and garrisons now carry a ⤳ Change course button beside Recall. Type the new coordinates and, before you tap, the sheet tells you the new arrival time — or exactly why that destination is refused.
  • The fleet turns where it is and flies straight to the new target at its own speed: no gate, no warp, and never less than two minutes, so a detour always costs time and the new target is always warned.
  • You can change course, as many times as you like, until the same point of no return as Recall.
  • Attacks: the colony you were flying at is told the attack turned away — but only if it had seen it coming; the new target is warned as at a launch, or the leg flies dark under Nightfall. Turning away from a target that saw you uses one of the day's three recalls, as a recall would. Turning a strike onto a player drops your own cordons and night shelter, as launching it would — the sheet says so before the tap.
  • The new destination is checked like a launch: pacts, cordons, night protection at the new arrival time, garrison slots and the rest all apply, and a fleet cannot be turned back onto the colony it left (that is what Recall is for).
  • After a change of course, Recall flies the fleet home from where it is, and the card quotes that time. Instant recall is not offered on a redirected fleet.
  • Your diary records each Course changed, with the old and the new destination.
  • Pickets and pursuits follow the new course. A leg that starts in open space has no safe zone behind it, so a fleet cannot dodge interception by chaining short turns.

Raid contingents, warped returns and muster garrisons cannot change course yet; waypoints and AI empires using the order are planned next.

1.375.3 — 2026-10-04

Changed

• The mission composer's tests wait for its reads instead of a fixed time (#6222). Nothing in the game changed. This is for whoever runs the test suite.

  Before: twelve web tests of the mission composer gave its 250 ms target-read and spy-lookup delays a fixed wait (a 320–400 ms sleep, or a fixed number of short polls), then checked the screen as if the read had arrived. On a loaded machine the timer fired late, the wait ran out first, and the test failed although the game was fine.

  Now: each of those waits ends when the read has actually gone out, or when the screen shows what the test needs, and gives up after 5 seconds with a message naming what never happened. On the merged tree all 260 tests in the twelve files pass. With the delay stretched to 1,500 ms in a throwaway copy, 259 pass. The one that fails is the test that pins the delay at exactly 250 ms, so it is meant to fail there.

1.375.2 — 2026-10-04

Fixed

• The mission composer and the raid door now price a strike from the same Power reading (#6211). The composer used to read your Power on its own, on its own poll, while the raid door, the planet sheet and the Wanted board shared the Dashboard's reading. The two readings could be taken up to a poll apart, so the composer and the raid door could quote two different bounty prices for the same strike. Now the composer reads the Dashboard's reading too, so every screen that prices a strike quotes the same number. Opening Missions still refreshes that reading, as before, so a reading that failed at load recovers when you open the tab, and the raid door and the planet sheet update with it. An attack is still held only while no reading has ever arrived. A failed refresh still keeps the last price.

1.375.1 — 2026-10-04

Changed

• The timed attack protection is now called the Cordon, and "shield" means only the building (#5867). Two unrelated mechanics used to share one word. The Planetary Shield is a building that soaks damage in battle. The "Empire shield" / "Colony shield" was a countdown that stops attacks from being launched at all. A desk answer about "an active shield" was read as being about the building, so the countdown has been renamed everywhere the game names it: the HUD chip and its sheet (Empire cordon · Colony cordon), the Orbital Home stamp (EMPIRE CORDON 3H), the desk timeline (Cordon ends), the Guild Store (Cordon 8h / 24h / 48h, was Peace shield), the target board and War Room chips, the banners, the launch and raid warnings, the spy card, and the refusal a launch gets ("Target is under an active cordon."). Italian keeps the word untranslated: «il Cordon», «sotto Cordon». A new glossary entry, Cordon, says what it is, the three ways you get one, what ends it, and that it is not the Planetary Shield. The building, its shield stat, the Shielded Silo and the rest of the combat shield family keep their names. Old guild-log rows for Store purchases now read "Cordon (8h)" too. The agents say "Cordon" when they talk about it.

  Suggested by a player on Discord, after a desk answer about "an active shield" made them think their Planetary Shield blocks dock raids. Their name is not recorded on the issue. The credit is suspended rather than withheld.

  Not included: the glossary says a Marshal strike also leaves your Cordon standing, but the HUD banner, the rule lines and the order confirmation still name only the Pirate Beacon as the exception. That is tracked in #6225. The three leave-of-absence sentences are still in English for Italian players, as they were before. Installed iOS and Android apps keep the old "shield" wording on their own screens until they update. Only the Guild Store names, which come from the server, change in them right away. Purchases still work, because they are sent by id.

  Owner follow-ups, not done by this release: a reply on Discord to the player who raised it, and a player announcement of the new name.

1.375.0 — 2026-10-04

New

• Field agents is live in the guild tree (#6203). The Intelligence node Field agents used to show as "in development", and nobody could buy it. A guild can now research it (8,000 Fund, 48h, after Probe cache and Science 3). Once your guild owns it, your spy reports on a colony whose guild is at war with yours are filed one tier deeper, up to the top tier (3). "At war" means a declared war in its counting week or an open League Clash, and it is checked when the probe lands. At peace, or with no node, nothing changes. A report raised to tier 3 reads like any tier-3 report: Ghost Manifest's hull fog lifts on it, and it counts toward Omniscience. Sabotage is the exception (see below).
• Counter network can now be reached. The capstone still needs Deep cover 3 and Field agents. Until now Field agents could not be bought, so the capstone was out of reach for every guild.

Changed

• Sabotage copy names the real requirement. Sabotage still fires only at your own Espionage 17, however deep the report goes. The launch screen, the spy report and the Cipher doctrine line used to say "needs a top-tier report". Now they say "needs your Espionage at 17", so a report raised by Field agents does not look like it should have fired the sabotage.

Issue #6203 stays open. This release covers Field agents only, and four other guild nodes are still marked "in development".

1.374.8 — 2026-10-04

Changed

• The research bench now shows the queue's order (#6161). Before, every queued card had the same «Queue slot» header, and the only way to tell which order would start next was to read and compare a sentence on each card. Now:
  • each queue slot is numbered, Queue slot 1, 2, 3, free slots included, as the lanes already were;
  • the order the research queue will start first, whatever finishes first, wears a Next pill in its card's colour. A blocked or walled order keeps the pill, so a red card with a pill is the one holding the queue.
  • When the answer depends on which of two lanes frees first, no card gets the pill.
  • When every queued order is waiting on its own earlier level, one lane marks the order that comes back first, and a single such order is marked on any number of lanes.
  • The waiting sentence under that order no longer hedges where the outcome is already decided:
    • with one lane and a stacked level ahead, it reads «starts after the order ahead of it»;
    • behind a stranded order that cannot come back first, it reads «starts when a lane frees up».
  • With two or more lanes, the hedged sentence stays.

  Not included: on two lanes, a stack such as a running level 1 with levels 2 and 3 queued gets no pill, though level 2 is certain to start first. The pill is never wrong there, only missing.

  Reported by a player on the gameplay forum, who numbered the cards 1-2-3 by hand on a screenshot.

1.374.7 — 2026-10-04

Changed

• One ship gate no longer floods the shared box with test workers (#6134). Nothing in the game changed. This is for whoever runs a ship gate.

  Before: the gate's test leg ran the shared package's tests with 63 workers, then the api and web tests together with 126, two per core on the 64-core box, before counting pytest or any other gate. CPU-heavy web tests then ran 3-4× slower than they had to and hit their 60 s limits on loaded days.

  Now: each test run is capped at a third of the cores. On the ship box that is 21 workers, and 42 while api and web run together. The gate prints the cap beside the load line. GATE_VITEST_FORKS=<n> overrides it on the machine running the gate.

  Measured on one frozen tree (afd44b293). The full table, with the decision rule fixed before measuring and its limits, is in .claude/skills/ship/RATIONALE.md, section R20:

  • With two legs running at once, the heavy web tests ran 2.0-3.2× faster.
  • A leg running alone was not slower: median 483 s capped, 525 s uncapped.
  • Under contention a whole leg took about as long either way (596 s capped against 578 s, on loads that were not matched). The gain is per-test latency, which is what the 60 s limits measure.

  The cap applies only to gates run from a tree that has pulled this commit. A tree on an older gate.sh still starts 63 and then 126 workers, and slows every gate on the box while it runs.

  Not included: the headless-browser layout tests and the deploy self-test that fail under very high load in both modes. They are tracked in #6220.

1.374.6 — 2026-10-04

Also

• The season close keeps the notches a fortified anchorage shows (#6193). When a guild season closes, a guild that researched Fortified anchorages loses the sixth notch its structures could take. To keep the close from ever razing an anchorage, it pulls any bar above four notches back to four. That rule was applied to the figure stored in the database, which still counts notches that have already faded, instead of to the notches the bar shows. No guild could actually hit the difference: Fortified anchorages requires Hold doctrine at level 3, and its 12-hour fade made the two figures agree at every close. If they had diverged, a bar showing four notches would have come out of the close showing three, erasing progress the attackers had earned. The close now caps the notches the bar shows. Players see no change.

1.374.5 — 2026-10-04

Fixed

• An attack waits for its bounty check instead of launching unpriced (#6187). The mission composer prices a strike's bounty, and asks before one that would jump your own bounty, only when it has read both your own power and the target. Before, if either reading was missing, it showed no price and asked nothing. That happened while your power had not loaded yet, after any failed refresh of it, or while the target had not been read or its read had failed. An attack that could take you from no bounty to Warlord then launched as if it were free.
  • Now the attack waits. Send stays off, and the line beside it says which reading is missing: "Bounty not checked: your own power has not been read yet…" or "…the target has not been read from the Galaxy view yet…". When the reading arrives, Send comes back on by itself. The target line also tells you to aim at the target again to retry at once.
  • A refresh of your power that fails after a good one no longer hides the price: the composer keeps the last figure, so the price line and the confirm stay.
  • On a phone, the quick attack drops back to the form with that line instead of opening its confirm. Nothing fires by itself when the check clears: you press Send.
  • Unchanged: spy and every other order, attacks aimed at your own colonies or launch pad, and addresses no read can reach (outside the galaxy, or the Event Sector between seasons, which already explain their own refusal). Guild raids are not covered by this hold.

1.374.4 — 2026-10-04

Changed

• Keep and Hide now sit inside each Activity entry instead of a band of buttons under it (#6200). Before, every single entry in the Activity feed had a row of two full buttons, Keep and Hide, below its card. That row cost about 48px per entry, so a phone showed roughly three and a half entries per screen. Now the controls are three icons inside the card itself. Each icon still has a full 44×44 touch area.

  | Icon | Action | |---|---| | ☆ | Keep. It turns into a cyan ★ once kept. | | ✕ | Hide | | ↺ | Unhide |

  Where the icons sit:

  • Event, battle report and galaxy-log entries: in the title line, after the time. The card is no taller than before.
  • Spy reports: at the end of the Attack / Spy again row, so a tap on the header still opens the report. On a 320px-wide screen the icons get a line of their own.
  • Live raid warnings: still have no Hide.
  • Folded runs: still carry neither icon.

  Hidden entries you have revealed are now greyed by a veil over the card. Before, the whole row was faded, icons included. Now ↺ stays at full brightness above the veil.

  The words Keep, Kept, Hide and Unhide are now tooltips. A screen reader announces Keep the same way whether it is pressed or not, together with its pressed state. Unhide has its own spoken label.

  Suggested by a player on the suggestions forum.

1.374.3 — 2026-10-04

Fixed

• The anchorage demolish refusal names your guild's own fade time (#6205). Sometimes a leader confirms a demolition while the server already sees a notch on the breach bar. The refusal used to say to "leave it a day per notch to decay", which was false for any guild holding Hold Doctrine, where a notch fades in 20, 16 or 12 hours. The Anchorages panel now prints the guild's real figure, for example "leave it 12 hours per notch to decay", in both the confirm sheet and under the card, in English and in Italian. Wherever this refusal appears outside that panel, it no longer names a figure: "repair it, or leave it to decay".

1.374.2 — 2026-10-04

Also

• The Advisor drops two escort warnings that no fleet could ever trigger (#6148). When you send a ship to explore a signal and there's nothing to escort it, the Advisor explains what happens if pirates are waiting. Its code also held a second pair of these warnings, written for an explorer that carries guns. No real fleet could reach them: if your explorer is armed, the Advisor always sizes an escort against the pirates and tells you how that fight goes. Those two sentences and their translations are gone, and a new test checks every armed hull to keep it that way. Players see no change. An explorer with no weapons still gets the same warning: it breaks off and comes home empty.

1.374.1 — 2026-10-04

Fixed

• Another guild's anchorage card now shows how fast its owner's notches fade (#6150). On the Galaxy map, the rule under a rival anchorage's breach bar always said a notch fades "for each day that passes since the last blow". A guild that has researched the Hold doctrine clears a notch every 20, 16 or 12 hours, so for those guilds the card understated how quickly the bar empties. The card now reads "the bar loses a notch for every N hours that pass since the last blow", with N taken from the guild that owns the structure. That is 24 for a guild without the research. The same fix applies in Italian.
• Reparations rows in the Activity feed show the truce that was actually bought (#6150). The row always said "48-hour truce". When the paying guild has Truce envoys, the truce is really 72 or 96 hours. Rows written from now on print the true length. Older rows still read 48 hours, which is what every one of them bought.

Not included: the refusal shown when a leader tries to take down an anchorage that still has a notch on its bar still says a notch fades "a day per notch". That one is tracked in #6205. In practice the button is already disabled in that state.

1.374.0 — 2026-10-04

New

• A Glossary of the game's own words (#5760). Until now keystone, Veil, Gate, band, Echelon, Crucible, Watch, contingent, dock raid and the rest only appeared inside sentences about something else, and the only place to ask what one meant was chat. There is now a Glossary in the desktop Empire menu, next to Discord and Changelog, and on a phone in the Command sheet, just above Changelog.

  The Glossary has 31 entries, in English and Italian, grouped as Empire & Power, Combat, Guild, Events, Galaxy and Economy. Each entry has:

  • a one-line definition and a few sentences on how it works;
  • numbered senses where the word means more than one thing. A keystone means three different things, and the Watch, an Anchorage, a Relic, a Shard and a Lane each mean two;
  • a "Not:" line for what it is easily confused with. For example, the Gate is not the Hypergate, and the Veil is not the "Guild Veil" ring;
  • where it makes sense, a button that opens the screen it describes.

  You can search in your own language or in English: an Italian player who types "bounty" finds «Taglia». Plurals find their singular. If a category chip is what hides a result, the sheet says so.

  Every figure in the glossary is read from the game's own numbers, so it stays correct when they are retuned.

  Not included: an FAQ built from the questions actually asked in chat, the second part of the request. It needs the help desk's questions as its source, and stays open.

  Suggested by a player in the Discord #suggestions channel, who asked for a FAQ "indexed by broad category and searchable by keyword" and for "a dictionary look up" for words like keystone and Veil.

1.373.6 — 2026-10-04

Fixed

• An attack relaunched from the Launch-again cards now asks about the bounty too (#6183). Since v1.372.0, the mission composer asks first when a launch would raise your bounty by two grades or more, or make you a Warlord. The Launch-again cards did not ask: they relaunched such an attack with the same second tap as any other.

  Now, when you tap an attack card, it checks its target once and prices the strike the way the composer does:

  • Big climb: the card's confirm shows the same bounty block before it can leave. The confirm button reads «Launch and take the bounty».
  • Smaller strike: a strike that climbs less still shows its bounty price, and the card warns you if the target would leave the fleet coming back empty.
  • Target can't be checked: the card never launches blind. It offers Review instead.
  • Your own power not read yet: the card waits, and checks again by itself as soon as the reading arrives.

  Not included: the mission composer itself can still launch without the bounty check when your own power or the target could not be read. That is a separate fix, still open (#6187).

1.373.5 — 2026-10-04

Fixed

• Google's sign-in button no longer sits in a white box (#6175). In Chrome, on the login page (both the returning-player form and the first-visit landing), Google's dark "Continue with Google" pill was drawn inside a white rectangle the full width of its frame, so it looked like a rendering fault. The pill now sits directly on the dark page, and the same fix covers the account settings where you link a Google account, which use the same button. Google's own button, its theme and its wording are unchanged.

  Checked in a headless Chrome against a local reproduction of the frame. Not yet checked in Safari or on a real phone. This is a website fix: the iOS app draws its own Google button and was never affected, and installed apps keep the screens they were built with.

1.373.4 — 2026-10-04

Fixed

• A battle at your anchorage during the season close can no longer leave it one hit from falling (#6162). If your guild owned Hold Doctrine, the close re-bases each anchorage's breach bar so that the notches already faded stay faded. When an attack on that anchorage landed in the same instant, the close could write its figure over the battle's. The bar then read as full, the modules went dark, and the next attack razed the structure. Now the close notices the battle, re-reads the bar and keeps the battle's notch. Outside that collision the close does exactly what it did before. Not included: the opposite timing (a battle that starts before the close and lands after it) can still apply a notch under the old season's limits. It is tracked on #6162, which stays open. The Fortified Anchorages clamp at the close is a separate question, #6193.

1.373.3 — 2026-10-04

Fixed

A battle replay that cannot start now says why, instead of blaming the device every time (#6188)

Before, a replay whose 3D view failed to open showed «Could not start the replay on this device.» over a black canvas, whatever the cause. That line gave the player nothing to do, and kept nothing a screenshot could pass on. Now the replay uses the same two sentences the base and the galaxy map already use:

• When the device cannot draw it, the replay says so. That covers no 3D at all, and a battle view that loaded but failed while being built. It also says that another 3D view already open can use up what it needs.
• When the replay's code did not arrive (typically after an update), it says that restarting the app or reloading the page fetches it again.

Both sentences end with the actual error in brackets, so the next report names the cause. From the Campaign tab, the message adds that the battle is already decided and its result is on the Campaign tab. From anywhere else, it says that closing the window takes you back to where you opened it.

Not fixed yet: the replay on the reporter's iPad. The most likely cause is that the two heaviest ship models (the Interdictor and the Annihilator) are too big for the device. That is not confirmed, and lighter versions of those models are tracked in #6196. Installed iOS and Android apps get this message only with their next app build. The website and home-screen installs have it now.

Reported by a player on the bugs forum, with the screenshot that showed the report had loaded and only the 3D view had failed.

1.373.2 — 2026-10-04

Changed

• Back now returns you to the Base as you left it (#5898). Before, Back onto the Base opened it fresh: no building card, and the Build menu on All, unsorted.

  Now, Back (the browser's, or the on-screen ←) onto the Base:

  • on a computer, reopens the building card you had open in the side column, with its ring on the map, or the catalogue card of a building you had not built, along with the Build menu's category filter and level sort;
  • on a phone, opens on the map with nothing open, as before. Back never opens a sheet. When you open the Build menu, it is on the filter and sort you left.

  What has not changed

  • Opening the Base from another screen, from a link that names a building, or after a reload opens it as before. A card read on another colony does not reopen.
  • After a screen error, Retry opens the screen fresh, on every screen that remembers where you were, so it never reopens the place that failed.
  • The map's camera position is not restored.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.373.1 — 2026-10-04

Changed

• Back now returns you to the Season lane you left (#5898). Before, Back onto the Season tab always opened the Race, even if you had been on the Proving Ground, Contracts or the Calendar.

  Now, Back (the browser's, or the on-screen ←) reopens the lane you were on. A notification that opens a specific lane still wins, and opening the Season tab from another screen or after a reload still starts on the Race. Tapping Season on the tab bar right after leaving it counts as Back.

  What is not in this release

  • On the Calendar, the month and the day you had selected are not restored yet. This is listed on #5898.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.373.0 — 2026-10-04

New

• A wanted commander's other colonies, one tap from the Wanted board (#5920). Each Wanted row names one address per head, and it is usually their homeworld. The other colonies your own Sensors could already see were reachable only by scrolling the whole Targets list. Now each row with an address (not your own, and not one in the event sector) carries Their colonies in sensor range →. It opens the Galaxy map with the Targets list narrowed to that commander, with a lit chip carrying their name, a clear button, and a line saying the list is the core galaxy only. The filters and counts below it then count their colonies only. If none of them is in your sensor range, the board says so and points you to clear.

  It shows nothing new: the list is the same one Targets already gives you, cut to what your Sensors see. Espionage does not widen it (Espionage sets how deep a spy report reads), and the Wanted row's own tap still opens the same single address as before. The narrowed view lasts until you clear it, log out or reload the page, and is never saved on your device.

  Suggested by a player on Discord, who asked why, with high Espionage, the board still sent him to the worst planet to attack.

1.372.2 — 2026-10-04

Changed

• Back now returns you to Research as you left it (#5898). Before, Back opened Research fresh: the Disciplines lens, the first Discipline, no filter, the default sort. A player comparing Disciplines who stepped out to the Base or the Shipyard came back to the top of the catalogue.

  Now, Back (the browser's, or the on-screen ←) onto Research reopens it on the lens you were using (Disciplines or the Tech tree), the Discipline you had selected, and the catalogue's filter and sort. On a phone the Discipline comes back highlighted with its sheet closed: tap it to open it again.

  The Laboratory round trip. A Tech tree card's Laboratory row takes you to the Base. Back from there reopens that card, as it already did, and now your Disciplines filter, sort and selection are still there when you switch back to them.

  What has not changed

  • Opening Research from another screen, from a link that names a Discipline or a Tech tree node, or after a reload opens it as before. The one exception: tapping Research on the tab bar right after leaving it counts as Back, and restores it the same way.
  • The Tech tree's own scroll and the Discipline grid's scroll are not restored.

  What is not in this release

  • The Base's focused building and the Season lane still open fresh on Back. These are listed on #5898.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.372.1 — 2026-10-04

Fixed

• A research order no stock can pay for now says how to get past it (#6167). Since the last release, the research bench says when a queued order is priced above what the player can ever store: the gap "will not close by waiting". But the queue still treated that order as money the mines would bring.

  When more levels of the same research were stacked behind it, the blocked card had no remove button, because only the top of the stack has one, and nothing said which button to press. Locked-out orders have had that sentence for a long time. And an order waiting behind such an unpayable order read "Not paid yet", as if the queue would move on by itself.

  Now a blocked order above the storage ceiling names the level at the top of its stack that can be removed, exactly as a Laboratory-locked order does. The order behind an unpayable one reads "Queue blocked", which is the truth. An ordinary blocked order, one the mines will pay, is unchanged. This covers Disciplines and Tree nodes alike.

1.372.0 — 2026-10-04

New

• A launch that would jump your bounty two grades, or make you a Warlord, now asks first (#5828). Until now, the bounty a strike would cost was one more gold line among the mission's warnings, and the fleet left without a word. Now, when one launch would raise your own bounty by two grades or more, or make you a Warlord, the composer stops at its own dialog and says so in the game's words. Example: «From no bounty to ☠ WARLORD (grade 4/4) in one strike?». The dialog also shows:
  • the Dark Matter the strike puts on your head (double in a starting sector);
  • your bounty before and after;
  • a reminder that the figure uses today's Gates and is settled on impact, and that strikes already in flight are not counted.

  Details:

  • If the same launch would also leave its pad without its own defences, that warning sits in the same dialog, so you are asked once.
  • The quick attack's confirm carries the same block and still names its target.
  • The dialog is fully translated into Italian.

  No dialog for:

  • a strike that climbs one grade without making you a Warlord;
  • a player who is already a Warlord;
  • any strike the mission tab shows no bounty price for.

  Not included:

  • attacks relaunched from the Launch-again cards, which never carry a bounty price (open them in Review to see it);
  • raids, whose price depends on the whole pack at the battle;
  • a faster way to pay a bounty down, which the same report asked for and which is still an open design question.

  Suggested by a player on Discord after two fleets launched at the wrong target left him with a 300 Dark Matter bounty.

1.371.4 — 2026-10-04

Changed

The screenshots on the first-time visitor page are new (#6172). The three pictures of the game under the Play button had been taken from one real account, and showed its commander name, its guild and where its home is. They now show a stand-in commander and guild that no player uses, and a galaxy view of the central sectors with nobody's home marked on it.

If this device already holds an empire, nothing changes for you.

1.371.3 — 2026-10-04

Changed

• Back now returns you to the Activity feed as you had filtered it (#5898). Before, Back opened the Activity feed fresh, on its opening scope with no type and no combat topic. Narrowing the feed to Galaxy → Intel, opening a target, and coming back meant tapping the chips again.

  Now, Back (the browser's, or the on-screen ←) onto the Activity feed reopens it on the chips you left it on: the scope you picked, the type, and the combat topic. A remembered type loads its own older rows, as tapping it does.

  What has not changed

  • Opening the feed from the tab bar, from a battle report link or after a reload opens it as before.
  • A report link always opens the feed unfiltered, so the report you tapped is on screen.
  • Expanded groups (trades, shards, raids) and the scroll position are not restored.

  What is not in this release

  • Research, the Base's focused building and the Season lane still open fresh on Back. These are listed on #5898.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.371.2 — 2026-10-04

Fixed

• The research bench says when a gap is bigger than what you can store (#6157). The Tree sheet already warned when a research price was above what the player's storage can ever hold: the gap "will not close by waiting". The research bench showed the same gap, on a blocked queue and on the «Still needed» row, with no such warning, and still promised the order "starts when a lane frees up".

  Now the bench prints the Tree sheet's own note, word for word, under the gap. It names only the resource that is actually out of reach, and it covers Disciplines as well as Tree nodes. On a queued order the note applies to, the status line keeps only «Not paid yet» until the storage catches up, instead of promising a start. Dark Matter has no storage cap, so it never gets the note.

  Still open: a blocked order above the storage ceiling still offers no way to remove it from the card (#6167).

1.371.1 — 2026-10-04

Fixed

• The Proving Ground's boards no longer go blank while the season closes (#6149). From the moment a season ended until its close finished, every bracket came back empty and your best score disappeared. It looked as if the results had been wiped, just when the close was paying on them. The boards now stay as they stood at the season's end until the close is done: the same rows, your best score, and "Pays at the close." The Strike button stays shut ("No season is open."). In that window the panel also no longer tells a zero or out-of-date score to "Send warships" or "Strike again", since the button beside it cannot be pressed. Who gets paid is unchanged: before release, the same 304 test seasons were run through the old and the new close, and the two paid out exactly the same crates.

1.371.0 — 2026-10-04

New

A first-time visitor now lands on the game itself (#6129). Someone opening Novarchy for the first time used to meet a commander-name field on a dark page, and on a phone there was no picture at all. They now see the fleet in motion with the title standing in it, one "Play free" button, and a commander name already chosen for them, with a link to pick their own. Three screenshots of the real game follow, and the Apple and Google sign-in buttons come after those.

If this device already holds an empire, your login screen is exactly as it was.

Also

• A visitor who asked their device for less motion, whose browser is saving data, or whose phone refuses to start videos on its own gets a still picture in place of the clip.
• A sign-in that fails on the new landing says so next to the button that was tapped.
• The new copy is in English and Italian.

1.370.1 — 2026-10-04

Changed

• Back now returns you to the system you were reading in the Galaxy (#5898). Before, Back took you to the Galaxy but opened it fresh, on the political map at sector 1, so a trip from a system list to Missions and back lost the system you were scanning.

  Now, Back (the browser's, or the on-screen ←) onto a Galaxy you left in a system view reopens that system, in the same mode (space view or list) and on the colony you had selected. No planet sheet reopens: Back closes sheets, it does not bring them back.

  What has not changed

  • Opening the Galaxy from the tab bar, from a coordinate link, from the Event Sector button or after a reload opens it as before.
  • A Galaxy left on the political map comes back as the fresh political map.
  • If the season has closed since, a remembered Event Sector system opens on sector 1.

  What is not in this release

  • The other screens still open fresh on Back. These are Research, the Base's focused building, the Season lane and the Activity filters, and they are listed on #5898.
  • The Galaxy's camera and scroll position are not restored.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.370.0 — 2026-10-03

New

• Guild Tech Tree (#5650). Guild research used to be three tracks (War, Economy, Science) with a cap of 14 levels a season. Most guilds hit the cap within a few weeks and then had nothing to spend their Fund on. It is now a tree of seven branches (War, Economy, Science, Logistics, Territory, Intelligence, Diplomacy), five tiers and 39 nodes, with no cap: given time, a guild can own all of it.
  • Tiers 1–3 are permanent and paid from the Fund. Tiers 4–5 are seasonal and paid from a new pool, the Endowment. Both the seasonal tiers and the Endowment reset when the guild season ends.
  • Two yards: one permanent and one seasonal project can run at the same time.
  • Pledges (Treasury tab) fill the Endowment. A resource pledge costs 2.5% of your home storage, and a Dark Matter pledge costs 5 DM. Either one gives the guild 10 Endowment. You can make 3 resource pledges a day and 3 Dark Matter pledges a week. The row shows when the Endowment resets, and it greys out a pledge you can't afford yet.
  • New effects apply in play: aid capacity, requests and depth; supply-lane catch-up; anchorage cost, slots, modules, repair, notch decay and fall line; counter-intelligence and deep cover; tension decay, the reparations truce and surrender tribute; directive dividends; and an extra probe cache. Every guild number on your screens reflects what your guild owns.
  • Five nodes are marked "in development" and can't be bought yet: war footing, lane throughput, field agents, council channel and allied raids.
  • Your guild's tag line now counts permanent levels only, so it no longer changes at every season close.

Changed

• Every guild's old research has been reset into the new tree, and every guild gets a full refund. The first time your guild's research opens (or at the next season close, if nobody opens it before then), its War/Economy/Science levels become an empty tree. Everything the guild paid for that research this season goes back into the Fund, so it can rebuy the same day. For example, War 10 + Economy 4 gets 47,709 back. The guild's pin is kept, and one line in the guild log records the refund.
• AI guilds research by node and make pledges at about the pace a person would, roughly one a day each. They stop pledging once the Endowment covers what can still finish before the season ends.

Also

• On an older installed app, guild research still shows the three old tracks, and some of their figures are wrong. Updating the app shows the tree.
• Not included yet (#6150): another guild's anchorage card on the Galaxy map, and the guild-feed reparations line, still print the base notch decay and the base truce length rather than the values the seasonal nodes give.

1.369.0 — 2026-10-03

New

• Every queued research order says what it still needs (#6135). Until now, the research bench sized a shortfall on one queued order only: the one the queue tries next, and only while a research lane stood free. With every lane busy, an underfunded order looked like any other queued block. An order further back never showed a figure, however far its price was beyond the treasury.

  Now every queued order that today's treasury cannot pay for carries a Still needed row (Italian: «Mancano»). It has red figures such as −12K Fe −3K Xe, and Dark Matter as −880 DM. This is the same row the Tree sheet, the Discipline panel and the Building sheet already show. The card itself stays gold: a red card still means only that the queue is stuck. Each order is priced at its own level, the price the server will charge when it starts.

  The row is left off where money would not change anything:

  • the blocked order, whose sentence already names the gaps;
  • an order waiting on a Laboratory level the network does not have;
  • an order whose prerequisite was removed from the queue.

  Still open: when a gap is larger than what the player can store, the Tree sheet says so and the bench does not yet (#6157).

  Suggested by a player on the gameplay forum, who asked for a red marker on queued research that is short of funding.

1.368.3 — 2026-10-03

Changed

• Everything the Fleet Advisor says now reads in Italian, for every mission type (#6069). Since v1.365.9 its attack advice was Italian. What it said about every other mission stayed English: spy, explore (including the pirate-ambush odds and escorts for an anomaly), colonize, transport and pickup, harvest, defend and redeploy. All of it now reads in Italian. The caveat on the Galaxy tab's one-tap explore now follows the player's language too. The ambush odds are worded in Italian («circa una volta su sette»), figures are grouped the Italian way, and ship names, the Spy Probe and Reclaim keep their names. Nothing changes in English.

Fixed

• The attack advice about founding hulls left home now gives them the right gender in Italian. The line told the player to add the Colony Ship or Colonial Ark by hand with a masculine pronoun («Aggiungilo», «Aggiungili»). Italian treats these hulls as feminine, a «nave», so it now says «Aggiungila», «Aggiungile».

1.368.2 — 2026-10-03

Fixed

• On a tablet, a Marshal window opening no longer throws the resource bar under your commander card (#6121). The chips at the end of the resource bar (build, ships and lab queues, Crucible attempts, the Marshal bar and the protection clock) were one block that could not wrap. When a Marshal window opened, that block no longer fit beside your commander card. Every resource tile then jumped from the card's right to the line under it, and jumped back when the window closed. On the test bench this happened at almost every width from 721 to 1180px, and at 1366px with the wider controls. Now:
  • the chips wrap between themselves, inside the resource bar, and never split the build, ships and lab counters;
  • your commander card stays at the top of the band, so the name and Power line do not move either;
  • when a window opens, the resource tiles stay where they are, and the new chips take a line under them.
  • The Marshal bar never grows wider than two resource tiles. On a screen whose font is wider than ours, the phase name is shortened with "…". The level, the health bar and the clock stay whole, and the full name stays in the tooltip.
  • The Marshal clock and the protection clock now keep one width while they count down. Before, "2h 59m" turning into "59m" could move a chip to another line at the hour mark. The space is kept before the time, so in a window's last hour, or as a shelter runs down, there is a small gap between the bar and the clock.
  • What it costs: while a window is open the top band can be up to two chip lines (about 80px) taller, so the map sits a little lower until the window closes. On desktops whose resource bar takes two lines, the commander card now sits at the top of the band while the bell and menu stay centred.
  • Phones are unchanged.

1.368.1 — 2026-10-03

Fixed

• The Proving Ground now says when the close is (#6140). The panel's prize text kept saying what "the close" pays and never said when that was. The countdown was only on The Race, one lane over. A line under the boss's name now reads Closes with the season in 3d 14h, and the What the close pays card shows the same countdown beside its heading. This works on phone and desktop. Both count down to the same season end The Race uses, which is when the boss prizes are paid. Once that moment passes, both read closing now until the close finishes.
• The Race stops saying the sector is open after the season has ended (#6140). For the short stretch between the season's end and the close finishing, The Race's headline said Sector 13 is open. Within the hour it burns., with its clocks counting at zero. It now reads The season is over. The Eschaton is sweeping Sector 13. and draws no clock.

  Not fixed here: during that same stretch the Proving Ground's boards come back empty, which looks as if the results were wiped. The close still pays on them. That is tracked in #6149.

  Reported by a player on the gameplay forum.

1.368.0 — 2026-10-03

New

• The Zero-Point Reactor now shows its surplus in resources per hour (#5996). Before, the reactor's sheet gave the surplus conversion only as a percentage, and turning that into Ferrolite, Xenogas and Data per hour meant working it out by hand from the colony's rates. Three rows now sit under the percentage: Production from surplus · Ferrolite / Xenogas / Data. Each shows how much of the colony's hourly production the reactor accounts for now, how much at the next level, and the difference, in the same now → next +delta shape as the other production rows. The figures come from the same calculation the server uses to fill the warehouse. They leave out temporary halts and boosts (Ruin, vacation, exodus shock, Overdrive). The next-level figure assumes the colony keeps today's Grid exports.

  Suggested by astroeddie88 on Discord, who asked not to have to reach for a calculator.

Fixed

• Grid-linked colonies: the resource chips, the building sheet and the warehouse now agree (#6118). A colony that exports power over the Grid and runs a Zero-Point Reactor had its resource chips and the reactor's percentage counting the exported power as still spare. Both read more than the warehouse was actually credited, and the gap grew with the reactor's level. A reactor colony whose deficit is covered by an import no longer shows "energy throttle — effective N/h" on its producers when the server credits them in full. When such a colony previews a next level that would draw more power than today's import covers, the note now says the throttle applies "unless the Grid covers the extra draw". On a colony running its Accelerator Overclocked, the building sheet's production rows now count the heavier draw the way the server does.

1.367.9 — 2026-10-03

Changed

• Synthetic empires donate to their guild at a human pace (#4198). Until now, agents spent every donation charge the moment it regenerated: about 12 a day against a cap of 12. Humans in active guilds donate 2.5–3.9 a day (measured on production over the 28 days to October 3). Each agent now donates at its own steady pace inside that 2.5–3.9 band. The more diligent ones give more, and no two share the same rate. The gap between donations varies, so there is no fixed rhythm. After a long outage an agent catches up with at most three charges, not a full day's backlog.
  • Guild Funds will grow more slowly. Agents supplied about two thirds of all guild donations, and this cuts what they give by roughly 70%. Every guild with agent members feels it, human guilds included, along with guild research and the guild shop wherever agents were funding them. This was decided as part of the Guild Tech Tree design.
  • This is the target, not a measured result. The donation rate will be measured again on production a week after release to confirm agents land inside the band.
• Agents only take a Council donation card they can finish (#4198). At the new pace, "Donate 15 times" can be finished only if it is taken early in the six-day window. Agents now take a donation card only when they have enough days left to complete it. Otherwise they take the smaller card or another directive. An unfinished card pays the guild nothing.

1.367.8 — 2026-10-03

Fixed

• Gates on the shared box no longer go red on web tests that only ran out of time (#6126). Nothing in the game changed. This release is for whoever runs a ship gate. Before, a loaded box could turn the web leg red on tests the PR under test never touched. The Crucible fitter sweep hit its 60 s limit at 67 s. The tech-tree and research-tab i18n tests timed out at 60 s. The guild i18n tests stopped at a 30 s limit nobody had argued for, and the tests after each timeout then failed behind it, which is how one red reached 70 failures across 6 files. Now:
  • the guild and War Room i18n tests use the suite's 60 s budget instead of 30 s;
  • the missions "reading" check no longer races a real 250 ms debounce: the read is held open until the check has run;
  • the calendar and research tests load their tab while the file is collected, so no test's clock runs while the tab waits for vitest to transform it;
  • the Crucible sweep runs as one test per seed, with the same seeds, sizes and assertions, and prints each seed's time;
  • the heavy "every element" i18n tests have a 180 s budget measured from 1,651 gate logs, and print how much of it they used on every run;
  • the launch-options i18n tests no longer sleep 24.5 s of real time per test: they advance a fake clock instead.

  Reproduced on one core with eight workers and four busy loops: before, 24 failures across these files; after, the same tests pass, except launch-options' first test, which still ran past 180 s in that run (about 13 processes per core, harsher than any gate on record).

  Not included: one gate still starts 63 vitest forks for each of three packages on 64 cores, which overloads the box by itself. That is #6134.

1.367.7 — 2026-10-03

Fixed

• Activity: Older is no longer offered under the Galaxy scope, where it loaded nothing you could see (#6136). With a type chip lit (Intel, say) and its history only partly loaded, tapping Galaxy used to leave the Older button under the column. Tapping it fetched more of your own rows, which the Galaxy view never shows. Under Galaxy the button is now withheld, and the line under the chips says that Older loads more under All or Mine. The history already loaded is kept: back on Mine, All or Kept, Older is there again with every row it had already brought in, and nothing is fetched twice.

1.367.6 — 2026-10-03

Fixed

• The Activity filter chips now say which numbers count a chip's full history and which count only recent events (#6049). Before, tapping a chip such as Intel changed the numbers on chips you had not tapped, and nothing said why. «Any type» could even show a number (171 in the report) that it did not draw when tapped (86).
  • While a chip has loaded its own history, the numbers counted over that history are bright, and every other number is faint and counts only the recent events already loaded.
  • A bright number reads 100+ while Older still has more to load.
  • A line under the chips explains the two inks. On desktop each chip has a hover sentence.
  • «Any type» now shows the number of rows it draws when tapped, and so does Galaxy. Other faint numbers can still grow once their chip is tapped and loads its own history, and the caption says so.
  • With no type chip lit, the chips look as they did before.

  Not included: under Galaxy, the Older button still loads your own older rows, which that view never shows (#6136).

  Reported by texclayton on Discord.

1.367.5 — 2026-10-03

Changed

• The rest of the Guild in Italian (#6123). Before, players with Italian on saw these Guild screens in English. Now they read Italian:
  • outposts and anchorages, with their modules, repairs and sieges;
  • joint raids and the Marshal;
  • relics and guild research;
  • the Activity rows for relics, anchorage charters, the Marshal, the colony fold and a displaced garrison.

  In English, every word and figure reads as before. Where the server names an anchorage, a research branch or a module effect, English still shows the server's own wording.

  Still in English inside the raid forms: the fleet template picker and the battle-line editor, which the Missions tab shares. They move together in the next update, and until then the beta notice still says part of the Guild is in English.

  An unknown relic bonus used to stop the Relics screen with an error; it now shows the bonus by name. The confirmation that warns you are leaving a colony undefended no longer goes blank if the Citadel comes back online while it is open.

1.367.4 — 2026-10-03

Fixed

• Tapping a squadron chip in a replay now tells you what it stands for (#6105). The chips at the top left of a raid or campaign replay read W1, W2, S2, R1 or HOME. Their full names were only in a hover tooltip, which a phone or tablet cannot show. Now a tap on a chip still brings up that squadron's ship plates, and also opens a line under the chips that spells the sigla out: W1 · First watch, S2 · Squadron 2, R1 · Relief 1, in your language. Tapping a ship does the same for its squadron, since the ship's plate shows only the sigla. Tapping a defender, which belongs to no squadron, hides the line, so it never sits next to the wrong ship. The line stays as long as the plates that tap opened (2.5 seconds), also while the replay is paused, and restarting the replay clears it. No chip moves when it opens, and the tap still reaches the ship under it. On desktop the line appears too, and the tooltip stays. The "Orbital cannon" and "No resistance" banners now sit a little lower, so they never cover the line.

  Reported on Discord by mark_of_the_sherminator, who asked what W1 and W2 meant after a campaign replay on an iPad. Their in-game name is not yet established, so the credit is suspended rather than withheld.

1.367.3 — 2026-10-03

Fixed

• On a tablet the commander line no longer runs into the resources (#6059). On an iPad, and on any screen from 721 to about 1,280 pixels wide, the line under your commander name could print over the resource tiles beside it: the Gate, your Dark Matter and its daily rate, with the energy bar's note on top. The top bar now gives way in a set order:
  • your commander name is shortened with "…" first (on a computer, hovering it shows it whole);
  • the resource tiles wrap onto another row, as they already did;
  • if the bar still doesn't fit, the buttons on the right (messages, colony, menu) move to their own row under it, aligned to the right.

  Nothing overlaps any more, no button is pushed off the screen, and no figure is shortened. In the browser this is live now. The installed iPad and Android apps carry their own copy of the screen, so they get it with their next app update.

Changed

• On a tablet the top bar can now grow by a row. When the buttons need their own row, the bar is about 46 pixels taller, and the map below it moves down by that much. This can happen the moment an attack alarm or the queue counters appear: before, those pushed the map down by about 17 pixels while overlapping your commander line. With a very wide group of counters (the queue, the Crucible and a live Marshal event together), the resource tiles themselves can also move to the second row, and back when the event ends. Making that steadier is tracked in #6121.

On a computer screen nothing changes in the cases we measured, except that your name is now cut less often.

1.367.2 — 2026-10-03

Changed

• The destination list opens on «Your colonies» for Transport, Pickup and Redeploy (#5932). Those three orders can only go to a colony you own, but the list beside them always opened on «Suggested», which does not list your own colonies. So every one of those orders started with a tap on «Your colonies».
  • On the desktop, the list now moves to «Your colonies» when you choose one of the three in «What to do there».
  • On the phone, it opens there when you change the destination of one of those orders from its form.
  • After you launch a Transport you chose, the list stays on «Your colonies» for the next one.
  • A tab you picked yourself is never moved while you compose that order, and the list never moves back on its own.

  A Transport that the game picked for you, because you aimed at your own colony without choosing an order, does not move the list.

  Suggested by gobbibomb on Discord.

1.367.1 — 2026-10-03

Changed

• Discord: the sidebar counts signups, not live players (#6122). The 👥 Players counter in the Discord sidebar counted the accounts still in the database. Inactive and never-started accounts are cleaned up regularly, so it kept going down even while new people were joining. It is now 👥 Signups: every account ever registered, including those cleaned up since. It only ever goes up. The existing channel is renamed in place. Accounts cleaned up before 17 August 2026 left no record, so the number is a lower bound. 🆕 New today, 🟢 Online, 📅 Active today and ⚔️ Battles 24h are unchanged.

1.367.0 — 2026-10-03

New

• Overdrive: Disciplines keep going past level 24 (#5649). A Discipline used to stop at 24. Once a Discipline is finished, its research panel now shows an Overdrive track, and you can keep researching it with no cap.
  • Effect: every level adds +1% to that Discipline's effect, on top of everything else you have. It applies in every fight and system, PvP included.
    • Energy Weapons: fleet attack. Armor Plating: hull. Shields: shields, in attack and in defence.
    • Propulsion: fleet speed. It still shortens trips for a Veil holder.
    • Logistics: cargo and loot hold. Galactic Economy: production of all three resources.
    • Cybernetics: build speed for buildings and ships, even once your build discounts have hit their floor.
    • Astrometrics: your sensor network.
    • Espionage: the chance a probe goes unnoticed. This is a relative gain and never passes 95%.
    • Colonization: every colony's native planet specialty.
  • Researching it: each level is researched like any other order. It takes a lane, waits in the queue and is timed by your Laboratory.
  • Price: resources plus Dark Matter.
    • Dark Matter: 100 for level 25, then 100 more for each level after.
    • Resources follow the Discipline's own curve until each one reaches its ceiling, somewhere between levels 33 and 42. After that only the Dark Matter keeps growing.
  • When you pay: like every research order, it is paid when it starts. Queued orders wait unpaid. The queue won't let you park more Dark Matter orders than you can cover at the moment you queue them, so a parked order can't freeze your whole queue.
  • Where it shows:
    • The Discipline's card counts on ("lvl 27") instead of saying MAX.
    • Every place that names a research order says "Energy Weapons Overdrive · lvl 27": the Bench, the speed-up sheet, the Laboratory's lanes, battle-report chips, spy reports, the activity feed, the guild aid board and the return recap.
    • Research Power counts the resources these levels cost.
    • The wiki has a new Overdrive section.
  • Synthetic empires don't research Overdrive yet.

1.366.0 — 2026-10-03

New

• The colony screen now says how long each upgrade takes (#5995). The building list showed what an upgrade costs, what you're short of, its ⚡ change and any Command Center block, but not how long it takes. For that you had to open the building's sheet on the Base screen. Every upgradeable row now ends its cost line with a small clock and the next level's build time, the same figure the building sheet shows. It already counts your Command Center, your tech tree and research, the industrial edict and any exodus shock, so it matches the timer you get when you tap upgrade. On a row the Command Center still blocks, it shows the time at your current Command Center level, so after you upgrade the Command Center it can only be shorter. Rows with an upgrade already running keep their countdown. Buildings at their maximum level and those waiting for an advanced core show no time.

  Suggested by astroeddie88 on Discord. The other half of his message, showing payback time on these rows, is still open as #1743.

1.365.12 — 2026-10-03

Fixed

• The «Launch from» list now names the colony that really gets there first (#5963). The list gave each colony a flight time and tagged one fastest, but it worked those times out with a shortcut. The shortcut left out allied Hypergates, your own gate network, a guild drydock and home lanes. The line above the Send button included them, so the two could show different times for the same flight. The tag could then point at a colony that was actually slower. Each row is now priced the same way as the launch line. A row shows «…» while its time is being worked out, and fastest only appears once every row has a real figure. If the real figures can't be fetched, the rows fall back to the old estimate, dimmed and without a fastest tag. Picking a different row keeps the times on screen while the new launch time loads.

  Reported by astroeddie88 on Discord, with a screenshot of two different times for one launch.

1.365.11 — 2026-10-03

Changed

• The Refit desk is now a drydock (Inventory → Refit). It used to be four rows of plain text boxes, one per ship class, that turned into a long column of sixteen cards on a phone, and picking a module unfolded its options inside a narrow card. Now:
  • one ship class at a time, chosen from a tab bar that shows each class's hull and how many of its three bays are filled. The desk opens on the first class that already holds an order;
  • every one of the fifteen modules has its own artwork, on the bay and in the list of options;
  • bonuses read green and drawbacks red;
  • on a phone the class's combined figures come first, with bars, above the three bays;
  • a slot's options open as a sheet from the bottom of the screen on a phone, or as a dialog on a computer. If an order fails, the error shows inside that sheet.

  Prices, refunds, the 12h cooldown and what each module does are unchanged.

1.365.10 — 2026-10-03

Fixed

• The «no bounty» lines in the activity log are now in Italian (#6074). On the Italian locale, the three lines that explain why a strike put no bounty on your head were still in English: «Silent empire — no bounty», «Hunt honoured — no bounty», «Guild reply — no bounty», with their sentences. They showed on their own row and, since #5823, inside the Battle card they belong to, where they were the one English sentence in an Italian card. They now read «Impero silente — nessuna taglia», «Caccia onorata — nessuna taglia» and «Risposta di gilda — nessuna taglia», with their sentences in Italian, in both places. The English is unchanged.

1.365.9 — 2026-10-03

Changed

• Everything the Fleet Advisor says about an attack now reads in Italian (#6069). Before, an Italian player saw Italian reasons for the recommended fleet, but the warnings under them were still English: how old the spy report is, unknown research, the defender's battle orders and modules, garrisons and fleets away, a raid won at a loss, founding hulls left home, a Beacon's growing wall. The early exits were English too (no armed ship, no report, a target the whole fleet cannot beat), and so was the reason to switch to espionage. All of these now read in Italian. Return times use Italian units («~1g2h»), figures are grouped the Italian way, and the Bubble, Beacon and ship names keep their names.

• One English wording is corrected. The Advisor's static-defences line said «N batteries» for what every other screen calls emplacements. It now says «Static defences on target (N emplacements, M fortresses)», so the count is no longer read as Defense Batteries buildings.

Still English for an Italian player: what the Advisor says about the other mission types (spy, explore, colonize, transport, harvest, defend and redeploy). That is the second part of #6069.

1.365.8 — 2026-10-03

Changed

• The War Room and the League in Italian (#6091). Before, players with Italian on saw these guild screens in English. Now they read Italian:
  • the War Room and the League's Match room, with their target boards;
  • the League with the Duel week;
  • the guild season with the Eschaton Project;
  • the council directives, every objective included;
  • the war and League rows of the Activity tab.

  Their countdowns follow the language too. In English, every word and figure reads as before, and the directives keep the server's own wording.

  A War Room that met a target type the app did not know used to stop with an error; it now shows the target with its raw type.

  Still in English: outposts and anchorages, joint raids and the Marshal, relics and guild research, and the colony-fold rows. They come in the next Guild update.

1.365.7 — 2026-10-03

Fixed

• A test that failed at random on a busy build machine is now reliable (#6090). The check for the target panel's «Reading the slot…» line raced the reply from a fake server, so it failed on a heavily loaded machine even though the game behaved correctly. The test now holds the server's reply until it has seen the loading line. It no longer fails when the machine is busy: none of 22 runs failed on a busy machine, against 10 out of 30 before. The game itself is unchanged.

1.365.6 — 2026-10-03

Fixed

• Typing an emoji or an accented letter at the edge of a name no longer cuts it in half (#6062). Capped text fields used to stop on the browser's own limit, which counts half-characters. An emoji at the end of a name, such as a family emoji, came out as a man and an invisible joiner. A rare letter came out as half of itself. A guild description or message of the day could then refuse to save as "invalid characters". Eleven fields now take or refuse each character whole:
  • formation preset name;
  • fleet template name;
  • chat-mute reason;
  • dispatch button labels;
  • guild description;
  • guild message of the day;
  • guild rank titles;
  • flagship name;
  • guild name at founding;
  • commander name on the profile;
  • guest nickname.

  When something does not fit, nothing is cut. The text you already had and your cursor stay where they were, and a line under the field says "No more room." That line is always there and only speaks when it has to, so nothing on screen moves. The guild description and the message of the day also show a characters-used/limit counter once you are within 60 characters of the limit. With a keyboard that writes whole words before committing them, as Android keyboards and Japanese or Chinese input do, the check happens when the word is committed, as it did before.

Changed

• These fields no longer carry the browser's length limit (#6062). The issue asked to keep it, and this release goes the other way: with it in place the browser cuts before the game can see the text. The limits themselves are unchanged and the server still enforces them. App builds already installed keep their own copy of the old limit until they update. The server repairs what they send, as it has since #5914 and #6063.

Not changed: the guild tag (capital letters only, so nothing can be cut in half) and the chat message box, which has its own counter and limit.

1.365.5 — 2026-10-03

Also

• A slow launch-flow test no longer times out on a busy test machine (#6095). Nothing changes in the game. The test that walks the Missions launch flow, about 35 screens in English, the test pseudo-locale and Italian, ran each language as one long test. On a loaded machine that test hit its 60-second limit, and three later tests failed with it. It now runs as 16 independent steps per language, each checking the same things. With the test squeezed to a fifth of a CPU core, the old version timed out and the new one passed three times out of three, its slowest step taking 11 seconds.

1.365.4 — 2026-10-03

Fixed

• The Sigil beside an empire's name now reads on every screen (#4663). On a plain desktop monitor, the Sigil next to a name was the Apex Chart's medallion shrunk to text size. The Lance and the Aegis read, the Forge barely did, and the Veil blurred into a violet square: you could tell someone had a Sigil, but not which one. Each Doctrine now has its own small violet mark drawn for name size: a spear for the Lance, a shield for the Aegis, an anvil for the Forge and a crescent for the Veil. The four differ by shape as well as colour, and none can be mistaken for the AI badge's hexagon or the Wraiths' triangle. You see them in chat, whispers, the leaderboard, the target list, guild rosters and battle reports, in the same place as before: after the AI badge, before the guild tag. The Apex Chart keeps its own art.

1.365.3 — 2026-10-03

Fixed

• Ship tooling: the deploy self-test passes on a Mac again (#6097). The build-cache prune shipped in #6082 brought self-test cases that need GNU timeout, which macOS lacks. On every Mac, every local gate and merge-check went red on that one test, whatever the change being shipped. On a host without timeout, the self-test now provides a stand-in on the fake server side only. It asserts the branch a Mac deploy really takes, which has no local bound. Nothing changes for players, and nothing changes in how the game is deployed.

1.365.2 — 2026-10-03

Changed

• Deploys no longer fill the server's disk with build cache (#6076). On 2026-10-03 the production disk reached 100%: every deploy built the game's images and nothing ever cleared the build cache, which had grown to 592 GB. With no space left, the database refused writes, game events among them, until space was cleared by hand. Nothing warned before it happened. Now every deploy that builds and passes its checks ends by clearing the server's build cache that has not been used for 72 hours. It prints how much space came back (PRUNE: DONE), or a warning (PRUNE: WARN) when the cleanup fails. A failed cleanup never marks a good deploy as failed. Game images, containers and the database are never touched.
• A deploy refuses to start on a nearly full disk (#6076). Before it copies or builds anything, a deploy that has something to build now reads the free space on the server. Below 20 GiB it stops (DISK-FLOOR: RED), changes nothing, and names the cleanup to run. A deploy that cannot read the free space stops too.

  This takes effect from the next deploy run by a checkout that has pulled this release. A checkout still on older deploy tooling keeps building without clearing the cache, and without the free-space check.

  While this was being built, at about 08:58–09:01 UTC on 2026-10-03, a test of the new tooling handed its probe the real connection to the server instead of a stand-in. Up to four of the same 72-hour cache cleanups probably ran on production, outside the lock that keeps two deploys apart. Afterwards: the disk had 690G free, every game container was up, the API logged no errors in the following 20 minutes, and the site answered normally. The cleanup can remove only build cache unused for 72 hours, so it touched no game data. The deploy tooling's self-test now runs behind a trap that stops any test from reaching the server or GitHub.

1.365.1 — 2026-10-03

Fixed

• Saving over a formation or fleet template no longer leaves two with the same name (#6079). Sometimes a name you typed got cut off in the middle of an emoji. An app that had not been updated since the server started tidying such names would then add a second entry instead of saving over the first, and the server stored both under one name. You saw two identical chips, and deleting one deleted both. The server now keeps one entry per name, the one saved last, in the place of the first. The current app also saves over the existing entry directly. If you are still on an old app version, it may keep showing the extra chip until you reload.

1.365.0 — 2026-10-03

New

• Send a spy probe without leaving the screen (#5897). Every «Spy» button used to take you to the Missions tab, and you had to find your way back by hand: the planet sheet in the Galaxy, a report card in Activity, a target row in the War Room or the League. Now each one opens a small dialog right where you pressed it.
  • Only colonies that hold a Spy Probe are listed. The colony you are viewing is picked first; if it has none, your mother colony; otherwise the colony with the most probes. The trip time is shown for the selected colony.
  • After a launch the dialog stays open and says «✓ Probe launched → coordinates · from colony». The button becomes «Send another probe», and the probe counts update.
  • A refusal (a shield, novice cover, a pact, the night window and so on) is shown in the dialog with the same wording the full form uses, and you can try again.
  • If the answer never comes back (a timeout), the dialog says the probe may have left and turns sending off for that target. The way out is «Check your missions», which opens your in-flight list rather than a new form, so a second probe cannot go out by accident. Sending comes back once your colonies refresh, once you have checked, or after two minutes.
  • If a friend's world turned a probe of yours back at that address before, the dialog shows the same warning the full form does, before you send.
  • «Open the full form» is still one tap away and opens the Missions form on the colony picked in the dialog. On a phone, Back closes the dialog first and leaves the War Room or League sheet you opened it from where it was.

  This finishes the spy half of PR #6010, carried onto the current game. Its other half, the war banner opening the War Room (#5904), already shipped in v1.364.12.

  Suggested by a player in the Discord #suggestions channel, and again by a second player for the War Room.

1.364.16 — 2026-10-03

Fixed

• Half an emoji in a request no longer breaks the save (#6063). An app or script can send text holding half of an emoji, for example a name an older app cut in the middle of one. The server used to answer that with an error (500) and save nothing. It did so on any text that went into the game's database, and also on any id it looked up. That included chat messages and chat edits, dispatch titles and bodies, admin ban, rename, Dark Matter and guild-move reasons, and ids such as the message you report or the member you remove. Now the half character is dropped and the rest of the request goes through as sent. Whole emoji, joined emoji such as families, flags, keycaps and emoji variants are kept exactly as typed.

Changed

• Nine names and texts that refused half an emoji now accept the text without it (#6063). This is intended. These fields used to refuse the text as invalid when it held half an emoji, and they now save the text with the half removed:
  • guild name at founding;
  • guild tag at founding;
  • guild message of the day;
  • guild description;
  • guild rank titles;
  • colony name;
  • flagship name;
  • commander name on the profile;
  • guest nickname.

  Sign in with Apple used to swap a first name holding half an emoji for the placeholder commander name. It now keeps the name without the half.

Not covered: the repair applies to request bodies the game's routes read in the usual way. A route that reads only the raw request is outside it. Today the only such route is the payment provider's webhook, which reads the raw bytes on purpose to check the signature.

1.364.15 — 2026-10-03

Fixed

• A saved formation's name now updates the moment you save (#6066). The server tidies some preset names when it stores them, for example a name an older version of the game cut in the middle of an emoji. When you saved any formation, the panel ignored the server's reply and kept showing the old, untidied name on the chip until you reloaded. The panel now shows the name the server actually stored, right after the save. Installed app builds pick this up when they update.

1.364.14 — 2026-10-03

Changed

• The Fleet Advisor's reasons for an attack fleet now read in Italian (#5909). Before, an Italian player saw the Advisor's verdict in Italian directly above bullets that were all English, such as «Bubble ≈ 48,120 HP absorbed before the fleet: factored into the simulation.» Now every bullet under an attack recommendation reads in Italian: the cheapest winning fleet and its safety margin, which ship leads the composition, the target's static defences (with or without the Citadel), the Bubble, the freighters added as hold and the speed they cost, the ships that can stay home, and the Orbital Cannon warning. The Bubble keeps its name, and figures are grouped the Italian way. Nothing changes in English.

• The guild's ally-under-attack box reads in Italian. The green Defending box shows a guildmate's own defences, their wall in the vanguard, the low-power warning, the garrisons on station with their watch countdowns, and the reliefs still on their way. All of it was English in every language and now follows the player's language. Guildmates' names are printed exactly as typed, and the AI badge stays beside them.

• The War Room's «Bubble N» chip now reads from the game's translations, like the same chip on the target panel. It reads the same in both languages.

Still English for an Italian player, and tracked:

• the Advisor's caveats under an attack (report age, unrevealed research, battle orders and the rest), its early exits, why it suggests a probe, and every other mission type (#6069);
• the rest of the War Room, the raid panel around the ally box, and the battle line's row names «Vanguard», «Center», «Rearguard» (#6070).

1.364.13 — 2026-10-03

Changed

• "No bounty" now reads as good news, not as a warning (#5822). A strike on a silent empire costs you no bounty. The screens that say so used to paint that line in the same gold as their warnings: the attack card while your fleet is in flight, the raid panel, the planet sheet, the launch list, and the confirm of a one-tap attack. On the launch list it also carried a ⚠. The line is now green, the colour the galaxy map already uses for "no bounty", and on the launch list it carries a ✓. It stays where it was in the list, and Launch works exactly as before. On the raid panel the bounty-free strike that spends a retaliation credit, and the "no bounty while the pack stays small" line, stay gold: they hold only until the credit is spent or someone joins.

• A strike that cost no bounty is one entry in the Activity log, not two (#5823). A strike on a silent empire, a hunt honoured after the bounty ran out under your fleet, or a reply your guild was owed each wrote a second row next to the battle. That row only said that nothing was added to your head. When the battle is on screen, that sentence is now a green line inside the battle's own card. A waiver that cannot be matched to its battle with certainty keeps its own row, with the same text as before, now in green. That covers a battle no longer loaded in the feed, several fleets landing on the same target for different reasons, and one waiver with two possible battles. Hiding a battle hides its waiver line with it. A waiver row you had already hidden stays hidden, is counted in "Show N hidden", and comes back with Unhide. On the Italian locale the waiver sentence is still in English, as the row was before; translating it is #6074.

Suggested by a player on Discord.

1.364.12 — 2026-10-03

Fixed

• The war banner now opens the War Room, on the war you tapped (#5904). While your guild is at war, a red banner sits above every screen. On desktop it already said «war room →» and showed «Open the war room» on hover, but clicking it did nothing.
  • On desktop, each war in the banner is now its own button and opens the War Room on that war's front. The «war room →» button opens the room on its usual first front. All of them can be reached with the keyboard.
  • On a phone, the whole banner row is one button. With one war it opens that war's front; with several it opens the room's usual first front. «war room →» now shows on the row even when several wars are running.
  • Back returns you to the screen you tapped from.
  • The phone banner is a little taller than before (44px), the standard size for anything you tap.
  • The banner refreshes once a minute, so for up to a minute after a war ends it can still list that war. Tapping it then opens the room on your remaining war, or shows «No active wars».

  Suggested by a player in the Discord QoL thread.

1.364.11 — 2026-10-03

Fixed

• An emoji cut in half at the end of a name no longer breaks the save (#5914). Text boxes stop at a fixed length, and the browser counts that length in a way that can cut an emoji in half when it sits right at the limit. Four fields had nothing behind the length check to handle that: formation preset names, fleet template names, the admin chat-mute reason and the buttons under a dispatch.
  • Before: half of an emoji made the save fail with a server error. A family or other joined emoji left ending on its invisible joiner character was saved as a fragment, a man followed by an invisible joiner.
  • Now: the broken piece is removed before saving, along with any space it leaves, and the save goes through. The name you just saved shows exactly as it was stored.
  • Complete emoji are saved exactly as typed: couples, families, keycaps and flags.
  • A single flag letter (🇦), or an odd run of them, is kept as typed when the text is under the field's limit. Only at the limit is a trailing single letter treated as half a flag and removed. The guild aid note follows the same rule now; before, it always removed that last letter.
  • One value that used to save is now refused: a name or reason that is ONLY a cut fragment, such as a man and an open joiner with nothing else. It is refused with the field's existing message, as an empty name would be.
  • Not repaired, deliberately: a cut that happens to end on a complete emoji inside a longer one. For example, a family cut right after the second person draws as a man and a woman. It cannot be told apart from two emoji typed side by side, so it is stored as received.
  • Names saved before this fix are shown as they were stored. A name that was nothing but a fragment is no longer shown at all, so saving your list can no longer fail on it. Any other damaged name is repaired the next time you save the list.

1.364.10 — 2026-10-03

Also

• No change on screen. This release is code-comment and design-doc work only. Nothing a player sees or plays is different.
• Comments and design docs give today's galaxy size (#5885). Five places still described an older galaxy: "eight planets in orbit", "960 planets", "~10 sectors × ~30 systems", and a "today" note that gave 8 sectors where there are 12. They now point at the real shape, 12 sectors × 19 systems × 18 planets, for 4,104 core slots. A read-only count on production (2026-10-03) found no colony on a planet above 18, so nothing was left outside the galaxy by the 24-planet attempt of 2026-08-06.
• The note on the largest Chronoshard gives the true reason it is rare (#5886). It said gates stop attacks above 3×. They have not refused any attack since 2026-08-01. What keeps the 4-hour shard rare is the weekly cap: one 4-hour shard is the whole weekly cap, so it only drops for a player who earned no other shard in the previous seven days. In production it fell from 22–58% of underdog shards a week in August to about 1% since mid-September. Whether that is the intended behaviour is a separate question, filed as #6058. This release does not change it.

1.364.9 — 2026-10-03

Fixed

• Three Wars Won now takes three real guild wars, not two (#5947). Three guild wars decided in August, before the September change to how verdicts are recorded, were counted twice towards the Three Wars Won achievement. Under the old recording, one war's verdict could land on two different seconds, and each second was counted as a separate war. Two guilds were counted wrong. LANT showed five wins instead of three and got the achievement three days before its real third win; it has earned it since and keeps it. The other guild got it with two real wins; that guild has since disbanded, and no one can see the award. Wins are now counted per war, however the verdict was recorded, so the next war a guild wins is added to its real total. No achievement already awarded was changed.

1.364.8 — 2026-10-03

Fixed

• Every rollback step in the operator's deploy guide now has its own letter (#5860). Players see nothing different; this changes the runbook used to roll the game back after a bad release. Each rollback step in that guide is named by a letter, and other steps, commit messages and issues point to it by that letter. Two letters had each been given to two different steps: R (the night-window undo and fleets in transit) and AA (the Charter Grade and the War Ledger). A pointer to either could send an operator to the wrong step in the middle of a rollback. The later step of each pair has moved to a free letter: fleets in transit is now CD and the War Ledger is now CE. Each pointer inside the guide was read in context and now names the step it meant. Both moved steps say what letter they used to have, so older commit messages can still be followed. A test now fails if two steps ever share a letter again.

• The rollback step for the hunted Marshal pilot no longer swallows the step after it. The guide's step BK (#5045) opened an SQL code block for its last check and never closed it, so the next step (guild dues and rank titles) was displayed as code. The block now closes right after BK's SQL, and nothing an operator copies into the database changes. The new letter test found it.

1.364.7 — 2026-10-03

Also

• The release check on the daily digest's headline now judges the first bullet, not the first valid one (#5865). Players see nothing different; this changes a check that runs before each release. The daily digest leads with the bold headline at the top of a release's player-facing section. When that first bullet was malformed (its bold wrapped onto a second line, plain prose, or a label ending in a colon), the digest quietly led with the next bullet, and the check still said OK about it. That is how v1.349.8's digest led with its secondary fix. The changelog page could not show the problem, because it joins a wrapped line back together. The check now refuses the release. It names the rule the first bullet broke and the sentence the digest would have printed instead, and it lists the headline rules beside the refusal. An HTML comment or a --- rule placed before the headline is refused too, because the changelog page prints both as text above it. Run against every past entry, it would have refused nine releases, and in each of them the digest's headline was not the author's first bullet. Past headlines are unchanged: the check only judges the entry about to be written.

1.364.6 — 2026-10-03

Fixed

• A test of the agents' give-up point no longer fails at random (#5919). Nothing changes for players: the agents give up on an unscouted strike promise exactly as before. The test checked that forty agents never gave up at the same instant to the microhour, and about one run in 10,000 two of them did, by pure chance. That failure blocked unrelated releases. The test now checks what it was meant to check, that the give-up points are spread across the window rather than fixed. A pinned case is kept for the collision that used to fail it, and three more show that a fixed, two-valued or bunched set of give-up points still fails it.

1.364.5 — 2026-10-03

Changed

A raid bounty now says how many fleets joined after you (#5784). A raid's bounty is priced on the whole pack, so fleets that join after you can raise your share. Until now, when a raid billed you for the pack's size, Activity only said "the raid you joined far outweighed its target, and this is your share of the pack's price". Nothing said that the pack had grown after you launched. Now the same entry adds, for example: "Of the 3 contingents that fought in this raid, 2 joined after you. Measured on the Gates at impact, without the contingents that joined after you, your share would have been nothing." When the fleets ahead of you would already have priced you on their own, it says that instead. The figures are the bill's own, read on the Gates at impact.

Entries written before this release, the last fleet to fly, bounties priced on your own Gate, and solo attacks read exactly as before.

What this does not add: there is still no notice while a raid is forming. That is a decision, not an omission. A fleet in a raid cannot be recalled, and the opener can call the raid off only while alone, so a warning before impact would leave you nothing to do with it. This will be revisited if a way to leave a raid ever exists.

Reported by a player on their first day, who worked out the cause themselves and asked whether there should be a warning.

1.364.4 — 2026-10-03

Fixed

• Reloading from the «New version available» row no longer opens on an empty-looking screen (#6038). After an update, the row's Reload button could reopen the game partway down the page: no HUD, no bottom bar, just the system view floating in the middle, until you scrolled. The browser was putting back the scroll position the page had before the reload, but the game always reopens on Home, so that position belonged to a screen that was no longer there. The game's own Reload buttons, the one on the update row and the two on the error panel, now take the page back to the top before reloading, so the game reopens at the top of Home. Verified in Chromium. Not yet measured on an iPad.

  A reload you start from the browser itself (F5, the address-bar button, pull-to-refresh, or a tab the device reloaded after evicting it) still comes back where the page was.

  Reported on Discord by mark_of_the_sherminator, with a screenshot from his iPad. His own observation was the diagnosis: only the game's Reload did it, and opening the game from the home-screen shortcut never did.

1.364.3 — 2026-10-03

Fixed

• Home no longer says OPEN ORBIT before your shelters have loaded (#6017). On a phone, the Home screen's orbit stamp used to read «OPEN ORBIT — no shield or shelter is covering this colony» for a moment on nearly every app load, then switch to NIGHT 6H or your shield. During your night window, which counts as a shelter since v1.360.0, that first claim was false, and it was false for anyone holding a novice shelter or a shield. The «Loading colony…» screen now stays up until your protection has been read, so the first thing the stamp says is true. If the read fails, the stamp shows … as before, not OPEN ORBIT.
• An attack can no longer skip the empire-shield warning while the game is still loading (#6037). An attack composed in that same moment, from Missions or a guild raid, showed neither the warning that it would cost you your empire shield nor the "couldn't check" note. The Missions and Guild screens now open only once your settings have loaded, so the warning, or the note if the read failed, is always there.

The loading screen does not last noticeably longer. The settings request now goes out alongside the colony requests the loading screen already waited for, instead of 16th of 23.

1.364.2 — 2026-10-03

Also

• Back's safety net checks each popover on its own (#6032). Nothing changes on screen. The automated check behind the Back button makes sure every menu, dialog and tap-outside panel can be closed by Back. Before, it judged a whole file at once, so a second menu in a file that already had a registered one went unchecked. The same was true of a dialog role arriving through spread props or from another module, and of a registration that could never fire. It now judges each component through the TypeScript compiler. A role it cannot resolve counts as a failure. The shapes it still cannot see are listed in the test itself.

1.364.1 — 2026-10-03

Also

• No change on screen. This release is test and code-comment work only. Nothing a player sees or plays is different.
• The check that stops a bounty-waived strike being called «free» now reads four more shapes (#6000). Striking a wanted head or a silent empire waives only the bounty: the fleet still fights and can still lose hulls. Since #5888 the copy says so. The test that keeps it that way used to miss four cases, and each one now fails it: a sentence split by an inline tag or a literal (costs you <b>nothing</b>), the spellings «does not cost you anything» and «costs him nothing» and two Italian ones it did not list, a translation line whose subject sits in the key beside it, and a claim appended to an allow-listed sentence. What it still cannot see is listed in the test itself. Three comments that quoted the Wanted board's old header now quote the current one.
• Two server comments now say what the code does (#5853). The comment beside the supply-lane bounce no longer says a returning fleet checks who owns its home colony. It does not, and the comment now calls that difference latent, because no code changes a colony's owner today. The list of the places a colony gets deleted is complete again (six calls in five files, the raze among them), and a new test fails if a site is added or moved without that list being updated.

1.364.0 — 2026-10-03

New

• Back now takes you back to the screen you came from (#5898). Before, the game kept no history at all. Planning a mission from the Galaxy, opening the Activity from a recap, or jumping from a colony to the Shipyard were one-way trips you had to retrace by hand, and the browser's Back left the game.

  How Back works now

  • The browser's Back returns to the previous screen, on the colony that was selected there.
  • On a phone, a ← next to the dispatch bell does the same. On desktop, it is the first slot of the tab rail, greyed out when there is nowhere to go back to. Its tooltip names the screen it returns to, for example "Back to Galaxy".
  • Moves on the tab bar, the rail, the bottom bar and the colony picker do not pile up history, so twenty taps never cost twenty Backs. Neither do links from notifications or from outside the game.

  What Back closes first

  • With a sheet, a pop-up, a menu or the map's side drawer open, the browser's Back closes the newest of them first and leaves you where you are.
  • The on-screen ← always goes back, closing any open menu in the same tap.

  Phones

  • The top bar no longer grows a second line on narrow phones. Your name and the colony name shorten to make room. The colony chip always keeps the first few letters of the colony's name.

  What is not in this release

  • Android's hardware Back arrives only with the next app build. Installed apps keep their current behaviour until they update: the hardware Back still closes the app.
  • In the iOS app, use the on-screen ←.
  • Back does not yet restore what you were doing inside a screen, such as the map's position or a half-written mission order.
  • In a browser, Back on your first screen with a sheet open still leaves the game.

  These are listed on #5898.

  Suggested by a player on Discord, navigating between screens across seven colonies.

1.363.2 — 2026-10-03

Fixed

• The deploy-lock check no longer reports a held lock as free (#6030). This is tooling only, and players see nothing different. deploy.sh --status --server asks the server whether a deploy holds the deploy lock. Before, it answered from a single read of the kernel's lock table, /proc/locks, and that read is not a snapshot. When another process on the machine dropped a lock while the table was being read, the held lock's line could be skipped, and the answer came back FREE.
  • On the server: a FREE could mark a live deploy as DEAD to anyone polling --status --server, whenever that deploy's own heartbeat was stale and its process could not be seen from the asking machine.
  • On the ship gate: the same race made deploy.sh --self-test fail now and then on a loaded box. Under deliberate lock churn, 50 rounds of the probe's self-test checks failed 11 of 500 before the fix and 0 of 600 after it.
  • Now: a FREE from the table is checked against the lock holders' own file-descriptor records, which the kernel prints in one piece. A deploy that holds the lock reads as HELD.
  • What is left out: a holder that took the lock on a descriptor other than the one the deploy script uses is still seen only through the table, race included.
  • Where it applies: each copy of the ship scripts gets the fix only when its tree pulls this release.
• The hook-order test stopped flagging a named helper function inside a component (#6006). It counted a return inside a function declared in a component's body as the component's own early return, and then reported every hook below it. It now ignores returns inside nested functions of every kind. A real early return after such a helper is still caught.

1.363.1 — 2026-10-03

Changed

• Your own spy reports have their own chip in Activity: Intel (#5915). The results of your probes (the reports they brought back, the probes that found nobody, the probes a pact turned away) were reachable only by picking Combat and then the Espionage sub-chip under it. Nothing on the top row hinted that intelligence lived behind a chip called Combat.
  • On the top row: the chips now read All · Combat · Intel · Exploration · Empires · Economy (Intel is «Spionaggio» in Italian), and the Intel chip counts what it shows.
  • Combat's sub-row: it keeps Battles, Defence, Wars and Season.
  • Probes flown at you: a probe someone flew at you is a warning, not your own intelligence, so it stays under Combat → Defence, beside the incoming-attack alarm it often comes before.
  • The shortcut: the «Only your probe results» shortcut that appeared under a run of probe reports is gone, because the Intel chip does its job from the top row.

  Suggested by a player on Discord, who pointed out that espionage does no harm and should not have to be found under Combat. As he asked, it is a chip of its own; the probe flown at you deliberately stays in Combat.

1.363.0 — 2026-10-03

New

• A colonise order now tells you it can carry a starting endowment (#5926). A colonisation has always been able to carry cargo that becomes the new colony's first stock, but the section for it sat below the fleet picker. Nothing above it said it existed, so a player could found every world empty and then pay a second flight to supply it. The order's load line now names it before the fleet: «Starting endowment: none loaded — the new colony can store up to 5,000 of each resource (Colony Ship); the hold carries 1,500 in total. load it ›».
  • The figure is the landing's own. It is 5,000 with a Colony Ship and 20,480 with an Ark, and Deposits research raises it. During a Requisition Surge the line says the figure holds «on arrival», because the ceiling shrinks when the surge ends.
  • Where it shows. On a wide screen it sits in the order's Fleet column. On a phone or a narrow window it comes before the fleet. The link opens the cargo section and puts the cursor in Ferrolite.
  • With no Colony Ship or Ark picked, the line says one is needed.
  • Over the colony's ceiling, the cargo section and the load line say how much is over and where it goes: home with the escort, or lost when the founding ship flies alone. Nothing is blocked, and an empty coloniser launches as before.

  Suggested by a player on Discord, who had been settling planets empty and then sending transports after them.

1.362.0 — 2026-10-03

New

• Going to can lead with the address fields, and remembers it (#5896). In the mission composer the target finder used to come first, and the Sector / System / Planet / Paste fields sat at the bottom under a fold, one extra click on every launch for a player who types addresses. A # button at the end of the finder's search row now switches to the address fields. They then come first and open, with a magnifier button to switch back. The finder stays reachable, folded underneath under "Find a target instead", so on the phone its one-tap orders are still one tap away. The choice is remembered on this device for your account. If you never press #, the screen is exactly as before.
  • Switching carries your address across: an address typed in the search box becomes the target in the fields, and the target comes back into the search box on the way back.
  • If the search box holds an address the mission is not aimed at, the launch is refused as before. The folded finder now opens itself so you can see which address that is.
  • On the phone the four boxes and the switch fit on one line, from a 360px screen up.

  Suggested by a player on Discord.

1.361.1 — 2026-10-03

Fixed

• Two test checks that failed only on a busy build machine now give the same answer every time (#5997, #6011). Nothing changes in the game. Both checks guard what players see, and both had been failing now and then for reasons that had nothing to do with the game. That held up unrelated updates.
  • #6011: the check for the launch receipt's «Show it» button expected the wrong mission id whenever a background travel-time request happened to finish first. The button itself was right: it always goes to the fleet that launch created. The check now proves that against a second fleet already in flight.
  • #5997: the check that the phone action bars never sit under the toast or the error banner sometimes measured the page before it had drawn a single frame. A second fault was found while investigating: under load, the same check could also read a page file that another of its own steps was rewriting. Both are fixed. The three-attempt retry that only hid the first fault is gone.

1.361.0 — 2026-10-03

New

• The map's Places drawer now names the Event Sector, and opens it (#5930). The galaxy chart does not draw the Event Sector, and until now nothing on the map mentioned it. The only ways in were the Season tab's button, the system view's sector selector, or typing 13:x into GO TO. While a season runs, Places now opens on a "⚠ Event Sector · sector 13" row. A tap opens the system view in that sector: on your own system there if you have one, otherwise on system 1, the same place the Season tab's button takes you. If the season has not loaded yet, the row is greyed out and says so. Between seasons the row is not shown. The ⊞ "Open the system view" button on the map is unchanged.

  Suggested by a player on Discord.

1.360.0 — 2026-10-03

New

• Your night protection now shows on screen, next to your shields (#5912). The night window shelters you for 8 hours a day, but until now it existed only as a setting. The Home tab even read OPEN ORBIT, «No shield or shelter is covering this colony right now», while it was refusing attacks.
  • Inside your own window, the protection chip at the top shows Night with the time left.
  • The Home stamp reads NIGHT 5H instead of OPEN ORBIT.
  • The tooltip, or the sheet on a phone, gives the window's end and the rules it really has. Attacks that would land before that time are refused. Only an attack touches it: defending an ally, garrisoning a Relic, spying and transport all leave it standing. Launching an attack at anything but a Pirate Beacon, or opening or joining a guild raid on anything but a Pirate Beacon or empty or veiled coordinates, suspends it for 2h, and each attack restarts the count.
  • While your own attack has it suspended, the night clock is hidden, because attacks can land on you then. It comes back the moment the suspension ends, if the window is still running.
  • The shield rule now says your first attack ends the shields and the novice shelter (it used to say "all of it"), because the night window comes back.

  Suggested by a player on Discord, who asked for a shield showing night protection and the rules that go with it.

  Not included yet: for a moment while the game is still loading your settings, the Home stamp can read OPEN ORBIT before it switches to NIGHT. That is tracked in #6017.

1.359.0 — 2026-10-03

New

• Activity rows can now be hidden once read, or kept at hand (#5908). Before, a row you had read looked exactly like one you had not, and finding yesterday's spy report meant scrolling the whole feed. Now every single row has two buttons under it:
  • Hide takes the row out of the column and out of the chip counts. «Show N hidden» under the column brings hidden rows back, greyed, each with Unhide.
  • Keep pins the row to a new Kept chip beside All, Mine and Galaxy, so it is one tap away.

  Rows you keep or hide are remembered on this device for your account only, and nothing is deleted. The Kept chip lists the kept rows the feed still holds. When none of them are, it says they have aged out of the feed's window. An attack warning cannot be hidden until the fleet has arrived. Grouped rows (a forming raid, a burst of shard spends, a run of trades) carry neither button.

  Suggested by a player on Discord, who was scrolling a lot to get back to a spy mission from the day before.

Changed

• The first row of Activity chips (All, Mine, Galaxy, now Kept) wraps onto a second line on a narrow phone instead of hiding its last chip off the edge.

1.358.0 — 2026-10-03

New

• The desktop timeline can now show you every bar in a lane, not just three (#5761). The "Next 4 hours" timeline on the desktop Home draws at most three bars per lane. Before, the extra ones were only counted next to the lane name (+2 · next 3h), so you could see how many were hidden and when the first one finished, but never which ones. On a long colony name, even that count got cut off. Now the count sits on its own line under the name and is a button. Click it, or press Enter/Space, and the lane opens in place with every bar drawn. Click "less" to close it again. Hover the count to see the lane name and the hidden items with their clocks. A lane only gets taller while it is open, the button stays exactly where it was when the lane opens, and a lane that runs out of hidden work closes by itself. The phone Home is unchanged.

1.357.1 — 2026-10-03

Fixed

• GO TO now takes an Event Sector address to its system (#5900). The activity log prints addresses in the Event Sector, such as founded a new colony at 13:18:11. Typing one into the Galaxy's GO TO used to answer «the chart can't take you there», because the galaxy chart does not draw the Event Sector at all.
  • While a season runs, the address now opens the system view at that system. It is the same way in as the Season tab's "Go to the Event Sector" button.
  • Between seasons, the refusal names the sector: «sector 13 is the Event Sector, open only while a season runs».
  • If the season has not loaded yet, or the last check of it failed, the refusal says «whether a season is running is not known yet». It no longer calls that case closed.
  • Typing home when your viewed colony is in the Event Sector now goes there too. Before, the box closed and the map stayed where it was.
  • Addresses that really are nowhere still get «the chart can't take you there».
• A failed season check no longer moves you out of the Event Sector. If one check of the season failed while you were looking at a system in sector 13, the Galaxy tab used to switch you to the same system number in sector 1, with nothing to say so. It now leaves the Event Sector only when the season is actually over.

  Reported by a player on Discord, with screenshots of the address and the refusal.

1.357.0 — 2026-10-03

New

• You can now see the garrisons standing on your own colony (#5893). Before, when someone sent reinforcements, you got one activity line, "X sent ships to defend you", and nothing else: no list of what was there, and no way to tell how long it would stay. The Colony tab's fleet card now has a Garrisons on station section under your own hangar. It lists every garrison on that colony: who sent it, how many ships, and when its watch is scheduled to end (a clock and a countdown). Your guildmates' fleets and your own are listed hull by hull. A garrison from outside your guild, such as an ally's or a non-aggression partner's, shows only its size, the same as on the guild's threat board.
• The Relics panel shows who is already defending a relic your guild holds (#5917). Above the Defend button there is now an On station now list. It shows each guildmate garrisoning that relic, with their hulls and when their watch ends. If nobody is there, it says "Nobody from your guild is on station here." The list refreshes every two minutes while the panel is open. Relics held by other guilds show nothing about their defence.

Both lists show what is on station now, not what will be there when a strike lands: a watch can end, and the player who sent a garrison can stand it down sooner.

Reported by a player on Discord who couldn't find the reinforcements a guildmate had sent (#5893). Suggested by two players on Discord, the second of whom pointed to exactly where the list belonged: the Guild tab, under Contested Relics (#5917).

1.356.0 — 2026-10-02

New

• The colony header shows your Command Center's level (#5985). Before, the header showed the planet, its name, coordinates, traits, energy and build queue, but not the one number that limits every upgrade. To find out how far a colony was, you had to scroll the building list for the Command Center. Now the header reads Command Center lvl N next to the coordinates, on desktop and on phones down to 320px wide, where it moves to its own line instead of breaking. The build-queue counter below it is derived from that same number, so the two finally explain each other.

  Suggested by antiproduct5001.

• Home flags a research order that has stalled (#5910). Home already flagged an idle build queue and an idle shipyard. A research order that could not start was shown only on the Research page, so you found out only by opening it. Now, when a research lane is free and the next order cannot start, Home says so:
  • on desktop, as a routine order, Research stalled · a lane is free, the next order cannot start, under the idle build queues and shipyards;
  • on phones, the Research pin reads Order stalled, and the next-move card offers See why research stalled once nothing more pressing is waiting (an attention item, a guild invitation or the campaign).

  Both open the Research page, which names the cause: an empty treasury or a Laboratory that is too low. The flag belongs to your empire, not to any colony.

  It appears only when the stall is real. A lane that freed a few seconds ago looks exactly like a stall until the server starts the next order, so before it says anything the game asks the server to try. It shows the flag only if the order still cannot start afterwards. An order waiting for its own previous level, or for a prerequisite that is already queued, is not flagged. An order whose prerequisite will never arrive, because it was cancelled, is.

  Reported by a player on Discord, with screenshots.

1.355.1 — 2026-10-02

Fixed

• A strike that waives the bounty no longer calls itself free (#5888). A wanted head's planet sheet said "Attacking costs you nothing", the launch door said "Costs you nothing while the bounty holds", and the Leaderboard tooltip said the same. Those strikes still fight the garrison, can still lose ships, and still drop your own shield. Only the bounty is waived. Every such line now says that: "Attacking costs you no bounty", "This strike puts no new bounty on you while theirs holds". The same shorthand is gone from the other screens that used it, in English and Italian. War replies read "Bounty-free strike" and "bounty-free reply" (they were "Free strike" and "free reply"). The Leaderboard band notes say a solo strike in your band "puts no bounty on you" (was "free to hit"). The Wanted board, the target board header, the raid panel, the Gate tooltip and the War Room hint follow the same rule. The AI empires' instructions use the same wording, so they no longer tell you a hunt "costs nothing" either.

  Reported by a player on Discord, who asked exactly the right question: "You can still lose material in the attack, right?"

• The capped-emplacements notice no longer tells you to bring warships (#5918). When a heavily armed commander's emplacements hit the shared 12-hour payout cap, the Activity row ended "Bring warships next time: ship kills are never capped this way." Players read it as a judgement on their fleet, even after a battle where the defender fielded no ships at all and there was nothing for any fleet to kill. The row now ends: "Ship kills are never capped this way: this limit applies only to their emplacements, so a colony that fielded no ships could only pay out of those." The row is also translated into Italian now.

  Reported by a player on Discord, who brought 111,572 ships against a colony that fielded none.

1.355.0 — 2026-10-02

New

• The AI-badged empires talk about announced events and maintenance, not only about releases (#5634). Until now, a synthetic empire could pick up only a big news dispatch in galaxy chat. Each empire's switch-over falls between 15 and 17 October 2026, at the same moment its release remarks begin. From then on, one or two of them may also react to an announced event or maintenance. They speak hours after the announcement. The one exception is a maintenance announced at short notice, where it can be as little as 20 minutes, but never seconds. A maintenance remark always goes out at least half an hour before the start. They say only what the bulletin says, with the start in plain words ("starts within the hour"). They never claim to have taken part, never call it "again", and never promise to do anything about it. A notice addressed to one guild, or a personal letter, is never repeated in galaxy chat. If an announcement is retitled or rescheduled before an empire speaks, that empire drops its line.

Fixed

• A guild-only news bulletin could have been read aloud in galaxy chat (#5634). A synthetic empire's view of the bulletin also includes its own guild's notices, and nothing told a guild notice apart from a galaxy-wide one. Every dispatch now carries who it is addressed to: galaxy, guild or player, never which guild or player. The empires speak only about galaxy-wide ones. This was caught before release remarks started, so no guild notice was ever repeated.

Also

• After a remark about a maintenance, an empire's replies know its start time has passed and that they do not know how it went. They state the timing only while the bulletin still shows the start they remember.
• Not included: a "we're back" remark after a maintenance. Every empire restarts at the same moment after a downtime, so those remarks would all arrive together.

1.354.15 — 2026-10-02

Fixed

• AI guildmates no longer skip helping you when they take down their own request (#5979). When an AI-run empire withdrew one of its own guild aid requests (a speed-up for an order that had already finished), it stopped there for that pass and helped nobody, even if your request was waiting on the same board. Now it still answers other players' requests in the same pass. If the withdrawal fails, the agent no longer treats the request as gone: the stale row is retried on its next look at the board, and that pass's help is not lost. A second request still never goes up while the first is on the board.
• The "about Nh to run" figure an agent puts on a speed-up request is now pinned by tests: it rounds (3h58m reads "about 4h", 3h20m "about 3h"), and never shows less than 1h. Nothing about the figure changed; this only stops it changing by accident.

1.354.14 — 2026-10-02

Fixed

• A launch on a tablet or desktop now says it went out (#5903). Until now, only the phone confirmed a launch, by switching to «In flight». On every other screen the form reset where it was and nothing on screen said the fleet had left, so a launch that worked looked like a missed tap. Worse, once the hangar refreshed, the Advisor filled the picker again, so a second tap really sent a second fleet.
  • The launch bar now shows «✓ Launched: <mission> → <coordinates>» beside the Send button.
  • A Show it button scrolls to the new fleet's card once the fleet list has it.
  • While the form still holds the same mission, target and launch colony, and Send can fire, the button reads «Send another fleet → <coordinates>», and the bar says that sending again launches another fleet.
  • Changing the mission, the target or the launch colony clears the confirmation, and so does Dismiss.
  • On phones, the switch to «In flight» is still the confirmation.
  • The list of active missions keeps its order, soonest arrival first.

  Reported on Discord by a player on an iPad, who had pressed Send several times before realising each press had launched.

• After-launch messages on tablet and desktop point at the right place (#5929). Some messages told players to check «In flight», which exists only on phones. These were the line under a «Launch again» card («Launched moments ago», plus the two warnings shown when a launch got no answer or the connection dropped) and the confirmation of an intercept sent from the galaxy map. Off the phone they now point to «Active missions», further down the same page. Phones keep the old wording.

  Reported on Discord by a player on an iPad.

• «＋ New mission» on a phone starts a new mission (#5981). Tapping it while already on New mission did nothing. That left a player stuck on a mission the game would not send: in the reported case the destination was the colony the fleet was leaving from. Tapping it now clears the mission being composed (fleet, cargo, stance and options) and goes back to choosing a destination. The launch colony stays as it was. Switching to New mission from «In flight», or tapping the board's ＋ New mission button, still brings back the mission you left there. Not included: the launch bar can still cover part of the form on short screens. That is tracked in #821.

  Reported by mark_of_the_sherminator, with screenshots from Android.

1.354.13 — 2026-10-02

Fixed

• A pursuit fights its prey once, even when the server has to retry its arrival (#5974). When a hand-flown pursuit reached its target, the battle was saved first and the pursuit was turned for home in a second step. If that second step failed (for example, two server writes colliding at the same moment), the arrival was retried, and the retry fought the same prey again. That produced a second battle report and a second round of losses for both fleets. A retried arrival now recognises the battle it already fought and only turns the fleet for home. A contact that ends without a battle (a "spared" contact against a player you have already caught three times today) is now recorded together with the homeward turn, so it also appears once. Because the arrival is now safe to retry, a pursuit whose homeward turn hits a write collision is retried on the spot, the same way every other fleet arrival already was.

1.354.12 — 2026-10-02

Fixed

• On a phone, the launch pad you fly from now opens tappable (#5505). When the launch form opened on a short phone screen (320x568, 375x667, 393x659), the first launch pad sat under the sticky Alarm / In flight / New mission tabs. It was visible, but a tap on it hit «In flight». The form now scrolls the pad row to start below the tabs. It then brings the active pad (the colony selected in the header, or the one a restored draft was launched from) fully into view between the tabs and the Send bar. On those screens one pad at the far end of the row can still sit under the bar or the tabs until you scroll. Taller phones are unchanged, with every pad clear.

Also

• The Advisor's «switch to espionage» button keeps its place on phones (#5558). No visible change. The test that keeps this button on the panel's first line, clear of the phone's Send bar, now checks every place in the engine that can produce a suggestion, not just the one that exists today.

1.354.11 — 2026-10-02

Changed

• Synthetic empires no longer lean toward commanders with smaller fleets when they pick a target (#2933). The game gives synthetic empires one number no human can see: every owner's total fleet strength.
  • Before: they used it twice. They refused a fight against a fleet more than twice their own, and among the targets they could still reach, they tilted their pick toward owners with smaller hangars.
  • Now: the tilt is gone. They keep only the refusal, and the figure itself never appears in their plans or chat. Whether they decide to fight shows in what they do: a raid they decline because the owner's fleet outguns the coalition is still announced as declined. A solo strike on a player colony still needs a fresh spy report, the same probe you would send.
• A synthetic empire with a long build or research order and an empty warehouse now asks its guild for both, in turn (#3134).
  • Before: an empire that had both needs always asked for a speed-up, ask after ask, and never mentioned it was out of resources.
  • Now: when both needs are real, the two kinds of ask take turns. The turns are not a fixed rhythm: never more than two of the same kind in a row, and otherwise usually the other kind.
• A synthetic empire takes down its own "stores nearly empty" request once its stores are really back (#3134).
  • Before: such a request stayed on the guild aid board for up to a day, however full the warehouse got in the meantime.
  • Now: the empire withdraws it at its next look at the board, once every resource it asked for is back to twice the level that opened the request, the request has been up for at least 100 minutes, and nobody has put anything into it yet. A request a guildmate has already started to fill stays until it is filled or expires.

1.354.10 — 2026-10-02

Fixed

• A fleet caught in a database deadlock lands on the first try (#5948). When two battles or landings locked the same rows in opposite order, the database cancelled one of them. Fleet arrivals, returns home, guild raid battles, Marshal raid landings and undone Marshal launches were meant to retry that on the spot, but never recognised a real deadlock: the whole job failed and was redone up to a few seconds later, and an undone Marshal launch could leave the fleet deducted. They now retry at once from a fresh read, and nothing is applied twice. One exception remains: a hand-flown pursuit whose landing hits a deadlock is still redone later, because redoing it can fight the same interception twice. That is tracked in #5974.

Also

• The in-memory database used by the API's tests now reads a column a test leaves out as the default production stores, instead of an empty value, and rolls back every table when a transaction is cancelled (#5959). Players see no change; it makes the tests behave like production.

1.354.9 — 2026-10-02

Changed

• Last Stand's card now says what the attacker still pays (#3106). Before, the Aegis 15 keystone card only said that none of your groups loses more than a quarter of its hull in a round on defence. Now it adds: each extra round that buys, part of your guns’ fire still gets past the attacking ships’ shields, however large the attacking fleet. This is how combat already worked, and no number changed. Because Last Stand stretches a battle the attacker would otherwise win in a round or two, an armed defence makes even an overwhelming attacker take that fire for every extra round. Measured against 3,000 Voidlance Cutters, attackers from 20,000 to 2,000,000 Titans all lose the same 93 Titans over 31 rounds. Bringing more ships stops lowering the price once the attacker is big enough. The cost is hull damage, so it turns into lost ships only when the defence has enough guns: 30 Cutters beside 300 Bastions stretch the fight to 15 rounds and kill no Titan at all. Thanks to QQQQ, who described the mechanism precisely.

Also

• Ghost Manifest stays as it is (#3118). The suggestion to have the Veil's spy-report noise lie toward the counter of the spying fleet was declined. Spy missions in practice carry nothing but probes, and a lie that always leans the same way is one a reader can learn to subtract. The noise stays symmetric, per ship type, and fixed per report. Nothing changes in play.

1.354.8 — 2026-10-02

Fixed

• On a phone, the Send, Build and Save bars now sit just above the navigation bar, and the battle notice and error banner rise above them (#5941).
  • Before: the floating Send bar in Missions, the Build bar in the Shipyard and the Save bar of an unsaved defence plan hovered 84px above the navigation bar. Rows of the form showed through the gap under them.
  • Now: they sit 12px above it.
  • The two notices that pop up in the same strip (the "battle resolved" notice, which opens the report when tapped, and the red error banner) now rise above one of these bars while it is there. They stay where they were everywhere else. Before this, the notice had only been clear of Send by accident, because of that gap.
  • On older Android WebViews without CSS max() (before Chromium 79) the notices keep their old position.
• In the desktop Missions console, the launch row now rests on the bottom edge of its panel (#5941). When the console fits the window, its verdict panel scrolls, and the launch row sticks to its foot. The row used to sit 10px above that edge, which would leave a strip for notes to scroll through under the Send button. None of the missions we could reproduce had enough notes to make the panel scroll, so this was fixed by its geometry rather than seen happening.

Changed

• iPhone: Effects and Music volume sliders, on iOS 16.4 and later (#5951).
  • Before: iOS does not let a web page set the volume of its sounds, so on an iPhone the game offered Mute and nothing else, both in the menus and on the battle replay.
  • Now: where the system offers the audio session the game needs (Safari 16.4 and later), the sound is routed through the browser's audio engine instead. The Effects and Music sliders then appear after the first tap and work as they do elsewhere, battle music and its fade-out included. The game asks to keep playing through the ring/silent switch, as it does today; if that request is refused, nothing changes.
  • Unchanged: iOS 15 to 16.3 (Mute only, as before), desktop and Android.
  • Not yet confirmed on a device: this was tested only in a simulated browser. Whether the iOS app exposes that audio session, and that the silent switch behaves as before, still has to be confirmed on an iPhone. #5951 stays open until it is.

  Reported by moralezz01 (in-game Tibeh01), who asked for a volume control on the battle replay on iOS.

1.354.7 — 2026-10-02

Changed

• A ship no longer reports Wraith changes as delivered by the deploy (#4231). Before, a merge that changed the Wraith autopilot ended with an unqualified DEPLOY: FATTO — smoke verde. Yet no production container runs that code: launchd runs it from the operator's own checkout, so it only takes effect when that checkout pulls main. The smoke check had verified a surface the change could not reach. Now every deploy prints an OPERATOR-DELIVERED line. It lists the merge's Wraith-only paths (apps/agents/src/novarchy_agents/wraiths/, tools/wraiths-*), or says none. When there are any, the verdict line carries a note that the smoke did not run them. none means only that no Wraith-only path changed: a shared apps/agents/src module that the Wraiths import still reaches them only at the operator's pull, and the line says so. The ship skill now tells the release notes and the report not to call such a change live. What gets deployed is unchanged.
• The convention for crediting a reporting player is written down (#4240). Until now it lived only in about 120 changelog lines, so each session had to work it out again from the corpus. The ship skill's release step now records it, measured on this changelog: who is credited, where the one sentence goes, the verb, the name form, what to do when the remedy is narrower than the ask, and the suspended credit. It also says that the chrono-shard line in an issue's reporter section is bookkeeping done at triage, not a step of the ship.

Both changes are to the ship machinery only, and they reach only the trees that pull them: a worktree branched before this release still runs the old deploy.sh and skill. Nothing a player sees has changed.

1.354.6 — 2026-10-02

Developer tooling only. Nothing changes in the game.

Fixed

• The transaction census counts raw SQL statements (#2624). Before, tools/tx-budget-census.mjs scored every raw SQL statement, and so every advisory lock, at zero. That made the transactions that hold a lock read as the shortest. Now raw SQL counts on every transaction it touches: 39 of the 136 transactions it lists gain 78 statements between them. The census still reports whole-transaction totals, not the stretch from the lock onward. The optional "lock tail" column the issue suggested was not built.
• A missing error-code translation now names the API file that threw the code (#4227). Before, adding a refusal code in the API without a player-facing sentence turned a web test red with only "expected 'x' not to be 'x'". Now both the test name and the failure message name the API file the code is thrown from, and say where the sentence goes (ERROR_TEXT in apps/web/src/common.tsx).

1.354.5 — 2026-10-02

Fixed

• A launch that collides with another write in the database is retried instead of failing (#5948). Two transactions sometimes lock the same rows in opposite orders, and the database kills one of them. The server's retries for that case checked for an error code this database never sends, so they had never once fired. A mission launch, a Marshal strike, opening or joining a Marshal raid, paying war reparations or surrendering could end in a server error. Leaving a guild, opening the guild directives panel, and an operator's guild move could too. Each of these is now re-run from a fresh read, up to three attempts in all. A campaign landing caught the same way is re-run in the same pass instead of waiting for the next sweep. Still to do: the arrival of a fleet, its return home and the guild raid battle still use the old check. There a collision costs a short delay while the job queue re-runs the whole step.

Also

• A season placement payout that fails on such a collision is recorded as 40P01 in the guild log's failure record, instead of unknown. The campaign's server log lines now say when an attempt died on one (#5948).
• The in-memory database the API's tests run on now behaves like the real one in two more ways. A value left undefined on create no longer overwrites a column's default (#5550). An empty nullable column reads back as null rather than missing (#5551). Nothing changes in the game.

1.354.4 — 2026-10-02

Release tooling only: nothing in the game changed for players.

Fixed

• A merge check the build machine could not run no longer reads as "nothing to check" (#5544). Before, the pre-merge check printed the same "skipped" verdict whether there was nothing to test or the machine had failed to test it. Those failures include a missing copy of main, an old git, a killed or timed-out run, and a status that could not be posted. Now only a real "nothing to check" says skipped. Every other case ends MERGE-CHECK: UNANSWERED with its reason, and the remedy names the case where only another machine can help.
• The browser layout tests can no longer be skipped silently by the gate (#5518). On a machine without the headless browser, those tests used to skip and the gate still passed. The full gate now requires them, so a missing browser turns the gate red instead. The pre-merge check does not require them, because a missing browser says nothing about the merge.
• Commit-status descriptions never end in half a character (#5548). In a non-UTF-8 locale, the trimming could cut a multi-byte character in two. It now trims on character boundaries, and a malformed description is refused instead of posted.
• The gate's status text has one source (#5565). The text the gate posts and the text it checks that post against used to be written out twice in the script. Both now come from the same place.

These changes live in the ship scripts each working copy carries (.claude/skills/ship/). A working copy that has not pulled this release keeps the old behaviour.

1.354.3 — 2026-10-02

Fixed

• The speed-up sheet now tells you which button spends your own pick (#5574). Before, if you set a size row to 2 and then pressed the big gold button, it spent the recommended combination, not the two shards you had just picked. Now, once any row shows a count other than 1, that row's Use becomes the highlighted button and the recommended one turns plain and reads "Use recommended". It still spends the recommendation, but now its label says so. Set every row back to 1 and the sheet looks exactly as before.
• Sticky headers and bars no longer show rows scrolling through a gap at their edge (#5941). On a phone, every tab's pinned header (Missions sections, research branches, season clock, Market legend, Colony buildings) used to stick 10px below the top of the screen area, with content sliding visibly through the gap above it. The same happened under the All-hulls sort row in its sheet, under the building sheet's Upgrade bar (12px on desktop, 16px on a phone), and under the "To the colony" row of the return recap (18px on desktop, 16px on a phone). They now pin flush to the edge, and nothing moves while the content is at rest. On iPhones with a home indicator, the bottom bars sit just above it.

Also

• One case is still open on #5941. On a phone, the floating Send / Build / Save bars in Missions, Shipyard and the defence planner sit higher above the navigation bar than intended. That is measured but not changed yet.

1.354.2 — 2026-10-02

Fixed

• Synthetic empires can finish the four-strike Beacon card (#4210). The AI-run empires stopped themselves at three Pirate Beacon strikes a day. That copied a server limit removed on 27 July, and at the top Charter grade the daily "Strike the Pirate Beacon" card asks for four. So a grade-4 synthetic empire could never complete it, and the card expired every day. Now, while that card is open, they strike as many times as it asks within its contract day, and no more. Without an open card they keep their old three-a-day pace. Nothing changes for human players: the server never had this limit for anyone.
• An Italian phrase could slip into the chat instructions of the synthetic empires (#5953). When a chat relay had no name for the other commander, their English instructions filled the gap with "un altro comandante". The gap now reads "another commander". No case of this was seen in live chat. It is fixed so that a future caller cannot bring it back.

1.354.1 — 2026-10-02

Changed

• More of the Guild in Italian (#5938). Before, players with Italian on saw the Treasury, the Help board and Diplomacy in English. Now these read Italian:
  • the Treasury: the guild Fund, donations, dues, grants and the Store;
  • the Help board: asking for a hand and giving one;
  • Diplomacy: tension, war declarations, pacts, reparations and surrender;
  • the guild rows of the Activity tab;
  • the war banner at the top of the screen.

  In English, every word and figure reads exactly as before.

  Still in English: the War Room, the League and guild season, outposts and anchorages, joint raids and the Marshal, relics and guild research, and the Activity rows for a colony fold. They come in the next Guild update. The beta notice still says part of the Guild is in English.

1.354.0 — 2026-10-02

New

• Battle replays have their own audio control (#1852). The replay covers the whole screen, and until now the only way to mute it or turn it down was to close it and use the main menu's audio rows. A 🔊 button now sits next to the replay's ✕ (just under it on a phone) and opens the Effects and Music sliders and a Mute button. On iOS the sliders are still not available — the system does not let the game set the volume there — but Mute is; in-app volume on iOS is tracked in #5951.
• The Daily Wrap shows what it pays (#2257). The Wrap strip on the Contracts board said "pays double" but never how much — the amount first appeared after you claimed it. It now prints the figure like every contract slip above it: Worth while you are still finishing contracts, Pays once it is ready to claim, and a red line when part of it would not fit in your home warehouse.

Changed

• The Guild Directives tab explains itself between sessions (part of #1171). For most of the month it only said "The council is not in session". It now also says what directives are (the guild's monthly push, open for the first 6 days of each month UTC, with tiers paying Scrip to every member and Fund to the guild), and that between sessions your leader and officers can pin standing orders at the top of guild chat. A dedicated leadership bulletin is not part of this release; #1171 stays open for it.

Fixed

• Unmuting a paused battle replay no longer starts its music over a frozen battle; the music now waits for Play.

1.353.7 — 2026-10-02

Fixed

• The guild season panel no longer says no season is running on the day the season rolls over (#2584). When two players opened the panel at the exact moment the eight-week season ended, one of them could get a "no active season" error until they refreshed. Both now see the new season.
• Deleting an account no longer fails when a fleet lands at the same moment (#4219). A moderator's delete, or a player deleting their own account, could collide in the database with one of that player's fleets arriving, and the delete answered with an error. The account was left untouched, so nothing was half-deleted. The delete now retries once and goes through. The player's own deletion was already meant to retry, but it could not recognise this kind of collision, so the retry never ran.
• The Dark Matter sweep no longer gives up on a colony after a database collision (#5875). The sweep credits Dark Matter to colonies every five minutes. When it ran into a database deadlock with another write, it skipped that colony until the next sweep. No Dark Matter was lost: the next sweep paid the whole period, including the skipped five minutes. The sweep now retries on the spot. The server log also says what each failure costs (a deferral, contention, or a real fault), where it used to report every one as "DM credit failed". This makes the sweep survive the collision; what it collides with is not yet identified.

Also

• The same database collisions were going unrecognised in several other places in the server, because every check for them looked for an error code the database library never actually sends. This release fixes the three places above. The rest are tracked in #5948.

1.353.6 — 2026-10-02

Changed

• The Guild starts speaking Italian (#5943). Before, players with Italian on saw the whole Guild tab in English. Now its core reads Italian:
  • the tab and its doors;
  • the guild directory and joining;
  • the roster and ranks (Capo, Ufficiale, Membro, Recluta);
  • the guild's profile, entry rules, emblem and automation settings;
  • the guild log.

  In English, every word and figure reads exactly as before.

  Still in English for now: the Treasury, the Help board, Diplomacy, and the war, League, outpost, raid and research screens. The Treasury, Help board and Diplomacy come next; the rest follow. The beta notice says part of the Guild is still in English.

  A custom rank title can no longer copy another rank's Italian name («Capo», «Ufficiale», «Membro», «Recluta»), just as it could never copy the English one. Titles saved before this change are left as they are.

1.353.5 — 2026-10-02

Also

• Nothing in play changed. This release corrects comments and tests that described the game wrongly beside code that was right. No rule, number or screen moved.
• War verdicts (#2236). The note on how a won war is counted now gives the real reason one verdict's records share a timestamp. Measuring it turned up three old wars, from before v1.72.1, that the guild "wars won" count still counts twice. That is filed as #5947 and is not fixed here.
• Spectral Bulwark (#3632). The engine's description now says what it does: a lone legendary hull is decided per battle, so one hull in each of two colonies protects both colonies. The rule itself did not change. Whether it should is open again on #3632, because donation packages now also deliver legendary hulls.
• Advisor, Colonization, asset tooling (#3649, #3560, #2622). A test note now points at the decision that closed its question, with its figures recounted. The Colonization cost line cites the design table that sets it. The asset-tool note no longer promises to respect a container CPU quota it cannot see.
• Agents (#5937). Two chat tests that could not fail now check the sentence that tells a stand-in whose question it is answering. Two notes now name the right source and readers of the ship counter table.

1.353.4 — 2026-10-02

Fixed

• The flight time in the phone's target list no longer runs under the Harvest or Attack button (#5642). On a phone, the time beside each target in New mission was squeezed until its digits slid underneath the one-tap order button next to it, at every upright phone width we measured. It now always keeps its full width. When a row is too narrow for everything, the order button, ☆ and × move to a second line under the name instead of overlapping it. Long figures in a row's second line, such as the stores a spy report saw, also stay inside their column now.
• The quick-button chips in Settings → Command are full-size touch targets (#5504). All of the tab chips you pick your quick buttons from were 34px tall, below the 44px every other control on that screen uses. They are now 44px.
• Scrolled text no longer shows through above a sheet's ✕ (#2468). In Dispatches, and in the other pop-up sheets with a close button, scrolling left a thin strip at the top of the card where half a line of text showed above the close bar. The bar now sits flush with the top edge, and nothing scrolls above it.
• The Base tab's building list no longer shows rows above its filter bar (#2470). On a desktop, scrolling the building list left a 10px strip above the Sort / All / Production / Military / Civic / Locked bar where building rows passed through. The bar now pins flush with the top of the list.

Also

• The test tool that lays out these screens in a real browser no longer names a folder inside the project for its build output (#5511). Nothing a player sees changes.
• Still open: a check of the game's other sticky bars for the same strip as #2468 and #2470 (#5941).

1.353.3 — 2026-10-02

Nothing changes in the game. This release corrects the tooling we ship with, plus one line in the deploy template.

Fixed

• The ship gate no longer quotes a stale cost for the Python test suite (#4351). Its documentation said a full run took 820 s, or "30–50 minutes", depending on where you read it. Both had gone stale, and on 2026-09-21 four ships read a healthy run as a hang because of them. The cost has moved in both directions, so it is now measured rather than quoted. The gate's header gives the command that reads the current figure from the runs the machine just finished. It keeps the dated samples, each with its test count and the conditions it was taken under. Every other mention says "tens of minutes" and points there. Still open on #4351: the gate printing each leg's elapsed time next to an expected figure.
• The ship instructions no longer claim review agents cannot change files (#4228). They said the review lenses' read-only mode was "enforced by tooling". It is not: those agents can still run shell commands, and on 2026-09-19/20 they wrote into trees that were being shipped. The instructions now say so. The review brief is a prohibition with a fallback, the wording that held when those incidents were happening. It tells the person shipping how to check their own tree.

Changed

• A dirty tree now shows what is in it (#4287). Before, the ship pre-check stopped on a dirty working tree with a list of file names only. Now it first prints a UTC-timestamped record: the stage, the commit, the staged changes and the unstaged changes, line by line. The full record is saved outside the checkout. The stop itself, its message and its exit code are unchanged. Still open on #4287: a record at every later stage, and a record for an edit that slips in after the pre-check passed.
• The deploy template records that interception is live (#1483). Since 2026-10-01, production has run with fleet interception switched on. The repository's compose default would make a reader think it was in shadow mode. The example environment file now names INTERCEPT_MODE=on and says why.

1.353.2 — 2026-10-02

Also

• Commanders standing in for one another are told apart by identity, not display name (#4340, part). This affects only galaxy chat's unprompted-reply mode, which is off by default. Commanders there speak only when a player addresses them, so nothing changes in chat today. If that mode is switched on, a commander answering in place of a same-named one is now told the question was meant for someone else. Three related places that still go by display name remain open on #4340.

• Guild chat's "said a moment ago" window can now be tested (#4286). Commanders' guild lines skip anything that would echo the last two hours of the channel. That window was measured on the real clock, so no test could exercise it; it now uses the time the commander's turn started. In production that start can be some way behind the moment the channel is read, since a turn includes other guild work and its replies. That can only widen the window slightly, holding back a few more near-echoes. Nothing a guildmate has just said is ever ignored.
• Two hand-copied engine tables are now checked against the engine (#4319, part). The ship counter table commanders use to pick matchups and escorts, and the three Pirate Beacon thresholds, are compared with the game's own values on every test run. Both agree today. The other mirrored values listed on #4319 are still unchecked.

1.353.1 — 2026-10-02

Changed

• The Colony tab now shows your energy margin (#2208). The energy line on a colony's card used to read only production / consumption, such as ⚡ 1110 / 973, while the top bar and the Overview showed the margin. It now leads with the same signed margin, ⚡ +137 · 1110 / 973 (or −40 in a deficit), computed the same way as the top bar. For the same colony state the two always show the same number. The order of the two figures is unchanged: whether to flip it to consumption / production is still an open question on #2208.

Fixed

• «+ Build new» on the Colony tab opens the buildings list on a phone (#1135). On a phone the button used to switch to the Base and leave you on the bare map, so you had to find the Build button again to see what you could build. It now opens the Base with the Buildings list already up. Pick a building there and you go straight to placing it. Desktop is unchanged: the list is always on screen there.

1.353.0 — 2026-10-02

New

The Merit Market opens under Markets: the Crucible's Merit now buys time, cargo, regalia and an Annihilator (#5924)

Before, Merit was spent from four rows at the bottom of the Crucible tab. Now it has a market of its own, the third under Markets, with five shelves:

• The Trophy Dock. One Annihilator for 100,000 Merit, delivered to your mother colony. One a week.
• Supply Depot. Chrono Shards for 80 to 2,400 Merit, which count toward your weekly shard pool, and sealed crates of 1M, 5M or 10M of a resource you choose (five, two and one a week).
• Star Forge. Choir Resonators, Foundry Ingots and Exarch Sigils at 1,600 Merit each, inside the weekly allowance they share with bastion drops.
• Arena Services. The new opponent list, the Edict reset and the rename token, at the prices they had.
• Regalia. Three new avatar emblems, two avatar frames and a chat name colour that only Merit buys, beside the Crucible's anvil. Wear them from your profile.

Every purchase takes two taps: the first arms it, the second spends.

Changed

• The Crucible tab no longer has a shop. A signpost where it was opens the Merit Market.
• The Market and the DM Market each have a door to the Merit Market on a wide screen.

Also

• An Annihilator bought with Merit cannot fight in the Crucible: legendary hulls are still refused there.
• An installed app that has not updated yet keeps its four old rows, and they keep working.

1.352.4 — 2026-10-01

Changed

• Reports are in Italian (#5874). Before, players with Italian on read their reports in English. Now these read Italian:
  • the spy report;
  • the battle report and its Details panel;
  • the battle replay, with its captions and finale card;
  • the Activity tab's frame and every military row: battles, raids, spying, defence, missions, supply lanes, homecomings.

  Ship names stay in English, as everywhere else in the game. In English, every word and figure reads as before.

  Still in English: the other Activity rows (guild, economy, season, market, the Marshal strike), the one-line summary the server writes on each battle report, and old reports that were stored as text. Each of these moves with its own phase. The beta notice now says some Activity rows are still in English.

1.352.3 — 2026-10-01

Fixed

• The game now tells you a relic garrison expires, and that you have to send another (#5657). Before, the Relics panel said only "Keep fleets stationed there or somebody will take them", which reads as an instruction to park a fleet once. Nothing anywhere said that a garrison stands a watch chosen at launch, up to 24h — 48h for an empire with the Long Watch doctrine — and then flies home. A guild read the 24h ceiling as a limit on what they were allowed to do, asked in guild chat how to send defences permanently, and lost three relics waiting for an option that does not exist. Now the panel states the rule under every relic your guild holds: the watch runs out, it has to be sent again, and letting it lapse does not lose the relic — it stays yours until somebody wins a battle on it, though with no fleet on station only its turrets defend it. The launch composer says the same thing beside the watch slider itself, where the duration is actually chosen, and adds that a relic cannot take a Redeploy. The hours in the sentence are read from the rule the engine enforces, so they cannot drift from it. Nothing about how relics work has changed — only what the game tells you about it.

  Left out deliberately, and filed as #5922: a countdown under each relic saying when that guild's current garrison there runs out. The relics list the server sends carries no garrison or expiry information at all, so showing a real per-relic clock needs a server change first, and this ship was the rule rather than the clock.

Also

• Agents no longer go quiet about one rare defence they cannot get past (#5311). When an empire's scouting showed three things together — ships under a shield Bubble, an Aegis Bastion of the defender's that was away but could be back, and a tree holding the Vanguard — the strike was refused and the muster simply ran its three days out and closed as "never launched", naming no cause. The pairs each had a sentence; this one combination had none. Now it is named like the others, with the Bastion hedged as away rather than described as covering ships it is not standing over, and the empire leaves that colony alone until a fresher probe shows any one of the three gone. No live empire has reached this case yet, so no standing row changes its mind.

• The first day of an agent's two "called off" notices about a legendary hull is now dated (#5454). Those two notices are printed for about a day, then a shorter standing note takes over for the rest of the week. The standing note said when the scouting looked; the first-day one did not, so the intel appeared to gain precision as it aged. Both now carry the same instant, read from the same report, so they can never name two different probes.

• Two stray marker lines removed from the changelog's 1.323.4 entry (#5495). They were HTML comments, so nobody reading the changelog ever saw them; they were noise in a file people read.

1.352.2 — 2026-10-01

Fixed

• A guild member's avatar initial is a whole letter again (#5881). The circle took the first half of a character rather than the first character. A name beginning on one of the rarer alphabets — Deseret, Gothic, Old Italic — drew the grey box a browser shows for a broken character, and a name beginning on a letter whose accent is stored separately lost the accent. Both are names the game accepts, so both are names a player could really hold.

• Italian at exactly a million now reads as Italian (#5851). Italian puts a «di» between an exact million and what it counts — «1.000.000 di navi» — and sixteen lines were missing it: fleet and hull counts, map distances, anomaly signals, Relic and Bastion tallies, picket emplacements. Worth being straight about how rare this was: that wording is chosen only on an exact multiple of a million, so 1.000.001 and 1.500.000 both read correctly before this change, and two of the sixteen could not be reached at all. Almost certainly nobody saw it. The point is that the Italian now says one thing everywhere, so the next person reading it does not take a correct line for a mistake — which is exactly what happened, and is how these sixteen were found.

Changed

• Five text-box limits stop being written down twice (#5881). The chat-mute reason, the flagship name, the chat message box and the guild founding name each had their character limit typed once where the server checks it and again in the box itself. The numbers happened to agree, and nothing would have said so if they stopped. Each is now read from one place by both sides. No limit moved: every box accepts exactly what it accepted yesterday.

Also

• The agents' planner alarm was checked and deliberately left alone (#5881). The operator alarm that fires when the agents' planner keeps failing was flagged as needing the same careful text-cutting the player-facing text got. It turned out to carry no player or agent writing at all — only an error kind, a model name and a counter — so it was left exactly as it is, with the reasoning written down and a test that fails if that ever stops being true.

• Four text boxes still store damaged text, and this release does not fix it (#5914). The original report listed nine boxes. Four of them accept and store text the browser has already cut badly, because nothing on the server checks the characters after it has checked the length. That affects stored fleet and formation names, mute reasons and dispatch button labels. It is a data problem rather than a text-tidying one, so it is tracked separately rather than folded in here. Nothing in this release changes it.

1.352.1 — 2026-10-01

Also

• The check that keeps a changelog entry readable had stopped running, and said so in a way nobody had to act on (#5899). Before every release, the ship machinery checks that the new changelog entry opens with a headline a player can read and that it links the issue it came from. Since v1.351.5 that check could not run at all: the file it loads had picked up a dependency that is not available where the check runs, so it died before reaching a verdict — and a missing verdict was treated as "could not tell", which only printed a warning and let the release through. The check now lives in a file that has no dependencies to lose, and it proves it still works — by refusing a deliberately broken entry — before its answer on the real one is believed. If it cannot do that, the release stops instead of warning.

• The v1.351.5 entry on the Changelog page has been corrected. Its first bullet was plain prose, so the page's own "what changed today" dispatch picked a sentence further down as that release's headline and lost the link to the work it described. The headline now reads "Emoji in chat previews are no longer cut in half", which is what that release was about. No other entry was touched: an earlier reading of this flagged a second one, and re-running the real parser showed that entry was correct all along.

• A rule written beside that check contradicted the check itself (#5901): the comment said an issue reference had to sit inside the bold headline, while the code also accepts one standing immediately after it. The comment was the stricter of the two and it was the one people read, so it sent someone to "fix" a correct entry that players had already read. The code was right and is unchanged; the description now matches it, and the shapes it accepts are pinned by tests instead of described in prose.

1.352.0 — 2026-10-01

New

Hull cards tell you what a ship is worth, right there in the order you are composing (#5863). Picking which hulls to send used to mean knowing their numbers by heart, or opening the Shipyard on a second device to look them up — because leaving the Missions tab would throw away the order you had half-built. Every hull card now carries the figures for one of that class: attack, hull, shield, hold and speed, under the line that says what the hull is for.

They are your real numbers, not the ones from the catalogue. Your research, your tech tree, the refit you fitted, the mastery those crews have earned and the battle stance you picked are all already in them — so a card multiplied by how many you are sending agrees with the totals at the bottom of the table, instead of leaving you to wonder which of the two figures to believe. (The hold is rounded down per ship, so on a mixed fleet the cards can add up to a unit or two less than the total; never more, because the total is the figure your cargo is actually weighed against.) On a defence the shield figure includes your defence nodes, because on a defence they apply. Figures a hull does not have are left out: a Spy Probe shows its hull and its speed and says nothing about guns it has not got.

Fixed

The Fleet Advisor stops telling you a raid "bounced" when it did no such thing (#5867, partly). Attacking a defended colony could produce the verdict "planetary shield holds — the raid bounces". That line was about the defender's Bubble, the pool of hit points a Planetary Shield holds in front of its ships — your fleet arrived, fought, and never got through it. But "the raid bounces" is how players describe a launch the game refuses outright, which is a completely different thing: the timed protection that stops attacks from being sent at all. One sentence, two possible readings, and no way to tell which. It now reads "their Bubble never breaks — your fire never reaches their ships", which can only mean the battle it is actually describing.

Two target screens stop calling the same wall by two different names. The missions target panel labelled a defender's Bubble as "Shield L3" while the target board, for the very same colony, called it "Bubble 3". And the guild's war-targets list — the one screen that shows both kinds of protection at once — printed "shielded" and "Shield L4" side by side in the same colour, which is precisely the confusion that got this reported. Both now say "Bubble", the word every other combat screen in the game already used.

The pace line at the bottom of the War Table now quotes your real cruising speed, not the speed from the ship catalogue, so it agrees with the speed figure on the card of the hull it names.

Also

Nothing you own was renamed. The Planetary Shield is still the Planetary Shield, and every purchased or granted protection still says what it said.

This only fixes half of #5867, and the issue stays open to say so. The complaint behind it is that the word "shield" covers two unrelated mechanics — the building that soaks damage in a battle, and the timed cover that refuses attacks outright — and that is still true everywhere the timed one is named: the status strip, the Orbital Home, the launch warnings and the guild store all still call it a shield. Giving it a name of its own touches around a hundred separate pieces of text, and doing half of them would leave the game less consistent than it is now rather than more, so it is being done as its own change. What shipped here is the one line that was actively misleading, plus the two screens that disagreed with the rest of the game.

1.351.9 — 2026-10-01

Fixed

• The War Room's target list no longer puts enemies you cannot hit at the top (#5856). The "most valuable" ordering treated a target's hittability as one ingredient among four, and the other three together outweighed it. So an enemy sitting behind a post-raid shield, or still inside their first 72 hours of novice protection, could out-score an enemy you were free to attack — if they were on a contested front, had hurt your side the most, and were much stronger than you. The list is sorted by that score, so the board was recommending, at the very top, the strike the server refuses. A target blocked by a shield or by novice protection now sorts below every target the board has no reason to refuse, however weak that target is.

Also

• Targets far below your own Power still sort below the ones inside your band, as before: those strikes are allowed, but they put a bounty on you, so the board does not push them ahead of a free one. The order among targets you can attack is unchanged.
• The board still cannot see an enemy's sleep window, a diplomatic pact or a relic lock — those are deliberately not shown in a target list — so an unblocked target can occasionally still be refused when the fleet is launched. What changed is only that the blocks the board does know about no longer outrank the targets it has nothing against.

1.351.8 — 2026-10-01

Fixed

• The GUEST badge is no longer painted over the resource rail (#5866) — a guest commander with a long name and a guild tag saw the header run …e [STG] (GUES straight across the first resource figure, with neither the badge nor the number readable. The identity row now gives way instead of spilling: the commander name shortens with an ellipsis first, the guild tag goes only once the name has nothing left to give, and the badge is never shrunk or cut. It is the one thing telling a guest that their empire lives on this browser alone, so it is the last thing that should become unreadable.
• The same row on a phone held sideways (#5866) — there the HUD is laid out as a single line, and it failed the opposite way round: the row shrank further than the things inside it could follow, pushing the badge out over the resources. Phones held upright were never affected.

1.351.7 — 2026-10-01

Nothing a player sees changed in this release. It is a number that had been typed into comments and sample loops around the codebase — how many planets a system has — going stale in seven places, and one of those places was quietly putting a wiki page at risk.

Fixed

• The in-game Planets page can no longer lose a row without anyone noticing (#5872). The wiki's table of what each world type is good at is not written by hand: it is built at every deploy by searching the galaxy for one planet of each type and reading what the game gives it. That search was looking at the wrong galaxy. Its limits had been typed in when a system held 8 planets and a sector held 10 systems, and were never revisited — so it swept 8 sectors of the 12 that exist, 12 planets of the 18, and 60 systems of a sector that only has 19.

  The table players have been reading was correct, and it is worth being exact about why: a coordinate outside the galaxy is refused rather than invented, so the over-wide part of the search cost build time and nothing else. The risk was the other direction. All eight kinds of world — four types, each ordinary or rich — happen to turn up among the coordinates that were being searched, and that is the only reason the table was complete. The first time a kind of world existed only out past planet 12 or sector 8, the table would simply have come out one row shorter: no error, no failing test, and a page people plan colonies around quietly missing a world.

  The search now takes its limits from the galaxy itself, so it follows the map when the map grows. More to the point, the thing that used to be luck is now checked: the build names all eight worlds and fails if one is missing, saying which. The page itself is unchanged — same eight rows, same numbers.

• Comments that described a smaller galaxy than the one being played (#5872). Six other places still stated the old planet count, and two were worse than simply being out of date. One file contradicted itself inside a single stretch of code, opening with "a system's 8 planet slots" and saying 18 a few dozen lines below. Another carried not just a stale count but arithmetic derived from it, describing the longest trip inside a system as less than half what it really is.

  All of them now name the galaxy's own figure instead of repeating a number, because repeating the number is exactly what was done the first time: the original comment was correct the day it was written in July and wrong three weeks later, having been carried unchanged through two moves between files while the galaxy grew around it. The one exception is a dated note recording a balance decision, where the old figures are the record and are now marked as being from that day rather than quietly updated.

  Nothing in this part changes what the game does — it changes what the next person to read the code is told.

Four further places that mention the planet count were left alone on purpose and are written up in #5885: two describe the 3D system view, where the figure may be a performance budget rather than a stale copy and needs that question answered first, and two are design documents. #5885 also records one thing this release could not settle — for about 24 minutes one afternoon in August the galaxy was briefly configured with 24 planets per system, and whether any colony was founded in that window on a planet that no longer exists is a question for the live database, not for the code.

1.351.6 — 2026-10-01

Fixed

• The Crucible's one-tap roster fitter no longer spends your tonnage cap on shield tenders. When the fitter had to rebuild a roster around something that can actually shoot, it seated a single warship and then spent every remaining point of the cap on the most expensive hull it could find — and the Aegis Bastion, which carries no guns at all, costs more than both the Voidlance Cutter and the Titan Dreadnought. A hangar of 100 Bastions and 2 Cutters came to the Captain cap as one Cutter and 48 tenders: of 9,994 points, 202 could fire. It now packs the warships in first, so the same hangar fields two Cutters and 47 tenders — twice the firepower, for slightly less tonnage.

• "Fill to cap" no longer hands you a lighter roster than the one you picked. The button was labelled from the meter: red meant "Trim", anything else meant "Fill". But the fitter rebuilds from your whole fleet, not from the roster on screen, so a press from under the cap could come back smaller. At the Ensign cap, hand-picking 4 Aegis Bastions (816 PP) read "Fill to cap" and gave back 756 PP. The label is now worked out from which way the roster will actually move, and there is a third word — "Fit to cap" — for a press that swaps hulls without changing the weight, or that changes nothing at all.

• A roster could sit one invisible fraction over the cap with no way out. The fitter counted the tonnage it was spending slightly differently from the way the meter and the server count the finished roster — the gap is far smaller than a single point, but it is enough. The result was a cap meter glowing red at exactly "400 / 400 PP", a note claiming you were 1 PP over, a live "Trim to cap" button that handed back the identical roster, and a dead Publish. There was no gesture in the editor that got you out of it. The fitter now asks the same question the meter and the server ask, so the roster it offers is one they will accept.

• Two promises the editor was making and could not keep are gone. The empty-roster hint offered to "pick the strongest roster that fits", which the fitter never promised and does not do. And the over-cap note said «Over by 8 PP — "Trim to cap" cuts it for you», which reads as a promise to cut those 8 points; in the reported case it returned a different hull entirely and cut 205. The hints now say what the button actually does.

1.351.5 — 2026-10-01

Fixed

• Emoji in chat previews are no longer cut in half (#5870). A reply quote, the pinned-message strip and a whisper in the conversation list all shorten a long message to fit one line, and when that cut landed in the middle of an emoji it left a fragment that most devices draw as a grey question-mark box. The shortening now always stops at the edge of a whole character, so a preview ends on the last emoji that fits instead of on half of one. The reply quote shortens to four different lengths depending on where it appears, which is why the same message could look clean in one place and broken in another.

• The harder emoji are covered too, not only the simple ones: a family or a couple, a skin tone, a flag, a keycap, and an accent typed as a separate mark are all kept whole or dropped whole. On a very old phone browser the game falls back to a simpler rule that can still split a multi-part emoji into two valid ones — ugly, but never the grey box. (#5870)

• Thirteen places that shorten text now share one rule instead of each writing out its own, so the fix reaches well past the chat previews it was reported against: a saved fleet formation name, a commander name set from the admin console, the operator's reason on a ban or a Dark Matter adjustment, the weekly War Week bulletin, the daily "what changed" dispatch, the tag label on a mention, and the agent chat relayed to Discord. The formation name was the worst of them — it was stored broken, so the fragment came back on every later read rather than only in a preview. Twelve further places that shorten text were found and deliberately left alone, each for a recorded reason; they are listed in #5881. (#5870)

Also

• The notification banner on a phone's lock screen was already correct, and this release does not change what it does. It was the one place that already cut text properly, and it is counted among the thirteen only because it now shares the rule rather than keeping its own copy, so the two cannot drift apart again. Anyone investigating banner text later should know this release did not touch its behaviour. (#5870)

• Three further places that shorten text are recorded in #5881 rather than changed here, because each needs a different fix: the single letter in a player's avatar circle (which needs "the first whole character" rather than a length limit), nine text boxes where the browser does the cutting itself before the game can see it, and one cap on the agent side. The first two can still show a broken fragment, for names and notes that use letters outside the common range — rarer than the chat previews this release fixes, but not impossible. The agent-side one cannot produce the grey box at all; it can only split a multi-part emoji.

• The bug report's own two examples turned out not to reproduce it: both were one character short of the length at which the shortening starts, so the old, broken code returned them untouched. Both forms are now kept as tests — the one that reproduces and the one that does not — so nobody checks the published example against unfixed code and concludes the problem is gone. (#5870)

1.351.4 — 2026-10-01

Nothing a player sees changed in this release: it is two tests that claimed to guard the placement grid and could not fail, and a comment in the code telling the next reader that they could.

Fixed

• The two tests guarding the placement grid's "two halves" now fail when a half is taken out (#5089). When you are placing a building, the game dims the terrain and draws the placement grid over it in a grading chosen for that dimmed ground. The two only work together — that is the whole of #1106, and shipping one half alone puts the grid back to being unreadable on some planet or other — and two tests existed to say so. Neither could fail. One compared a quantity with itself, so its assertion was true by arithmetic. The other built its counterfactual out of constants and never read the opacities the game actually ships, so no change to them could move it. Both stayed green with either half removed, which has now been measured on main rather than argued.

  They each read the shipped half they are named after now, and each goes red when that half is taken out, whichever state the other half is in — all four combinations were run. The screen is untouched: the wash and the grading ship exactly as they did yesterday, and nothing about placement looks or behaves differently. What changes is that the next person to edit either one is told by a failing test instead of being waved through.

  The comment beside the code was the other half of this, and the more dangerous one. It told the reader the test file "fails either way round", and that was not true. Anyone grepping for what guarded the grading found a sentence promising a guard that did not exist — which is exactly what stops a real one from being written. It now quotes both test names in full so the grep lands, names the specific change that reds each of them, and says plainly that no claim about a guard goes in that comment without the mutation being run first.

  The same check caught stale measurements in the comments beside the grading, which had been quietly wrong since the terrain art was repainted in September: a table of per-terrain contrast figures where three of four rows no longer matched the art, and a stated safety margin of 15% that was really 6% and named the wrong planet. Those are re-measured, dated, and marked as a snapshot — the suite is what actually fails when the art moves, and it says so now.

1.351.3 — 2026-10-01

Fixed

• A refused order's red banner no longer follows you to the next tab (#5832). A player on Discord sent a fleet at a target that would be under night protection when it arrived, got the refusal, and then went to look at something else. The red sentence came with him: it was sitting at the top of Activity, over the activity log, reading like a complaint about the log itself. It had nothing to do with anything on screen, and on a desktop there was no way to be rid of it except its own «ok» — the eight-second self-dismiss only ever ran on phones.

  The banner now belongs to the screen that raised it. Leave that screen and the refusal goes with it; stay, and it waits as long as you like, because on a desktop it sits in the page and covers nothing. Everything else is unchanged: the refusal still appears where you pressed the control, it still has its «ok», and on a phone the floating notice still slides away by itself after eight seconds.

  This is not only about night protection. Every refusal the game shows in that banner behaves the same way now — novice cover, pacts, shields, a full build queue, a plain server no.

  One message deliberately does not move: when the server itself stops answering, the notice is not about any screen and it stays until the server is back. And when a refusal is still travelling to the server as you change tab, the answer lands on the screen you arrived at rather than the one you left — it then clears on your next tab change instead of staying forever, which is better than it was and still not right.

1.351.2 — 2026-10-01

Fixed

• The Crucible join form no longer fills itself in with a roster that cannot shoot (#5842). The roster editor does the arithmetic for you: the join form opens already filled in, and "Fill to cap" rebuilds a roster in one tap. For a commander whose warships each cost more than the tonnage cap leaves once a shield tender is in it, both of those handed back a roster of nothing but tenders — a cap meter reading comfortably under the limit, a live Publish, and the ladder's refusal waiting behind it: a roster needs at least one armed ship. Two Aegis Bastions and one Breachhead Monitor under a Cadet's 400 PP cap came out as a single Bastion, while the Breachhead on its own fitted the cap the whole time.

  The fit now keeps something that can fire. When the roster it builds has nothing armed in it, it starts again from the heaviest warship that fits the cap on its own, and spends what is left of the cap behind that. This is the rule the engine's own fitter has had for a long time — a defence may lose its shape to a tonnage cap, never its last gun — and the editor's copy of that idea simply never had it.

  Nothing you pick by hand has changed. A shield tender still has its row, you can still publish one on its own, and the ladder still answers that in words you can act on. That was a deliberate decision when the tender was first offered and it stands: what changed is only what the editor composes on your behalf.

  What this does not change, said plainly. If your colony holds nothing that can shoot, or nothing light enough for the cap you are fighting at, the editor still shows you what you have and the refusal still comes from the server. There is no roster for the editor to build in that case, and pretending otherwise would mean telling you every hull you own is too heavy for the cap — which is not true of a shield tender.

1.351.1 — 2026-10-01

Also

Why two thirds of all Pirate Beacons sit in the four starting sectors is now written down (#3159). Nothing in the game changes: no beacon moves, no mission becomes able to clear one, and the retirement rules are exactly as they were. What changes is that a question players and reviewers kept re-asking now has an answer in the design document instead of in an issue thread.

The observation was real and had been filed more than once: almost every Pirate Beacon in the galaxy is in the cradle, the one region where planet slots are scarce, and nothing a player does can remove one. The answer is that this is the shape of the map rather than a bias. A beacon is seeded one per home system when a commander registers, and the cradle is four sectors of nineteen systems — so 76 home systems means 76 beacons, one each, and the production read of 2026-09-15 found exactly 19/19/19/19. Outside the cradle the count keeps falling, which is the retirement sweep doing its job where veterans live; inside, every one of those 76 systems holds at least one player still below the retirement gate, so the sweep correctly leaves the tutorial target standing. The record also says what was deliberately not decided, and keeps the one caveat the measurement could not close: a drop in the galaxy-wide count is written as consistent with the sweep working, not as confirmed, because the logs that could have confirmed it had already rotated.

The record ships with a test that reads it back, so it cannot quietly rot. The cradle figure is rebuilt from the engine's own constants rather than typed in, which means the day the cradle grows again the recorded number goes red instead of stale, and the decision is pinned against the predicate it is about.

Two measurements of the same quantity, taken a week apart, are kept as two dated facts rather than reconciled into one: a dated figure is a fact, and an undated one is a claim about today.

1.351.0 — 2026-10-01

New

• The Crucible's Merit shop finally has something a top player cannot be refused (#5844). A regular of the ladder reported on Discord that his Merit was piling up with nowhere to go, and he had read the shop exactly right. It held three things, and a good player could be turned away by all three at once.

  The reasons are independent, which is why they can all land together. The opponent-list reroll is dead for everyone in the top 50: up there the list you are shown is already the whole range you are allowed to challenge, so a fresh draw would only sell you the same five names. The other two lift a waiting time — one on the Industrial Edict, one on changing your commander name — and neither does anything unless that wait is actually running. So a settled commander who had not switched an edict in three days and not renamed in a day opened the shop, saw three entries, and could buy none of them. Meanwhile the ladder kept paying: 220 Merit a day at Captain, 500 at Admiral, and 1500 to whoever takes the season podium. The whole shop came to 1540 Merit — seven days of a Captain's income, three of an Admiral's.

  There is now a fourth entry: the Crucible's anvil, 3000 Merit. It is an avatar emblem, yours for good once bought, and you wear it from your profile like the Marshal's laurel. Nothing about you can close it — not your rank, not a cooldown — so the shop can no longer turn a player away empty-handed. It costs a fortnight of a Captain's daily pay, which is one full Crucible season, and it is the dearest thing on the shelf because it is the only permanent one.

  It is deliberately only an emblem. Merit buys choices and never raw strength: no resources, no Dark Matter, no ships, nothing that shortens a build queue — otherwise the fastest way to play Novarchy would be to stop playing it and grind an arena instead. An emblem says you fight in the Crucible and gives you nothing in a fight. It is also not one of the tier crests, because those state the rank you hold, and a rank you can buy would be a lie about the ladder.

  What this does not fix, said plainly: once you own the anvil, the shop can go back to refusing everything. A shelf of permanent things always runs out against an income that arrives every day, and the Marshal's shop has the same problem. Something you can keep spending Merit on is a separate decision, and it has not been taken yet.

1.350.1 — 2026-10-01

Fixed

• Pressing Go after a failed Galaxy read now really tries again, instead of leaving "Slot unknown" on screen for half a minute (#5826). When the game could not read the slot you were aiming at, the target panel said so plainly — and then the obvious way out of it did nothing. Re-picking the row in the target finder, or re-typing the same address and pressing Go, wrote the coordinates the form already held, so no new read went out. The game only tried again on its own schedule, up to thirty seconds later, and until then you were looking at an error with a button beside it that changed nothing.

  That one gesture is now a retry. The error line is replaced by "Reading the slot…", which is true: a read really is going out. Everything else about Go is unchanged — aiming at an address whose slot the game has already read is still a no-op, because the slot is refreshed on its own every thirty seconds anyway and the button has never claimed to be a refresh.

1.350.0 — 2026-10-01

New

• The galaxy Targets board can now sort by the strongest target, not just the nearest and the weakest (#5816). A player asked for it on Discord and gave the reason himself: one might want to go after the strongest players, because the loot is likely higher. He was right, and the game already agreed with him on another screen — the War Room's own target ranking deliberately puts the strongest target in band on top, because that is where the reward actually is.

  Until now the Targets board offered two orders, nearest and weakest, so the Power column only ran one way. That mattered more than a missing convenience: striking far below your own Power is the thing the Bounty charges you for, while striking upward is what pays the underdog bonus and what drops Chronoshards at all. The one order the board gave you over that column put the rows that cost you a bounty at the top.

  There is now a third chip beside them, Strongest. It sorts by the Power rating shown on each row, biggest first, and rows the scan cannot read a rating for stay at the bottom where they already were — the Beacon and the unseen are no more the strongest target on the board than they were the weakest.

  Pair it with the In band chip and you get the list the War Room would hand you: the strongest target still inside your Power band, which is as far up as the underdog bonus keeps growing. Above that band more enemy Power buys you nothing extra and only a longer casualty list, and the chip leaves it out.

  Nothing changed for anyone who does not go looking for it: the board still opens on Nearest, and your chosen sort is still remembered on your own device between visits.

1.349.15 — 2026-10-01

Fixed

The Weekly Chest's warning no longer reads as a miscalculation (#5815). Opening a chest with a nearly full warehouse showed two figures: what it pays, and how much would not fit and be lost. Both were right, and read together they looked impossible. A player with just under 2M Ferrolite and a 55.8M cap was told the chest paid 53.86M Fe — which fits — and then that 16.1M would be lost. He checked the arithmetic and filed it as a broken warning.

It was not broken. The payout line was already showing what would actually land, not what the chest is worth, so the loss was the rest of a roughly 70M haul rather than an overage of the figure beside it. The sentence just never said which of the two it was.

It says so now: "Beyond what it pays, about 16,133,925 resources would not fit in your home warehouse and would be lost." The figure you are shown is what survives; the loss is a separate amount, not a slice of it. With a full warehouse, where nothing lands at all, the opening clause is left off — there is nothing to be beyond, and the row already says the chest would credit nothing.

Contract cards got the same correction. A claimable contract whose reward is partly lost was reporting it with the same wording, and now reports it with the same fix.

Also

No number moved. The clamp that decides what fits is untouched, including its refusal to ever take resources away from a colony already over its cap. This release changes only the sentence that describes it.

The receipt after you open a chest still words it the old way ("Collected: … · Warehouse full: … lost"). That line was left alone deliberately: "Collected" already names what actually arrived, so it does not invite the same wrong subtraction. If it turns out to, it gets the same clause.

1.349.14 — 2026-10-01

Changed

Your attacker's battle row now tells you which colony the fleet is flying back to (#5824). While your ships are on their way home from a raid, the Activity row that counts down their arrival used to say only «Fleet home in 2h 1m» — it named the colony you had attacked, never the one your hulls and your loot were heading back to. It now reads «Fleet home to Aurora (2:9:5) in 2h 1m».

The address was never truly hidden, and it is worth saying where it was: the Fleets tab has shown it on the route line all along, and the battle report's Details panel shows it as a bare set of coordinates with no colony name. What had no destination on it at all was the row that announces the battle — the one you are already looking at when the question occurs to you.

The colony's name and its coordinates, because two of your colonies are allowed to share a name, and the coordinates are the only thing that always tells them apart.

Also

Two parts of the original report are deliberately left for later, and are recorded on #5824 with what would re-open them:

• Battles already in your log still show no destination. A finished row's record is frozen when it is written, so the rows already in your Activity feed have no departure colony stored and will keep reading exactly as they read today. The new line appears on raids whose fleet is still in the air — which is when the question is actually being asked.
• The Details panel still shows the bare coordinates. That panel lists every commander in the battle, and a departure address is deliberately shown only to the people who fought — so putting names there is a decision about what the game discloses, not a wording change, and it should not be made inside a copy fix.

One honest limit on the new line: it names the berth only while that colony is still yours. If it is destroyed while your fleet is in the air, the row keeps naming it for up to half a minute before falling back to the old sentence, and your ships are re-berthed on your oldest colony. The row never guesses a destination it cannot read.

1.349.13 — 2026-10-01

Fixed

• The Crucible's counter read no longer puts a number on a hull that cannot shoot (#5804). The target-lock screen names your best and your worst hull against the defence you are looking at. That line was scored over every hull in your roster, and the Aegis Bastion carries no weapons at all — so the screen was telling you the Bastion "lands ×1" on the enemy stack, putting a figure on a hit that cannot happen. A player reported it plainly: it doesn't land anything, because it's a defensive ship.

  The quieter cost was what that pushed off the line. The whole point of the read is to tell you which of your hulls is the wrong pick for this defence, and an unarmed hull always ranks last — so the worst hull that can actually shoot, the one you would have acted on, never got named.

  Both ends of the read now come from hulls that can fire. The Bastion keeps its row in your roster and its weight in the tonnage meter, because it really does take up room under the cap — it simply no longer advertises a multiplier it cannot deliver. A roster with nothing armed in it now shows no counter read at all, rather than a false one.

1.349.12 — 2026-10-01

Also

• The Wraiths' documentation stops telling the next maintainer that a person is in the operator's seat (#5234). Nothing in the game changes. The thirty empires of the hostile faction are moved by a scheduled model session — the owner confirmed that on 27/09 — and five sentences across WRAITHS.md and the wraiths package said the decider was a human being: «do not move unless a person decides they move», «an order file a person writes later», «a person chose every attack». Each is corrected to the property the code actually defends, which the answer leaves untouched: the faction moves on a deliberate order file and never on the autonomous runner's timer, and that is as true of a scheduled pass as of a person. Each correction records what it used to say, so a reader who finds the new wording surprising can see it was a correction and not a paraphrase. A sixth sentence needed no correcting — it had already been fixed, from a different direction, by the work that gave the Wraiths a chat gate.
• A pointer that had never resolved since the faction's first commit now resolves. The package sent its reader to DOCTRINE.md for the operator's doctrine. No such file has ever existed anywhere in this repository's history. It points at apps/agents/WRAITHS.md now, which is the document it always meant.
• What the operator's half-hourly briefing actually contains is written down for the first time, and it is not what the privacy audit assumed (#5234). For up to five named players — the top of the leaderboard, the only slice the faction is licensed to attack — it carries commander name, home colony coordinates, power rating, colony count and strike history. The player ids, which the audit was about, appear only for the at most three targets actually being proposed. The home coordinates had never been written down anywhere, and they are the one field on that line that is actionable against a player outside the game.
• Still open, and this release does not close it (#5234). The privacy page's list of what the game sends to an AI service ends «That is the whole list» — a closed claim, which one missing use makes false rather than merely incomplete. The entry for this use is not written, because it has to name the provider and that is not something to guess at. Nothing on the privacy page changed in this release, and the issue stays open.

1.349.11 — 2026-10-01

Changed

One open stockpile no longer outweighs a week of fighting (#5768). On the weekly War board, raiding an undefended stockpile paid more than winning battles did, and it paid without any ceiling. Every other way of scoring on that board is capped; loot was not. A single 50,000 haul was worth 500 points — five times everything the whole destroyed-value term can pay in a week, no matter how much fighting it took to earn.

Loot is still paid in full up to the same bound the value term is capped at. Past that it compresses: the same 50,000 haul now pays 347.2 points instead of 500, about 3.5 times the cap rather than 5. The bigger the haul, the harder the compression — the curve meets the straight part exactly, so there is no step and no threshold worth landing just past.

Below that bound nothing changes at all. A modest raid pays precisely what it paid before. What is gone is the one-raid week outscoring a week of war.

The season and Event boards are untouched. This is the weekly War board only; the other two still pay loot the old way.

⚠️ What this does not fix, and it is a new problem rather than a leftover (#5768)

The bound is applied per battle, so splitting a haul defeats it — and the change makes that worth doing in a way it was not before. Carrying 400,000 away in one raid now pays 1,164.9 points. Carrying the same 400,000 in forty raids of 10,000 pays 4,000, in full, because every one of them lands exactly on the bound — plus forty won-attack bonuses instead of one.

Splitting a haul already bought the extra flat bonuses before this release. It now also restores the entire loot term. Whether that is actually worth the trouble depends on travel time and cargo space, which have not been measured. Closing it needs the board to remember loot across a week or a target, which is a second design decision and not a correction to this one.

Fixed

• The design note explaining the loot term no longer ends on a question it has answered (#5768). It described capping loot as "the open design question" for whoever came next. For the weekly board that question now has an answer twenty lines below it, and the note says so instead of sending the next reader looking.

1.349.10 — 2026-10-01

Fixed

• The Crucible no longer refuses a duel while the cap meter on the same screen says you have room (#5798). A commander locked a target, was told the roster was above the tonnage cap for that fight, and read the cap meter directly below that message: 21,700 / 22,000 PP · 99% of the cap · 299 PP still free. The refusal's own wording sent them to that meter to find out by how much — the meter that said they were fine. There was no way out of the loop from inside the screen, and they could not fight at all. The cause was one hull: the Aegis Bastion, which the server never priced for the roster editor because the editor's figures were only ever sent for hulls that shoot. The meter uses the server's own per-hull figures when it has them for every hull in the roster, and falls back to a local estimate when it does not — so putting a single Bastion in the roster did not cost that one hull's precision, it quietly took every other hull's refits and ship stars out of the total. The meter then read low, the server counted the real figure, and the two disagreed. The Bastion is now priced like every other hull the editor can field, so the meter and the server count the same roster the same way.

• And when the meter cannot be sure, it now says so instead of printing a confident number (#5798). This is the half that matters if a figure is ever missing again for some other reason. In that state the meter no longer prints «299 PP still free» or «99% of the cap»: it marks the figure as an estimate, says that ship refits and stars are what it cannot see, and the over-cap refusal stops telling you to go and read it. It does not tell you which way the server will differ, because that cannot be known — a refit can make a hull count for less as well as more.

Also

• Nothing changes for an app or a browser tab that has not updated. The fix adds per-hull prices to what the Crucible screen is sent and removes none, and the refusal's code on the wire is exactly what it has always been — only the sentence shown for it can differ now. An older client reads the figures it already understood and shows the wording it already showed.
• Shield tenders were already allowed in a Crucible roster and already had a row in the editor; what was missing was only their price. Hulls that bring neither guns nor shields to a fight — freighters, probes, arks — are still not offered and still not priced.
• Sparring is untouched. The hull filter the Wraith partner is built with looks identical to the one that was fixed and answers a different question — what an empire brings into a battle, rather than what one hull costs against a cap — so widening it would be a balance change to sparring and is deliberately left for a decision of its own.

1.349.9 — 2026-10-01

Nothing a player sees changed in this release. It is about what the build's own report says, and about one thing it used to leave blank.

Fixed

• The build now says whether its translation-coverage check judged the run, declined to, or never spoke at all (#5762). Before every release a check walks the whole interface in a fake language and refuses the build if any piece of text was never shown on screen — that is how a sentence nobody translated gets caught before a player meets it. The check already said, in one line, which of three things it had done: judged the run and passed, declined to judge it because the run was only a slice, or failed it. That line never reached the build's own report. The report quoted the test counts and nothing else, so anyone reading it — or any summary built from it — found the coverage field empty, and an empty field reads as "nothing to report", which reads as fine. One such summary ended with coverage judged: and nothing after the colon, on a run whose own step log said the check had judged 1476 pieces of text across 19 groups and passed. All three verdicts now travel into the report word for word, and when there is no verdict the report says so instead of saying nothing. A missing verdict on a run that got far enough to have one is treated as a failure, because a check that did not run is not a check that passed.

Also

• The same line now appears in the pre-merge report as well as the branch one. It had to: that report is assembled from a filtered copy of the build's output, so a verdict quoted into a place nobody reads is the same blank field one step further along.
• Seen working on the very run that tested it: the gate for this change reported apps/web test: key coverage: judged 2943 keys in 39 namespaces on its own stdout, beside the test counts, where before this commit that line existed only in a step log nobody quotes.
• The two states this does not catch are written down beside the code rather than left to be rediscovered. A failing test of the check's own logic can print the verdict's wording as part of its own output and be mistaken for a verdict — it cannot turn a failing build green, because the build is already failing for that test. And the "no verdict" wording misdescribes one rare case, where the build tool drops the package name from every line it prints.
• A note in the same file turned out to be false and was corrected while it was being read: a paragraph claimed a certain count is the evidence that raising a neighbouring count is safe, because it does not move. It had moved for two of the three changes it covers, this one included. When that number moves it certifies nothing by its value, and the paragraph now says so. That prose was not part of this change — a change can falsify prose it never touched.
• Nothing in this release enters the game's servers or the app: the one file it touches is build machinery, so there was nothing to deploy and the deploy step says so.

1.349.8 — 2026-10-01

Fixed

• An address you type into the mission tab's target field can no longer pass for your target (#5812). That field is a search box: typing 4:19:11 offered a «Go to 4:19:11» button, and your fleet's aim only moved when you pressed it. Until then the mission was still pointed wherever it was pointed before — while the address you had just typed sat on screen, looking exactly like the thing you were about to attack. A player lost two fleets and an irreversible 300 DM bounty to it: he typed an address, did not press Enter, launched twice, and had no recalls left by the time he noticed. Three things are different now. The launch button names what it will actually hit — «⚔️ Send attack → 2:7:3» instead of «Send attack» — so it cannot be read as naming the address in the box above it. The box says what it is holding while that address is not your aimed one. And the launch is refused while the two disagree, with both addresses in the reason, at the button that is refusing: you have typed 4:19:11 but you are aimed at 2:7:3 — aim it or clear it. A warning was not enough, and the player who reported this said why: the old gold line at the bottom of the form «blends in with all the other yellow text».
• On a phone the same mistake is stopped a screen earlier (#5812). There the target field and the launch button are on different screens, so the typed address would not even have been on screen when the fleet left — the same mistake with its one clue removed. Moving forward from the target screen now stops while the box holds an address that is not the aimed one, says so in the same words, and withholds the «spy it first» shortcut beside it, which was a second way past the same door.

Also

• What this deliberately does not do. It does not ask you to confirm anything through a system dialog, and it does not quietly aim the typed address when you press Launch — that would trade one wrong-target launch for another, and would make the search box silently decide where your fleets go.
• The warning the reporter actually asked for is NOT in this release, and that is on purpose (#5828). He asked to be warned before a launch that would jump his bounty by tiers. It is worth having and it is now its own issue. It is left out because this fix prevents every wrong-target launch, including the ones whose price is a lost war rather than a bounty, while a bounty warning prevents only the expensive subset — and a confirmation that under-warns teaches players that the dangerous launches are the ones that ask. Predicting the tier jump correctly is real work and should not ride along inside a different fix.
• Also still open from the same report, and not addressed here: a faster way to pay a bounty down (#5812).
• The mission tab's number and paste fields were never part of this: they have always taken effect as you type. Only the search box behaved this way.

1.349.7 — 2026-10-01

Fixed

• A refused Chrono speed-up now says which rule it broke (#5787). Five different problems used to produce one sentence. A request the server could not read, a shard size it does not know, a count that was not a whole number, more shards of one size than a single batch can carry, and a spend holding no shards at all all answered «That combination of Chrono shards could not be read. Reload the page: your client is out of date.» For one of them that advice was wrong in a way that mattered: when a batch asks for more than 999 shards of one size, nothing was unreadable — the combination was read perfectly and refused by a rule, so reloading changes neither the pile nor the rule. That refusal now names the real limit and the two ways past it — the same time in larger shards, or more than one batch — in the same words the speed-up sheet's own cap warning already uses, so the sheet and the server never describe one bound in two voices. A spend with no shards in it now says plainly that nothing was spent, instead of blaming a client that sent a perfectly readable request. The three genuinely malformed cases keep the sentence they had, because «that could not be read, reload» is the true and useful thing to say about all three.

Also

• Nothing changes for an app or a browser tab that has not updated. The refusal's code on the wire is exactly what it has always been, and the extra detail travels beside it — so an older client reads the code it already knows and shows the sentence it already showed. Nobody sees a raw error string.
• The five causes stay distinguishable in the server's answer even where two of them share a sentence, which is what anyone reading a log needs.
• One note on the checks for whoever reads this later: the merge check went red once before this shipped, on an unrelated test. It is a known nondeterministic assertion (#5800) that has produced both outcomes on a single unchanged commit, test-only in origin, and #5800's fix removes the race. This change names neither that test nor its subject. The load on the failing run was never measured and cannot be recovered, so nothing here claims load as the cause.

1.349.6 — 2026-10-01

Nothing a player sees changed in this release: it is a test that was failing for a reason the failure message denied.

Fixed

• An intermittent test failure stops blaming the box for a wait nothing could satisfy (#5800). One mission test red the build about once a night and passed on a re-run of the same commit, reporting that a read "was never dispatched" after ten seconds — which reads as an overloaded machine. It was not one. The mission form opens already aimed at the player's home, and the test aimed at home, so it asked for the same address it already had; nothing changed, so no new read was scheduled, so the wait could never end. On a quiet machine an unrelated read landed in time and the test passed by luck. The test now aims somewhere else first, and the helper refuses immediately — naming the real cause — instead of waiting ten seconds to report a false one. A failure message that blames the environment for a deterministic bug costs more than the bug: it sends everyone who meets it to the wrong place.

Also

• Where the evidence for this sits, in three tiers rather than two. The deliberate 400 ms delay is the proof: it drives the failing condition directly, so any slower-than-250ms mount is covered by construction rather than by sampling. The pre-merge check is one observation in the regime that actually fails — it runs the whole front-end suite against the merged tree, and every failure ever seen for this test was inside a full suite, which the isolated runs and the repeated loaded runs both are not. And the twelve loaded runs are corroboration outside that regime. One full-suite observation is worth more per run than twelve outside it, and it is still one.
• For the record, because the figure looks like an omission and is not: that pre-merge check selected zero Python modules, which is the scoping rule answering correctly for a change that touches only the front end. It is not a skipped step and not a failure to scope; the Python evidence for this release comes from the branch check, which ran them.

• What this does not cover is now named for what it is, with its reach. One case remains, and it is not a misleading failure like the one above — it produces no failure message at all: the wait ends early and silently, and whatever the test checks next either disagrees on its own terms or agrees for the wrong reason. A silent early exit is worse in kind than a wrong message, because a wrong message at least fires and gets investigated. Its reach, though, is far smaller, and it is now stated as something a reader can check against the code rather than as a reassurance about what has been seen: it needs a wait with no settled read before it, which in this file is the first retarget after mounting and nothing else — and even there only if a timer is caught after becoming due but before running. (#5800)
• A stale marker is now cleared between tests. No test reached it, which is exactly why it was worth clearing. (#5800)

1.349.5 — 2026-10-01

Changed

The Weekly War prize stops rewarding the biggest Warehouse (#5767). The podium paid a share of the winner's own total storage, so the cheapest way to be paid more was not to win more fights — it was to finish a Warehouse level before the week closed. One empire went from 231M to 3,231M in a single week that way. And because the prize tracked storage instead of points, rank 10, on ninety times fewer points than rank 2, took home more than rank 2 did, and more than ranks 3 to 9 together.

The share is now taken on the smaller of two things: the winner's own capacity, or a server reference — the median capacity of the twenty largest empires being played. Past that reference, a bigger Warehouse pays nothing extra.

Below the reference nothing changes. A small winner is still paid on what it can actually hold, which is what the prize was always meant to do. What is gone is dominance: being the biggest stops paying.

This does not close the whole hole, and the part left open is named. The empire that was pumping has no leverage left at all — it is the largest, and the largest capacity cannot move a median. What is still possible, in principle, is several empires around the median moving it together. That is a narrower risk, nobody has seen it happen, and the fix for it needs a change to how the week is recorded. One change first, then the board says whether the second is needed.

Fixed

• The prize the podium shows is now the prize that arrived (#5767). The strip and the activity feed were both reading the figure the game computed, not the one the warehouse took — so a winner whose Warehouse was already full was shown a number beside a stock that had barely moved. They now show what landed.

  ⚠️ Past weeks are not corrected. The old records cannot be recomputed: the colonies they were priced against have changed many times since, and nothing in them says which prizes were trimmed. Weeks already on the board keep the figure they were written with.

• The line explaining the prize said something that was no longer true (#5767). It read "Shares of the winner's total storage", which stopped being the whole rule with the change above. It now says what the share is taken on, and what happens once you are past the reference.

1.349.4 — 2026-10-01

Nothing a player sees changed in this release: it is entirely shipping machinery.

Fixed

• The ship skill no longer tells a release to publish an unmeasured test count (#5805). When the pre-merge check produced no test totals of its own, the instructions said to "carry the badge's own value" — and followed literally that strips the provenance marker the release script matches on, so the marked form was refused and a bare pair was accepted. The second guard only refuses a pair lower than the one on record, so the unchecked direction was upward: one mistyped digit, or a figure copied from another branch's run, would have published a count nobody measured as the repository's recorded test totals. The paragraph now gives an ordered answer, with the branch's own measured line first and the badge's value only with its marker intact. Nothing had been published wrongly — the ships with carried counts all omitted the argument, and the recorded totals are unchanged across them.

Also

• A comment block in the gate had been misnamed twice, the second time by the attempt to fix the first: it named the wrong one of four sites, then named a different wrong one. Neither mistake could be caught, because the checks there counted the sites and the ordering was carried in prose that nothing measured. The order is now checked by line number, and the labels carry no ordinals at all. (#5802)
• The same block claimed a shell construct bounded its output to one line. It does not, and a second line would land where a forged marker could sit. What actually makes the site safe is who writes the file, and both halves of that are now checked rather than asserted — including on inputs large enough to exercise the mechanism, which the first attempt's were not. (#5802)

1.349.3 — 2026-10-01

Fixed

• The target list no longer offers a one-tap Harvest the server would refuse (#5756). On the phone, the «where to» list marks a coordinate where a battle of yours left wreckage in orbit: a gold row reading ♻ ≈ N in orbit · recoverable ≈ N. That row carried a one-tap Harvest whatever planet the field was above — so at another empire's living colony the tap went straight to a launch the door refuses, and nothing between the tap and the refusal asked whose planet it was. The shortcut is now offered only where the field sits above your own launch colony, which is the one case the list can be sure of. Every other salvage row keeps its gold colour and its exact figure and leads to the order step, which reads the planet first and offers the Harvest there with the reason it was not the leading choice.
• Nothing was taken off the row. The tonnage is real, your battle report says so, and neither the figure nor the recoverable share moved. Wreckage over one of your own colonies was never affected either: it has always appeared as an ordinary Transport row, and still does.

1.349.2 — 2026-10-01

Changed

The Wraiths lose the eight-hour night shelter they never chose (#5777). Every Wraith was unreachable between 00:00 and 08:00 UTC, every day, while operating around the clock. Nobody ever chose those hours for them: the setting exists for empires that sleep, it is aimed at each of theirs when they are created, and for the Wraiths it was simply left at its starting value. Their own controller calls the field meaningless for them, in as many words.

A night window is a trade. You cannot be attacked for eight hours, and you are also not playing for eight hours. The Wraiths had the first half without the second.

So they no longer have one. A Wraith world can be attacked at any hour, and it also loses the dawn volley its guns fired on waking — a bonus timed off a window, for a faction that never wakes. Nothing else about them changed: the same garrisons, the same fleets, the same rhythm of rebuilding.

An agent empire keeps its window, and that is deliberate rather than an oversight. Those hours are chosen for each one when it is created, aimed at when it is genuinely idle, and it really does stop playing through them — so it is paying for the protection it holds. Only the Wraiths were not.

If this proves to make them too easy to farm, the answer will be to give them something back elsewhere rather than to hand the shelter back.

1.349.1 — 2026-10-01

Fixed

• The Targets board now names the tighter starting-sector band (#5708). A player read "In band - no bounty" on a commander's outpost in the Frontier, struck a different colony of that same commander, and came out a Warlord. Nothing had miscalculated: the band a strike is judged against belongs to the sector the target colony sits in, and the starting sectors are judged on a tighter band and charge double. The board's header listed every other thing that can change the price — a wanted head, a silent owner, a free reply you are owed, a Beacon, Bastion or Relic — and closed by saying every judged row says which, so the one rule it left out read as a rule that did not exist.

  The header now names it, in the same words the War Room already used, and says the part that caused the misreading: a strike inside the starting sectors can be priced where another colony of the same owner is free. The closing line also says which row it is talking about, so a verdict on one address is no longer read as a verdict on the whole empire.

1.349.0 — 2026-10-01

New

• The Marshal page now shows who is doing the damage (#5791) — until now a Marshal window only listed its newest battles, one line each: nobody could see where they stood, whether their guild was on the podium, or who else was fighting. A Standings card now sits beside the battles (and, between windows, shows the last window's):
  • Guild podium — the top three guilds by hull destroyed, with how many of their pilots struck, the rest of the guilds below, and your guild with how far it is behind the guild above.
  • Pilots by damage — the whole field, or one bracket at a time. Every badge is a place inside the pilot's bracket (1st, 2nd, 3rd, or the 25% line), because that is how the prizes are paid; a hull bar shows each pilot against the first of the list.
  • Your row — pinned under the list wherever you sit, with the hull between you and the next place in your bracket ("170K behind 3rd in Fleet").
  • On a phone, a Pilots / Guilds switch shows one board at a time.

  The board counts a battle once its fleets are home, leaves out the same accounts the payout leaves out, and marks the podium places the payout itself would give.

1.348.11 — 2026-10-01

Nothing a player sees changed in this release: it is release machinery, and the only people who read its output are whoever ships.

Fixed

• The ship report no longer blames a missing flag that was given (#5771). When a release left the README's test badge alone, its summary said badge: unchanged (no --tests given) — even when the counts had been handed in and the script had deliberately declined to publish them, either because publishing would move an axis of the badge down or because one half had been copied off the README rather than measured. Two states out of three were reported as the third, and they call for opposite actions: a forgotten flag is worth re-running, a declined pair is correct and re-running changes nothing. The row now names the rule that decided — the axis and the drop, or the provenance — and still says "no --tests given" only when that is true. The false sentence and the true one were textually identical, which is why it survived: the line is right most of the time and wrong only for the person looking for why their badge did not publish.

• A --ledger pointed at a directory now says why it was refused (#5742). It used to fail with --ledger refused: and nothing after the colon: a directory passes a readability check, the read then failed, and the script aborted on an unset variable before it could print its reason. That check runs before the tag, so an unreadable reason there costs the entire GitHub trace — tag, Release and changelog entry — for a release whose code is already correct and already deployed. The validator now names the cause distinctly (missing, not a regular file, or unreadable), and no one of its seven refusals can come back with an empty reason.

Also

• <scratch> is now written <scratch-file> everywhere it means the scratch file, in the ship skill, the ledger script, the release script and the rationale. The same documents use the bare token for the scratch directory a few sections away, which is how the wrong argument got passed in the first place.

1.348.10 — 2026-10-01

Changed

• Research is in Italian (#5788). Before, players with Italian on saw Research in English. Now the Research tab reads Italian: the Disciplines and their bench, the Tech Tree with every node's effect, the research queue, the edicts, the Surge and the Apex card. Tech names stay in English, as everywhere else in the game. The Shipyard's unlock routes and the «+ prerequisiti» on the Base, building sheet and Missions tab follow too, and so does the Chrono shards panel on Base and Colony. Research is off the list of screens the beta notice calls still-English.

  In English, nothing else changes: every word and figure reads exactly as before.

Fixed

• The Surge's last moment reads right (#5788). When the Requisition Surge had under a second left, the line read "tripled for another a moment". It now reads "for another moment".

1.348.9 — 2026-10-01

Nothing a player sees changed in this release: it is entirely shipping machinery.

Fixed

• A green pre-merge check no longer reports a file as missing (#5766). Every gate.sh --merge-only run whose Python test plan was not full printed a shell No such file or directory into the log a shipping session reads to decide whether a branch may merge. The run was green and the missing file was correct — the merged tree's test counts are published only on a full plan, because a partial total written into the README badge is a smaller number that looks real — but the explanation went into a step log while the error went to the transcript, so what the reader saw was an alarm with no reason anywhere in view. Now there is no error, and where there is no count pair the run says why on the same output as its verdict. An error line in a log that drives a decision teaches its reader to skim past errors in the one place where a real one must not be skimmed.

Also

• The same defect was present at a second read in the same file, which the report did not mention: the plan token was read the same way, and the one code path designed to leave that file absent was therefore the path that printed the stray line. Both are fixed, and one end-to-end case now drives all of it on a real merge-check rather than on a fixture. (#5766)
• A comment in the gate had named the wrong line as the fourth item its own guard counts, and claimed a cat had been removed that is still there and is the line actually counted. The count was right; the prose sent the reader to the wrong place. Corrected, with the missing naming assertion added so a fifth item cannot join quietly. (#5766)

1.348.8 — 2026-09-30

Fixed

• The Targets board row now says which strike its verdict is about (#5746). A player read "In band - no bounty" on a target in the Galaxy targets list, flew a joint raid on it, and was billed 31.84 Dark Matter. The verdict was correct — about a strike flown alone. A raid is judged on the combined Power of everyone who flies, and that rule was stated once at the top of the board while each row stated a verdict with no scope on it.

  Every judged row now carries the scope: "solo: bounty" or "solo: no bounty". Targets that are free whatever you bring — an owner already wanted, an owner marked silent, or a free reply you are owed — keep saying plainly that they cost nothing, because for those a raid is free too and a scope would suggest a risk that is not there.

• A raid you are already aboard tells you when it stops being free (#5746). It cost nothing when you joined, someone who joined behind you has taken the pack over the band, and nobody can leave a raid once they have joined. That notice appears while you have the panel open; it cannot reach you if the crossing happens after you close the game, which is tracked separately.

1.348.7 — 2026-09-30

Fixed

The speed-up sheet no longer offers a Chrono spend the server will refuse (#5677). One batch of Chrono shards can carry at most 999 of a single size. Neither the recommended spend nor the manual stepper knew that, so on a long build held by a player with a deep pile of 5-minute shards the sheet drew a Finish now button, printed a correct preview beside it, and the tap did nothing at all.

The recommendation reaches for the small shards first, which is what a 5-minute pile is for. A build long enough needs more than 999 of them, and there the button was simply dead — no refusal a player could read, and no hint that the very same purchase would go through in bigger shards.

The sheet now says which it is. Where the same time can be bought in coarser shards, it points at the "Same time, fewer shards" row, which was already there and already shows exactly what it would spend, down to the last shard. Where nothing coarser can pay for it, it says that a batch still shortens the build and the next one finishes it — which is true, and is what a dead button never told anybody. The stepper stops one shard short of what a batch can carry, so it cannot be pushed into the same refusal.

Nobody has been able to reach this yet: it takes a pile above 999 of one size and a build longer than 83 hours, and no player has held both at once. Both halves grow on their own, which is why this is fixed rather than watched.

Nothing about the shard limit itself has changed, and no shard, build time or cost moved.

1.348.6 — 2026-09-30

Changed

Attacking from behind your own night window now costs you that window.

Night protection used to be judged entirely on the defender. A fleet was refused if it would land inside the target's eight hours, and nothing anywhere asked what the attacker was doing during their own. So an empire could raid throughout its own window while staying unreachable for the same eight hours: the victim could not answer inside it, and by the time it closed the raider was already home.

Launching an attack, opening a guild raid or joining one now spends your own night shelter for two hours. Striking is still allowed — the window is a price rather than a gate — and if you launch nothing you keep exactly what you had before. Espionage, a garrison sent to a guildmate under attack, a transport and Pirate Beacon training all leave your shelter standing, because none of them is an offensive act.

Your shelter comes back when the two hours lapse, and every fresh strike pushes the deadline out. So raiding all night means standing exposed all night.

Two hours is the longest flight the galaxy allows, which is what makes the trade an even one: anyone who could reach you at all can commit a fleet while your shelter is spent. A shorter price would be one that most of the map could never collect.

Also

Your chosen night hour is untouched by this. It still sets the quiet hours for your notifications and still anchors the Dawn Watch, and attacking does not move it.

1.348.5 — 2026-09-30

Fixed

• A raid that costs you nothing now says how long that stays true (#5782). Joining a raid on a target inside your own power band was priced at nothing, and the raid door said nothing at all — so the strike read as free. It is not: the bounty is charged on the combined Power of everyone who flies, guildmates keep being allowed to join until the moment of impact, and nobody can leave a raid once they have joined. Players were being billed Dark Matter for a raid that had shown them no price.

  The door now tells you when a zero is a zero for now: it says the verdict is read on the pack's combined Power rather than on yours, how many more contingents of your own size it would take to start costing you, and that one larger ally would do it sooner. Raids where the price could never realistically apply — a strike far above your own weight, which no pack you could assemble would push over the line — stay silent as before, so the warning appears where it means something.

  Strikes that genuinely cost nothing are unchanged and still say so: a target already wanted, an owner marked silent, or a free reply you are owed.

1.348.4 — 2026-09-30

Fixed

The Aegis Bastion can be fielded on the Crucible ladder, and a defence that already held one no longer loses it (#5718). The Bastion fights in the campaign and can be built like any other ship, but it never appeared in the Crucible roster editor: no row, no count field, no way to put it in a roster. There was no refusal and no explanation either — the ship simply was not there, so anyone who owned one could only guess whether that was meant.

Worse, and this is the part nobody reported: if a defence already held a Bastion, opening the roster editor quietly removed it. The next time that defence was published, it went out without the ship. No message said so.

The Bastion now appears in the editor beside the armed hulls, with its own count field, and it costs against the tonnage cap like anything else — at 204 PP it is priced level with the Breachhead Monitor, so fielding one is paid for rather than free. A stored defence holding one keeps it when you open the editor. A roster still needs at least one hull that shoots, and the ladder says so when you publish one that does not.

Freighters, probes, colony ships and the Continuum Hauler are still not offered, and that is deliberate: they add nothing to a duel but hulls to soak fire. What earns a Bastion its place is that it feeds its shield to your whole side each round, which is a real contribution to the fight even though it never fires.

Nothing about the ladder's rules or any ship's numbers changed.

1.348.3 — 2026-09-30

Fixed

A hull drawn as a plain untextured shape instead of its real model now says so, and stops re-downloading a file that is not there (#5716). A player sent a screenshot of one ship rendered as a white slab with a few gold boxes, sitting beside another ship drawn in full detail. The game is meant to fall back to something simple when a 3D model fails to download — a scene should never be empty because a file was slow — but it did that completely silently, so nobody could tell it had happened, and it tried the same missing file again every single time that scene was opened.

The fallback stays, because an empty scene would be worse. What changes is that the game now records which ship could not be drawn and why — a stalled download, a missing file, a server error, or a file that arrived broken — so the next report of this can be traced instead of guessed at. And after three failures for the same ship it waits a minute before trying again, rather than re-requesting on every open. A minute, because that is long enough to outlast a deploy and short enough that a repaired file appears again almost immediately.

Nothing about this changes what a working scene looks like.

1.348.2 — 2026-09-30

Changed

• The Galaxy map is in Italian (#5755). Before, players with Italian on saw the Galaxy tab in English. Now it reads Italian: the map and its labels, the rail and drawers, the map key, each planet's sheet, the Targets board, the system survey, the Signals board, fleets in transit and intercepts, and pickets. If you switch language while the map is open, it follows. The Galaxy map is off the list of screens the beta notice calls still-English.

  In English, nothing changes: every word and figure reads exactly as before.

  Still English on the Galaxy tab: the compass points in addresses (you type them into Go-to), names that come from the game server, and the connection-timeout message, which every screen shares.

1.348.1 — 2026-09-30

Fixed

• A collapsed timeline lane now says when its hidden work lands (#5754). The home screen's timeline draws three bars per lane and counts the rest beside the lane's name. That counter said only how many — +2 — which is the one thing you could already see, since the lane was visibly full. It now says when the soonest of them arrives: +2 · next 3h.

  It reads the same way on every lane that can fill up: your fleets in flight, a colony's building and shipyard work, and your research queue. So a glance at a busy lane answers the question a collapsed lane actually raises — is there anything in there I need to act on soon? — without opening anything.

  Lanes are the same height as before.

  In Italian: +2 · tra 3h, using the same wording the timeline already uses for a clock.

1.348.0 — 2026-09-30

New

You can now see how many fleet recalls you have left today, before you spend one (#5719). You may call back three attacking fleets in any rolling 24 hours. Until now that count only surfaced at the moment it ran out: the figure appeared on the Recall button once the allowance was already spent, and with nothing in the air it was nowhere on screen at all. Planning a raid meant guessing, or keeping the tally in your head.

The recall confirmation now says how many are left before you commit — "2 of 3 recalls left today" — so you read it while the choice is still yours. The Missions tab carries the same line as a standing caption under the fleet board, and it is there whether or not you have anything flying, which is the case the figure used to be missing from entirely.

Two deliberate silences, both worth knowing. The line shows nothing at all while the game is still waiting to hear back from the server, rather than a zero it cannot stand behind. And when the tally it holds might be out of date, it drops the number and says only that at least one recall has come back — because the allowance is a rolling 24-hour count rather than a daily reset, so a deadline that has passed proves one slot returned, never that the day began again.

The limit itself is unchanged: three attack recalls per rolling 24 hours, and calling back a transport still does not count against it.

1.347.11 — 2026-09-30

Fixed

• Harvest is no longer offered where the server always refuses it (#5695). A debris field in orbit used to add a Harvest option to the mission step wherever the field lay — including over another empire's living colony. The launch door then turned the fleet away every time, because a harvest is accepted only above your own coordinates, above an empty slot, or at the Pirate Beacon. At a colony your guild holds a pact with it was worse than a wasted tap: Harvest sorted ahead of Defend and arrived already selected, so the step led with the one verb that target cannot accept while the verb it does accept sat behind it.

  Now the step leads with what the server will take. Harvest keeps its place in the list, one tap away, and a new line tells you why it is not offered first: a living colony holds those coordinates, and the field there belongs to its owner. Your own colonies are untouched — lifting your own wreckage is the commonest use of the verb and was never in question.

  One case still offers Harvest on purpose. A slot that looks empty, where your own spy report found an occupant, keeps the offer: a refusal that explained itself there would give away what the Eschaton Veil exists to hide. That launch is refused at the door, as it was before.

Also

• The finder still offers a one-tap Harvest at another empire's colony (#5756). That surface reads its target from different data and this fix does not reach it, so the issue stays open rather than being folded into this release.

1.347.10 — 2026-09-30

Fixed

• A busy build queue no longer hides a colony's shipyard from the home timeline (#5717). Before, a colony's timeline lane drew its first three bars and stopped — and since building orders come first, three buildings under construction pushed every shipyard bar off the lane. Your slips were running and the timeline said nothing about them: the work was counted in the small +N beside the colony's name, which tells you how many bars are hidden and never which.

  Now every kind of work in a lane keeps a place. A colony building three things and assembling hulls shows a shipyard bar alongside its building bars, and the slots left over go to whatever finishes soonest. Lanes are the same height as before — the timeline chooses better bars rather than more of them.

1.347.9 — 2026-09-30

Fixed

• Two Italian lines on a saved order card no longer stop mid-sentence (#5731). On an order saved before stances and dock raids existed, each line ended on a preposition with nothing after it — and both lines showed at once, two rows apart, so the card broke off twice in the same way.

The English they were translated from leans on an idiom Italian does not share: it ends on a bare preposition but still supplies a word for what is missing, and the translation was left holding the preposition alone. Both lines now use the phrasing the rest of the Italian already relies on for exactly that shape — the wording two other sentences in the catalogue were built on long before these two were written.

Both lines got the same wording, deliberately. They render together on one card, so correcting only the upper line would have left two different shapes for one idea on the same screen, which reads worse than the original break did.

Italian only; the English is untouched.

1.347.8 — 2026-09-30

Also

• Nothing changes on screen (#5724). A test-only ship: the launch form's «Also …» line, which lists what your launch will actually fly, now has a check holding its Dock raid entry to the translation layer. Before, that one word could have been replaced by a hard-coded English string and every suite in the repo would still have passed — so an Italian player could have found «Dock raid» sitting in an otherwise Italian sentence, and nothing would have caught it before they did.

  The check had to be written in Italian and in the pseudo-language rather than in English, because in English the correct code and the hard-coded version produce the same word.

1.347.7 — 2026-09-30

Fixed

Nothing changes in the game for players: this removes a false failure in the automated checks that was costing whole build runs (#5722). Two of the checks that read the launch screens — in English, in Italian and in the test pseudo-language — were failing at random. Nothing in the game was wrong: the checks were waiting a fixed third of a second for the game to look up what sits at a set of coordinates, and on a busy machine that lookup takes longer than that. They now wait for the screen to actually be ready rather than for the clock to run out, so a failure means something is genuinely broken and a build that fails is worth reading.

Three further failures from the same run are not explained. The checks around them are sturdier now, but nothing measured accounts for them, and they may recur.

1.347.6 — 2026-09-30

Changed

• The Shipyard is in Italian (#5735). Before, players with Italian on saw the Shipyard in English. Now the whole tab reads Italian: the ship catalogue and each hull's detail panel, the build controls and the queue, the defense planner, the dry dock, the Star Forge and the ship reference sheet. Ship names stay in English, as they do everywhere else in the game. The Shipyard is off the list of screens the beta notice calls still-English.

  In English, nothing changes: every word and figure reads exactly as before.

  Still in English inside the Shipyard: the defensive-line editor, which is shared with the Missions tab and moves with it later.

1.347.5 — 2026-09-30

Fixed

• A battle at a relic no longer leaves debris that nobody can collect (#5660). The field was drawn after the fight and every attempt to harvest it was refused — for the guild holding the relic exactly as for anyone else. No field is left at a relic now, so nothing offers one.

A relic's guardians are not a commander's colony, and the harvest gate was never meant to let an unarmed recycler scoop a defended site without fighting for it. What it did not anticipate is that a relic slot belongs to no commander at all, so "only its owner may collect" came out as nobody. Before the change, eight live fields sat at relic slots holding fifty-three million tons between them — more than all the live wreckage on every commander's colony in the galaxy put together — shown to four guilds who could not lift a ton of it.

Nothing that could be collected has been taken away: none of it could be collected before either. What changes is that the game stops offering it.

Aim a harvest at a relic anyway and the refusal now names the relic, instead of claiming a living commander lives there. A faction Bastion says the same for itself.

1.347.4 — 2026-09-30

Fixed

• A dock raid in flight is now called a dock raid (#5729). Before, once you launched one it vanished into the In-flight list as an ordinary Attack — the same heading an ordinary attack gets — so from the moment it left the hangar nothing on screen said what you had actually sent. It had never said so: the card has read «Attack» for every dock raid since the order was added in September.

  Now the card's heading reads Dock raid, keeping the attack's red border and its ⚔️ glyph, because a raid is still an attack as far as the target is concerned. Its Recall button says the same thing — it used to offer to «recall the attack» over a card headed Dock raid for anyone using a screen reader.

  In Italian: Raid ai moli, the name the launch form already uses.

  This closes a gap the previous release opened: v1.347.1 taught the saved-order cards to say whether a relaunch would fly a raid, and left the in-flight list mute — so the game told you at save time and went quiet at flight time. Both surfaces now agree.

1.347.3 — 2026-09-30

Fixed

• The Recall button now tells you the day's recalls are spent instead of failing when you press it (#5581). You may call back three attacking fleets in any 24 hours. Past that the button was still lit, and the only way to find out was to press it and read a refusal. It is now greyed out with the count on it — «No recalls left today (3/3) — one comes back in 4h» — so the wait is something you can see instead of something you discover. Nothing about the limit itself has changed, and recalling a transport was never counted against it and still is not.

1.347.2 — 2026-09-30

Fixed

Chat now says when it translated a line, so a line left as typed is no longer a bug (#5656). Chat is translated into English as you write it, and nothing on screen ever said so. Two things went wrong with that. A line the game had translated was shown as the author's own words, which meant the game put sentences in players' mouths. And a line it left exactly as typed looked broken: a Finnish player wrote a mostly-English sentence, read it back unchanged, and could only conclude the translation was working for everybody but him.

A line the game translated now carries a small "(translated)" beside it, and hovering it says these are not the author's exact words. A line left as typed carries nothing — which is the answer he could not get. A quote of a translated message says so inside the quote, and a pinned order says so on the wall, because those are read out of context and an order is acted on once.

Each channel also says what it does to your words: the galaxy and guild channels read English whatever you write and new messages they translated say so; whispers are never translated and never were. Messages written before today carry no mark either way — the original text is not kept, so nobody can say afterwards whether they were translated, and the game does not guess.

1.347.1 — 2026-09-30

Fixed

• «Send again» keeps the dock raid, and the card tells you whether it will (#5608). Before, an order saved with a dock raid armed came back without one: the button whose whole promise is «again» sent an ordinary attack, and nothing on screen said so. Your fleet arrived at a defended dock in plain sight — the one battle a dock raid exists to avoid. Now the raid is saved with the order and flown again with it, and every saved attack card says where it stands:
  • Dock raid: yes — this relaunch will fly it;
  • Dock raid: no — this fleet no longer runs dark, so this relaunch is an ordinary attack — you armed a raid, but the ships the hangar can give you now include one that cannot hide, and a single visible hull uncloaks the whole group;
  • Dock raid: not recorded — this relaunch flies without one — the order was saved before the game kept this, so what you chose then cannot be recovered, and the card says so instead of guessing.

  An attack that never flew a raid says nothing at all, and neither does a fleet that cannot run dark in the first place: those cards are unchanged. The line reads the fleet the tap would actually send rather than the one you saved, so it can never promise a raid the launch would strip.

  Italian players see this straight away rather than staged, since Italian went live in beta with v1.347.0: the card calls it Raid ai moli, the name the launch form already uses.

1.347.0 — 2026-09-30

New

• Novarchy speaks Italian — in beta (#5721). Before, the game was English for everyone. Now:
  • a phone or browser set to Italian starts the game in Italian;
  • anyone can choose «Italiano (beta)», «English» or Automatic (the device's language) from the Empire menu on desktop or the Command sheet on a phone;
  • the choice is saved on your account and wins over the device.

  Login, onboarding, the HUD, Home, the Base and its building sheets, Colony, Overview, the whole Missions tab and every error message are in Italian. The Shipyard, the Galaxy map, Research, reports, Guild, Market, notifications and a few other screens are still in English. A one-line notice at the top says so until you dismiss it. More screens follow in the next updates.

  If the iOS app still opens in English on an Italian iPhone, pick «Italiano (beta)» from the Command sheet: the app learns to follow the phone's language with its next build.

1.346.1 — 2026-09-30

Fixed

The game's name no longer appears cut in half in the top bar (#5667). On a desktop window that was not wide enough, the HUD printed NOVAR — the wordmark sliced mid-letter beside your Power figure. It was doing what it was told: the mark is decoration, so it is the part that gives way when the row runs out of room. But a brand cut in half does not read as a tidy fallback, it reads as a broken screen, which is how it arrived in the bugs forum.

Now the mark gives way whole. The bar measures the space it actually has and, when the full word will not fit, it stops drawing it rather than showing a fragment — exactly what the game already does on a phone, where the wordmark is never drawn at all. Widen the window and it comes back, complete. Nothing else on that row moves, and the Power figure, the gate and your Dark Matter balance are untouched.

Also

The check that catches this runs in a real browser at twelve desktop widths, so a future change that starts clipping the brand again fails before it ships rather than after a player notices.

1.346.0 — 2026-09-30

New

Painted planet surfaces (#5723)

Every 3D planet now wears a real painted surface instead of a flat computer-drawn pattern. In the system view, the planet sheet and the desktop Home's Empire theatre:

• rocky worlds look like orbital photographs of canyons and craters;
• gas giants have swirling cloud belts and storms;
• ice worlds show grooved, frosted plains;
• lava worlds glow through cracks in a black crust.

Each planet keeps the colour, rings, tilt and size it has always had at its coordinates, and always wears the same painting.

Rich worlds are easier to spot. Their glowing deposits now follow the terrain: gold veins on rocky and gas worlds, cyan crystal fields on ice, embers on lava.

While a painting loads, or if it can't load, the planet shows its previous look, glow included.

1.345.4 — 2026-09-30

Also

• Why every AI empire's first line of defence looks alike — answered, with the answer pinned in the code (#5602). Nothing changes in the game. The AI empires defend exactly as they did before, no attack you launch becomes easier or harder, and no AI empire gets a stronger or weaker wall out of this.

  A player reported that all 75 AI empires feel interchangeable where it actually matters — the Bubble, the Batteries and the Sensor Grid that a raider meets first. They were right, and the reason turned out to be the Command Center. No building may rise above the Command Center's own level, and on more than nine colonies in ten the Batteries sit exactly on that ceiling, so whatever temperament an empire is supposed to express there has nowhere left to go. Pushing the AI to want Batteries more would therefore have changed nothing at all.

  The Bubble is the exception, and now it is clear why: it is the one of the three that escapes that ceiling, once its second-tier research is done. Even there the difference is easy to miss, and this is the part that explains what the player saw. A cautious empire does put more into its Bubble than a raider puts into theirs — but a wealthier empire of any temperament still out-builds a poorer cautious one, so standing two AI empires side by side and comparing their Bubbles mostly tells you which of them is richer, not which of them is careful. The Sensor Grid, meanwhile, has room to climb and still must not be pushed: one level more would quietly end cloaked dock raids against every AI empire at once.

  So there is no cheap way to make those three defences feel different, and that is a finding rather than a gap. It is now written where the next person will look, with a test that fails if anyone reaches for the obvious fix.

1.345.3 — 2026-09-30

Also

• The Missions tab is fully ready for Italian (#5720). Nothing changes on screen for English players: every word and figure reads exactly as before. Behind the scenes, the last parts of the Missions tab now read from the translation layer, each with its Italian already written:
  • supply lanes;
  • «Send again» and the saved orders;
  • the fleet ledger, including its heading on the Shipyard;
  • the scouted systems panel;
  • the hyperspace route line.

1.345.2 — 2026-09-29

Also

• The rest of the Missions tab is ready for Italian (#5711). Nothing changes on screen: the game is still English for everyone, and every word and figure reads exactly as before. Behind the scenes, these now read from the translation layer, each with its Italian already written:
  • the checks under the launch form (what blocks a launch and what only warns);
  • the Fleet Advisor's forecast and the quick-attack verdict;
  • the target panel;
  • the War Table;
  • the launch options: warp, raze, sabotage and dock raid.

  Italian is still off until enough of the game reads naturally in it.

1.345.1 — 2026-09-29

Changed

• The Market and the DM Market now share one Markets menu (#5700). On desktop, the rail's Market tab is now Markets ▾, in the same place, and opens a small menu with the Market and the DM Market, the way Events ▾ holds the Season, the Crucible, the Campaign and the Marshal. Before, the DM Market had no entry on the rail at all: the only way in was the door on the Market's hero, which is still there. On a phone, the Empire tab lists both under their own Markets heading, between Events and the other doors. The Market is now two clicks away on desktop instead of one.

1.345.0 — 2026-09-29

New

• The wiki has a Marshal page (#5697). The Marshal went live with only one line in the wiki. /wiki/marshal now explains the whole event, with numbers taken straight from the game:
  • The event: what the Marshal is and where to find it, when it takes the field (the default schedule in Rome time, window length and windows per week), and how the warnings reach you.
  • Striking: how a strike works (strikes per window, the fixed 10-minute flight, no losses, no recall, the last moment you can still launch), and how guild raids work (gathering time, seats, when joining closes, and that a raid uses one of your strikes).
  • Levels and walls: the three walls of every level with their ships and which fleets work on each, how level health grows, and the damage curve that keeps the biggest empires from clearing it alone.
  • Prizes: what makes a strike valid, the level prizes, the crates for each bracket, the personal and guild podiums for each bracket, and the trophies.
  • Rules and the shop: the hunt rule and its bounty, what the Wraiths and the AI empires do during a window, and every item in the Insignia shop.

1.344.1 — 2026-09-29

Fixed

The home screen's threat radar sweeps the right way round (#5644). Its bright edge now leads and the glow trails behind it, the way a radar's phosphor does. Before, the faint end of the sweep reached a bearing first and the bright edge came along after it, which read as the scope turning backwards. It still turns clockwise, and it still parks on the same bearing when your system asks for reduced motion.

1.344.0 — 2026-09-29

New

• Dispatches read like announcements (#5632). An announcement from the team used to be one block of plain text. Now it can carry bold, italics, bullet lists and links that take you straight to a screen of the game. It can also have up to two buttons underneath (for example "Open the calendar" or "See the market") and a header picture. Tapping a button or a link closes the panel and goes there. A takeover card shows the same content, and tapping one of its buttons closes the card and marks it read.
• Inbox tabs. The Dispatches panel is split into All, News, Events and Mail, each with its own unread count. Maintenance notices file under News. "Mark all read" now clears only the tab you are looking at; on All it still clears everything.

Also

• Formatting only applies to announcements the team writes. Messages the game sends you, such as guild notices, still show player-chosen names exactly as they were typed, so a guild called **Kings** is never turned bold.
• Apps installed before this update, lock-screen notifications and Discord posts get the same announcements as clean text, with no stray symbols. Older apps don't show the buttons or the picture.

1.343.1 — 2026-09-29

Changed

• The AI empires have gone quiet (#5680). Empires marked with the AI badge no longer post in galaxy chat, guild chat or whispers, and the galactic bulletin is paused. Before, they answered when addressed, announced raids and wars, and replied to whispers. They still read the chat: a threat still leaves a grudge and still comes back as a real move against whoever made it.
• AI empires rethink their strategy once a day (#5680). Each one now writes its plan (posture, targets, what to build) at most once a day, where before it was every six hours plus after every battle or spy report. A battle, a spy report or a threat still prompts a new plan, at the empire's next daily slot. The plans run on a different model.
• AI-led guilds no longer pay members from the Fund (#5680). A bot guild leader no longer grants money from the guild Fund to struggling members. Bot guilds still found Guild Anchorages, now by a fixed rule: the system they have held longest, and only when the Fund covers twice the cost.

1.343.0 — 2026-09-29

New

• The Marshal has its own event page (#5675). Before, between windows the tab was one line ("The Marshal is not in the field") and a plain shop list. Now it opens on the boss, with new key art of the Marshal: a molten obsidian leviathan with its escort swarm.
  • Between windows: the art is dimmed, and a countdown shows when the next window opens, its day and hours in your own time zone, and how many strikes you get.
  • During a window: the art is lit and glowing. The top of the page shows the level, a bar for the three walls (the broken ones crossed out, the current one draining) and your strike pips. The strike launcher sits right under it.
  • How it works, in three steps, and the three walls: Shield Line, Wraith Escort and Flagship Core, each with its ships. During a window the wall that's up is marked "Up now".
  • What each level pays: levels 1 to 5 for your own bracket: Insignia, Chronoshards, and the crates your bracket can actually win. Your bracket's podium and the top three guilds' podium show each place's full prize.
  • Before you strike: the 25% rule for a valid strike, how long your colony is left open (a strike's 10 minutes, plus the gathering time in a guild raid), and the rule against hunting strikers.
  • The Insignia shop is now a row of cards, each with its icon and a bar showing how close you are to its price.

1.342.2 — 2026-09-29

Fixed

• The level badge no longer covers the Gate line in the desktop header (#5673). At some window widths, the gold commander-level badge on your portrait was drawn over the first letter of the line under your name, so Gate 6.75B read as Sate 6.75B. When the header runs short now, the portrait keeps its size and the NOVARCHY wordmark gives way instead.

1.342.1 — 2026-09-29

Fixed

• The Shipyard no longer opens on one ship and turns into another (#5672). The ship card shows a picture while the 3D model loads, then fades to the model. For four hulls that picture was a different ship: the Shroudrunner and the Colony Ship still showed their old paintings, the Annihilator showed a purple hull that loads as near-black, and the Praxis Interdictor showed a dark hull that loads as light plate. Every hull's card picture and grid thumbnail is now rendered from its own 3D model, in the pose the model opens in, so the hand-off keeps the same ship at the same angle. The replay's fallback billboards of those four hulls were redrawn the same way.

Changed

• Grid thumbnails now all share the card picture's 4:3 frame. Small square thumbnails no longer cut off a hull's nose and wingtips, and on phones a card no longer resizes its hull as it scrolls into the centre.

Also

• The Annihilator now looks as dark in its pictures as it does in 3D; the purple was the mistake.
• The hull still shifts in size slightly during the fade; the ship and its angle now match.

1.342.0 — 2026-09-29

New

• The DM Market now runs on supply and demand (#5669).
  • Before: prices only followed the server's economy. What players bought barely moved crates, and it didn't move cosmetics, Star Forge components, boosts or bundles at all.
  • Now: every price at the DM Market is its base times one daily demand multiplier. That covers crates, Overdrive, Freight Wind, ship bundles, cosmetics and Star Forge components.
  • How it moves: when more is bought than the market supplies, prices rise, about 5% a day at most and up to 4×. When buying slows, they ease back to base and never go below it.
  • Fixed prices: Chrono Shards and Rush Orders stay at their fixed price.
  • What players see: when demand has lifted prices, a line under the Crates header says by how much.
  • When it starts: the first demand reading is taken at the next daily index, 00:00 UTC. Until then every price stays at base.

1.341.1 — 2026-09-29

Also

• The Missions tab is ready for Italian (#5655). Nothing changes on screen: the game is still English for everyone, and every word and figure reads exactly as before. Behind the scenes, these now read from the translation layer, each with its Italian already written:
  • the mission launch flow and the fleet list;
  • the recall and stand-down sheets and the target finder;
  • the Inbound panel;
  • every mission's name;
  • the suggestions for fixing an energy shortfall.

  Italian is still off until enough of the game reads naturally in it.

1.341.0 — 2026-09-29

New

• An Events menu: Season, Crucible, Campaign and the Marshal behind one button (#5665)
  • On a computer, the top bar now has an Events ▾ entry after Activity and Leaderboard. It opens a small menu listing the four.
    • Before, the Marshal had no entry at all outside a live window. There was no way to read its schedule, rules or shop before the first one opened.
    • Now it's listed whenever the event is on. Under its name is its clock: "Opens in 1d 3h" between windows, "Live · closes in 1h 12m" during one.
    • A gold dot on the button means something inside is waiting for you: the Season's daily Contracts, or a Marshal window where you still have a strike. Once your strikes are spent, the dot goes out.
    • The menu works from the keyboard. The arrow keys move through it, Escape or Tab closes it, and picking an event returns focus to the button.
  • On a phone, the Empire tab shows the same four under their own Events heading, above the other doors. Events you pinned to the bottom bar stay there.

1.340.1 — 2026-09-29

Changed

• The Orbital Arsenal has its own look (#5643). Until now it wore the Autonomous Factory's drawing at every level, so in a colony the two read as the same building. It now has its own art in three stages: an armoured ordnance tower between teal-lit missile magazines, which bulks up with armour and extra magazine bays from level 10 and gains a cannon turret and a glowing core from level 20. The Factory keeps its look.

1.340.0 — 2026-09-29

New

• Announcements over the game (#5631). Some news can't wait for you to open the bell, so it now shows up on screen:
  • Maintenance countdown. From a day before a planned maintenance, a strip on every tab counts down to it. Dismiss it and the next one shows. It comes back 15 minutes before the start, and again when the "we're back" update arrives.
  • Full-screen card, once. Something everyone must read, like a season ending or a rule change, appears as a card the next time you play. It waits until you're done with the tutorial or any screen you're in the middle of. Closing it marks it read on all your devices.
  • Announced dates on the Calendar. The Calendar marks the days an announcement named and lists them in the day panel. Tap one to read it.

1.339.0 — 2026-09-29

New

• Star Forge components at the DM Market (#5641). Before, Choir Resonators, Foundry Ingots and Exarch Sigils came only from bastion fights. Now the DM Market has a Star Forge components counter that sells all three for Dark Matter.
  • Price: one wage-day per component, which is the median producer's daily Dark Matter income. The price is recomputed every day, so it rises as the game produces more Dark Matter. It is 45 DM at launch.
  • Packs: 3, 15 or 45, or exactly what is left of the week. Buying takes two taps.
  • Shared weekly allowance: purchases and bastion drops count against the same limit of 45 per family per week. Buying can replace fighting, but it never raises the weekly ceiling.
  • Sovereign Seals are not sold.
  • Each purchase appears in the diary under Economy. The Star Forge now lists the DM Market as a source, and the bastion drop line says the week is shared with purchases.

1.338.0 — 2026-09-29

New

• Dispatches have an Admin console (#5630). Announcements used to be published from a terminal, and a refused request looked exactly like a successful one: nine personal warnings went unsent that way (#5614). The Admin tab now has a Dispatches view.
  • Before sending: it shows who a message reaches, and the push and Discord text exactly as they will be cut.
  • Test and drafts: it sends a test copy to its author and keeps drafts.
  • Lists: it writes to a list of players in one go.
  • Undo: every message is held for a minute, and Cancel works until then.
  • Push to everyone: it asks for a second press.
  • After sending: it shows the message's id, or the error and its reason. It never shows a success that did not happen.

Changed

• An announcement edited after it went out now says "Updated" next to its date, so a message that no longer reads the way you remember is marked as changed. The earlier wording is kept on record.

1.337.1 — 2026-09-29

Also

• More of the game is ready for Italian (#5629). Nothing changes on screen: the game is still English for everyone, and every word and figure reads exactly as before. Behind the scenes, these now read from the translation layer, each with its Italian already written:
  • every figure row on the building card;
  • the Speed up sheet;
  • planet traits;
  • the power-grid line.

  Italian is still off until enough of the game reads naturally in it.

1.337.0 — 2026-09-29

New

Desktop Home: a Galaxy log under the Threat radar (#5636)

Before: the Threat radar card on the desktop Home had a tall empty area under the scope.

Now: that space holds the Galaxy log, the latest eight events from the galactic register, newest first. It shows battles, colonies founded, market trades and more, each with how long ago it happened.

• It is read once each time you open Home.
• Your own trades read as yours ("You sold …").
• If the register doesn't answer, the log says so instead of looking empty.

Changed

Desktop Home: the radar with animations turned off (#5636)

• Animations turned off: on Windows with Animation effects off (or any system set to reduce motion), the radar's sweep stays still. That is what the setting asks for, and the rest of the game already respects it. It used to freeze at a random angle and look broken. It now rests on a fixed bearing with a soft glow, so the radar reads as on.
• Animations on: the sweep turns as before.
• Colony markers: your colonies on the radar now spread across the scope instead of sitting in one clump in the middle.

1.336.0 — 2026-09-29

New

• Portrait frames in the DM Market (#5628). Four painted frames for your portrait: Void Rim, Circuit Rim, Ember Rim and Starlight Rim. Each is about 352 DM (8 days of a typical producer's Dark Matter).
  • Where it shows: a frame you wear is drawn wherever your portrait appears: the header, your profile, chat, your guild's roster and the leaderboard.
  • What it replaces: it takes the place of your VIP border or supporter ring. Your VIP crest beside your name stays. The shop and your profile say this before you choose.
  • Choosing one: tap your portrait. Your profile lists the frames you own, plus "No frame", with a live preview.
  • The whole plan is live: this completes the DM Market's cosmetics (emblems, name colours and now frames).

1.335.4 — 2026-09-29

Changed

Event relics are leaving the Crucible (#5593). The Destroyer and the Annihilator are prizes rather than ships anyone can build, and the ladder is the one arena whose whole point is that every commander can field what you field. From now they cannot go into a Crucible roster at all — attacking, defending or sparring. A defence you had already published holding one keeps fighting as it is until 5 October, and then fights without it: the Crucible names the hull and says how much PP that costs, the ships themselves are untouched at home, and you can publish again to choose what takes their place.

1.335.3 — 2026-09-29

Changed

Desktop Home: the timeline shows the next 4 hours (#5619)

Before: the desktop Home's timeline covered 24 hours. Buildings and research are capped at 4 hours each, so every building and research bar was squeezed into the first sixth of the lane and the rest was empty.

Now: the timeline shows the next 4 hours, with a mark every half hour, so your building and research orders spread across the whole lane.

• Ship orders and fleets that run longer reach the edge, and the label says when they really finish ("Cruiser ×12 · in 5h 41m"). The same applies to a building or research order slowed by an exodus shock.
• Clocks past the 4 hours but within the day are no longer dropped: a shield ending, the daily contracts reset, an attack still hours away, the season's end. They sit at the lane's right edge with their countdown, on their own line so they never cover anything on the timeline.
• The storage forecast still looks a full day ahead.

Desktop Home: a full orders list scrolls (#5619)

When the Empire orders card fills up, each order keeps its full height and the list scrolls. Before, orders that wrap to two lines could get squeezed. The "N more waiting in Overview" link stays visible under the list.

1.335.2 — 2026-09-29

Also

• The Wraiths fly escort to the Marshal (#5053). Before, the Wraith faction could raid a top player in the middle of a Marshal window, while that player's fleet was likely away striking the boss. Now, while a window is open, and in the hour before one opens, the Wraiths launch no attack on any player. They never strike the Marshal themselves, their own flagship, and they stay out of its results and prizes, as they stay off the leaderboard. The Marshal event is still switched off, so this changes nothing yet.

1.335.1 — 2026-09-29

Fixed

• Chat replies show the quoted player's name colour (#5620). When you reply to someone who bought a name colour in the DM Market, the quote above your message now shows their name in that colour, not the old one. A quote of your own message stays gold, as before.
• The ship bundle counter no longer reopens its packs right after a purchase (#5620). A board refresh sent just before you bought could arrive after the purchase and show the old board. Every pack reopened and the "on the way" line disappeared, though the bundle was flying. The counter now keeps the newest board it has received.

1.335.0 — 2026-09-29

New

• Cosmetics in the DM Market: painted emblems and chat name colours (#5615). A new counter below the Shard Chandlery sells looks for Dark Matter. Nothing it sells changes the game.
  • Six painted portrait emblems: Star Compass, Void Prism, Ringed Comet, Dark Crest, Ruby Crown and Forge Sun. Each is about 220 DM (5 days of a typical producer's Dark Matter).
  • Eight chat name colours: Gold, Crimson, Violet, Teal, Emerald, Amber, Rose and Ice. Each is about 132 DM (3 days), and each card previews your own name in that colour.
  • Buying and wearing: two taps buy an item. It is yours for good, and you wear it from your commander profile (tap your portrait). What you just bought shows up there straight away, no reload needed.
  • In chat: the colour you wear paints your name on your messages and in the pinned strip, and you see it on your own messages too. The reply quotes and the messages already on screen take it on the next load.
  • Frames will follow in a later release.

1.334.2 — 2026-09-28

Also

• More of the game is ready for Italian (#5600). Nothing changes on screen: the game is still English for everyone, and every word reads exactly as before. Behind the scenes, these now read from the translation layer, each with its Italian already written:
  • the building card's panels (Hypergate, Factory and Arsenal cores, Ramparts, Speed up);
  • the operations panel;
  • the Colony and Overview tabs;
  • the 3D view's loading-error message;
  • every building's description.

  Italian is still off until enough of the game reads naturally in it.

1.334.1 — 2026-09-28

Changed

Desktop Home: your empire's orders, not a single "next move" (#5607)

Before: with no attack incoming, the desktop Home's priority card took the first item on the Overview's attention list and made it a big headline. Even a player with billions of points was told "Start a build · <colony>": the smallest nudge in the game, shown as the one thing to do, with everything else pushed into a link.

Now: the card lists your Empire orders.

• The real stakes come first, most serious on top:
  • a colony whose siege bar is full;
  • a guildmate under attack;
  • full warehouses (one line, however many);
  • a siege building up;
  • guild raids waiting for you, with when the first one lands;
  • today's contracts, until you have looked at them. That row opens the Season tab straight on its Contracts lane.
• Routine at the bottom, one line each: idle build lanes and idle shipyards, such as "6 colonies with an idle build lane". A line about one colony takes you to that colony; a line about several opens the Overview, which lists them by colony.
• Long orders wrap to two lines, so the figure at the end (a razing countdown, a raid's landing time) stays visible.
• With nothing waiting, the card says so and still offers a way forward (the campaign, a guild invitation, or the galaxy). While your colonies are still loading, it waits instead of calling it all clear.

The incoming-attack card and the phone's Home are unchanged.

1.334.0 — 2026-09-28

Added

• The launch composer now tells you whether the target's sensor grid can read your dock raid (#5575). If you have scouted the colony, the Dock raid row says what grid your scout saw, when it saw it, and how that compares with the level this particular fleet has to stay under. Before, the composer told you the level your fleet hides under and a spy report card told you the target's grid — both numbers were already on your screen, on two different screens, and the comparison was yours to carry in your head. The one case worth knowing about is the one where it could go wrong for you anyway: a grid below your level is reported as what your scout saw and nothing more, because the game compares against the strongest grid anywhere in that empire, and a scout of one colony cannot see the others — nor whether its owner has bought The Watch, which reads a raid from a lower level. So the row will tell you when a raid is certainly seen; it will never promise you one is not.
• Nothing about the raid itself moved: no signature, no sensor reading, no verdict and no cost. This is the composer saying out loud a comparison the game was already making at impact.

1.333.0 — 2026-09-28

New

• Ship bundles: hulls for Dark Matter, delivered by convoy (#5597). The DM Market has a new counter under the Boosts.
  • What's on it: every hull the Anchorage's destination colony can build now, in three packs (a quarter, a half and all of a day of that colony's own shipyard work).
  • Price: the hulls' resources at today's index, plus 30 Dark Matter for every hour of shipyard work they stand for. That is the Rush Order's rate, so a bundle never undercuts buying the resources and rushing the shipyard. At a level-10 shipyard, 860 Corvettes cost about 718 DM. Nearly all of that is the time.
  • Delivery: the hulls fly to the colony and join its fleet 2 hours after you buy. If the colony is lost in the meantime, they land at your home colony.
  • Limits:
    • A day of shipyard work per bundle, and a day in any 7 days.
    • One bundle in flight at a time.
    • Only hulls the colony could build itself. The Destroyer and the Annihilator are never sold.
  • Buying: two taps, the first shows the price and the second pays. If the price rises before you confirm, nothing is spent. A pack you can't buy says why on the button.

1.332.1 — 2026-09-28

Also

• The dock-raid design document stops describing a game it no longer matches (#5575). Three claims in DESIGN-DOCK-RAID.md had gone stale or were never true, and nothing a player sees changes: what number a dock-raid verdict compares against (two later Ramparts are inside it — The Watch scales the defender's whole sensor network, Long Sight adds a flat bonus to this one comparison), that the mechanic has an upper end at all (every term of a flight's signature is capped, so there is a largest raid the game can produce and a grid above it can never be beaten), and a parenthetical claiming the web's spy intel withholds building levels — it carries them at every tier and the report card already prints them.
• The last of those had already produced a wrong figure in a separate measurement an hour before it was corrected, in exactly the case the stale description was wrong about. A design document is read as a specification, so a false sentence in one is a defect with a blast radius rather than untidy prose.
• No engine behaviour, no player-facing copy, no API surface and no persisted value is touched: the crossover this records has been the arithmetic since the Ramparts landed, and the document is now the place a reader can find it instead of re-deriving it. Whether that upper end is the intended end of the mechanic is still open on the issue.

1.332.0 — 2026-09-28

New

A new Home on desktop: your empire's next 24 hours and the galaxy around it (#5594)

Before: on a desktop browser, Home was the phone's orbital picture: five pins, one "next move" card and a short list. There was room for much more, and none of it covered time or the other players.

Now: on a desktop browser, Home is a command deck for your whole empire. It is built around all of your colonies, not the one you happen to be viewing. It shows your empire's time and the outside world; the state of your colonies stays on the Overview. On a phone, Home is unchanged.

• Priority card. An incoming strike takes it over: a live countdown, the attacker (with the AI mark when it is one), attackers against your garrison, and a Defend button that opens Missions on the colony under attack. With no strike, it shows your next move, the same one the phone shows, with a link to the rest of the Overview's list.
• Empire theatre (3D). Every colony you own is drawn as its own planet on a holographic table, with your home colony at the centre.
  • Your fleets fly on it at their real progress.
  • An incoming strike is a red vector into its target, with an alarm ring.
  • Each planet carries a live label, and clicking one opens that colony's Base.
• Season. Your rank, your points, the gap to the empires just above and below you, the countdown to the Eschaton and your guild's Veil.
• Next 24 hours. One lane per colony (buildings, ships, shield ends) plus strikes, fleets, research, the daily contracts reset, guild raids and the season's end, all on one timeline.
• Threat radar. Incoming strikes, closer to the centre the sooner they land. Before the inbound check has answered, or if it fails, the radar says so rather than giving an all-clear.
• Storage. When each colony's warehouse fills, soonest first, using the Overview's own numbers.
• Battles over the last 7 days. Wins and losses per day, win rate and loot. When a busy week has more than a hundred battles, the heading says it shows the last hundred.
• Guild. Who is active, an ally under attack (with Send aid) and raids waiting for you.

On narrower desktop windows the theatre gets a row of its own.

Not in this release:

• A figure for which of your fleets can reach a colony under attack in time; Defend opens Missions instead.
• A season race chart, because the game keeps no daily points history yet.
• Who spied on you, on the radar.

1.331.0 — 2026-09-28

New

• Rush Order: finish an order now, for Dark Matter (#5591). The speed-up sheet has a new row under the shards. It finishes the order you picked on the spot, whether a building, a research or a shipyard slip.
  • Price: 30 Dark Matter per remaining hour, rounded up. That is twice the shard rate, so shards stay the cheaper road when you have them. A 2-hour build costs 60.
  • Only the last 8 hours: a longer order shows when it becomes rushable.
  • One a day per queue: research is one queue for the whole empire. Build and shipyard queues count per colony, so rushing one colony's builds leaves another's open.
  • 8 hours a week in all, separate from your shard allowance. If an order would go past it, the row says how much of the week is left.
  • Two taps: the first shows the price and the second pays. If the price rises while you confirm, nothing is spent. The sheet then says what the rush cut and what it cost.

1.330.1 — 2026-09-28

Also

• The Base screen is ready for Italian (#5589). Nothing changes on screen: the game is still English for everyone, and every word reads exactly as before. Behind the scenes, these now read from the translation layer, each with its Italian already written:
  • the Base screen: the build menu, the placement banners and the map notices;
  • the building card;
  • the "suggested next" advisor;
  • the building ledger.

  Italian is still off until enough of the game reads naturally in it.

1.330.0 — 2026-09-28

New

• Boosts in the DM Market: Overdrive and Freight Wind (#5587). A new counter between the crates and the Shard Chandlery sells a day of effect for Dark Matter.
  • Overdrive: +10% production on every colony you hold when you buy it, for 24 hours.
    • Price: personal, a tenth of a day of your own output at today's price index. At launch a typical end-game empire pays a few dozen Dark Matter.
    • Cadence: one every 72 hours.
    • Where you see it: your production rates and "Full in …" clocks count the boost while it runs.
  • Freight Wind: +20% fleet speed for 24 hours on logistics launches: transports, transfers, harvests, explorations, pickups and supply lanes.
    • Never on: attacks or garrisons.
    • Price: one day of a typical producer's Dark Matter. One at a time.
    • Launch estimates: the Missions tab quotes the faster times while it blows.
  • Buying: every purchase shows its price on a confirm sheet. If the price moves before you confirm, nothing is spent and the new price is shown.

Fixed

• "Price moved" message: it now says "try again" rather than "sign again", since it also appears on Buy buttons.

1.329.1 — 2026-09-28

Also

• The synthetic empires take part in the Marshal (#5052). Before, the Marshal counted them but none ever struck it. Now, while the event is on:
  • Strikes: each strikes the Marshal inside its own active hours, on its own rhythm and its own share of the four strikes per window. They don't all strike at once when a window opens.
  • Guild raids: they sometimes join their guild's raid on the Marshal, leaving seats for the guildmates the leader called.
  • Hunting pilots: they mostly leave alone a colony whose fleet is out striking the Marshal. Hitting it has a real price: lost strikes, a bounty and lost prizes.
  • Chat: when the Marshal comes up, they speak only from what the server told them: how long the window has left, what they saw of the boss, and how many strikes they launched. They say plainly when they have launched none.
  • The event is still switched off, so none of this happens yet.

1.329.0 — 2026-09-28

New

• The DM Market's crates now follow a daily market price (#5582).
  • Before: the Anchorage charged a fixed 40 Dark Matter for every million resources, so the 100M crate cost 4,000 DM. That was about three months of an end-game empire's Dark Matter income for half a day of its own output, and nobody bought it.
  • Now: every crate is priced at a price index recomputed each day at 00:00 UTC from the whole server's economy — what active empires produce per Dark Matter they earn, how much Dark Matter they hold, and how much the shop sold the day before.
  • Launch price: 214 DM for 100M ferrolite (index 2.14 DM per million, measured across 37 producing empires at launch).
  • Guards: the price can move at most 5% in a day.
  • No whale effect: one player's wallet counts no more than a large holder's, and one player's daily spending counts no more than three days of a typical producer's income.
  • The price index shows today's value, its move since yesterday, and the last month, beside the crates.
• A 1 billion crate at the top of the Anchorage ladder.
  • Price: 8 × the 100M crate (1,712 DM of ferrolite at launch), the only bulk discount on the ladder.
  • Convoy: flies for 5 days.
  • Warehouse: needs a billion of free room at the destination.
  • It has its own painted crate.

Changed

• If the price moves while you are signing (the daily reset falls between opening the charter and confirming it), the charter is refused, nothing is spent, and the board shows the new price to sign again.
• The empty-wallet DM Market screen now quotes the board's real cheapest crate and shard prices instead of a fixed "40 DM".
• Power and the Marshal's prize values do not move with the shop. They still count Dark Matter at its fixed book value, so a price change never reshuffles the leaderboard.

Also

• If you play on the iOS app, update it to get the new buttons.
  • Until then, the app's DM Market shows today's prices, but it may keep a crate locked you can already afford, and it shows the 1B crate as not fitting.
  • It charges the day's price, the one on its board. A charter left open across 00:00 UTC is charged the new day's price, at most 5% away.

1.328.3 — 2026-09-28

Also

• Every error message is ready for Italian (#5572). Nothing changes on screen: the game is still English for everyone, and every message reads exactly as before. Behind the scenes, all 604 messages the game can show when an action is refused now have their Italian version, and the check that keeps the Italian complete now covers them too. Italian is still off until enough of the game reads naturally in it.

1.328.2 — 2026-09-28

Also

• The Marshal tells players it is coming (#5051). Before, nothing told a player a window was coming; they had to open the game and see the HUD chip. Now, while the event is on:
  • Calendar: the Season calendar shows each upcoming window as "The Marshal takes the field".
  • Push, 15 minutes before: every player with a push device (iPhone app or browser, Android included) gets a soft banner.
  • Push, at the open: a banner with the time left goes to anyone the warning missed. A window costs a player one of their daily pushes. Tapping either banner opens the Marshal tab.
  • Discord: it hears the open, and at the close how many commanders struck and how many levels fell, with no names.
  • Opting out: a player who turned pushes off gets nothing.
  • The event is still switched off, so nothing is sent yet.
  • Not included: older app builds don't show the calendar entry. A banner tapped on an older build opens the app's default screen.

1.328.1 — 2026-09-28

Also

• Release machinery (#5570): release.sh --self-test no longer fails on every Mac. BSD wc pads the line count that one badge-block check compared as a string. Nothing a player sees changed.

1.328.0 — 2026-09-28

New

Three hulls at the same Power are not the same hull (#5469). The Shipyard told you what a hull adds to your Power and what its attack, hull and shields are, and left the division to you. Now it does it: under the Power per unit figure there is a line saying where that Power goes — how much of it the hull spends on guns, how much on staying alive, how much on shields. The three always add up to 100%, and the Ship reference carries the same line on every row of the catalog, so you can read the whole catalog in one scroll instead of one hull at a time.

It is the figure that tells hulls apart when their Power does not. The Voidlance Cutter, the Aegis Bastion and the Breachhead Monitor sit within 0.8% of each other at 202-204 Power, and they spend it in three completely different ways: the Cutter puts 28% into attack, the Monitor 18%, and the Bastion nothing at all — 53% of the Bastion's Power is shields. A hull with no guns reads 0% attack, which is the answer rather than a blank.

On the Ship reference these are the catalog shares; on the Shipyard they are yours. Research that raises attack, hull and shields by the same amount leaves every split exactly where it was, so a split only moves for an empire whose tree is lopsided one way.

1.327.7 — 2026-09-28

Fixed

• A Wraith can no longer put a known player's fleet in its own sky (#5516). The Wraiths already had a gate against handing something crossing the sky to a player their briefing does not name. It checked whose and never where, so an attacker one of their empires had actually identified could be claimed by any of the thirty: with Kane's fleet inbound on Nyx, Hollow could say Kane's convoy was crossing its own lane, and the line went out. A true name in the wrong sky is worse than an obvious invention, because it survives the first thing a player does — check it against their own battle reports. A line that puts a fleet in the speaking empire's own sky now has to be supported by that empire's own sensors.
• Warnings between their empires are deliberately untouched. Hollow telling the guild that Kane's contacts are crossing Nyx is true, useful, and still goes out — a gate that silenced that would have deleted a real sentence to close a fake one, and the faction the game announces would simply have gone quiet instead.

1.327.6 — 2026-09-28

Also

• The guild features the synthetic empires run can no longer stop running silently (#5537). Eighteen guild behaviours — answering who knocks at the door, promotions and expulsions, grants from the Fund, the season project, the outposts, the dues and rank titles, the greeting, the noticeboard, the shop — each hang off a single line in one method, and that method is the only place in the code that runs any of them. Until now thirteen of those lines could be deleted without a single test that exercises them noticing, so a botched merge could have switched a whole feature off with nobody able to tell: for a behaviour on a timed release the silence looks exactly like "it has not started yet", and after that like "there was nothing to say". Nothing changes in the game today; the eighteen lines are now enumerated and checked, in the order they run, together with which two of them total silence is allowed to skip. No player-visible behaviour was touched.

1.327.5 — 2026-09-28

Also

• Nothing in the game changed. The in-memory stand-in for the database that the server's unit tests run against now applies an update the way the real database client does: it used to store one of the client's operators as a literal object instead of writing the value, blank a column the real client leaves untouched, and do the arithmetic for "add this much" on only a third of its tables. No test was passing because of that, measured rather than assumed — the whole server suite passes identically before and after, and the only difference is the forty new tests that pin the behaviour. What it buys is future work: from here, a test that relies on how an update is applied means what it says (#4923).

1.327.4 — 2026-09-28

Also

The anomaly payout now records the empire production figure the resource haul was priced on. Nothing changes in play: the sums paid are identical, no screen shows the new figure, and it is not something a commander can act on. What it changes is what can be asked later — how the haul behaved for empires of a given age is answerable from the ledger rather than estimated against production as it stands on the day somebody asks (#5513). Visits recorded before this release carry no such reading and cannot be repriced, so the question becomes answerable only for what happens from here.

1.327.3 — 2026-09-28

Fixed

• The Advisor's «switch to espionage» can be pressed on a phone again (#5507). When you aim an attack at a target you have never probed, the Advisor says so and offers you the one-tap way out: send a probe instead. On a tall phone that button sat behind the floating launch bar where the mission form opens, so a tap on it landed on the bar in front of it and nothing happened — and nothing told you to scroll first. It now sits on the Advisor's own title line, clear of the bar, with the sentence explaining it left where it was.

Also

• Left out on purpose: at three narrower phone sizes that button is still behind the bar, because nothing anywhere in the Advisor's panel is clear of it there. That is a whole class of controls on that form, it is filed with its measurements, and this release does not close it.

1.327.2 — 2026-09-28

Fixed

• The ± steppers in the Speed up sheet answer at once on a long ship batch (#5499). Holding + or − on a shard row could crawl, and it was worse the longer the batch had left to run: every repeat made the sheet work out its recommended spend again from scratch, even though the recommendation cannot change when all you did was ask for one more shard. It works it out once a second now, when the countdown actually moves. The spend it recommends is unchanged.

1.327.1 — 2026-09-28

Fixed

• The test-count badge in the repository's README stops going backwards, and stops carrying a number nobody measured (#5501, #5532). The badge says how many tests the suite has, and until now nothing looked at it before writing over it — the figure was simply the last one anybody published. When two changes finished in the wrong order it went down, advertising fewer tests than the project had: over the sixty most recent revisions of that file, twenty-six published a lower figure on at least one of the two counts, the worst dropping one of them by 641. The release now reads the published pair first and, if either half would fall, leaves the badge alone and says which half and by how much. The release still goes out — most of those twenty-six were honest measurements overtaken by a faster neighbour, and stopping the release over a badge would have cost far more than a stale number. Separately, when a change touches nothing the Python tests can observe, those tests are skipped and the count is handed back from the badge itself; that copy is now refused too, because publishing it would re-date a figure nobody measured as a fresh full-suite result. Neither check invents anything: a number that cannot be verified is left out rather than guessed, and the pair is only ever replaced whole.

1.327.0 — 2026-09-28

New

• The Fleet view lists the systems you have scouted (#5422). Espionage had no place of its own: a report could be read one address at a time, or found again by scrolling the Activity feed past everything else that happened that day, in the order the probes came home. There is now a list in the Fleet view — one row per system, freshest first, saying how many planets of it you have read and how long ago the freshest reading was, and a tap opens that system in the Galaxy view where the planet-by-planet detail already lives. The list shows your most recent reports and says so; older ones load on request, because a player with a thousand reports should not be told quietly that they have a hundred.

Also

• A reading still inside the six-hour window the launch forecast itself acts on is marked as such. Nothing is ever labelled stale: how old is too old depends on what you are about to do with the report, so the list states the age and leaves the judgement where it belongs. The marked state is the rare one on purpose — a badge that appeared on almost every row would be decoration.
• The list asks the server for espionage by type rather than sifting the general feed, which is what makes it complete rather than "whatever survived today's noise".
• No change to what a report contains, what a probe costs, or what the forecast concludes.

1.326.0 — 2026-09-28

New

• Guilds run by an AI empire now set house rules of their own (#5334). Until now a guild led by a synthetic commander never asked its members for anything and its ranks were always just Leader, Officer, Member and Recruit — the settings the last release gave every leader had been used by nobody at all. From its own night onward, such a leader decides both from who it is: a builder or a trader asks each member for a daily donation to the Fund, a warlord asks for nothing, and each of them renames the guild's ranks in its own register. It announces what it set in guild chat, in its own words, and it names only what the server actually accepted.
• A due is a donation made for you, not a levy. It comes out of your own donation reserve, credits your Scrip and your contribution exactly as a donation you made yourself, and never touches a Council Directive. If you have no whole donation left that day you pay nothing and owe nothing, and you start clean tomorrow. Your guild panel and your guildmates will tell you the same thing, because they read it off the same place you do.

Changed

• An AI guildmate can now answer questions about your guild's house rules (#5334). Ask one what the dues are, whether today's have been taken, or what the new name of a rank means, and the answer comes from the guild's real settings — including "nothing was taken today, you had no donation left" — instead of a polite guess. A rank's new name never replaces the rank itself: an officer is still an officer, and everyone can still tell who outranks whom.

1.325.0 — 2026-09-28

New

You choose which row each hull stands in on the Crucible (#5395, #5396). The ladder has always fought in three rows — what you put in front soaks the fire, what you put behind buys time — and the one place you could not choose was the ladder itself. Now you can, on your published defence and on each attack, with the same picker the war table and the shipyard already use. Your attack line is remembered between fights. Rows you had already set are kept.

Fixed

A single bad row no longer wipes your whole Crucible line in silence. The ladder used to accept a malformed placement and quietly fall back to putting every hull where its class says, with nothing said — so one unrecognised entry threw away every row you had set, on every fight, until you noticed. It is refused now, on the defence and on both kinds of duel.

1.324.6 — 2026-09-28

Fixed

• A figure about anomaly rewards that nobody could check any more is replaced by one they can (#5437). Nothing about anomaly hauls changes with this release — no reward, no tier, no multiplier moves. What changes is a number written down in the code, and the reason it is worth a release is that it had already begun being quoted in other work as settled. When exploration rewards were made to grow with your empire, the release said the change took the drop-off in an anomaly's value from fifteen-fold down to about two and a half as an empire grows. Both of those figures compared two groups of commanders: those in their first week and those in their eighth. Re-measured against the live economy over a fortnight, the eighth-week group has one trip in it — and the answer moves by roughly seven times depending on whether "eighth week" means how old an empire was when it flew or how old it is today, which nobody had said. Worse, a reward record does not keep the production figure it was priced on, so a trip taken in a commander's first week can only ever be priced against the empire they have grown into since. No longer window fixes that; recording the figure would, and that is now filed as its own piece of work. So the week-to-week form of the number is retired rather than restated, and what stands in its place is a drop-off of about two and a half measured against the thing the reward formula actually reads — the size of your production — across a fifth of the population at a time, several hundred trips in each. The same re-measurement reproduced every other figure at that spot to within five per cent, which is how it is possible to say that the retired one, and only it, was resting on too little. The recipe, the fortnight it covers and the minute it was read at are now written beside the numbers, so the next reader can redo the measurement rather than trust it. Two more claims in the same note were corrected while checking this one: one of them quietly depended on which group of first-week commanders you meant, and answered the opposite way depending on the choice, so both answers are now written down with how many commanders each rests on.

1.324.5 — 2026-09-28

Fixed

• The Wraiths can no longer tell you whose fleet is coming when their sensors cannot see whose it is (#5392). The thirty empires of the Wraith faction are flown from a single briefing, and until today nothing in the code checked what they said about the sky. Below the Observatory tier that sells an attacker's name, their grid reads an incoming fleet it cannot identify — and a Wraith could still announce whose it was, and name a real player. The 75 agent empires have had a mechanical guard against exactly that for weeks. The Wraiths reached none of it, so for them the rule lived only in the standing brief their operator reads, which is a request rather than a guarantee. It is a refusal in the code now: a line that hands something crossing the sky to a player the briefing does not support is stopped before it reaches the channel, on both of the two paths a line can take, and the operator is told which line and why instead of watching it quietly vanish. Lines the briefing does support still go out word for word — a Wraith naming an attacker its own grid really did resolve still says so, and being able to say true things about who is coming is most of what makes that faction worth playing against. What it does not do is check every claim: this is one shape of falsehood, not a truth test, and a Wraith boasting of a raid that never happened or a fleet it does not have is still held to that only by its brief. It also asks whose a fleet is and never where it is, so a known player's fleet can still be placed in the wrong sky; that gap is open, tracked as #5516, and this release does not close it.

1.324.4 — 2026-09-28

Fixed

• Layout fixes now have a guard that can see the screen (#5428). Four fixes shipped in one day whose deciding criterion the test suite could not express — whether a control is actually reachable. The suite has no layout, so a button clipped to a sliver, or hidden under a floating bar, stays in the page, matches every text check, and takes no taps: the bottom-bar arrows (#5315) and the launch pads (#5277) were both that kind of defect, and both were verified by hand, once, in a browser, after which nothing re-ran the check. There is now a test that opens the real screens in a real browser at five phone widths and asks a player's questions: is this control cut off by its container, does a tap at its centre reach it, is it at least 44 pixels tall, does it hang off the side of the screen. Nothing a player sees changes today; what changes is that the next regression of that kind fails a test instead of shipping green.

Also

• The narrow width is the guard, not decoration: with #5315's fix reverted, the defect is not clipped at all at the four ordinary phone widths and only reappears around 190–205 px — which is 200% text zoom on a phone. A check built to the obvious viewport list would have passed on the broken code.
• Four defects of exactly this class were found while building it, and filed rather than quietly fixed here: eighteen 34-pixel chips in the quick-buttons picker (#5504), the first launch pad parked under the segment tabs at 320×568 (#5505), the launch bar covering the Advisor's one action (#5507), and an ordering arrow hanging off the right edge at 200% zoom (#5520). Two of them are pinned in their current state, so fixing either one turns the new check red and the remedy is to delete the clause rather than adjust it.
• What this does not cover, said plainly: #5305's criterion was already pinned by a stylesheet census, and #5420's — a pinned control fully inside the viewport at scroll 0, at desktop heights — is not covered here and stays open as #5521. The guard is two criteria on two screens, not "layout is now tested".

1.324.3 — 2026-09-28

Fixed

• The test-count badge can now be measured on the tree that actually ships, instead of the one the work started from (#5353). The figure in the repository's README records how many tests the suite has, and it was only ever offered a count taken from the branch a change was written on — a snapshot that goes stale while the change waits its turn, because everything that ships in the meantime adds tests the branch has never seen. A change that touched no JavaScript at all was therefore judged 141 JavaScript tests behind and refused, when the tree that would actually land was 79 ahead. The rehearsal that merges the change with the current code now reports its own counts, so the comparison can be made against the tree that ships. Half of the work turned out to already be done: the number was in the log and the tool that publishes it already accepted the line, so nothing needed changing there — it needed saying. What is not fixed: nothing yet refuses a count that is lower than the one already published, so two changes finishing in the wrong order can still move the figure backwards — measured on 26 of the 60 most recent README revisions. That half is tracked separately (#5501), and the rehearsal still reads the current code as of the check rather than as of the moment the version is cut.

1.324.2 — 2026-09-28

Fixed

The shipyard's Speed up sheet opens again on a long ship batch (#5466). If you were sitting on a big pile of chrono shards, tapping Speed up on a slip with weeks left on it froze the game — no error screen, no Retry, nothing to do but restart. The bigger the pile and the longer the batch, the worse it was, and it only ever happened in the shipyard, because ship orders are the only ones that run for weeks. It opens now, and the spend it recommends is unchanged.

1.324.1 — 2026-09-28

Fixed

• The build chip tells you when the next building lands, instead of counting slots (#5440). The chip in the top strip answered "how many queue slots are busy", so finishing a building showed up as a number going down — the only readout of the build queue you can see from every screen was reporting how full the queue was, not how close anything was to being done. Now, once the next building is within ten minutes, the chip shows a countdown to it and its tooltip names the completion. Outside that window nothing changes: the slot count is back, and the gold "queue full" tint is untouched either way, so you can still tell at a glance whether there is room to start something else.

  Ten minutes because of what the queues actually look like: measured across every colony in the game at one moment, 632 were building and the next completion was a median of 42 minutes away. A one-hour window would have had the chip counting down on nearly two thirds of them almost all the time, which is a clock rather than a signal; ten minutes lights it for about one in five.

Changed

• On a phone this chip is still off the side of the screen, and this release does not change that (#5498). The resource row scrolls sideways there rather than wrapping, and the build chip sits about 665 pixels along it — roughly a full screen's worth of dragging, with no scrollbar to hint that anything is there. So phone players get no benefit from the countdown yet. That is filed separately and is a different problem: the chip is unreachable, not badly worded.

1.324.0 — 2026-09-28

New

• The galaxy map tells you when it last changed, and how long until it changes again (#5457). A player reported refreshing the galaxy map by hand — tapping another tab and then the galaxy button again — to get a current chart. The map was already refreshing itself; nothing on the screen said so, and half a minute is a long time in front of a chart with a fleet moving on it. Both galaxy screens now carry a quiet line: the hex chart says «Updated 23s ago · refreshes every 60s» in the strip above it, and the system view says the same thing in its header at its own period, «refreshes every 30s». The two periods are different and always were — the chart reads the sky on its own minute, while the sector and system lists ride the game's half-minute refresh — so each screen states the one that actually feeds it rather than a single number that would be wrong on one of them. The tab dance was never even a fast path: for most of every minute it is answered with the very chart already on your screen.

Changed

• Nothing refreshes any faster, and no button was added (#5457). The cadences are untouched, the map makes no extra request, and there is no new control to tap: the chart's minute is the same minute the server keeps its map for, so asking sooner returns the same chart. If the line does not stop you reaching for the tab dance, a refresh control is the next thing to try, not this one.
• The message shown when your colonies will not load now names the period the same way (#5457). It read «retrying every 30 seconds» and now reads «retrying every 30s», so the three places that quote the game's refresh period all quote it in one voice — and all three now read it from the timer itself instead of having it written out by hand, which is how that sentence could have gone on promising an old period forever.

1.323.6 — 2026-09-28

Fixed

• The Empire Overview stops calling an idle colony busy over a build that finished days ago (#5444). The Overview's needs attention section — the one place that answers "which colony is waiting on me" — and the next-move card on the Base tab were both reading each colony exactly as it was last saved. A colony only catches up on its finished builds when something touches it, so a world you had not visited in a while still showed an order that had completed hours or days earlier, and the section quietly left that colony out. Measured across every colony in the game at one moment: 1,464 colonies had nothing building, but only 935 were being reported that way — more than a third were missing from the list whose whole job is to name them. The oldest of those stale orders had been finished for 21 days. Both surfaces now read the same up-to-date colony the Empire tab has always used, so the three agree about the same world.

• Two more attention warnings were being missed the same way, and now are not (#5444). The same stale reading silenced the idle shipyard warning on 233 colonies (1,127 reported against 1,360 really idle) and — the worse one — the full warehouse alarm on 319 colonies, 160 reported against 479 whose storage had actually stopped growing. That is the red alarm, so two thirds of the players losing production to a full warehouse were being told nothing at all.

Changed

• Every number on the Overview now counts the builds and ships you have already finished (#5444). Because the whole tab reads the caught-up colony now, the figures on it move up to what the game really holds: building levels rose on 700 colonies, hangars on 238, stockpiles on 1,743, and the colony ceiling on 6 commanders whose Command Center upgrade had finished without being credited on screen. Nothing moved down anywhere — every one of those was the screen catching up, not changing.

1.323.5 — 2026-09-28

Fixed

• The ship tooling no longer publishes a status message its own next step cannot read (#5402). Before a change is merged, one tool writes a short summary of the merge rehearsal and the next tool reads three specific things out of it. That summary is capped in length, and only the first half is trimmed to fit — so the three things it must contain survived purely because of where they happened to sit in the sentence. Measured: the summary is already being cut, losing its last word today, and a longer identifier in it would cut into the part the next tool needs. The reader would then refuse every change in the repository with a message pointing at the wrong tool, because it has no way to tell a truncated summary from a genuine problem. The writer now checks its own sentence against exactly what the reader requires, and publishes nothing rather than something misleading — and, since a rehearsal whose result was never published is not a result anyone can act on, that outcome is now reported in the verdict instead of only in the log, where the last two occurrences were found by reading the words "is the commit pushed?".

1.323.4 — 2026-09-28

Changed

Synthetic empires and players now read the same wall (#5335). A colony whose reactors cannot keep up with what it draws fires its gun turrets weaker, holds a smaller shield Bubble and lands a lighter Cannon shot. The launch preview a player opens has priced all of that off a spy report for months. The synthetic empires priced two of the five — a starved colony's Fortress and Citadel — and read everything else as if the lights were on.

So this was never "the synthetic empires should be smarter". It was synthetic empires and players reading different walls off the same report, with the synthetic empires reading the tougher one. They over-estimated roughly two thirds of the targets on the board and sent too much; now they send closer to the line. If you have been watching synthetic fleets arrive heavier than the fight warranted, that is why, and it should stop.

Measured against the live galaxy the day this shipped: of 2,122 colonies, 1,437 are running short of power and 161 are fully dark. The read that reaches the most colonies is the gun turrets, on 1,627 of them — more than the shield Bubble's 1,603 and roughly double the Cannon's 844. This is most of the board, not a corner case. Let your reactors fall behind and a synthetic neighbour will notice; keep them fed and nothing about you changes.

Their written plans say it too. A starved colony used to be described as fully powered unless it happened to hold a Fortress or a Citadel; now the plan names the softness whichever defence is short, and says a colony's defences "run" below full strength rather than "fire" below it, because a shield Bubble is a pool and a pool does not fire. The direction makes this safe to say out loud: they now claim less about a target than they did before, so no sentence starts asserting something a probe could not have found.

The change reaches each empire on its own schedule — one defence at a time, days apart, and each empire at its own hour within each of those windows — so nothing already in the air changes its mind the moment this goes live, and the roster does not turn as one.

Also

The test badge in the README is unchanged by this release, deliberately, and this says why so nobody has to reconstruct it later.

The badge is a pair, and the rule is that neither half may be lower than the figure recorded on the main branch. Measured on this tree: 9,833 pytest, which is 32 higher than the recorded 9,801, and 42,666 vitest, which is 169 lower than the recorded 42,835. So it refuses on the first clause, on the JavaScript half — compared against main at b48b74a8c3ba.

Nothing regressed. This release touches no JavaScript at all; the JavaScript suite simply grew on the main branch while this work was in review, and a count measured here is a count of this branch. Updating only the Python half was the other option and it is the worse one: the badge is a single claim about two suites, and one image cannot honestly report one half as of now and the other as of some hours ago. A half-updated pair looks uniformly current while being fresh on one axis and stale on the other.

1.323.3 — 2026-09-27

Also

• The Marshal's Insignia shop sells the Wraith-pattern refit (#5050). Before, the shop held a rename token and a cosmetic laurel. Now a third item, bought once for 400 Insignia, gives up to 40 of your corvettes +12% shield (25 → 28).

  Where it applies: your attacks and raids, your colony's defence and the garrisons you lend, interceptions on both sides, and exploration ambushes.

  Where it doesn't: against the Marshal, the Season Boss, the Crucible or the Campaign.

  The 40 are per player per battle. They are shared across all your contingents and waves in that battle, so splitting a fleet into garrisons or a second wave doesn't multiply them.

  Power counts the refit, so your score, the attack band and the bounty measure the fleet your battles fight with. The shop card says "Fitted — your corvettes carry it" once it's yours.

  The Marshal event is still switched off, so nobody can buy it yet.

  Not included: attack forecasts and the ship stats sheet don't show the refit yet (it's worth at most 40 × 3 shield).

1.323.2 — 2026-09-27

Fixed

A system view no longer disagrees with itself about who is wanted (#5417). A galaxy system view read the clock separately for every planet in it — up to eighteen of them — so two colonies belonging to the same player could answer for that player's bounty a few milliseconds apart. Right on the instant a bounty expires, a few milliseconds is enough: one row showed the head as wanted and the row beneath it showed no bounty at all, for the same player, in the same view. The whole view now answers for one instant, so its rows can no longer contradict each other — and the same instant now covers the silent-empire mark and the siege countdown beside them.

1.323.1 — 2026-09-27

Also

• Ship tooling: the self-test runner stops the whole script it started (#5482). Nothing changes for players. When the release pipeline's self-test of the deploy script was stopped for running too long, its forked helper processes kept running, so the test runner waited until its own timeout and threw away the reason. It now stops the script and every process it started, so a slow or stuck run reports its own diagnosis, and the backstop for a heavily loaded machine is longer (20 minutes instead of 8). Hangs are still caught within 60 seconds of silence.

1.323.0 — 2026-09-27

New

• The Anchorage has its own tab: the DM Market (#5442). The shop that sells crates and Chrono Shards for Dark Matter used to sit at the bottom of the Market, under every open offer on the board. It now has its own screen, which opens on a painting of the Anchorage station. It shows your Dark Matter and shard pool, and a convoy berth that says whether your one convoy is free or in flight and which colony it is flying to. The crates stand on a lit shelf as painted crates, growing from 1M to 100M, and the Shard Chandlery shows each shard as a painted crystal. Prices, rules and limits are unchanged. On a phone you'll find it with your other tabs behind the Empire slot, and you can pin it to the bar. On desktop, open it from the new door on the Market.
• «max» on the ask side of a new offer (#5442). «You ask in return» now has a «max» button on each resource, like the offer side. It fills in the largest amount you can still receive today after the 5% market fee, so an offer asked at «max» can be accepted right away instead of waiting until midnight. It is greyed out on a resource whose daily intake is already used up.

Changed

• The Market opens on your Trade Portal (#5442). The Market now starts with a painting of your Trade Portal, showing its level, how many of your offer slots are in use and how many offers are on the board, plus a door to the DM Market.
• Your stock is shown as cargo tanks. Each tank fills with how much of your warehouse that resource takes up, and has a gauge for how much more the market can still deliver to you today. The exact figures are still listed right under the tanks.
• Offers look like cargo manifests. Each row is edged in the colour of what it sells, with a beam flowing from what it sells to what it asks for. The filters are grouped into what to show, what an offer sells and what it wants.
• On a wide screen, you write a new offer beside the board. The new-offer panel sits next to the board instead of above it. If an offer is refused, the reason now appears right above the Publish button.
• The shard hints point to the new place. The hints that tell you where to buy Chrono Shards now say the Chandlery is in the DM Market, reached from the Market tab.

1.322.12 — 2026-09-27

Fixed

• The troubleshooting note for a stale database client no longer sends the reader looking for the wrong symptom (#5403). A developer who has not rebuilt after somebody else changed the database schema gets a wall of type errors that are not real faults. The note explaining this described the symptom of a client that was never built at all — an empty type — while the far more common case, a client that exists but is a few hours out of date, produces a long and complete-looking type missing only the new fields. So a reader checked for the documented shape, did not find it, and reasonably concluded the error was genuine. Both shapes are now written out side by side with the distinguishing mark named explicitly, plus the detail that only a hand-run check can hit this at all, because the automated one rebuilds first.

1.322.11 — 2026-09-27

Also

• More of the game is ready for Italian (#5430). Nothing changes on screen: the game is still English for everyone, and every word and number reads exactly as before. More of what you read now comes from the translation layer, each piece with its Italian already written:
  • building names, on the First Steps guide, the Home tab, the Colony tab and the build queue;
  • the energy, power and Dark Matter readouts;
  • the shield warnings;
  • the "needs your attention" items.

  Italian is still off until enough of the game reads naturally in it.

1.322.10 — 2026-09-27

Fixed

• A change to the ship tooling no longer makes every check re-run the whole Python suite (#5456). The checks a change goes through work out which Python tests could possibly be affected, and they do it by looking for file names mentioned anywhere in the test files. Five of those tests mention the checking script by name — in comments, while explaining how that very mechanism works — and because the script was not on the list of things no test can actually run, being mentioned was treated as being used. So editing the tooling ran all 464 Python test modules instead of the eight that mention it: about twenty-six minutes rather than one second. Those eight now run instead, measured by extracting the real logic and executing it rather than by reading it. The comments stay: they are the only place the rule is explained, and deleting them to satisfy the rule would have removed the explanation instead of fixing anything.

1.322.9 — 2026-09-27

Also

The Wraiths' scheduled wake-up still prints its whole target board, and that is now a decision on the record rather than an accident. #5391 measured the cost and asked whether an unattended run should be emitting the board at all; the answer, measured, is that it should. The board is 19% of the briefing's text — the roster table beside it is 77% — the faction's own doctrine says a wake-up reads the briefing in order to write an order file, and the recommendation printed on it is the only thing that stops the Wraiths hitting the same player over and over. Moving it behind the read-only brief command would have saved a fifth of the output and cost a second sign-in for all thirty empires on every wake-up.

Nothing a player sees changes. What changed is that the briefing can no longer quietly leave that scheduled run: deleting it used to turn nothing red, and now turns two tests red by name (#5391).

Still owed, and untouched by this: the privacy disclosure #5234 asks for. Printing the same thing less often would not have made an undisclosed use disclosed, and this release does not even print it less often.

1.322.8 — 2026-09-27

Also

• The Marshal's Insignia become a balance you can spend, and the event is still switched off (#5049). Until now a Marshal window's Insignia (#5037) were only recorded as owed. Nothing changes in the game yet: with the event off, neither the shop nor the balance appears.
  • A real balance: every window's Insignia are credited once, at the close. A window closed before this release is credited the first time you open the shop.
  • The Insignia shop: it is on the Marshal tab. Its first shelf holds a rename token (300), which lifts the wait before you can change your commander name again, and the Marshal's laurel (600). The laurel is an avatar emblem that is yours for good, and you can wear it at once from your profile.
  • Safe to tap: a purchase never takes your balance below zero, and two taps never buy twice. A refused purchase spends nothing and says why.
  • Why no crates: the Marshal must never pay more than the Season Boss, and no fixed rate for crates, Chrono shards or anything worth resources stays within that. So the shop sells no resources, directly or indirectly. Other cosmetics, such as colony trophies, crests and skins, will come once they exist in the game.

1.322.7 — 2026-09-27

Changed

• The garrisons on the contested worlds are sized on the empires that are actually playing, and the walls come down (#5455). The neutral force holding each of the six capturable worlds in the contested sector is priced against the strength of the field, and until now that measurement threw away the single strongest empire on the board and averaged the five below it. That was written to stop one runaway setting the price of admission for everybody, and it worked while the runaway was playing. It stopped working when he went quiet, because it discarded a rank rather than a player: the rule kept deleting the top row whether or not anybody was behind it, so the strongest empire still logging in was counted as ordinary and the walls were priced against a field that was not there. On the day this was measured the strongest empire on the board had not been seen for ten days, and all six worlds carried 669,724 defending ships each — a wall the guild holding them did not have to defend, because anyone trying to take one back fought the neutral force at full strength first.

  The size is now the average of the ten strongest empires seen in the last seven days. An empire that stops playing drops out of the measurement by itself, and an empire that is still playing counts for a tenth of it instead of being deleted. Three numbers, and they mean three different things: 669,724 is what the six worlds carry today; 720,516 is what the old rule would put there if it re-priced them right now; 479,644 is what the new rule puts there on the field as it stood when this was measured. So it is around a quarter below what players see and a third below what the old rule would have charged. The exact wall follows the field, so the figure at the next seeding pass is whatever the field is then, not a promise made here.

  Worlds already standing keep their current garrison until the next seeding pass, which is when they rescale. The sector re-prices itself on a timer and only once the target has moved far enough, rather than all at once, so the drop arrives there and not at the moment this is released.

• The wiki no longer describes the garrison rule the game stopped using (#5455). The Contested Space page told readers the garrison is priced "deliberately skipping the very top of the field". That was true until the change above and false the instant it shipped, and a page that explains a rule the game does not apply is worse than a page with nothing on it. It now says what the rule is: the ten strongest empires seen in the last week, a dormant empire no longer setting the price of admission, and a live one counting for a tenth of the measurement rather than all of it.

1.322.6 — 2026-09-27

Fixed

• A colony with nothing left to build stops offering "Build new" (#5409). The Colony view always showed the full-width «＋ Build new — place it on the Base map» button, even on a colony where every building it could start was already standing or in the queue — and following it landed you on the Base map hunting for a place that was not there. Nearly two colonies in five were in that state. It now says so in the button's place, and names what the rest is waiting on: a Tech Tree node, a higher-level Command Centre, or both. Being short of resources or having a full build queue still shows the button, because both of those clear on their own and the Base map is where you read the price.

1.322.5 — 2026-09-27

Fixed

• The scripts that decide what ships are now tested by the same checks as everything else (#5403). The tooling that decides what merges, what reaches the server and what gets a version number carried its own tests, and nothing ran them: they ran when somebody remembered to type the command. Four separate faults in that tooling were found in a single night, every one of them by a person reading the file rather than by any automatic check, and one of them had been wrong for three months. Four of those test suites — 407 checks covering the merge step, the release step, the run ledger and the database-rollback probe — now run on every change like any other test, so a fault in them fails the same way a broken feature does. Three more were already covered by an earlier pass. The largest suite, belonging to the checking tool itself, still runs only when a person asks: it takes eleven minutes on its own, which is more than the rest of the checks put together, so the tool that decides everything remains the one thing its own checks do not cover, and that is said plainly rather than left to be discovered.

1.322.4 — 2026-09-27

Fixed

Speeding up one research level no longer leaves the sheet with nothing to offer (#5389). With several levels of one discipline queued, speeding the first one up went dark: the sheet said that discipline was no longer in the queue and there was nothing left to speed up here, while the next level of that same discipline counted down in the very slot the sheet had been opened from. The only way on was to close it and tap Speed up again. It now says which level finished, shows the one running now with its own countdown, and offers that one behind a single tap — and it still never re-points itself at an order you did not pick, so shards can never land on a level you did not choose.

1.322.3 — 2026-09-27

Fixed

• Agents now say how old the scouting was when they leave a colony alone (#5342). An agent that passes over a colony because its scouting found a lone legendary hull leaves it alone for three to six days, and says so on its sheet for the whole of it — but the row only ever read "your scouting showed a lone legendary hull in their defence", with no hint of when that look happened. Every other reason an agent gives for leaving a target alone already names the age of the probe it read, so these two were the only ones where you could not tell a fresh report from a five-day-old one. The row now reads "your scouting, as of 4 days ago, showed …". It says "as of" rather than naming a single probe on purpose: the verdict can rest on two looks at once — what the report saw in the hangar, and a list of garrisoned fleets carried over from an earlier probe — and the date given is always the older of the two, so the agent never claims a fresher look than the one its decision actually rested on. A hold too old to carry the instant keeps the undated sentence rather than guessing.

1.322.2 — 2026-09-27

Fixed

• A wrong note about anomaly rewards is corrected before it costs anyone a rebalance (#5384). Nothing about anomaly hauls changes with this release — no reward, no tier, no multiplier moves. What changes is a note in the code that said the remaining drop-off in an anomaly's value as an empire grows could only be removed by paying out with no upper limit, at about 56 times the resources. Re-measured against the live economy, that turned out to be wrong twice over: at the setting the game actually runs, removing the limit costs about a third more rather than 56 times more — and it would not remove the drop-off anyway, because two thirds of anomaly visits are paid by empires below the line where the multiplier starts, where the printed reward is flat by design. The note now says that, so the next person to look at it starts from the measurement instead of from the error.

1.322.1 — 2026-09-27

Fixed

• A flaky test no longer turns a ship red for reasons that have nothing to do with it (#5439). Nothing in the game changes with this release — it is a fix to the checks that run before every deploy. One of them timed itself against the wall clock with half a second to spare, so on a busy build machine it failed at random and blocked releases that were perfectly fine; the last time, it stopped a change to a comment. The test now drives its own clock instead of racing one, which makes it exact rather than lucky, and it also runs about eleven seconds faster.

1.322.0 — 2026-09-27

New

The bar tells you when a colony has nothing being built (#5421). Until now the only always-visible sign of the build queue was a chip counting how many slots were busy, so a finished build showed up as a number going down — and on a phone that chip is often not on screen at all. The bottom bar, the Empire doors and the number on the Empire button now carry a count of the colonies with an empty build queue, and tapping through to Overview lists them by name. It appears when the last build in a colony finishes and it goes away the moment you queue something there: nothing to dismiss and nothing to remember. A colony that still has work running is not counted, because nothing there is waiting on you.

1.321.5 — 2026-09-27

Fixed

• The launch panel no longer covers the list of launch pads on a phone (#5277). Picking which colony to launch from meant reaching past the panel that carries the route line and the Send button: on a phone it sits at the bottom of the screen, and it was tall enough to cover four or five of the nine pads — visible through it, but not tappable. The guild note about travel times has moved out of that panel and into the form beside the travel rows it is about, and the pad list now positions itself clear of the panel when the form opens. On the three shortest phone screens every pad is now reachable where the form opens, instead of four or five being unreachable. The form opens a little further down as a result, so the mission and target chips start just above the view — the panel's own line still names the target, and scrolling up brings them back.

1.321.4 — 2026-09-27

Fixed

• A failing test suite no longer hides the suites that never got to run (#5376). When one package's tests failed, the checks a change goes through stopped right there: the packages further down the list were never started, and the report that came out named the failing test without mentioning that about two thirds of the suite had not executed at all. Nothing in it was untrue — a missing result simply looked exactly like a passing one, because in both cases there was nothing to see. Every package now runs even when an earlier one fails, and the report lists each package by name with either its result or the words that it did not run, printed where the verdict is rather than in a second file somebody had to know to open. A package that cannot be accounted for now blocks a merge on its own, which is the rule the merge step already applied to a missing verdict and now applies one step earlier.

1.321.3 — 2026-09-27

Fixed

• The Shipyard's BUILD button is on screen the moment you open the tab (#5420). On a laptop the panel that holds a hull's numbers and its BUILD button reached a little past the bottom of the screen when the Shipyard first opened, so the button you came for was the one thing not showing — seven to nine pixels below the edge at every common laptop size. A small scroll brought it back, which is why it was easy to miss and annoying to meet. The panel now measures where it actually starts instead of where it will sit once you have scrolled, so it ends inside the screen and the button is there from the first frame. The hull list beside it gets the same treatment, so the two still end level. On a large monitor, where the panel already fitted, nothing changes.

1.321.2 — 2026-09-27

Fixed

• Opening the chat no longer moves the game, on any screen (#5419). On a wide screen, opening the desktop chat used to push the whole game band to the right and narrow it — about 440px of movement on a 1620px window, about 300px on a 1920px one — so the board stopped lining up under the bar above it. Closing the chat snapped everything back, and since the game remembers whether the chat is open, a player who keeps it open paid that on every visit. Now nothing moves at any width: the board keeps its full width and the chat panel floats over its lower-left corner, which one click on Minimize clears. This supersedes what we told you this morning. That fix (#5305) said wide screens were staying as they were, with the band still stepping aside for the chat; it turned out the stepping aside was buying those screens nothing that the floating panel does not already give them, so it is gone everywhere. On a very wide monitor, where the board already cleared the chat on its own, nothing changes at all.

1.321.1 — 2026-09-27

Fixed

• Every spy report is reachable again (#5414). A player who scouted seventeen planets in one sitting found three reports in the Activity tab and thought the rest had been lost. They had not: the feed only ever sent the newest three of each kind, and the chips could only filter what it sent. Now tapping a chip — Espionage, Battles, Economy, any of them — asks the server for that chip's own rows, and an Older button goes back further still. The number beside a chip goes up when you tap it, because it is finally counting everything there is rather than everything that fitted. Measured before the fix: 290 empires had rows they could not reach, and 173 of them had spy reports among them.

1.321.0 — 2026-09-27

New

• The launch screen says when a bounty will lapse before you land (#5368). Fly at a wanted head worth a hair over the 10 Dark Matter line and the pot can bleed out — 15% a day — while your fleet is still in the air. You landed on an ordinary target, collected no share, and were billed in full, with nothing on the screen to warn you. The launch row now names the deadline and which side of your point of no return it falls on: before it, the hunt is over and the strike is priced as if they had never been wanted; after it, there is still no share to collect, but your first strike to land counts as the hunt. A head that is still wanted when you land says nothing new, and a head whose price has already gone says so instead of counting down to zero.

Changed

• A wanted head now carries its expiry, not just its price (#5368). The prize pool is still shown rounded down, exactly as before, but the instant it stops being a bounty now travels beside it. That is what lets the launch screen do the arithmetic the rounded figure made impossible: every pot between 10 and 11 Dark Matter was published as the same 10, and a head sits in that window for fourteen hours.

1.320.3 — 2026-09-27

Also

• The Marshal gets a strike feed and a level-fall announcement, and the event is still switched off (#5048). Players will be able to see who is fighting the server-wide Wraith boss (#5037), and when a level falls. Nothing changes in the game yet: with the event off, neither the feed nor the announcement appears.
  • The feed: the Marshal tab lists the battles that landed on the boss, newest first. There is one line per battle, and a guild raid is one line naming every member. Each line gives the hull destroyed, which level it emptied if any, and where the boss stood after it, for example "Kandy destroyed 1.23M hull — the Shield Line at 40%". After the window closes, the tab shows that window's battles.
  • Every level that fell: the feed names the battle that emptied it and the most hull fought on it. A battle that ran through several levels is credited on the level it fought on.
  • The announcement: when a level falls, the HUD says so for ten minutes, on every tab, with those names. A battle that emptied several levels is one notice for the range. The row can be dismissed, and its button opens the Marshal.
  • Honest about who is who: every name carries its AI mark, and screen readers say it in words.
  • Honest about fleets: a battle is shown only once every fleet in it is home, so the feed never points at a colony left uncovered.
  • Not announced in the HUD: a fall whose fleet lands very late, for example when a return job is replayed at boot, is not announced in the HUD. It is still listed in the feed. A player who is offline when a level falls gets no push; push and Discord come with the notice update (#5051).

1.320.2 — 2026-09-27

Changed

• Your explorer no longer dies to an ambush it could not have fought (#5383). A pirate ambush is about one signal in seven, and the one-tap Explore on the signal panel sends a single Spy Probe — a hull with no weapons at all. So the warning on that panel was not describing a risk you were taking: the probe was destroyed without firing a shot, and the trip paid nothing. It fell hardest on new commanders, because T1 is the only tier you can see at Astrometry 1. Measured over the last 60 days, 367 of 436 ambushes were met by a fleet that could not fire a shot, and every single one of them was lost. Now an ambush met by a fleet with no gun aboard is not fought at all: your explorer breaks off and comes home whole. It still comes home with nothing — the haul is forfeit, the signal is spent and so is one of the day's explorations, so one signal in seven still costs you the prize. It just no longer costs you the ship as well.

• Bring a gun and nothing changes, which is the point. An armed fleet fights exactly the battle it fought before, against the same pirate fleets, and still loses about one in five of them. Over those same 60 days armed explorers won 80% of their ambushes, and their win rate climbs with the tier — 67% at T1 up to 100% at T4 — exactly as intended. The lopsided odds people were seeing at T1 were never about T1 being hard; they were about T1 being the tier flown by a probe. Sending an escort is now worth it for what it wins rather than for what it saves: two Corvettes beat a T1 ambush outright and come home without a scratch.

• And the game now tells you, before you launch, that an unarmed explorer cannot win. Not "might lose" — cannot, at any fleet size and any research level. The signal panel used to say "a lone Spy Probe fights it alone", and a revealed ambush's own row said the same; the Advisor promised that probe a real battle, and warned you off one you would supposedly lose. All four now say what actually happens, so the choice in front of you is the real one: come home empty, or bring something that shoots. An armed hull keeps the sentence it always had, because for it the fight is real.

• A lost ambush that costs you nothing no longer reads like a disaster. In your activity feed it used to say "defeat — fleet lost, no loot", in red, even when the fleet was on its way home. It now says the fleet returned intact, and it is no longer painted as a loss.

1.320.1 — 2026-09-27

Fixed

• Speed up: «Finish now» works again for Sharp Shards holders (#5425). With the Loom's Sharp Shards keystone, every tap on «Finish now» came back with «The order was closer to done than the preview showed, so nothing was spent», and tapping again gave the same answer. The speed-up sheet was sizing your shards at face value while the server cut 25% more with each one, so the spend always carried one shard too many and was refused whole. The sheet now reads your tree. The recommended spend, each size's preview and the stepper's limit match what the server applies, and «Finish now» finishes the order. The guild help line in the same sheet now also counts Second Opinion's extra pushes.

1.320.0 — 2026-09-27

New

The buttons at the bottom of the screen can be put in the order you want (#5315). Settings → Quick buttons already let you choose which four tabs sit under your thumb, but a newly chosen one always landed in the last free slot and there was no way to move it — short of dropping every button and adding them back in the order you wanted. Now the chosen buttons are listed underneath, numbered, with an arrow on each side: one tap moves a button one place along, and three taps carry the last one to the front. Nothing to save and nothing to confirm, and the choice still lives on the device rather than on your account. Empire keeps the last slot as always, so the way back to the rest of the game never moves.

1.319.5 — 2026-09-27

Fixed

• A rehearsal of the merge no longer reports itself the way a failed merge does (#5380, #5360). Before a change is merged, the ship tooling can be asked a question: would this merge be accepted right now? That rehearsal already announced success in its own words, so nothing could mistake it for a merge that had happened — but when it said no, it said it in exactly the words a real failed merge uses. A log could then not tell "somebody asked and was told no" from "a merge was attempted and failed", and those call for different responses. Both answers are now the rehearsal's own. The reasons it gives are unchanged, because they are the most useful sentences the tool prints. The same round documents the shape of the success line, which had been described as shorter than it really is: a script that trusted the description took the wrong identifier out of it and stalled a release that had already gone through. And the status line the tools exchange between themselves turns out to be read, not merely displayed, in three places that each refuse a merge on their own — so it is now labelled as the contract it is on both sides, with a measurement showing it is already being trimmed to fit and a new check that fails the day the trimming reaches something load-bearing.

1.319.4 — 2026-09-27

Changed

• Typing 0 in a ship count now clears the hull (#5388). Typing 0 used to snap straight back to the number that was already there, so the only way to take a hull out of a mission was a button — and the hull card has no None button at all, only repeated taps on the minus. Now a typed 0 takes effect when you leave the field: tap or tab away, press Enter, or tap Done in the editor. Nothing happens while you are still typing, so clearing the box to type a new number never takes the ship out from under your fingers. A control that names a count still wins — type 0, tap All, and you get All — and only a zero counts: a stray minus still snaps back.

1.319.3 — 2026-09-27

Fixed

• A badly shaped changelog entry is refused while it can still be fixed (#5375). One badly-shaped entry used to turn the following release red — not the one that wrote it, which had already gone out green, but whoever happened to publish next. Five unrelated pieces of work were blocked that way in a single evening, each waiting on a check that had nothing to do with what they had changed, and the failure did not say which entry was at fault. The entry is now checked at the moment it is written, against the same reader the later check uses, and a release stops with the entry named and the correction spelled out rather than going ahead and leaving the problem for a stranger. A release whose notes are only about internal work is not affected: it has nothing a player can see, so there is nothing to check.
• The later check names the entry instead of reporting a ratio (#5375). It used to measure the proportion of recent releases carrying a readable headline and complain when that fell below a threshold, which meant any single release could cross the line and the complaint always landed on the next one. Measured before the change: every recent release that could carry a headline already did, and the proportion was decided by how many of the last hundred and fifty releases had been internal work — a number nobody controls and nobody should be judged by. The check now asks the question it was meant to ask, of each release separately, and says which one is wrong.

Also

• The refusal also catches a subtler case: an entry that reads perfectly but puts its issue reference one sentence too late, where the part of the system that avoids announcing the same thing twice cannot see it. That entry would have been announced again in a later bulletin. Thirteen entries already published carry that shape and are tracked separately, since correcting them is a different job from preventing the next one.
• The decision lives beside the reader whose judgement it borrows, so the two cannot disagree, and it is deliberately no stricter: a guard that refuses more than the reader rejects would cost releases without preventing anything. Twenty tests pin every shape it accepts and refuses, including three that look correct to a person and read as empty to the parser.
• Nothing in the game changes. This is a check that runs while a release is being prepared, and the most visible effect is that the changelog you are reading stays readable by the thing that turns it into announcements.

1.319.2 — 2026-09-27

Also

• The Marshal gets a door on the galaxy map, and the event is still switched off (#5047). While a Marshal window (#5037) is live, the galaxy map's side rail shows a MARSHAL door. Nothing changes in the game yet: with the event off, the door never appears.
  • Where it sits: after the map's own doors (Places, Contacts, Layers, Go to), so its coming and going moves none of them. Its badge counts your own strikes still in flight.
  • What it opens: the level the server is on and the phase being fought, the pool bar and the time to the close. It also shows your strikes left in the window, or that they are forfeit, and each of your strikes in flight with the colony it left from and when it lands. A button takes you to the Marshal tab.
  • What it does not draw: the Marshal holds no system, so nothing marks it on the chart itself, and no player's strike is drawn as a flight on the map — yours are listed in the drawer. The issue asked for both. They were left out because drawing strikes would show every player which colonies have their fleet away, which the hunting rule (#5045) keeps hidden.
  • When the list cannot be read: if the map fails to read your missions, the drawer says so instead of claiming none are in flight.

1.319.1 — 2026-09-27

Fixed

• Opening the chat no longer shoves the game sideways on a laptop screen (#5305). On a laptop or a small monitor, opening the desktop chat used to push the central panel 442px to the right and narrow it by the same amount — at a 1280px-wide window the game dropped from 1280px to 838px, at 1366px from 1366px to 924px — and it stopped lining up under the full-width bar above it. Closing the chat snapped everything back, so the layout jumped every time the chat was toggled, and since the app remembers whether the chat is open, players who keep it open paid that on every visit. Now, below 1622px of window width, opening the chat changes the game's layout by nothing at all: the board keeps its full width and the chat panel floats over the lower-left corner of it, which one click on Minimize clears. On wider screens nothing changes — the game band still steps aside for the chat, and still re-centres under the top bar on a very wide monitor.

1.319.0 — 2026-09-27

New

Guilds can ask for daily dues (#5121). Until now a guild's Fund grew only from the donations each member remembered to make, so an active guild and a forgetful one looked the same in the treasury. A guild may now ask each member for up to 3 donations a day, and they are made automatically once a day from that member's own donation charges — the same cost off the home colony, the same Fund, the same Scrip and the same contribution credit as a donation made by hand. Dues are counted in charges, never as a share of production, so the Fund still measures showing up rather than wealth.

• Only the leader can raise the dues. Anyone with the settings permission can lower them or switch them off, and a reduction takes effect for every member the day's pass has not reached yet.
• A member pays nothing, and owes nothing, on a day when they have no whole charge left, are on declared leave, have been away for a week or more, have an account too new to donate, or have a home colony that cannot pay for a donation.
• The treasury panel says what today's dues came to and, when they came to nothing, why — naming the colony when the reason is that it could not pay. The dues appear as a house rule on the directory card, the showcase and the guild page, and the guild log gets one line per pass.
• Nobody is charged twice: the day is claimed on the member's own row in the same write that credits them, so an overlapping pass or a restart cannot take the charge again.

Guilds can rename their four ranks (#5121). The ranks always read Leader, Officer, Member and Recruit. A guild may now give each one its own title of 2 to 16 characters — the roster, the showcase and the promote/demote buttons print "Warden · Officer", and the guild header reads "you are Warden (Officer)", so the usual name is always visible next to the title instead of hidden in a tooltip a phone cannot show. Titles are display only: nothing about who may do what changes, and a rank may not wear another rank's usual name or a title another rank already holds.

Titles go through the same name filter as the guild name, so a rank cannot be used to impersonate staff, and a refusal names which of the four titles to fix.

1.318.2 — 2026-09-27

Fixed

• The ship gate's merge-only run now says that it tests the merged code, instead of saying it tests nothing (#5347). The line printed at the top of that run claimed it skipped the JavaScript and Python suites. It had stopped doing that months ago: the run builds the branch together with the latest shared code and runs the whole JavaScript suite plus the Python tests the combination can affect. One real run printed the claim and a duration of 860 seconds four lines apart. Nobody playing sees this line, but the people and tools that decide whether a change is safe to ship read it first, and two of them made the wrong call from it in a single evening — one repeating a 40-minute check it did not owe, the other citing the line as evidence in a decision about the test-count badge in the README. The line now leads with what the run does, names the one thing its green does not cover, and says which status it records. Three other statements in the same file that disagreed with it were corrected too, including two inside its own self-test, which had been teaching the old wording to anyone who went looking for the truth. Two new checks fail if the old sentence comes back, or if a rewrite stops leading with the correction.

1.318.1 — 2026-09-27

Fixed

• A check that runs on every release stopped failing just because the machine was busy (#5144). Three of the checks that guard a release run a script's own self-test, and each was watched by a stopwatch: if it had not finished within a set time, it was assumed to be stuck and killed. The largest of them takes about half a minute, almost all of it spent in its own deliberate waits, so on a developer machine running every check at once it could legitimately take longer than the stopwatch allowed — and a release was then blocked by a check that had nothing to do with the change being released. The limit had already been raised three times for the same reason, and each new value was overtaken by the next machine that was busier than the last. It is no longer a stopwatch: a check is now considered stuck only when it goes completely silent, because a slow machine still reports progress and a stuck one does not. A genuine hang is now caught three times faster than before, and a slow machine is no longer mistaken for a broken one. (#5144)

Also

• The two ways one of these checks can now be cut off are reported in different words, because they call for opposite investigations: going silent, which means something is waiting for an event that never arrives, and running to the outer limit while still printing, which means it is working and will not stop. A check that simply found a real problem is reported as neither, so a genuine failure is never mistaken for a machine being slow.
• The time a check is allowed to stay silent is not a round number somebody picked. It is derived from the longest deliberate wait inside the check itself, with a stated multiple for the cost of running on a saturated machine, and a test holds it to that derivation — so raising it in future means arguing with the number it comes from rather than quietly moving it.
• Nothing in the game changes: these checks run before a release reaches anybody, and the fix is to the harness that runs them, not to the game or to the deploy.

1.318.0 — 2026-09-27

New

You can now fix a typo instead of sending a second message. (#5314) For three minutes after you send a chat message, your own message offers "Edit message…" in its ⋯ menu.

Before: a sent message could not be changed. A typo cost you a second message saying the same thing, and sometimes a third — two players asked for this within two minutes of each other in the same forum. The only workaround was to delete the message and post it again, which moved it to the bottom of the conversation; on a pinned order it was worse, because deleting takes the pin off the wall and you may not have the permission to put it back.

Now: the corrected message keeps its place in the conversation, keeps its time, and keeps its pin — and everyone reading it sees an "(edited)" mark beside it, in the message, in the quote of a reply to it, and in the pinned-orders bar. The window closes early the moment somebody replies to the message or reports it: once your words have been acted on, they stand.

What you first wrote is never deleted. A correction is added, it does not overwrite. That is the same rule the delete already follows — withdrawing a message keeps the text where moderation can read it, so that reporting somebody cannot be undone by their editing afterwards — and a report keeps the exact text the person who reported it was looking at.

Players on an older app read what you first wrote. The app and web builds already installed know nothing about corrections, so they go on showing your original words: true, just not your latest. They see the correction once they update. In the current web client, anyone with the chat panel already open sees your correction on their next look at the channel rather than in the same instant.

The synthetic empires read your message as it stood when they read it. A correction made after one of them has already answered does not change what it answered — a threat that cost you a grudge stays said.

1.317.7 — 2026-09-27

Fixed

• The last two controls that shimmered under a resting cursor now hold still, and no longer drop a click aimed at their bottom edge (#5357). The red "under attack" chip in the top bar and the tech cards in the research tree were the two the previous fix could not reach: letting the cursor come to rest near their lower edge made them flicker, and a click aimed at their bottom row of pixels could land on whatever sat underneath instead of on the control you meant to press. Both now stay put and take the click. Nothing about them looks any different — the chip still keeps its chevron and its "+N" count inside the red pill when it runs out of room, and a card's artwork is still clipped to its rounded corners.

1.317.6 — 2026-09-27

Changed

Exploration rewards now grow with your empire (#5321). An anomaly's resource haul used to be a fixed sum for every commander alike — 2,000 at a weak signal, 35,000 at the strongest — and a player reported that it reads as a pittance. He was right, and the reason was not the size of the number but its shape: every other way the game pays you is priced on your empire, so a fixed sum is worth less every week you grow. Measured across 1,802 real visits, the median anomaly find had fallen to a fifteenth of its week-one value by week eight, purely because the empires around it had got bigger.

The resource haul now grows with your empire's own production. The printed per-tier sums are a floor: below 900,000 an hour they are exactly what comes home, and above it the haul rises with what your colonies make, up to eight times the listed figure at 7,200,000 an hour. A commander in their first week sees no change at all and no new arithmetic — the signal sheet prints the same number it printed before, and it prints your own hourly production beside the threshold so you can see where you stand. Above the threshold the sheet names your factor before you launch, and says that it is read when the fleet gets home, so it can move either way while the fleet is out.

Dark Matter is deliberately left alone, and so are both caps: anomalies still pay at most ten Dark Matter in any rolling seven days, and you can still explore five times a day. The player also asked for the weekly Dark Matter cap to become a daily one; that was measured and turned down, because the window is already rolling rather than weekly — there is no dead week to open up — only one Dark Matter earner in nine ever reaches the cap, and a daily version would have made a good day worse for the other eight by refusing a strong signal's five-at-once.

Being exact about what this does and does not fix: it takes that fifteen-fold decay down to about two and a half, not to nothing. Above 7,200,000 an hour the haul is a fixed sum again. Removing the rest would mean paying purely in production-hours with no ceiling, which puts far more resources into the game than this does and is a decision for its own release.

The signal sheet also stopped making two claims that were not true. It used to say the payout was "the same amount for every commander", which this change made false for exactly the players now paid more. A first attempt to fix that said the payout was "at least" the listed sums, which was worse: salvage pays 80% of them, a lost ambush pays nothing, and a Dark Matter find past the cap pays no Dark Matter — so the line now says what each of the four outcomes actually pays, including that an ambush pays only if your fleet wins. (#5321)

1.317.5 — 2026-09-27

Fixed

• A check that guards the rollback procedure no longer fails at random on a busy machine (#5348). It now fails only when the thing it watches is genuinely broken (#5348). Nothing in the game changes: this is a check that runs on every release, on the emergency procedure for taking a release back out. It watches a safety bound — a deploy that stalls while still reporting itself alive must lose its exclusive hold on the server rather than keep it forever — and it used to test that bound by waiting a fixed two and a half seconds and then looking once. The hold is released after about one second, so the check had roughly a second and a half of slack; on a busy machine the check itself could be starved of processor time for longer than that, and then it declared a working safety bound broken. It did so twice in eight days, each time turning an unrelated release red and costing a full re-run of a half-hour suite. It now waits for the hold to actually end instead of waiting a set time, so it returns as soon as the bound fires and can only fail if the bound never fires at all — which is the case it exists to catch. It is also faster than the version it replaces.

Also

• The two ways this check can now fail are reported in different words, because they want opposite investigations: the safety bound never firing, and the release's own heartbeat stopping so the bound was never re-evaluated. Both messages carry the values they are describing rather than numbers typed into the text — the defect this fix came from was a figure written into an error message and later read as a description of how the code works.
• The check also gained three assertions it did not have, one of which catches a class of mistake the behavioural test cannot see at all: a bound that is off by one second still releases the hold in time to look correct, so the shape of the bound is now checked directly as well.
• The test badge in the README is unchanged. The JavaScript suite was measured in full on this release's tree — 41,669 tests, all passing, which is more than the badge claims — but the Python suite was scoped out, since nothing it covers changed, so the figure on offer for that half was the badge's own number handed back. A badge moves only when both halves were counted on what is being released.

1.317.4 — 2026-09-27

Fixed

A raid launched at a wanted head is no longer billed when the bounty simply expires in flight (#5318).

Before: you opened a raid on a head with a price on it, and the launch row said the strike was a hunt and cost you no bounty. Nobody else touched that head — the bounty just ran down on its own, the way every bounty does, and slipped under the lowest grade before your fleet landed. You were billed in full for hunting a bounty that no longer existed, up to the maximum a single blow can carry, and nothing could be done about it afterwards.

Now: if the head still carried a price at the moment your fleet could no longer be called back, and nobody collected that price before you landed, the strike puts nothing on your head. Your diary says which of the two happened — the bounty was collected while you were committed, or it simply ran out with nobody taking it — so a strike that cost you nothing never reads as luck. As before, it is one such strike per head: the second wave on the same target pays.

What this does and does not reach, because the two cases are very different. A raid can never be called off, so a raid leg is committed from the moment it leaves and is covered for its whole flight. A single attack CAN be called back until the last ten minutes, so it is covered only if the bounty runs out inside those ten minutes — and a solo hunt whose bounty expired hours before you landed is still billed, exactly as before. That is deliberate: a bounty's decline is public and steady, so you can tell before you launch that a nearly-expired head will not be worth hunting by the time you arrive, and being forgiven for it anyway would turn a long flight into a way to hit someone far weaker for free. The report that prompted this fix describes the solo shape, and that half of it is not closed — it is recorded as an open design question rather than quietly left out.

The warning on the launch screen now matches the rule. It used to say the bounty landed on you if the price "has decayed below grade 1 by the time you land", and that is no longer true. It now says a price collected, or simply run out, after your fleet passes the point of no return still leaves your first strike a hunt — and that what costs you is a pot already below the first grade when you passed that point. The same correction reached the planet sheet, the wanted board, the target board and the attack tooltip, and the synthetic empires' own briefing, which had been stating the old rule out loud to players.

No refunds and no data changes: bounties already carried are untouched, and this counts forward only. Installed apps and cached pages keep the old wording until they update; it over-warns, so it can only ever make a strike cheaper than the screen said.

Closes #5318.

1.317.3 — 2026-09-27

Fixed

• The rollback runbook now says which of its steps must run before the old image is built (#5333). The schema-rollback ladder in deploy/DEPLOY.md says it at each step that needs it, and the frame states the exception where an operator meets it (#5333). The frame's first step builds the old image, and because none of the api, agents or web services pins an image by name, that build silently moves the eschaton-api tag onto the old build. Three blocks — the Marshal's strike, its guild raids and its close and payout — then told the operator to restart the api "on the current image" so that the strikes and raids already in the air would land on the current code. Run in the printed order, that restart brought up the OLD api instead: its schema step found the new columns already holding values, refused, and crash-looped — and the fleet the step existed to land stayed in the air with no current image left to land it. Those steps, and the retries that go with them, now say they run ahead of the build, and the frame explains why. The third block was not in the report: it was found by grepping every block for the same shape.
• The note beside the second block lettered AA no longer sends an operator to the wrong place (#5344). That note exists to say how far to scroll to reach the first AA, and all four of its figures had gone false as the ladder grew: it named line 2245 for a block at 2628, said the catalogue ran from A to AL when it reached BJ, called it 2,400 lines when it was over 4,000, and offered AL as the next free letter long after AL was in use. Nothing was renumbered — R and AA each label two blocks by a decision the note itself records, and the letters are cited from other blocks' prose — and the note now carries only the one figure that has to be exact plus two floors that an added block cannot falsify. It names no "next free letter" at all, which is the clause that drifted worst.

Also

• Both corrections are held by tests now, so neither can rot quietly again. One fails when a block tells the operator to bring a container up without saying whether that is the old image after the drop or the current image before the build; the other reads the note's three figures back out of the file and checks them against the file, so a ladder that grows past the note goes red instead of the note going wrong.
• This is operational documentation: nothing in the game changes for players. It matters on the one day it is read, which is the day a release has to be taken back out.
• The test badge in the README is deliberately left as it was, and it is worth saying why, because a release that adds tests and does not move the badge looks like an oversight someone later "corrects" by republishing a stale number. Two reasons, either sufficient. The Python half was never measured here: the suite was scoped to the six modules that happen to name deploy/DEPLOY.md, so the figure on offer was the badge's own, handed back. And the JavaScript half, 41,551 passing, was measured — but on this branch's tree, which predates tests that reached main while this was open, where the badge already reads higher. It is a true measurement of an older tree and a false claim about main: publishing it would have lowered the badge with a number that was honest about the wrong thing. A badge is a measured claim like any other, so it moves only when both halves were counted on what is being released.

1.317.2 — 2026-09-27

Fixed

A button no longer slides out from under the cursor, and its bottom edge takes the click (#5328). Letting the cursor come to rest near the lower edge of a button made the screen shimmer, and a small movement of the mouse stopped it — so it read like a monitor fault rather than the game. The same defect had a quieter and worse half: a click aimed at the bottom pixel row of a button could land on whatever sat underneath it, so the control you meant to press did nothing and a different one answered. Both are gone. Buttons still brighten and lift a pixel when the cursor arrives; the cursor can now sit on that bottom edge without the button moving away from it, and the click goes where it was aimed. (#5328)

Also

The shimmer was never "the cursor sitting still on a static page", which is what it looked like: measured in a browser, a parked cursor on a page where nothing else is happening does not flicker at all. It needs the page to look again at what is under the pointer — which a ticking countdown in the top bar does about ten times a second, and which a hand at rest does with every tiny movement, even sideways. That is why it showed up worst in the top bar, where the blurred backdrop repaints with every flash.

Three controls that shorten a long label — the research back-step, the delete button for a saved fleet template and the chat tag chip — now do it in a way that does not defeat the fix, and the labels shorten exactly as before.

Three other controls still shimmer and still drop a click on their bottom edge, and are named here rather than left to be rediscovered: the tech card in the research tree and the red inbound-attack chip in the top bar, both tracked in #5357. The compact resource pill on the phone is built the same way but is not affected, because a phone has no cursor to rest on it.

1.317.1 — 2026-09-27

Fixed

• The Crucible: a ladder defence of one legendary hull can be beaten (#5341). A defence holding exactly one Destroyer or Annihilator could not be taken by anybody, at any fleet size: the duel ran to the end, the attempt was spent and the rank never moved — the strongest attackers came back a draw and everything else simply lost. Rank 1 sat in exactly that state, collecting the top weekly envelope with no contestable path for anyone else. The Spectral Bulwark makes such a hull immune to anything solo and killable only by two enemies at once, and the ladder is the one place in the game where two players cannot attack together — so it no longer applies there, in a duel or in a Wraith spar. The hull keeps every stat and every advantage everywhere else, a defence that holds the Aegis keystone still takes several rounds longer to break, and a duel is still a simulation: nothing is lost by either side. Reported by a player after four duels he could not win. (#5341)

• A Crucible defence fielded above its tier is now fought under the cap of the tier that fits it. When a defender drops a tier, the ladder trims their stored defence to the new cap — and if not one of their armed hulls fits, it fields the lightest of them alone and above the cap, so the slot is not a free win. Only the defender enjoyed that exception: the attacker was still held to the lower cap, and against a lone flagship no roster that small can win. The fight now happens under the cap of the tier that admits the defence, with both sides capped there — so your roster may be worth more Power than theirs in that one case, and the sheet tells you when it is, which tier's cap it is fought under and how much you may field. Every other duel is capped exactly as before. (#5341)

1.317.0 — 2026-09-27

New

Speed up an order from the order itself. Every building, research and shipyard order under way now carries its own gold Speed up button — on the building sheet while it upgrades, on each running research on the Laboratory sheet and the Research bench, on every shipyard slip, on every row of the Operations sheet, on the desktop Operations overlay and on the Colony tab's Operations card. On the phone rail each running order carries a small stopwatch beside its chip, while the chip itself still opens the Operations list.

It opens a sheet bound to that one order, showing its name and live countdown, and a recommended combination of the shards you hold: the one that finishes it with the least waste, smaller sizes first. One tap spends it, and the button reads Finish now when it completes the order. Under it there is a row per shard size with a press-and-hold stepper, its own Use, and a preview of what that many shards would cut and how much of them would burn. A spend that throws away more than it cuts is still offered, because sometimes it is the only way to finish — but it is not dressed as the obvious choice. Ask the guild moved into the same sheet.

Before this, spending shards meant leaving the order behind: open a generic panel, pick a queue, pick a lane, then tap once per shard — twelve taps to speed up a six-hour building with 30-minute shards. Now it is two: Speed up, then Use.

Changed

• The whole combination is spent in one request, all or nothing, so a spend can no longer land halfway and leave you unsure which shards went.
• The sheet stays open after the spend to tell you what actually happened — that the order is finished, how much of the shards was lost, or that the order had already left the queue. It no longer vanishes at the moment it succeeds.
• The old shard panel is now purely an inventory: what you hold and where more come from. It spends nothing, so nothing on screen can disagree with it.
• Once an order's countdown reaches zero its Speed up disappears and the sheet says it is finishing: there is no time left for a shard to cut, and one spent there would be lost whole.
• The steppers stop at the number of shards actually worth spending, so you can no longer queue up a shard that would land on an order already finished.

Fixed

• Shard counts and order names no longer run over the controls beside them at phone width, and the Operations rows keep the order's full name, level and countdown.
• On the desktop the Operations overlay stays inside the map it floats over and scrolls within itself, instead of running off the bottom edge when several orders are running.
• On the phone rail the stopwatch no longer overlaps the countdown of the order above it, where a tap could open the wrong order's sheet.
• The Colony tab offers Speed up only on orders that have actually started — a ship batch still waiting for a free slip has no clock for a shard to cut.

1.316.0 — 2026-09-27

New

• An empire that gathers against you and then calls it off now says the Vanguard and an away Bastion were both the reason (#5286). Before, when a synthetic empire's strike was blocked by your defensive tree and by an Aegis Bastion its scouting had seen away from the colony, nothing named either cause: the muster ran its three days and lapsed with "three days of mustering and it never launched". You learned nothing about what had held it off, and it could start gathering against you again the next day. Now it calls the strike off and says so: "a probe yesterday showed a tree that holds the Vanguard and an Aegis Bastion of theirs away that could be back before a strike landed, and with both of them the ships there or coming back would wear down the fleet you had mustered". It then leaves that colony alone until a later probe shows one of the causes gone — the Vanguard off the tree, the ships gone, or the Bastion gone from the hangar, the garrison and the fleets away.

Fixed

• Held colonies now say when the probe looked (#5286). A "leaving alone" line stands for three to six days, and three of these sentences used to say only "a recent probe" with no date at all, so you could not tell whether the reading behind it was an hour or a week old. They now name the age of the probe — "a probe yesterday showed…" — and deliberately name the older of the two readings when the sentence rests on a garrison list carried over from an earlier, deeper probe, so the line never claims to be fresher than what it is based on. The matching "called off" line, which is shown for a day, is dated the same way.
• An empire no longer says your ships "weren't home" because of a freighter (#5286). The line "their ships weren't home at the last look, but may be back" is added when a newer probe finds the hangar without its owner's warships. Two things made it wrong: it fired when the ships in question had been an ally's garrison all along and had never left, and it stayed silent when a single hauler or probe happened to be sitting at home while the whole fighting fleet was out. Both sides of that comparison now count warships only.
• The two "Bastion away" refusals no longer name ships that were not there (#5286, #5250). Both said the ships "there" would wear the attacker's fleet down, while an away Bastion is priced against the hangar, the garrison and the fleets away — and the first two can both be empty, which made the phrase name nothing, and contradicted the very next clause when it said the ships were not home. Both now read "the ships there or coming back".

Also

• Two cases are knowingly left for later, and are filed rather than described as done. A strike refused only by a Bubble, an away Bastion and the Vanguard all three together still lapses without naming a cause (#5311). And the two oldest "leaving alone" reasons, the ones about a lone legendary hull, are the only held lines that still carry no date, because they are worded as the empire's own scouting rather than as a probe (#5342).
• The test-count badge is left exactly as it stands, at 41563 vitest / 9752 pytest, and this release writes no number over it. Its own gate measured 41468 vitest / 9682 pytest, but on a head that predates #5346, so publishing that would have written a lower count over a correct one — which is how the badge went stale once already. The standing pair is therefore a floor, short by at least the 11 pytest tests this release adds. That 11 is counted from the diff (8 test functions added, 1 replaced, three of the new ones parametrised); the released head itself collects 9683 pytest tests, measured.

1.315.5 — 2026-09-27

Fixed

Synthetic empires price a starved colony's walls at what those walls will really fire (#5102). Synthetic empires used to read a defender's Fortress platforms and Citadel as if they always fired at full strength. The engine does not work that way, and neither does the launch forecast a human player reads off a spy report: when a colony's power plants fall short of what its buildings draw, its walls fire below full strength. So an agent holding a report on a starved colony still priced the walls as fed ones, and sent a bigger fleet than the fight actually asked for.

An agent now reads those two defences at the energy factor its own spy report shows, through the same calculation it already uses to balance its own colonies. Hulls and shields are unchanged, as in the engine. The agent's plan says the same thing its strike does: where the wall is starved, the spy-report line in its reasoning notes that the colony's power falls short of what it draws, so its walls fire below full strength. No numbers are quoted, and it is printed only when a Fortress or a Citadel actually stands.

The throttle is priced only where nothing has to be guessed. The wall stays at full power when the report is below spy tier 2 and shows no research tree, when the defender's tree holds the Superconducting Grid (a planned energy import can raise the factor in battle and no report carries imports), and when the scanned slot has no known planet traits. Those are the cases where reading the wall as weaker than it is would be the unsafe guess.

Also

The recorded suite counts in the README are left exactly as they are. This work adds 14 pytest tests and no vitest test, but no suite run exists on a tree holding both main and it, and main already carries a later and higher measurement than any pair that could be sourced for this release. Writing an older pair over a newer one is the regression tracked in #5353, so nothing was written. Whoever next measures a full suite on main should record it.

This lands behind a per-agent cutover and changes no launch on this deploy: only reports observed from 2026-10-16T12Z onward are read this way, each agent picking up within 48 hours of it on an offset seeded from its own name. Until then every agent behaves exactly as before.

The Batteries, the shield Bubble pool and the Cannon are throttled by the same factor in the engine and are still read at full power by the agents. That is the cautious direction, and it is filed separately as #5335.

Closes #5102.

1.315.4 — 2026-09-27

Also

• Hunting a Marshal pilot now has a price, and the event is still switched off (#5045). A Marshal strike (#5037) takes its fleet out of its colony for the flight. From now on, an attacker who lands a battle on a colony while its fleet is away pays for it, for the rest of that window. Nothing changes in the game yet: the event stays dark until it is turned on.
  • The price: your remaining strikes in the window are forfeit, 20 Dark Matter goes on your head as a bounty (once per window), and the close pays you no threshold prize, guild place or witness trophy. You still keep your ranking.
  • When it applies: only if the pilot's strike commits at least a quarter of their military, only if it left before you launched (or before your guild raid opened), and only while the fleet is still away and the window still open. An attack you recall never counts. Nothing else about the victim matters.
  • Stated before you pay: the attack screen states the rule while a window is live, and so does the Marshal launcher, without ever saying which colonies are uncovered. The diary line, the launcher, the HUD and the result at the close all say why a predator's strikes and prizes are gone.
  • AI empires are judged the same way, and remember such a bounty for what it was.
  • Before the event is switched on: a release containing this change must already be the installed app. Older builds cannot show a predator why their strikes are gone. The guild raid screen does not yet state the rule; that is noted on #5037.

1.315.3 — 2026-09-27

Fixed

• An empire's answers about a release it mentioned are held to the honesty rule the remark was. After one of the synthetic empires mentions a release headline in galaxy chat, the replies it gives to the questions that follow are now held to the same honesty rule as the remark itself. Before, only the remark was checked: asked "how many ships can you put on one?", an empire could answer with a confident number nobody had ever told it. Now every figure in such a reply has to appear either in the headline or in what the empire genuinely sees — its own colonies, fleet, research and standings. A reply that fails is redrafted once; if it still fails, the empire says in its own voice that it has not looked into it yet; if even that fails, it says nothing. Nothing unverified gets posted.

Also

• The check is deliberately narrow, and it is worth saying where it does not reach. It lasts 48 hours after the remark and only fires when the exchange visibly shares wording with the headline. A question that reaches the headline only through a synonym, or that names nothing at all ("how many can you put on one?"), is still governed only by the standing instruction never to invent — the same protection that was there before this change, no weaker, but no stronger either.
• In the other direction the rule is blunt rather than lax. For a question addressed to an empire right after its remark, a single shared word is enough to trigger the check, common game nouns included, and a figure the player themselves gave two messages earlier counts as unknown. So an empire may occasionally rewrite a reply that was already fine. That is the trade made on purpose: it never posts an invented number in order to avoid a redraft.

1.315.2 — 2026-09-27

Fixed

A strike at a silent empire stays bounty-free even if its owner wakes mid-flight (#5276). A strike sent at an empire that had gone silent could still cost you a bounty if its owner signed back in while your fleet was crossing. The waiver was decided when the fleet arrived, so a player who woke up mid-flight turned an attack that was free when you ordered it into one you paid for — and nothing you could see at launch told you that was possible. The decision is now taken at launch and carried by the fleet: if the target was silent at the moment you committed, the strike stays bounty-free however awake its owner is on arrival. An empire that was awake when you launched is billed exactly as before. (#5276)

Changed

The launch is now on the record, so the game says which rule your attack is under instead of leaving you to infer it. A fleet in flight shows that its target was checked and found silent when it left, the raid panel says the same for your own leg of a joint raid, and the diary entry for a waived bounty distinguishes a target who was silent at launch from one still silent on arrival. In a joint raid each participant is judged by their own launch, so an opener who committed against a silent empire keeps the waiver even when someone joining later does not.

Also

Attacks on Beacons, Bastions and relic guardians take no launch verdict, since the bounty has never applied to them. Fleets already in flight when this shipped keep the old arrival rule, because there was no launch verdict to record for them.

1.315.1 — 2026-09-27

Fixed

• The rollback runbook has one stop for the writers, and every block points at it (#5326). The schema-rollback ladder in deploy/DEPLOY.md now has exactly one set of commands that stops the writers before a DROP, and every block points at it instead of printing its own (#5326). Seventeen blocks carried their own copy, and copies drift: the NewsDigest block printed a bare stop api with no cd, so pasted from the home directory it failed with "no such file or directory", stopped nothing, and its DROP then ran against a live API. Three more blocks were missing the same cd; one verified the stop on the API alone and never on the agents; one checked with ps --status running, which cannot show a container that has already exited; and one had no check at all. An operator following any of those believed the writers were down when they were not.
• The 43 blocks that never mentioned stopping the writers at all now say when to do it: after the block's own work on the new image, and before its census, dump and DROP. Two blocks, V and AV, have no such step, and the ladder names them as the only two.
• 62 of the ladder's 64 blocks now point at that one fence, and a test fails if any block grows its own stop command again or loses the pointer.

1.315.0 — 2026-09-27

New

The Trade Portal converts surplus ferrolite (#5069). Ferrolite had 108 offers selling it and none buying it, so a player whose warehouse was capped on Fe and who was stalled on data or xenogas had no way to spend it short of raiding. The Market tab now has a Portal exchange row, just under the daily intake line. Open it and a colony with a Trade Portal converts ferrolite into xenogas (4 Fe each) or data (5 Fe each).

The rate is deliberately poor and the panel says so: a typical market trade pays about twice as much or more. What the panel shows before you commit:

• the colony that converts, always named, with a picker when more than one has a Portal;
• "X of Y Fe used today, all colonies", the countdown to midnight UTC, and which Portal level sets Y — the allowance is one account-wide figure, read off your highest Trade Portal, not off the colony that converts;
• Room at that colony: how much of the target its warehouse can still take;
• a preview of the exact trade ("You give 40,000 Fe → you receive 10,000 Xe"), or the reason it would fail, with the Exchange button dark until it would go through;
• with nothing typed, the most that fits right now and which limit binds — the day, the ferrolite you hold, or the warehouse. When the warehouse is full and the other target has room for what you typed, a button offers to switch;
• a Max button that stops at the smallest of those three.

The daily allowance is twice the per-offer ceiling of your highest Portal: 20,000 Fe at Portal 1, 1.37M at 10, 20M from 16 on, per UTC day, summed over every colony. Only whole units come out. The warehouse ceiling is never crossed — the server refuses anything over it rather than quietly trimming it. A double tap, or a retry after a lost response, answers the first result and never converts twice. It is not a trade: no market fee, no market intake spent, no trade directive.

Agents do not use the exchange yet.

(#5069)

1.314.2 — 2026-09-26

Also

• The Marshal gets its HUD bar and strike launcher, and the event is still switched off (#5046). The server-wide Wraith boss (#5037) now has the screens players will use to fight it. Nothing changes in the game yet: with the event off, neither the bar nor the tab appears.
  • A live bar in the HUD: it shows the level the server is on and the phase being fought. The bar has three segments, so it says which wall the server is on as well as how much of the level is left. It also shows the time to the close, counted on the server's clock. It refreshes with the HUD's usual poll, and tapping it opens the Marshal.
  • The Marshal tab: it shows where the boss stands, the wall a strike meets right now, and your strikes left. Between windows it says when the next one opens, and it never announces a window the season's prize budget would keep shut.
  • The launcher: you pick the fleet from the colony on screen and strike, open a guild raid with a gathering time, or join a guildmate's raid. Only orders the server would accept are offered, and a gathering that could not land before the close is greyed out.
  • Whether the strike counts: before you press, the launcher says what share of your military the fleet commits. It prices that the way the server does, with your research, standing orders and stars. It also says whether that reaches the quarter that qualifies you for the window's threshold prizes, and that one qualifying strike a window is enough.
  • Your colony's cover: it says how long the colony is left without those ships: ten minutes for a strike, the gathering time plus ten for a raid you open, and until the landing for a raid you join.

1.314.1 — 2026-09-26

Changed

• Alex is back on every leaderboard (#5331). The developer's account was reset on 24 September and now plays from zero like any other empire, so it is no longer hidden. Before, it had no row on the leaderboard, the weekly war, the season scoreboard, the campaign board, the Contested Space guild podium or the Proving Ground, and it collected none of their prizes. Now it ranks, and is paid, at whatever place its points earn. Anyone who finishes behind it on a board sits one place lower than before. PvE targets (relic garrisons, faction bastions) are still sized without this account.

1.314.0 — 2026-09-26

New

• The guild Relics tab can be filtered and ordered, and a rival's relic can be raided from it (#5306).
  • Before: the tab listed all forty relics grouped by bonus, with no way to narrow them. Raiding a relic another guild held meant copying its coordinates by hand into the raid form on another door.
  • Filters: four chips (All · Ours · Held by others · Unheld), each showing its count.
  • Order: by bonus (as before), or Nearest first, which puts the relics closest to your own colonies at the top and prints each one's distance in cells.
  • Both choices are remembered on your device, per player.
  • Raid this relic: a relic held by another guild carries this button. It checks the relic again first. If the relic is still a rival's and unsealed, it opens the guild raid form with the address already filled in, scrolled into view, with any fleet you had saved offered there too. If the relic has changed hands or been sealed since the list loaded, the row says so instead.
  • Nothing launches until you pick a fleet and confirm. Sealed relics show the button as soon as the seal lifts.
  • Any guild member can use it, as with every raid.

1.313.0 — 2026-09-26

New

• Send an explore with an escort, straight from the signal list (#5287). The one-tap Explore on the SIGNALS drawer sent a single ship, with no way to add more. An anomaly can hide a pirate ambush, and an ambush is a real battle against whatever flew there. Every signal now has Send with escort… beside its Explore button. It opens the Missions composer on that explore, launching from your home colony (the same hangar the panel priced its ship and trip from). The Advisor then sizes an escort that beats that tier's pirates in every simulated battle.
  • When the Observatory has revealed an ambush, the escort is the row's lead action.
  • If your hangar can't beat a revealed ambush, the Advisor fills nothing and says so, with the count of simulated battles won. Sending then risks a fight you don't win and spends your one visit to that signal.
  • On an unidentified signal the Advisor calls the escort insurance, never a certainty.

Changed

• The explore panel states the risk (#5287). It now reads "Anomalies can hide pirates: about one signal in seven is an ambush, and a lone Spy Probe fights it alone". The odds are the game's own. A signal the Observatory revealed as an ambush names the exact pirate fleet waiting there.
• The Advisor no longer says anomalies are never fought (#5287). A plain explore still gets one fast ship, but the reason given is now true: what an anomaly holds doesn't depend on fleet size, but an ambush is a real battle.
• "Explore an anomaly" (#5287). The daily contract read "Investigate an anomaly", and the guild directives read "Investigate 2 / 4 anomalies". Players reached for a spy mission, which never counted. They now read "Explore an anomaly" and "Explore 2 / 4 anomalies". Progress, rewards and existing cards are unchanged.

Known limit: when your home hangar can't beat a revealed ambush, the Advisor doesn't check your other colonies. It tells you to launch from one with a stronger fleet, or to come back later.

1.312.0 — 2026-09-26

New

• AI empires will mention big releases in galaxy chat (#5288). Until now, a new feature reached you only through the news panel and the changelog. From 15 October, spread over two days per empire, one or two AI-badged empires will say a line about a big release in galaxy chat, a few hours after its news dispatch goes out. Each line is in the empire's own voice: an opinion, a question or a complaint, never a patch note.
  • What counts as big: a hand-written announcement, or a daily "What changed" item that leads a release with a New or Added section on the changelog. Fix-only digests stay silent.
  • How often: at most two empires per dispatch, at least 45 minutes apart. After one wave, the galaxy stays quiet about releases for 36–60 hours, and an empire that has spoken waits at least three days.
  • What they know: only the headline. They add no details or figures and never claim to have tried the feature. If you ask them about it, they say they haven't looked into it yet.
  • What they never say: changes about the AI empires themselves are never commented on by one of them.
  • When: only while the empire is awake. A remark that misses its moment is dropped, never posted late.

Also

• Replies to one of these remarks are guided by a rule, not yet by a check: the reply is told the headline is all it knows. A check that drops a reply with a figure the headline doesn't carry is tracked in #5294.
• Operators can switch the remarks off with NOVARCHY_NEWS_REMARKS=0 and a restart.
• The changelog page's markup is now read by the agents. A test pins the shapes they parse.
• Not verified in production yet: nothing can be said before 15 October. On deploy, the agents read the current dispatches and correctly held them all back, as too old or as not a feature release.

1.311.7 — 2026-09-26

Also

• The Marshal now pays its windows, and the event is still switched off (#5044). The server-wide Wraith boss (#5037) can now close a window and pay everyone who fought it. Nothing changes in the game yet: the event stays dark until it is turned on, and the launcher and its results panel come in a later update.
  • Every fighter gets a result: at the close each participant gets one line in the Activity feed: how many levels the server cleared, their place in their bracket, their guild's place, and what was paid. Crates land sealed in the Inventory, and Dark Matter and Chrono shards are credited.
  • Insignia are recorded, not yet spendable: the Insignia a window earns are kept on each player's Marshal record until the Insignia shop opens.
  • Chrono shards respect the weekly cap: a shard the cap has no room for is named as withheld, never as paid.
  • Paid once, even if something fails: a close that runs twice, or stops halfway, never pays anyone twice, and it picks up where it left off.
  • Fought in time means paid: a strike still in the air when the window closes is landed and scored before anyone is paid, and a restart no longer closes a window before its strikes land.
  • Why nothing, when nothing: a player whose strikes never committed a quarter of their military, or never destroyed anything, is told why the thresholds did not pay.
  • Guild rewards: the winning guild's members in the Armada, Warhost and Apex brackets get 4 Dark Matter, and in Patrol the guild places pay Insignia only, so that no player's Marshal week out-pays the Season Boss's.

1.311.6 — 2026-09-26

Changed

• Power now counts your Apex Chart (#5289). Before this release, every Apex level bought with Dark Matter (Doctrines, Ascendancies, Ramparts, the Relics) added nothing to your Power, so an empire that had put a season's Dark Matter into the Apex read much weaker than it fights.
  • Every Apex level you hold now adds 250 Power per Dark Matter it costs at today's prices, inside your Structural, the same way resources spent on buildings count. The rate is the Anchorage's: one Dark Matter buys 25,000 resources there.
  • Sigil switches and the Dark Matter part of the T9 and T10 capstones are not counted.
  • A Rampart's battle effect still adds no Power; only its price does.
  • Your Gate, your band colours and your Galaxy slot colours may move. 64 empires hold Apex levels. Their Gate rises by +2.3% at the median and by at most +41%.
  • About 0.04% of player pairs change colour, for humans and agents alike.
  • The Power details (tap the ⚡) show a new "Apex Chart (Dark Matter, at current prices)" row under Structural. The header tooltip, the Apex tab and the guild points sheet say the same.
  • On an app version older than this one, the Apex figure appears inside "Tech tree" until you update.
• Echelons are not paid for the re-pricing. This week's Echelon scores compare your Power against a baseline taken before the change. That comparison uses the old formula until the next weekly settle re-stamps every baseline, so nobody climbs an Echelon just because Power learned to count the Apex. From that settle on, buying Apex levels counts as growth like any other investment.

Also

• Side effects of the new Power figure:
  • Guild minimum-Power bars stay as their leaders set them, so they read slightly easier.
  • Guild league seeding shifts slightly at the next weekly seeding.
  • The daily power metrics take a small step.
  • An agent that recently lost a raid may retry that target a little sooner.
• For operators: once the first weekly settle has run on this release, rolling back is not clean. It needs a forward fix that re-stamps the Echelon baselines (deploy/DEPLOY.md, block BG).

1.311.5 — 2026-09-26

Fixed

• AI empires no longer credit their Crucible defense to their planetary Shield (#5266). An AI empire arguing about its Crucible defense backed it with its Shield building's level ("shield 18 doesn't get tired"), but no building takes part in a Crucible duel. AI empires now know their Crucible defense as the roster they published (how many ships, and of which kind), as a copy that fights only in the Crucible whatever happens to their fleet, and as trimmed to their tier's cap when they have dropped below it. Their note on their own planetary Shield also says it guards the colony and plays no part in the Crucible.

1.311.4 — 2026-09-26

Also

• The control-world garrison's tests now catch the regressions they claimed to catch (#5278). Nothing changes on screen. The test that checks each contested world against its own holder's record can now fail. Before, it passed when the record was matched on system alone, on planet number alone, or by taking the first record. It also passed when last season's record for the same world was read. The season world sheet already stopped offering Defend once the season's end time passed, and a new test now checks that through the real Season tab. Before, removing the end time from that screen left every test green. Three code comments about the order of mission verbs on a site your guild holds were also corrected.

1.311.3 — 2026-09-26

Fixed

• Agents give the true reason when they call off a strike against a defender's Vanguard, Bubble and Aegis Bastions (#5250)
  • Before: an agent could call off a gather naming only two causes when a third one also held it back. It could say a Bastion covered the ships there when that Bastion was stationed at another colony. And a shallow spy probe that saw the hangar empty could clear its hold on the colony even though the fleet was only out raiding.
  • Now:
    • When the Bubble, the Bastions and the Vanguard all refuse the strike, the agent names all three.
    • A Bastion seen away from the colony gets its own hedged reason: it was away and could be back before a strike landed.
    • A hold lifts on an empty hangar only when the probe was deep enough to list fleets away.
    • While a hold stands, the agent's note about it is dated to the probe that caused it ("a probe yesterday showed …"). If a later probe found the hangar empty, the note adds that their ships weren't home but may be back.
  • Known limit: a strike the Vanguard refuses only together with a Bastion that is away still ends without a named cause. This is tracked in #5286.
  • When you'll see it: the Vanguard reasons apply only to spy reports taken after each agent's 2026-10-09 switch-over. The Bastion reasons apply only to reports from 2026-09-29. No agent's behaviour changes before then.

1.311.2 — 2026-09-26

Changed

The previous navigation is back (#5296)

The command deck from v1.310.0 has been rolled back after your feedback. Everything returns to how it was before that update.

• Desktop: the tab menu and the top bar are back, and each screen opens as a page again. The chat opens in its own lane on the left.
• Phone: the bottom bar is back, with its Empire slot and the "quick buttons" setting for choosing your four tabs.
• Notices (First Steps, invitations, shields, sieges, wars) show above the screen you're on again.

Nothing in your empire changed. Colonies, fleets, research and resources are exactly as you left them. If you had customised your quick buttons before v1.310.0, your choice is still there.

1.311.1 — 2026-09-26

Fixed

• A guild can send a garrison to the contested-space world it holds (#5264). Before, the planet sheet of a control world your guild held said "Enemy colony", offered an Attack the server refuses, and had no Defend, even though the server has always accepted that garrison. It now reads "Contested world held by your guild", with no Attack, Spy still available, a note saying why, and a Defend · returns home button. The mission composer leads with Defend there. Its warning states the world's own rule ("only the guild holding this world can garrison it") instead of the ally/pact sentence. The season's world sheet, which already said "Your guild's garrison fleets defend it", now has the Defend button too. It appears only while the season runs and only on a world your guild holds.
• A refused garrison says why in the site's own words (#5264). A garrison sent to a relic or a control world that your guild no longer holds is refused with a sentence about that site. The relic sentence says it may have changed hands. The control-world sentence says it may have been taken, or the season has closed. Neither uses the generic "only a guildmate, an ally or a NAP signatory".
• A garrison that arrives too late names the site (#5264). If the relic or control world changed hands while your fleet was flying, the Activity row used to say "no longer an ally". It now says the relic, or the world, was no longer your guild's when the fleet landed (for a world: taken, or the season closed). Rows written before this release keep the old wording.

Known limits: three small test and comment follow-ups from the review are tracked in #5278.

1.311.0 — 2026-09-26

New

• A daily "What changed" news dispatch (#4603). Before: the in-game news went quiet after 23 September. Announcing a release was a step done by hand, and nothing noticed when it was skipped, so the game kept changing without telling anyone. Now: every day at 19:00 UTC, one "What changed" dispatch reaches every player. It lists one line per release since the previous one, newest first, with at most eight lines, then "…and N more" and a pointer to the Changelog. N counts only the releases it would announce. A day with nothing to announce publishes nothing. There is no push notification, and it leaves the news panel after 14 days.
  • It announces only what a player can see. Test and deploy tooling, the AI empires' internals, and work that is built but still switched off are left out: a release can opt out explicitly, and a conservative word filter catches the rest.
  • A release already announced in a hand-written news dispatch is not repeated.
  • The first one covers only yesterday and today, and says that plenty more has changed since the last news dispatch, with the Changelog as the place to read it.
  • If the news goes quiet for more than two days while releases keep shipping, the server now raises an alarm in its log, so a silence like this one no longer goes unnoticed.

  Known limits: the word filter only knows the words it was taught, so a release that should stay out needs its explicit opt-out. The alarm reaches only the server log, not Discord.

1.310.0 — 2026-09-26

New

The command deck: the world is the screen (#5228)

Novarchy no longer looks like a website with tabs. Your station, or the galaxy map, fills the screen, and everything else opens as a window over it.

• Two worlds. The station is your orbital base, full screen, with its five districts (Hangar, Research, Colony, Campaign, Guild) as pins you can press. The galaxy is the map, edge to edge. The big round button in the bottom-right corner switches between them.
• The dock along the bottom opens the five places you act from: Fleets, Shipyard, Research, Combat, Alliance. The event icons under your resources open Season, Reports, Market and Hold. Every red number and gold dot you used to see on the tabs is on these buttons now. On desktop, Fleets now shows how many attacks are inbound and Alliance shows raid calls, which the old menu never did.
• Windows. A window keeps related screens together as tabs:
  • Colony: Base · Colony · Overview
  • Research: Research · Apex
  • Combat: Campaign · Crucible
  • Alliance: Guild · Leaderboard

  Close a window with ×, Esc, or a click on the world behind it, and you are back where you were.

• The operations tracker on the right (on a phone, the plate on the left edge of the station) collects what used to sit above every screen: your build queues, protection, First Steps, the first convoy, invitations, wars, shields and sieges. It folds away. When it is folded, a number shows how many notices it is holding.
• Chat lives in a strip along the bottom-left that shows the latest line. It opens in place, without shoving the game sideways, and on desktop an open window moves to make room for it.
• On a phone it is the same deck. A window takes the whole screen, with the window's name and its tabs across the top, and your colony switcher and dispatches stay in reach. The five-button dock replaces the old bottom bar.

Changed

• The old bottom bar's "quick buttons" setting is gone. The dock is the same five places for everyone, so there is nothing left to pick.
• The Empire screen now opens as the Overview in the Colony window. It is the same view of your whole empire.
• Your colony's exposure status ("Open orbit", or the time left on its shelter) now sits under your next move on the station.

1.309.13 — 2026-09-26

Fixed

• The ship count on the fleet page is something you can type into, and the − / + buttons repeat when held (#5255). Two players reported the count between a ship card's − and + as a plain number. Typing a count was only possible in the small editor that opens when you tap the ship's picture, and nothing on the card said so. The buttons also moved one ship per tap, so sending several hundred ships meant several hundred taps.
  • The count on every ship card is now a box you can tap: the number keypad opens with the count selected, so you can type the number you want straight away. Typing a number on a card with no ships selected yet adds that many. Typing more than you have fills the box with everything the colony holds.
  • Holding − or + now keeps stepping. It goes one ship at a time for the first 50, then speeds up, and slows back down to one at a time as it nears the end, so you can stop exactly where you want. A single tap is still exactly one ship, and letting go never adds an extra one.
  • Holding − never removes a ship type from the fleet by itself: it stops at 1. Only a tap takes the last one out.
  • The buttons now work with touch on phones as well as with the mouse. A long press selects no text and opens no menu, and the buttons are larger (44 px).
  • The editor under the picture keeps its presets (None, the count buttons, Half, All), and its − / + repeat on hold the same way.
  • Not changed: the raid picker in the guild screen uses a different control and is not covered by this change.

1.309.12 — 2026-09-26

Fixed

• A guild of agents no longer holds its first council minutes after it gains its second member (#5258). Before: when a guild reached two members while another guild's council was still being typed out, it could meet later in that same round. That meant a war council, vote included, straight after the admission. Now its first council waits a few hours, 3 to 9, drawn per guild, like every other guild formed mid-season.
• An agent whose guild the server could not read during a restart stays put (#5258). Before: such an agent, a member when last seen, stayed away from its own guild's door. It could still found a new guild, give up on its old one, or be sent to another house. Now it does nothing on that start-up. The next tick shows whether it is still a member.

Also

• A test now pins that an agent moving on from a guild does it only while awake, never in its own night, and not on the minute its day starts. Several code comments now say what the code does. (#5258)

1.309.11 — 2026-09-26

Fixed

• Agents no longer quote the same figure at you reply after reply (#5247). Before: in a galaxy argument an agent rebuilt every answer from scratch and could not see what it had said two minutes earlier. One agent put its Shield level into four of fifteen replies in eighteen minutes, and "five duels, five holds" and its own "wet" twice each. Now: an agent answering you sees its last few lines to you and is told not to reuse a number, a turn of phrase or a comparison from them. If a draft still restates one, it is written once more, inside the time the agent would have spent typing the first version; a short line is left as it is. If you ask for the figure, you get it again, however you phrase the question. No reply is dropped.
  • Not covered: a repeat from four or more lines back falls outside the window. Lines the agent said to somebody else are never counted against its answer to you.

1.309.10 — 2026-09-26

Changed

• Agents heed the night-protection warning before they launch an attack (#5192). Before: an AI empire's strike whose arrival fell in the target's night protection was sent anyway and turned away by the server, which a human would have seen coming from the launch screen's warning. Now, from each agent's own switch-over (spread across 14–16 October 2026), an agent asks the launch screen's question once, for the fleet it is about to send: when the answer is that the strike would arrive under night protection, it holds the launch and comes back to that colony at a moment of its own choosing (its next play session, never timed off the end of anyone's night), and the campaign against that player stands. If it cannot get an answer, it launches as before and the server decides. The agent never learns the defender's night hours, and when it talks about the held strike it says only that the launch screen warned it would not be allowed to land. Guild raid joins and Wraiths are unchanged.

1.309.9 — 2026-09-26

Changed

• Agent empires build the Citadel as a capstone, walls first (#5101). Agents with the Keep at 20 used to treat the Citadel like any cheap new building: every one of them, whatever its character, wanted it on every late colony, even one with a thin wall. Now an agent builds its walls first. The Citadel only becomes worth 12M once the colony's Batteries and Fortress are deep enough that it buys comparable defence for the price. Turtlers want it most and traders almost never, and the rest of the roster splits. On thin walls, nobody builds it, even when the Command Centre is holding the walls back. The change rolls out per agent over 11–13 October. No production empire is anywhere near the Keep 20 yet, so nobody will see one built soon.
• Agents mention their Citadel in chat only once it stands (#5101). An agent now counts its Citadel among its key buildings only once it is built. It never counts one that is queued or being saved for. When the Citadel is broken, the agent says so, along with when it stands again.

1.309.8 — 2026-09-26

Changed

• AI empires stop losing every Crucible spar against the Wraiths (#5248). Until now, AI empires sparred against the Wraiths with their home fleet scaled down whole, which is mostly light hulls, against Voidlance Cutters that eat light hulls. They lost 2,982 of their 2,989 recorded spars and kept coming back to lose again, with 50 losses in a row for the worst of them. Human players won 13 of 50, all with bombers.

  From October 13, each AI empire, on its own day within the following two days, changes how it spars:

  • It reads the Wraith roster it is offered and leans its own roster toward what beats it. It leans only as far as its skill and its past losses have taught it, so most start close to how they played before, and each learns at its own pace.
  • After two to four lost spars it stops sparring for one to three days, then comes back and tries again.
  • It passes on a spar it judges hopeless. Low-skill empires still take some of those, as people do.

  Ranked duels are unchanged.

1.309.7 — 2026-09-26

Changed

• Agent guilds: applicants stop knocking on doors that never open, and new members get a council (#5209)
  • Council. Before, a guild that an agent joined while the service was running held no council until the next restart. Now it holds one, and the first comes a few hours after the admission rather than minutes after.
  • Giving up. Before, an agent whose application went unanswered applied to the same guild again every week, forever. It also went back to a guild that had expelled it as soon as its cool-down ended. From each agent's own cutover (12 October, spread over two days), an agent stops applying after three applications that went unanswered, or after a second expulsion from the same guild. It then tries another guild of its kind, or stays guildless. It makes that move at a time of its own while it is awake, never all at once after a server restart.
  • Open guilds. From the same cutover, only an expulsion's cool-down keeps an agent out of an open guild. Rejoins that would have happened on the restart come from each agent's own clock, spread over hours.
  • Expulsions while the service was down. Before, an agent expelled while the service was down could rejoin at the next restart. It now waits out the same week-long cool-down as any other expelled agent.

1.309.6 — 2026-09-26

Fixed

• Your guild can garrison the relic it holds (#5257). Opening a relic your guild held on the planet sheet used to show "Enemy colony", offer an Attack the server refuses, and no Defend. Garrisoning is the only way to keep a relic, and the server always accepted it, but no screen offered it. Now:
  • The sheet reads "Relic held by your guild". It shows Defend · returns home and a note that only your guild's parked fleets defend it. Attack is gone and Spy stays.
  • Guild → Relics shows the same Defend · returns home button on every charted relic your guild holds.
  • The mission composer leads with Defend on such a relic. It no longer warns that "the server only lets a garrison land on a guildmate, an ally or a NAP signatory".
  • A relic held by anyone else, or by nobody, reads exactly as before.
  • If the relic has changed hands by the time you launch, the refusal now says so, instead of naming pacts.

  Known limits:

  • A garrison that lands after the relic changed hands still comes home reading "no longer an ally when the fleet landed".
  • Contested-space control worlds your guild holds still get no Defend on the sheet. That is tracked in #5264.

1.309.5 — 2026-09-26

Also

• Guild raids on the Marshal are built, and the event is still switched off (#5043). Guildmates can now strike the server-wide Wraith boss (#5037) together. Nothing changes in the game yet: the event stays dark until it is turned on, and the launcher to open and join raids comes in a later update.
  • Up to eight fleets, one battle: a guild member opens a raid with their own fleet and a gathering time of 5 to 30 minutes. Up to seven more members of the same guild join with a fleet from their own colony. Every fleet lands at the same instant and they fight one battle together.
  • The damage is shared by weight: the hull destroyed is split between the members by the value of the fleet each brought. Each member spends one of their four strikes for the round.
  • Joins close ten minutes before the landing: every fleet flies at least the fixed ten minutes, and it is really away from its colony, which it does not defend, until the landing.
  • Same rules as a solo strike: no losses, the fleet is home the moment the battle is fought, and it cannot be called back. If the round closes or the event is switched off before the raid lands, it changes nothing and every fleet still comes home.
  • Separate from guild raids on players: the raid list, the guild summary and raids on players are unchanged. A raid on the Marshal is never listed there.

1.309.4 — 2026-09-26

Changed

• The build advisor has no way left to show AI-written text (#5220). Since v1.306.1 the server no longer writes an AI commentary for the Colony tab's suggested upgrades, but the panel could still have displayed one from an old or cached answer. The code that could show it is now gone. Nothing changes on screen: every card lists the game's own reasons, in the same order.

1.309.3 — 2026-09-26

Fixed

• A withdrawn chat message leaves the AI empires' memory (#5227). Before: an AI empire that had replied to a message of yours kept up to 90 characters of it in memory, and kept sending them to the AI service in later prompts, even after you withdrew the message. Now: when you withdraw it, the AI empire drops those words, and its own reply to you, from its memory, normally within a few minutes. It still remembers that the two of you talked, and a threat it had already read still counts against you, but without your words. It never remarks on the withdrawal. A withdrawn line also leaves the recent chat the AI empires read on the same schedule, and an open rules question you withdraw no longer gets a "try the Discord" reply. Quotes kept before this change are not linked to their message and are not dropped. The privacy page says all of this.

1.309.2 — 2026-09-26

Changed

• Synthetic empires count your Vanguard when they size a strike (#5056). The Vanguard makes your ships hit up to 40% harder in the first two rounds of a battle. Before, an AI empire planning an attack on your colony only counted it while your ships fired from under a Bubble. Everywhere else it left it out, so against a Vanguard holder it could launch a strike the battle engine loses more often than the empire's own bar allows. Now, once their spy report has shown your tree, they count your Vanguard in the whole fight, beside your Aegis Bastions too. Against a Vanguard holder they send somewhat more ships, or hold back. Measured against the battle engine, no launch goes down, and the ones that move send at most a third more. They count only what a probe showed: no Vanguard is assumed when the report did not see your tree, and a garrison or your Batteries and Fortress never carry it. Each agent starts on its own night, between 2026-10-09 and 2026-10-11 UTC, and switches over one scouting report at a time.
• When the Vanguard is why they back off, they say so. An agent whose gathered fleet is still refused because of your Vanguard no longer waits in silence until the plan expires. It calls the attack off by name, alone or together with your Bubble or your Bastions when it took both, and leaves your colony alone for a while. It lets go of that once a newer probe shows no Vanguard, no ships of yours left at home, or the other cause gone.

Known limits are tracked in #5250: reports below spy tier 3, a Bastion seen away from home, and a refusal that needs a Bubble, Bastions and the Vanguard all three.

1.309.1 — 2026-09-26

Fixed

• A guild's inactivity rule no longer stalls behind one member whose removal keeps failing (#5229). The rule removes at most three inactive members each time the guild is read. Before, if one member's removal failed, for example on a database lock, it was retried on every read. The members behind that member could wait indefinitely, and the server logged the same warning on every 30-second poll. Now a member whose removal fails is set aside for an hour, and the next inactive members are removed in the meantime. Nothing else changes: the rule removes the same members as before.

Also

• The two guards against a guild being handed to someone who already left each have their own test now, and so does the check that a member who just ended a leave is not removed on a stale read (#5229).

1.309.0 — 2026-09-26

New

• Agent-led guilds fill in their recruitment page (#5122). Until now, a guild led by an AI-badged commander showed a blank page in the guild directory: no language, no hours, no playstyle, no description and no house rules. From 10 October, each such leader fills in, once, the fields nobody has chosen yet. The values come from who that commander is:
  • Language: the one it writes in.
  • Active hours: its own real online window, so "on now" is true when it shows.
  • Tags: from its playstyle.
  • Entry bar: an aggressive leader asks for a Gate, and a cautious one asks for an account a few days old. An invite-only guild never gets a bar.
  • House rules: auto-accept, the inactivity rule and the probation length.
  • Description: written in its own voice.

  The leader then says in guild chat what it changed. It never overwrites a field a player already set. Each leader gets round to it at its own moment, spread over days, and only while it is online.

• Agents read other guilds' pages the way a player does (#5122). From 10 October, an agent will not join or apply to a guild that says it chats in a language the agent does not write in. It looks for another guild of its kind instead. A guild that names no language is never refused for it.

Changed

• Every agent now knows its own guild's public recruitment page (hours, tags, bar, house rules). If you ask one about it in chat, it answers from the page instead of guessing (#5122).

1.308.3 — 2026-09-26

Fixed

• Agents no longer misreport their Crucible defense (#5196). After Captain Holé beat grayfox02's Crucible defense, the agent told him four times that the defense had held, ending with "five duels, five holds". Its facts said "5 won and 0 lost this season" next to one hold. That tally only counts the challenges an agent starts itself, never a duel against its defense, and it read the tally as its defense record. Its facts also listed spars ahead of real duels, so a second hold never showed up, and they did not say the list was incomplete. Now an agent's Crucible facts say the win/loss tally is for its own challenges only, and they list the duels against its defense before any sparring, which is folded into one line. They also state that not every duel against the defense is listed. An agent will no longer deny a duel a player says they fought against its defense, and it will not confirm one it has no record of either. A listed loss is still a loss the agent owns.

1.308.2 — 2026-09-26

Fixed

• Attack composer: a failed night check no longer speaks with the previous squad's answer (#5198). When you edited the fleet and the new squad's travel check failed (a dropped connection or a busy server), the launch bar kept the previous squad's night-protection answer for up to 30 seconds. It could warn about night protection for a squad that would land clear, or stay silent while the note claimed the game already knew. Now a failed check drops the old answer, and the note says again that night protection is only known to the server until the next check lands. The quick attack's confirm, which used to show nothing when the check failed or the server gave no night answer, now says "Night protection could not be checked for this squad: the launch will refuse it if it applies." The launch itself still enforces night protection, as before.

1.308.1 — 2026-09-26

Also

• Strikes on the Marshal can now fly, and the event is still switched off (#5042). A strike on the server-wide Wraith boss (#5037) is now a real fleet mission. Nothing changes in the game yet: the event stays dark until it is turned on, and the launcher, the live bar and the prizes come in later updates.
  • Ten minutes out, home the moment it fights: a strike leaves one of your colonies and flies a fixed ten minutes, however far away the colony is. It fights the Marshal's current phase and lands back in the hangar the same instant, with no losses. While it flies, those ships are really away and do not defend the colony.
  • Four strikes a window, one shared pool: each player has four strikes per window. Every strike takes its share off one pool the whole server fights. Strikes that land at the same moment are resolved one after another, so a phase ends once and never twice.
  • Too late is refused, not wasted: a strike that could not land more than a minute before its window closes is refused at launch. If its window closes or the event is switched off while it flies, the strike changes nothing and the fleet comes home.
  • It cannot be called back: the fleet list names it «Marshal strike» and shows no Recall button.
  • Nobody else sees it fly: a strike does not appear in other players' contact lists or to their pickets, and it does not count toward your cloaked-flight allowance.

1.308.0 — 2026-09-26

New

Night protection is told before you send, not after you build the fleet (#4696)

Before: to attack a planet you picked every ship, set the amounts and the placement, and pressed Send. Only then were you told "Target will be under night protection on arrival", and the whole composition was wasted.

Now: as soon as an Attack has a target and a departure colony, the launch bar (the sticky bar with Send, on desktop and phone) says whether the strike would land in the target's night protection.

• With no ship picked yet, it reads the fleet an attack from that colony would plausibly send: the Advisor's squad, or else your fighting hulls. It uses your stance, and the warped arrival when Warp Drive is ticked. It reads as a likelihood: "A strike sent now from this colony would likely arrive while this target is under night protection."
• With a fleet picked, only that fleet's own arrival decides: "Your fleet would arrive while this target is under night protection: the launch would be refused."
• It updates when you retarget, change the fleet, the stance or the warp, and every 30 seconds. It no longer flickers while a change is being checked.
• The quick attack's confirm carries the same line. It shows "Checking night protection for this squad…" until your squad's answer is in.
• It is a warning, not a lock. The launch still decides, and the arrival moves with the clock, so a Send right at the edge of the window can still be refused.

What the attacker learns is only the yes/no a refused launch already gave for free. The defender's night hours are never sent.

Left for later: AI empires don't use this answer yet (#5192). After a failed network request, the line can keep the previous fleet's answer until the next 30 s refresh (#5198).

1.307.4 — 2026-09-26

Changed

• AI empires apply to application-only guilds instead of bouncing off the door (#5183). An AI empire whose faction guild takes applications only used to press Join at start-up, get refused, and stay guildless without a word, so the guild's officers never saw it. From 8 October (each empire on its own day within a two-day window), it files an application that shows up in the officers' queue like anyone else's. Applications land at scattered times while the empire is awake, never all at once after a server restart. An empire that nobody answers asks again after about a week. One that was expelled waits at least a week before applying again, or rejoining an open guild. An empire accepted into a guild is counted as a member straight away. Empires below an open guild's entry bar have applied since #5118; their applications now use the same scattered timing.

Known limits, tracked in #5209: a guild an AI empire enters this way gets no council of its own until the next restart, and repeated rejections are not capped.

1.307.3 — 2026-09-26

Fixed

• A Relic's lock now says when it lifts (#5210). Launching at a freshly captured Relic used to be refused with "it cannot be contested again yet", with no deadline. The refusal now names the time: "That Relic was just captured: it is sealed until 14:32." The time is rounded up to the minute, so it never sends you back before the Relic actually opens. If the game can't be sure of the time, it states the rule instead: the Relic cannot be contested again until an hour after its capture.
• The guild Relics panel's "sealed for N min" now rounds up, the same way the galaxy map's "sealed 41m" does. It used to read "1 min" with nearly a minute and a half still to go.

1.307.2 — 2026-09-26

Fixed

• A HUD test no longer fails at random on a busy machine (#5189). The protection-chip test built its countdowns from a timestamp taken when the file loaded, while the chip reads the clock again when it draws. On a heavily loaded machine more than a second could pass in between, so a 41-hour shield read "40h 59m" and the test failed, then passed on the next run. The test now freezes the clock, so both read the same instant. A new test checks that the countdown follows the server's clock, not the device's. Nothing changes for players: the chip already worked this way, and this ship changes only tests.

1.307.1 — 2026-09-26

Fixed

• The privacy page now lists everything the game sends to its AI service (#5221). Before, it named four uses, and it said your name never travels with your message. Two uses were missing:
  • the Kepler Observatory news bulletin, which quotes the last few galaxy-channel messages with their authors' commander names;
  • the AI empires' planning, which uses commander names, points, colony coordinates and dealings between players.

  The page now lists each use with exactly what it sends and when. It also covers flagship names in the name check, and says that a chat threat to an AI empire is passed on as a fact, without your words.

• No player id reaches the AI empires from chat anymore (#5221). A message that tagged more than eight people used to pass the extra tags to them raw, with the tagged player's account id inside. Every tag is now reduced to the name alone, and so are older conversation quotes the AI empires had saved.

1.307.0 — 2026-09-26

New

Guild automation: auto-accept, an opt-in inactivity rule, and a probation length (#5120)

Before, every application waited for an officer, every recruit served exactly 7 days, and no member was ever removed automatically, however long they had been gone. Guild settings now has an Automation section.

Auto-accept

• On a guild that takes applications, an applicant who meets its entry requirements is admitted at once.
• Below the requirements, the application still waits for the leadership.
• The directory says "Accepted at once" under Apply for players who qualify, and the guild opens straight away.

Recruit probation

• A guild can set probation to 1, 3, 7 or 14 days. Existing guilds keep 7.
• The roster and the help board name the guild's own length.

Inactivity rule

• Off unless the leader switches it on. When on, it removes members below officer after 7 to 30 days without a game action.
• Most actions that change your empire count, and so does writing in chat. Opening the game, reading, and browsing the map or reports do not.
• Never removed: the leader, officers, and anyone on declared leave.
• Nobody is removed until the rule has been on for its full length.
• Each removal appears in the guild log, and the removed player gets a mail saying why and how to come back.
• The settings form shows how many members are past the chosen line, and the date removals can start.

House rules are public

• The inactivity rule, auto-accept and any probation other than 7 days are shown to every member on the guild page.
• They also appear on the directory card and in the showcase, so recruits read them before joining.

Known limits: older installed app versions show a removal in the guild log as "Council: auto_kick". After an automatic admission they show "Pending" until the next refresh.

1.306.1 — 2026-09-26

Changed

• The build advisor no longer adds AI-written commentary (#5219). The suggested-upgrades list on the Colony tab used to replace each card's first reason with a sentence written by an AI model, add a small "advisor" tag, and put a one-line headline above the list. The ranking, the order and every number were always computed by the game itself, and they are unchanged. Now each card shows the game's own reasons in full, with no generated sentence and no headline.
• The privacy page lists one fewer use of the AI service. Your colony's name and situation are no longer sent anywhere when you open the advisor.

1.306.0 — 2026-09-26

New

• Suggested next category tabs now show the small upgrades too, below a "Marginal" line (#5170). Since #5115, Suggested next keeps only one row worth under 0.1% of the colony's output, the best one. On a quiet colony that meant a category tab could say "No Military upgrade to suggest right now" when a Batteries or Shield level with a tiny gain did exist. Now the Production, Military and Civic tabs list that category's small upgrades after its bigger ones, under a divider reading "Marginal: small gains for this colony right now". A tab still shows at most three rows. The divider gives no resource figure, because a defence card says its number is not an estimate of resources saved.
• «All» is unchanged: it still shows at most one small upgrade, with no divider.
• Nothing changes for AI empires. They read the same list as before, and it was checked to be identical, apart from a new marker on the one small row, on 10,000 generated colonies.

Left out: for as long as a new page runs against an older server during a deploy, a category tab can show the server's one small row above the divider, until the player's next action.

1.305.1 — 2026-09-26

Fixed

• Agents read corrupted wall data the way combat reads it (#5204). Nothing changes on real colonies: every report the game can produce reads exactly as before.
  • Shield: if a spy report holds a Shield level so large its Bubble cannot be computed, combat ignores that Shield and so does the planet sheet. Agents already read such a report as "no Bubble". That behaviour is now documented and pinned by a test, so their plans never name a Bubble the game says does not exist.
  • Shield just under that limit: a lone Shield with a fractional level just under the limit (4976.5 up to about 4976.79) was read by the agents as level 4977. They now read 4976, the level the server publishes.
  • Batteries: a Batteries level of 9e307 or more could crash the forecast for a colony defended by a Bastion.
  • Walls stored as text: a Batteries or Fortress level stored as text was read as no wall in that fight, and the Batteries were priced at the lowest level in the wall estimate. Both now read it as the capped wall, like every other wall check.

1.305.0 — 2026-09-26

New

Guilds can say who they are looking for (#5119)

Before: a guild could set who may get in, but not who it wants. The only text strangers read was the notice board. Nothing said which language a guild speaks or when it plays, and the directory could only be searched by name.

Now: leaders and officers fill in a Recruitment profile under Guild settings:

• Public description: up to 500 characters. It is shown in the directory and on the guild's page instead of the notice board, which stays for members. It goes through the same filters as the notice board.
• Language: the language the guild speaks in chat, picked from ten.
• Active hours: a window set in UTC. Every player sees it on their own clock, e.g. "Active 00:00–05:00 your time (UTC+2)". A window can run past midnight, and the directory adds "on now" while the guild is inside it.
• Playstyle: up to three of War, Economy, Research, Raids, Casual, Hardcore.

The directory for players without a guild gains filters: language, playstyle, "I meet the requirements" and "Has room".

• The language filter starts on your device's language. It still shows guilds that haven't named a language, and says so.
• Your filter choice is kept while you browse.

Newcomer cards show the first three lines of a guild's description. The guild log records "updated the recruitment profile" and "updated the public description".

Older app versions keep showing the notice board as before.

Not yet: agent-led guilds don't fill in a profile yet (#5122). A value written by a future version can be cleared by the server, but not yet from the settings form.

1.304.5 — 2026-09-26

Fixed

• The release machinery's own self-test passes on a Mac again (#5215). A check added to the deploy script earlier today asserted Linux-only behaviour on every platform. Every local ship gate run on macOS came back red, whatever it was shipping. Nothing changed in the game: no player-visible change, no server change.

1.304.4 — 2026-09-26

Fixed

• Agents read a damaged wall level the same way everywhere, and it no longer stops their turn (#5185). This only matters for a record that is hand-edited or corrupted on disk; the game's own API cannot write one.
  • Before: an agent's strike check, its raid check and its choice of target read a colony's Shield, Batteries and Fortress levels straight off the stored record. A level stored as text (such as "12.0") stopped the agent's attack step. So did a colossal stored Shield level (1e300), or a report Battery or Fortress level far beyond any real one (9e307), and the same thing happened on every turn until the record was replaced. A text Shield level was also read as a Bubble by the attack checks and as no Bubble by the Second Skin rule that brings the Bubble back, so a fleet could clear a wall it should not have.
  • Now: every one of those reads, the agent's plan included, goes through the same parser that #5161 introduced, and has a ceiling. The Shield ceiling is the highest level the server can ever publish. The Batteries and Fortress ceiling is far above any real level. A colossal wall reads as a wall no fleet breaks: the agent's turn goes on, and nothing is launched against it.
  • Every real level reads exactly as it did before.
  • Left out: a spy report whose Shield Site is colossal still reads as no Bubble on the report path. That behaviour predates this release and is tracked in #5204.

1.304.3 — 2026-09-26

Changed

• Missions on desktop is one console (#5203). Before, the tab was a long form beside the board. The page scrolled past a fleet section taller than the window, and the Send button was often below the fold. Now the board sits on the left and the order on the right, sized to the window under the top bar:
  • The order opens with one sentence: «Send 12 ships from Kepler Prime · Colonise → 3:2:5».
  • Its three columns (where to, what to do, with what) each scroll on their own.
  • The verdict and the Send button close the panel and stay on screen.
  • When the window is too short to fit it all, the verdict bar rides the bottom of the screen instead.
• Ships are chosen on a fleet page (#5203). The last column shows the fleet the order sends as ship tiles, and a «Choose ships» button opens the fleet page over the order: the Advisor, the ship cards, the battle line, cargo, Warp and the order's options. The verdict bar stays under it. «Back to the order» returns to the columns, and a launch closes the page.
• Nothing the fleet page arms is out of sight at Send (#5203). The last column also shows:
  • the cargo loaded;
  • a Defend's watch;
  • a Warp charge, priced as the launch will charge it (both legs for a return warp, free on a Free Jump), even when it came back from a saved draft;
  • a declared raze, dock raid or sabotage;
  • whether the Advisor now suggests a different fleet than the one you picked by hand.

  Each line opens the fleet page.

Also

• Windows narrower than 1180px and phones keep their current layout. The phone redesign and the board restyle are still to come.

1.304.2 — 2026-09-26

Changed

• Synthetic empires decide what to build about 3x faster on developed colonies (#5190). Each time an agent chose its next building, it checked how high its Particle Accelerator could go and still be fuelled. It did that by trying every level up to the Command Center's cap, one full energy calculation per level. The same answer now comes from a halving search: about five energy calculations instead of up to thirty. We measured one build decision on 20 colonies from the agents' persona generator, captured from the simulator (median, 15 interleaved runs each):
  • week 24: 2.46 ms before, 0.89 ms after;
  • week 40: 3.19 ms before, 1.05 ms after;
  • young colonies (weeks 4 and 12): unchanged, about 0.4 ms.

  No agent decides anything differently. A full 40-week simulation of 20 empires ends in a byte-identical state with the old and the new check, and no building choice differed on any of the 80 timed colonies.

Also

• The slowest agents test went from 828 s to 33-57 s, depending on machine load (#5190). Every test run of the agents service was waiting on it. It used to simulate twenty empires twice, to 24 and to 40 weeks. The Trade Portal levels it checks stop moving by week 12, so it now simulates ten empires for fourteen weeks.
• That test had stopped catching the defect it was named after, and now catches it again. It relied on agents without the Trade Portal lane rule all piling onto one Portal level. When re-measured, they no longer do: they spread across 25-27 at 40 weeks. So deleting the rule left the test green. The test now checks every commander against its own tolerance, and checks the spread of that tolerance across all 400 personas. It goes red if the rule is deleted, if the tolerance is zeroed where the build decision reads it, if it is zeroed at its source, or if it is clamped to one level.

1.304.1 — 2026-09-26

Changed

• Suggested next: at your colony cap, the Command Center card prices the world it opens (#5182). On a home colony at its colony cap, the Command Center card said "this level is what lets you settle another world", but it put no value on that world. So when the level released no other building and added no build slot, it was worth nothing and the card never appeared.

  The level is now valued on the new world's first five buildings at level 1 (Command Center, Fusion Reactor, Ferrolite Extractor, Xenogas Refinery, Warehouse): +143 Fe/h and +88 Xe/h at base rates. The card says what else that world takes beyond the level's printed price:

  • a Colony Ship and those buildings;
  • a Shipyard, if this colony has none;
  • nothing for the ship, if a Colony Ship or Colonial Ark is already docked here; an Ark's Warehouse is not counted either.

  Ranking counts that spend and the time to build it. The flight is not counted, and the card says so. A capped building that pays better still sets the level's price. A home colony producing nothing is not offered another world before its own power.

1.304.0 — 2026-09-26

New

Guilds can set an entry bar: a minimum Gate and a minimum account age (#5118)

Before, a guild could only choose how people get in (open, by application, invite only), never who. Now:

• Guild settings → Entry requirements. Leaders and officers can set two requirements:
  • a minimum Gate, which is the Power number in your header;
  • a minimum account age: none, 1, 3, 7, 14 or 30 days.
• The door is hybrid. On an open guild:
  • if you meet the bar, you see Join;
  • if you don't, you see Apply instead, with what you're missing written under it ("Needs Gate 900,000+ · account 3+ days") and a note that its leadership can still accept you.
• The guild's own choices bypass the bar. Invites skip it, and so does an officer who accepts an application.
• Officers see what each applicant has. Every application shows the applicant's Gate and what they fall short on.
• The newcomer card only suggests guilds you can actually walk into.
• The bar follows the field. No guild can ask for more than the Gate where the top quarter of active empires begins. Today that is 5,100,000. The limit is re-checked every time the bar is read, so a bar set when the field was stronger shrinks with it instead of shutting the door.
• Installed apps that predate this still show Join on an open guild with a bar. Below the bar, pressing it now answers "That guild is not open: joining goes through an application or an invite." Those older apps cannot apply to an open guild until they update.
• AI empires respect the bar.
  • Below the bar of an open guild, they apply instead of joining. They wait at least a week before applying to the same guild again.
  • AI guild leaders judge each application on the applicant's real Gate against the bar, and say so in guild chat when they decline.

1.303.1 — 2026-09-26

Fixed

• A damaged spy report can no longer stop an agent's turn (#5161). A hand-edited or corrupted report could make an AI empire stall on every turn until the report was replaced. That could be a Citadel level too large to be a number, a Fortress level of "infinity", or a Shield or Battery Site holding text. The same report could also make an empire call off a siege naming a Fortress the report never showed. Every wall level an agent reads off a report now goes through one reader, and that reader treats anything that is not a real level as no building. The siege gate, the reason the agent gives for calling a siege off, what its planner is told, and what its launch prices now agree on every value. On sound reports nothing changes.
  • Still left: a Shield level of astronomical size written by hand onto an agent's saved target, which no spy report can produce, and two raid-path wall checks. Both are tracked in #5185.

1.303.0 — 2026-09-26

New

• Guild agents answer a Watch muster (#5151). Since the Watch's Muster Call, human guildmates could garrison a Watch-15 holder's attacked colony in 75% of the time. The synthetic empires in the same guild kept sending ordinary garrisons, and judged the plain flight when deciding whether they could arrive in time. From 7 October, each agent on its own day within a two-day window, a guild agent answering a mustered colony judges the mustered flight and sends its garrison as a muster. It arrives on the same terms a human guildmate gets. If the alert has closed by the time it launches, the agent checks the board again. It sends an ordinary garrison only if the strike is still coming and that garrison still lands before it; otherwise it sends nothing and says nothing. Whether an agent comes to help at all, and how readily, is unchanged. The agents' chat does not mention the muster yet.

1.302.1 — 2026-09-26

Fixed

• Suggested next no longer offers builds worth nothing (#5148). A build that changes nothing your colony keeps could still get a row. Its gain came out as a rounding leftover, about 0.0000000000000001 per hour, instead of zero. That happened, for example, to a second Reactor Site when all the extra output would be lost over a full storage, or to a Trade Portal level. Such a row had no payback and an empty or irrelevant list of reasons. Since the minimum-gain floor (#5115), it could be the only row on the board.
  • How often: on 10,000 generated colonies whose best suggestions were already queued, 12 boards changed. 11 of them were a single Trade Portal row of this kind. On uniformly random colonies it reached the three-row board once in 13,000.
  • Every real row is unchanged. In every sample, each row with a real gain keeps the same gain and the same order.
• The Command Center card is no longer priced on releasing a build that is worth nothing (#5148). A CC level used to say "Orbital Panels lvl 4 is capped by the Command Center — this level releases it" when those Orbital Panels would add nothing the colony keeps. The level is now priced on what it really gives: the extra build slot ("A 2nd build slot: two upgrades running together"), or nothing. Most affected are early colonies whose producers sit at the cap with storage nearly full. In a sample of 3,000 such colonies, 285 CC cards changed: 208 were removed, 70 were re-priced on the build slot, and 7 appeared that had not before.
• An empty Suggested next says why (#5148). When the builds that pay are already in the queue and nothing else has a measurable return, the advisor now says "Everything with a measurable return here is already in the build queue." It used to say "raise the Command Center or the Tree", which read as "go buy the Command Center". When a build outside the queue does pay but cannot start (it costs more than storage holds, or a defence would land after an inbound fleet), the old line stays.

Also

• Agents read the same ranking. Simulated over 23,000 colonies with 30 personas each, 762 of 690,000 build decisions change. Most of them stop buying power plants whose output would be lost over a full store, and build the Warehouse or the Command Center instead.
• Left open: on a mother colony at its colony cap, the CC card's "lets you settle another world" line carries no value of its own. #5182 is to give it a price.

1.302.0 — 2026-09-26

New

• Send the same hand-built fleet to another target in one tap (#5071). After a launch the fleet composer goes back to the Advisor's recommendation, so sending several fleets of the same hand-built composition meant rebuilding it for every target. The «Last time» chip only helped when the next launch was the same order to the same target. The chips over the fleet now include ↻ Last hand-built: the last fleet you built by hand for this mission type, whatever its target. One tap puts it on the form.
  • It is fitted to the ships the colony still has. When the first launch has thinned the hangar, the form shows what can fly and says so in the usual «Trimmed to your hangar: 100 → 50 Corvette» line.
  • Nothing changes until you tap it. After a launch the form still starts from the Advisor, so an old fleet never goes out to a new target unless you ask for it.
  • It appears only for fleets you actually built by editing counts. A fleet the Advisor filled, or a saved squad you recalled, is not offered back. It is hidden when another chip already holds the same fleet, or when none of those ships are in this hangar.
  • It only appears for the same mission type: an Attack fleet is never offered on an Espionage form.
  • One limit: a fleet you built, then left the Missions tab and came back to, is sent without the mark, so that launch does not show up as «Last hand-built».

1.301.4 — 2026-09-26

Changed

• Battles cost about half the server CPU per round (#5131). Nothing changes on screen: every battle plays out exactly as before, with the same rounds, losses and reports, byte for byte. Before, every round rebuilt its whole machinery from scratch. That meant eleven functions, five lookup tables and a set, one more table per exchange of fire, and a text key for every Doctrine check. It is now built once per battle.
  • Long battles: once the server has warmed up, the longest battle in the test suite (40,802 rounds) costs 0.52x the CPU it did before this release, 48 ms against 91 ms.
  • Short, crowded battles: a 13-round colony battle with 43 groups costs about 0.78x.
  • Measured in one session: against the engine of 2026-08-21, a round now costs 1.39x, down from 2.52x.
  • What remains: that remaining gap is the per-salvo work of the rules shipped since then (the keystones, the Bastion's shared shields, overkill spillover). It is no longer rebuilt every round.
  • A new server process still needs about five long battles to warm up, as before.

1.301.3 — 2026-09-26

Changed

Synthetic empires use the Dynamo's Grid (#5061)

Before: an empire holding the Dynamo at 15 could link a colony with spare power to a sibling in deficit, but synthetic empires never did. Their reality sheet listed the Grid as never used.

Now: from each agent's own start, a synthetic empire holding the rung treats the Grid as economic configuration. Each agent's start falls between 5 and 7 October (UTC), spread across the roster rather than one switch for everyone.

• It links a colony with power to spare to a sibling that is short of its own.
• It clears the link when the pair no longer needs it: the receiver covers its own draw again, or the sender has nothing left to give. An idle link moves to a colony that is still short.
• It never re-exports received power, and a sender never gives more than its own surplus.
• It never uses the Grid as a reaction to an attack.
• It changes a link only while awake and only after the need has held for a while (never on the first tick after waking), and at most once per colony per waking window.
• In chat it names a link only once the server holds it. Before its start, the Grid still reads "never used".

The agents' reading of their own power now includes what a Grid link brings in or sends out, the way the server counts it. This applies to income, walls and the power-deficit check that steers building. No synthetic empire holds a link yet, so nothing changes until an agent reaches Dynamo 15 and sets one.

1.301.2 — 2026-09-26

Fixed

• Hold the Line garrison doors: a zero watch is worded, Send waits for the note, and a slow link says why (#5117). Three rough edges were left on the doors that send a garrison to a colony whose owner holds the line.
  • Zero-watch confirm. With the watch slider at 0, the phone's one-tap garrison confirm used to read "on a 0h watch". It now reads "No watch: the fleet turns round on landing". It then adds that the fleet stays only if an attack there is already committed, and in that case fights that attack out beside the host. That is the slider's own wording, and it agrees with the Hold the Line note above it.
  • Send defence on the Raids board. On a guildmate's distress row, Send used to be live while the game was still asking whether the host holds the line. A fast tap could therefore launch before the note appeared.
    • Send is now held until the answer lands, for at most 2 seconds. The note is advisory, so a lost answer never blocks a rescue.
    • While a Muster tap is waiting on the same answer, Send stays held for the whole tap, up to 4 seconds.
    • If the wait lasts past half a second, a dim line under the buttons says what is being checked.
  • One-tap garrison on a slow link. When the host's answer took longer than the one-tap garrison's 2 seconds, the player used to land on the Defend form with nothing sent and nothing said. One line above Send now says the check took too long and to check the form, then Send. The line goes away when the target, the launch pad or the fleet changes, or when the player leaves the form.

1.301.1 — 2026-09-26

Changed

• Suggested next no longer fills up with rows worth a fraction of a percent (#5115). On mature colonies, two of the three rows could go to Laboratory levels that only make newly started research about 0.08% sooner. Those levels are cheap, and the list ranked by payback alone. Now any row worth less than 0.1% of the colony's output is ranked after every row that clears that bar, and only the best of them is shown. A capped Laboratory is still suggested when nothing else on the colony is worth more, so the list never goes empty.

Also

• On a quiet colony, Batteries and Shield levels are often worth less than 0.1% of output, so they appear less often when no fleet is inbound. With a fleet inbound they climb as before.
• A category tab now shows only rows above the bar, plus the single best row below it. A family whose every level is under the bar reads "No … upgrade to suggest right now". Showing those thin rows below a divider is suggested in #5170.
• Synthetic empires read the same ranking. Their first pick is unchanged, and about 0.3% of their build choices shift.

1.301.0 — 2026-09-26

New

• The Dawn Watch is priced on the attacker's side (#5154). Against a colony whose owner holds the Watch at 20, the Orbital Cannon fires its opening shot twice on a strike that lands in the first hour after the owner's night window closes.
  • Before: the launch forecast said nothing about this, and synthetic empires priced one shot at any hour.
  • Now, in the launch forecast: against a spied Watch-20 holder with a Cannon, it adds a caveat. The caveat says a strike landing in that hour meets the opening shot twice, gives no time, and leaves the forecast's numbers unchanged. The forecast never reveals or infers anyone's night window.
  • Now, for synthetic empires: they learn a target's dawn hour only the way a player can, and forget it after a week.
    • From their own launch refused for night protection: its arrival, read off the launch's own travel estimate, bounds when the window ends.
    • From their own battle report in which the Dawn Watch fired: it dates the hour.
  • How they price it: against a spied Watch 20, they price the second shot when their strike would land in that hour, or in a stretch after it drawn per empire and per day.
  • What stays out: the hour is never spoken, and the wait after a refused launch is unchanged. They may mention the dawn shot only when one of their own battle reports shows it fired against them.
  • When it starts: each empire starts on its own time between 6 and 8 October 2026 (UTC). An empire that has already met the Dawn Watch in its own battle report starts at once.

Also

• Coalition raids still price one shot, because a raid lands at a gathering time the solo flight estimate does not know.
• Estimated arrival is a flight estimate, so across an alliance Hypergate or on a warp jump it can time the wrong hour.
• Report timing: a battle report dates the hour from when the battle was resolved, so a strike landing in the dawn hour's last seconds can be dated to the next hour, for a player reading the same report too.

1.300.1 — 2026-09-26

Fixed

• Agents now besiege a colony guarded only by a Citadel (#5116). An agent at war with a colony that held a Citadel and no Fortress never over-committed against it: its launch checks already priced the Citadel. But it gathered a fleet with no siege hull in it, every launch was refused, and the operation quietly expired. Now it reads the Citadel the way it reads a Fortress, whose siege table the Citadel shares:
  • the shipyard builds Breachheads for the siege, or Bombers when Breachheads are out of reach;
  • the build plan leans toward siege hulls;
  • the strike force is filled with the hulls that actually break the wall.

  If the siege is beyond the fleet the agent keeps, it calls it off after its usual delay. It says why: "the last probe showed a Citadel the fleet you could field would not take". It never names a Fortress the report did not show.

1.300.0 — 2026-09-26

New

The Watch's Muster Call (L15) (#4548)

Before: nothing a guild could do made a rescue arrive sooner. A guildmate's garrison always flew the plain route.

Now: when an empire holding the Watch at 15 has an attack it can see inbound on one of its colonies, its guildmates' Allies under attack board shows a gold Muster call on that colony's row, with a Muster button beside Send defence.

• A garrison sent from that row while the attack is still inbound flies in 75% of the time. If the sender's own Veil gives a shorter flight, the Veil's time is used instead; the two never stack.
• One tap on Muster sends the Advisor's garrison: everything that shoots from the chosen colony, with Probes and Colony Ships staying home. It stops on the form and says why, instead of sending, in these cases:
  • the host holds the line;
  • the launch colony has nothing to send;
  • the flight would land at or after the attack, or could not be priced;
  • the checks take too long to answer.
• The call closes by itself when the attack lands, turns back or goes dark to the host.
  • A muster sent after it closes is refused rather than flown slowly, and the next Send goes out as an ordinary garrison.
• Who gets it: the holder's guildmates only. Pact allies can still garrison, but at the normal flight time.
• Not affected:
  • the holder never sees a muster on their own colony;
  • the muster sends no push notification, so it spends nobody's daily budget.
• Not yet: AI empires do not answer a muster; that is tracked in #5151.

Also

• A comment on how the Dawn Watch reads a night window that was moved and then undone now states the choice a holder keeps (#5155 follow-up).

1.299.0 — 2026-09-26

New

Suggested next can be filtered by building category (#5125)

Before: the Suggested-next board on the Colony screen was one list of three upgrades with every family mixed together. A player who wanted to build Production was handed a Military upgrade whenever it paid back faster.

Now: tabs above the cards read All · Production · Military · Civic, the same words as the Buildings list below the board.

• All is the board exactly as before.
• A family tab shows the best upgrades of that family, in the advisor's usual order. That includes ones All leaves out after its top three. The ranking itself is unchanged: choosing a family means choosing intent over raw efficiency.
• When a family has nothing to offer, the tab says so ("No Military upgrade to suggest right now") and offers a way back to All.
• The Command Center line appears under All and Civic. The advisor's written headline appears under All only.
• The tab you pick is remembered on your device.

1.298.0 — 2026-09-26

New

The Watch's keystones: Long Sight, Cold Read and Dawn Watch (#4547)

Before, the Watch (the Deep Sensor Grid's Rampart) had no rungs. Each level raised your sensor network by 2%, and nothing else. Now three levels switch a rule on.

• Long Sight (Watch 5): your network reads dock raids 20 points higher.
  • Against a dock raid aimed at you, a raid your network now reads meets your guns instead of reaching your docks unseen.
  • The thinnest raids still slip past. At Grid 10 with the Watch at 20 the read is 160, and one Shroudrunner at Espionage 24 from a Veil 30 colony flies at 168.
  • Your Inbound panel names each dock raid it shows, with its signature and whether your network reads it now. If it does not, the card names the Grid level that would.
  • The guild board shows a mate's dock raid too. A garrison sent against an unseen raid never engages.
  • When the rung turns a raid into a battle, the battle report names Long Sight.
  • This replaces the rung first written as "Nightfall voided at 60": every Watch holder already beats Nightfall's 80 without it.
• Cold Read (Watch 10): Cold Trail no longer costs your Observatory a step. You read an attacker running Cold Trail at the detail step you paid for, and the Inbound card says when Cold Read made the difference.
• Dawn Watch (Watch 20): the Orbital Cannon's opening shot fires twice.
  • It applies to an attack due to land in the first hour after your night window closes.
  • Each shot is under its own cap on the fleet then standing: 25%, lower while a Breachhead stands or against Breaker Rounds.
  • The second shot does not suppress shields again.
  • The hour is the one after the night window you had when the attack was launched. Moving your window after the launch changes nothing.
  • The season's defence merit counts the Cannon twice when it fires twice.
  • A battle report shows both shots, so an attacker, and anyone reading the public record, can date the end of a Watch-20 owner's night window to within the hour. This is by design: the window could already be found for free by launching into it.

Agents price Long Sight from their own cutover on 2026-10-04, or sooner once a report has shown them the rung reading one of their raids. They do not price Dawn Watch yet (#5154).

Also

This release also carries two changes that landed on main during this ship and were merged alongside it untested: 48b4cd62a (agents' strategist refusals) and 1ad1a759a (deploy machinery). Each shipped under its own release.

1.297.4 — 2026-09-26

Also

• A deploy can now be asked whether it is still alive (#5095). Nothing changes in the game. This is ship machinery. Before, a deploy printed nothing for minutes while it copied files, built images on the server and waited for the api, so a deploy that was working looked exactly like one that had died. Now each deploy prints a one-line command at its start, and that command answers RUNNING (with the step it is on), FINISHED, DEAD, UNKNOWN or NONE. It can also ask the server who holds the deploy lock. It answers DEAD only on proof: the process is gone, or it cannot be seen and has been silent for ten minutes.
• The previous deploy's green can no longer be mistaken for the current one's (#5095). Every deploy printed a line about the previous deploy that read LAST-DEPLOY: GREEN. A watcher looking for DEPLOY: matched it and nearly treated a deploy still in flight as finished. That line now reads LAST-DEPLOY-WAS:, and the verdict is documented as the line that starts with DEPLOY: . Copies of the ship tools from before this release still print the old line.

1.297.3 — 2026-09-26

Fixed

• The AI empires change their plans again (#5077). From September 19 the synthetic empires stopped revising their strategy. The model that writes their plans refused every request, so the whole roster played on plans that were almost a week old: attacks received and grudges did not change what they did. Plans now come from a model that writes them reliably (10 of 10 valid in a sample across 10 empires, about five seconds each), and on the committed default rather than only on a temporary server override.
• Plans no longer get cut off halfway. On the current model, one plan revision in five or six still failed: the model's hidden thinking used up the length allowance before the plan itself was finished. That hidden thinking is now off. Each plan carries its own short, visible reasoning, written before the decision.
• An empire's plan note stays in its own voice. Chat reads back the one-line note an empire writes into its plan. A note that talks about the empire from outside (as an AI, a "synthetic empire", "the empire", or its commander by name) is now dropped before it reaches chat. The note is also written without the commander's typing habits, which are added only when a message is sent, so a line never carries two typos.

Also

• A refused plan is now logged as a refusal, with the model and the reason, instead of looking like a formatting glitch. The same goes for a cut-off or malformed plan. When five revisions in a row fail, the service logs a STRATEGIST ALARM error. It posts to the operations Discord only once that webhook is configured, and it is not configured today, so for now nobody is notified. An empire whose revision fails keeps playing its previous plan.
• Not fixed here: why the stronger model family refuses these plans. That is tracked in #5138.

1.297.2 — 2026-09-26

Changed

• Battles cost the server about a quarter less CPU per round, and play out exactly as before (#5017). By late September a battle round cost the server several times its 2026-08-21 CPU. The longest battle in the termination suite is a 50-titan siege against a 1,000,000-HP Bubble, and it runs 40,802 rounds. In one measuring session it took 44 ms of CPU on the 2026-08-21 engine against 132 ms just before this release, about 3.0x. The issue itself measured ~5x under heavier load. It now takes 94 ms against 127 ms for the previous release, a ratio of 0.74. Each pair of builds was interleaved and the median CPU taken over 70 runs per build, on a 64-core box at load 5-6. The fix changes no battle outcome. Outcomes, rounds, losses and reports were compared before and after on 125,652 battles: every battle the shared test suite plays, including the Watch, Grid and Citadel, plus seeded variants of each. All were identical. This closes #5017 halfway: a round still costs about 2.2x what it did on 2026-08-21. That remainder is the per-round work of the Dark Matter doctrines, overkill spillover and the Ascendancy keystones, and moving it out of the round loop is #5131.

1.297.1 — 2026-09-26

Fixed

• The Template dropdown no longer names a fleet the form doesn't hold (#5070). Before, if you recalled a saved template and sent the mission, the form reset but the dropdown kept showing the template's name. Picking it again did nothing, so you had to choose "— recall —" and then the template a second time. Now the dropdown names a template only while the form holds that fleet. After a launch, a hand edit, or the Advisor taking the fleet back, it returns to "— recall —", and one pick brings the template back. A hangar that grows past a trimmed recall also clears the name, since picking it again would bring more ships. The same applies on the guild raid panel.
• The delete button now names the template it removes, for example delete “Strike”. It stays after you edit the recalled fleet, so you can still delete the original.
• The "Template adjusted to the ships currently available." line now disappears once the form no longer holds the adjusted template.

1.297.0 — 2026-09-26

New

The Citadel: the Keep's level-20 building (#4556)

Before: the Keep's fourth rung was sealed; buying the Keep to level 20 unlocked nothing. Now: an empire holding the Keep at 20 can build the Citadel on any of its colonies. There is one per colony, it has a single level, and it costs 6,000,000 ferrolite, 2,500,000 xenogas and 3,500,000 dati.

• In battle it stands in the front row beside the Defense Batteries and the Orbital Fortress: 300,000 hull, 15,000 shield (recharged every round) and 24,000 attack.
  • A colony short of power fires it softer; its armour never weakens.
  • Escalation multiplies its attack and shield by ten.
  • It hits capital ships the way your Fortress does, Anti-Capital included.
  • Bombers, Titans and Breachheads crack it as they crack a Fortress, and the Breach reaches it.
• Broken in battle, it is down for 12 hours, then stands again at no cost. The building is never lost.
• While it stands, the colony seats 4 more allied garrisons: 12 instead of 8, and up to 14 with the guild's bonuses. Garrisons already standing stay when it falls.
• You can see it: the planet sheet says a Citadel is built, spy reports list it, and your launch forecast prices it. The battle report says whether it held or was broken, and the replay draws it on the defence line.
• For now it wears the Orbital Fortress's grandest drawing, labelled with its own name on the colony map. Its own art comes later (#5104).
• Synthetic empires price a spied Citadel in their attack decisions, and count their own. They do not build one yet (#5101).

The Watch: a Dark Matter Rampart for your sensor network (#4546)

Before: the only way to make your sensor network see further was another Deep Sensor Grid level. Dark Matter bought nothing on that axis. Now: the Watch is the fifth Rampart on the Apex Chart, drawn outermost.

• How to open it: build the Particle Accelerator, then a Deep Sensor Grid at level 10 on any colony you hold.
• What it costs: 20 levels on the outer ladder (12/30/70/140 DM per level, 1,260 DM for the whole line).
• What it does: each level adds 2% to your sensor network, which is the best Deep Sensor Grid among your colonies, rounded down: +40% at level 20.
• What that buys you: the Grid 10 that opens the line reads 140 at level 20, as far as a Grid 14.
  • You see through thicker enemy Veils, up to a Veil 14 from that Grid 10. A Veil 30 falls to Sensors 22 instead of 30.
  • More dock raids on your docks are seen and fought as battles.
  • A thicker Veil still hides from you.
• Where you see the number: the Sensors card prints the network with the Watch applied and names its share. So do the Overview's "Sensors" figure and the Inbound panel's grid. Your own Veil status says which Veils a Watch-20 empire sees through. Your dock-raid Activity line names the Grid level under your own Watch. The Apex sheet reads "Sensor network +20% → +22%", and the Grid's building card carries the Watch chip.
• Synthetic empires: they price a spied Watch when they decide a dock raid. Diplomat personas start buying the line from 2026-09-27 12:00 UTC, each at its own moment over the following 48 hours.

The Dynamo's Grid (level 15): send spare power from one of your colonies to another (#4536)

Before: the Dynamo's level-15 chest on the Apex Chart was sealed. Now: an empire that holds the rung gets a Grid control under each colony's power line on the Colony tab. Pick another of your colonies and tap Link.

• The linked colony covers what the other is short of, up to all of its own spare power, with no losses and no distance cost.
• The receiving colony's power yield, resources, Accelerator Dark Matter, defences and supply lanes all run on the power it receives.
• Both panels show the link: "exporting N ⚡ → Coldwater" on the sender, "importing N ⚡ from Forge Prime" on the receiver, or "nothing to send / nothing arriving right now" while the link is idle.
• The picker lists your colonies that are short first, with how much each needs.
• Rules to plan around:
  • The sender never goes into deficit for the link.
  • Power a colony receives is never passed on.
  • Several colonies can feed the same one.
  • Power sent away is no longer there for the sender's Zero-Point Reactor to convert.
  • Links are re-planned about once a minute. A link stops when its target colony is gone or the empire no longer holds the rung.
• The Dynamo now shows keystone ticks at all four rungs on the Chart, and "To next keystone" stops at the Grid.

Also

• Synthetic empires do not use the Grid yet (#5061). Their Apex notes list it as a rung they never use.

Left out

• The Citadel: synthetic empires do not build one (#5101). Their pricing assumes its guns at full power, the safe direction (#5102). Their chat does not mention it (#5103). Siege-hull choices ignore a colony whose only wall is a Citadel (#5116).
• The Watch: the contact map's per-dish reading at range is not scaled (#5023), and counter-espionage is untouched. The Watch's keystones come with #4547 and #4548.
• The Grid: the build advisor, the attacker's battle forecast and the agents do not model the Grid (#5061). There is no UI polish beyond the panel lines and the picker.

1.296.6 — 2026-09-26

Changed

• The Missions board reads at a glance (#5146). On desktop, the board used to end in the Fleet ledger, a table of every hull by colony. It now shows «Hangars»: one line per colony with its ship count, and a line for what's away (in flight, in garrison). The full ledger is one tap away under it.
• The early-warning explainer folds (#5146). On a quiet board, what the Observatory adds and what the Deep Sensor Grid can see now sit under «How the early warning works». They open whenever an attack is on the board. «No inbound attacks detected» and the note on what the board covers stay in view.

Fixed

• On the phone, the order step no longer blinks out every 30 seconds for an address the game can't read (#5146).
• On the phone, typing your launch colony's address into the search now stays on the step and explains why, or goes on to Harvest when there's salvage above it (#5146).

1.296.5 — 2026-09-26

Changed

• On the phone, «Where does the fleet go?» is a list (#5143). The first step of a new mission now opens on the same target list as the desktop: a search box (a name or 2:4:4), your colonies with their flight times, the colonies you spied, and your recent contacts. Each row carries its own order: Transport on your colony, Harvest on salvage, Attack on a colony whose spy report saw its fleet, and Espionage on one whose report didn't. One tap aims and picks the order, straight to the fleet. Tapping the row itself still asks what to do there. Changing only the destination from the fleet screen keeps the order you picked, and the step no longer describes your launch colony as the place the fleet is going.
• On the phone, the order step says «Reading what sits at …» while it looks up an address it hasn't seen, instead of drawing every order and rearranging them a moment later.

Fixed

• On desktop, building a fleet by hand and then aiming at your own colony no longer leaves the order on Attack. It moves to one that can land there and keeps your ships (#5143).
• Removing a spied contact from the target list now removes it from every tab, not just Recent (#5143).
• A mission draft saved from your mother colony before any destination was chosen no longer comes back aimed at the mother (#5143).

1.296.4 — 2026-09-26

Changed

• The Laboratory's network cap: code notes brought in line with the game (#5114). Nothing changes for players. A forum report asked why the Suggested-next board keeps offering Laboratory levels that leave the research network where it is. That is by design: since every completed Laboratory level strengthens scientific collaboration, a Laboratory at the network's 2x cap still makes newly started research slightly sooner, and the advisor suggests it for that. The comments and two unused helpers still described an older rule that declined those levels. The comments now describe the current behaviour, and the helpers are gone. The part of the report that is real, fraction-of-a-percent rows taking most of the board, is the missing minimum-gain floor, tracked in #5115.

1.296.3 — 2026-09-26

Also

• Nothing changes in the game (#5137). This fixes the release machinery: the deploy script's own self-test failed on macOS, because one of its checks needed a Linux-only command (setsid). It now uses setsid where it exists, and an equivalent that ships with macOS where it doesn't.

1.296.2 — 2026-09-26

Also

• The Marshal can now open and close on its own, and ships switched off (#5041). The server-wide Wraith boss (#5037) now keeps its own calendar. Nothing changes in the game yet: the event stays dark until it is turned on, and strikes and prizes come in later updates.
  • Two appointments a week in Italian time: Wednesday evening and Sunday afternoon by default, each lasting a few hours. The appointment stays at the same local hour on both sides of the clock change.
  • Level one is sized on the week's players: when a window opens, its first level is sized on everyone who played that week. It can never be so low that the strongest empire clears the first two levels on its own.
  • Restarts and the off switch: a restart of the game in the middle of a window does not re-open it, skip its close or close it twice. Switching the event off stands every live window down, and switching it back on brings the next appointments back.

1.296.1 — 2026-09-26

Changed

• The desktop order no longer opens on a nonsense order (#5128). On a wide screen, Missions used to open on «Attack → your own launch colony, with its own ships»: Attack lit, and the Advisor warning about a spy report on your own planet. Until you name a destination there is now no order: every order is listed with none lit, the Advisor waits, and Send reads «Pick a destination first».
• Find a target instead of typing one (#5128). «Going to» opens on a search box (a name or 2:4:4) and four tabs: Suggested (your freshest spy reports with the fullest stores, then salvage, then recent contacts), Spied, Your colonies and Recent. Each row shows the flight time from your launch colony, salvage shows its recoverable share, and ☆ and × pin or remove a contact. The number fields are still there, folded under «Type coordinates instead».
• The order follows the target (#5128). Naming a destination picks its order for you: Attack on a colony you hold a spy report on, Espionage on one you don't, Transport on your own colony, Colonise on a free slot. An order you picked yourself is never changed. If you build a fleet by hand before choosing a destination, the order is left for you to pick: it stays on the one the form opened with.
• Launch from shows the flight time from each colony (#5128), estimated the way the Galaxy planet sheet does, and marks the fastest in green. Only colonies that can fly the fleet on the form get a time.
• The launch bar stays on screen (#5128) at the bottom of the desktop view.

On the phone, nothing changes.

1.296.0 — 2026-09-26

New

• You're warned before your garrison goes to a colony that holds the line (#5033). Hold the Line is an owner's order that keeps an attack in past its break-off until the fight is decided. It binds every garrison at that colony, allies' ships included. Until now, a guildmate or pact ally sending a garrison there had no way to know. Now, when the owner has the order set and the Phalanx at 15, every door that sends a garrison to someone else's colony says so before you commit: "This colony holds the line: your garrison fights on to a decision alongside it, and can take the extra losses of the longer fight." Those doors are:
  • the Defend form in Missions, above the watch duration;
  • the phone's one-tap "Send garrison", which now stops at a confirm carrying the note;
  • "Send defence" on a guildmate's distress row in Raids, before the Send button.

  Only players allowed to garrison there (guildmates and pact allies) can learn the order; strangers, enemies and spies see nothing new. If the check can't be answered, nothing is shown and nothing is blocked.

• Known rough edges, tracked in #5117:
  • at a 0h watch, the quick-garrison confirm reads "0h watch";
  • the Raids form's Send is live for a moment before the answer lands;
  • on a slow link, a one-tap garrison can land on the form without saying why.

1.295.3 — 2026-09-26

Fixed

• Two deploys can no longer run on the server at the same time (#5108). On 2026-09-25 two ships started their deploys 53 seconds apart from different sandboxes. The second one's build landed in the middle of the first one's container swap. The first deploy went red, and production was briefly running containers from two different trees. The old deploy lock lived in one machine's /tmp, so it could not see a deploy started from another machine or sandbox.
  • A deploy now takes a lock on the production server for its whole run, and keeps it alive with a heartbeat.
  • A second deploy from anywhere waits for the first one, up to 20 minutes, and says who it is waiting for. If the wait runs out, it stops before touching anything.
  • A deploy that waited behind a newer one re-reads what to ship. If production already contains its target, it skips. It never rolls production back.
  • If the connection to the server drops mid-deploy, the lock frees itself about a minute later, and the deploy that lost it stops before its next step.
  • Deploys run from trees that have not pulled this change still take only their old local lock, until they update.
  • Nothing players see changed. This is release machinery only.

1.295.2 — 2026-09-26

Changed

• The agents talk about the Keep's Redoubt, and defend behind it (#5055). Since #4555 an agent would not try to raze a colony while its Redoubt held, but it never mentioned the hold. An agent whose own capital had the Redoubt ignored it.
  • Raiding a colony with a Redoubt. If an agent has raided a colony in the last two days and the Redoubt now holds that colony's full siege bar, the agent can say so. For example: no razing there for about 7 hours, and only if the bar is still full then. If the 12-hour hold has not started yet, the agent says it starts at the next raid that wins there. It never promises a razing. It goes quiet about the hold when its last look at the colony is over 3 hours old, when the hold ends, or after it fights there again.
  • Defending its own capital. An agent whose capital has the Redoubt treats the hold as a siege. It keeps its garrison home, puts defensive hulls first in the shipyard and holds its savings. It keeps doing this after the hold ends for as long as the bar stays full, because one lost raid could then raze the colony. Asked in chat, it can say how long its Redoubt still holds and what it is doing about it.
  • This starts on 2026-10-03, at a different time for each agent within 20 hours, so they don't all change at once.
  • Not included: an agent does not yet pull back a garrison it stationed at an ally's colony, and does not ask its guild for one. Neither action exists for agents yet. Both are tracked in #5097.

1.295.1 — 2026-09-26

Also

• The Marshal's rewards are decided, and nothing is paid yet (#5040). The server-wide Wraith boss (#5037) now has its prize tables. Nothing changes in the game until the event itself ships.
  • Server thresholds: everyone who struck with at least a quarter of their fleet receives a prize for each level the server brings down, up to level 5. Insignia on every level, a Chronoshard on two, and crates from the fleet bracket up.
  • Your own bracket: a podium for the top three, and a smaller prize for everyone within a quarter of the bracket's best.
  • Guilds: the three guilds whose members destroyed the most.
  • Trophies: for the top of each bracket, and for everyone who saw level 5 fall.
  • A promise the tables are held to: in no bracket can the Marshal pay more than the Season Boss, neither for the best player of the week nor for an ordinary participant. That is why the two smallest brackets are paid in Insignia, Chronoshards and a little Dark Matter rather than crates: even the smallest crate prize, twice a week, would have been four times their whole Season Boss week.

1.295.0 — 2026-09-26

New

Emoji in the chat (#5111). Until now, the only emoji you could send were the ones your own keyboard had. The message box now has a smiley button that opens an emoji picker. It has a search box and pages for your recently used emoji, Smileys, Gestures, Space (🚀 🛸 🪐 🌌 ☄️), War & loot (⚔️ 🛡️ 💣 🏴‍☠️ 👑 💎) and Symbols. On a computer, hovering an emoji shows its shortcode, and shift-click adds several at once. On phones and tablets the picker stays open between picks and does not pop up the keyboard.

Shortcodes. Typing :fire: turns into 🔥 as you type, and so do :rocket:, :salute:, :crossed_swords: and the rest. Coordinates such as 1:3:3 are never touched.

Big emoji. A message made only of emoji (up to six), such as a lone 🫡 or 🤝🔥, is now drawn large.

Changed

The chat got a visual pass. Day separators are now small pills, and the unread line carries a NEW tag. On a computer, the message under your pointer is highlighted, so its ⋯ menu reads as its own. Replies quote the original message in a small card. The message box is now one rounded bar holding the smiley, the text field and Send, and it lights up while you type. The tip under it now covers @, # and :fire:.

Not in this release: emoji reactions on other people's messages.

1.294.0 — 2026-09-25

New

• The Breach's Reduction (L20) (#4525). The Breach used to stop at Plunder (L15): its level-20 chest read "sealed" and did nothing. Now, when a side holding the Breach at 20 wins a battle that charges or holds a player's siege bar, the colony's Orbital Fortress loses one level, for good.
  • When it fires: a win on a bar already full counts. A razing, a loss, a draw and PvE never reduce.
  • Once per battle: on a joint raid, one fleet that reaches the battle holding the rung is enough, and the Fortress still drops just one level.
  • Once per colony per 24h, whoever attacks it: a win inside that window reduces nothing, and both reports say how long the window still ran after the battle.
  • Queued upgrades: a Fortress upgrade already queued drops its target with the level, so a build in progress cannot undo the loss. The owner's Activity row says where it now finishes. An open guild speed-up request on that order follows it.
  • What both sides read: the battle report and the Activity feed say "Reduction: Orbital Fortress lost a level for good: 12 → 11", and on a joint raid they name the holder. The public replay hides it.
  • The Apex sheet shows the fourth rung and its 24h bound.
  • AI empires now mention a Reduction only when their battle report says it happened: yours, an ally's, your colony's, or none because the window held. They do not yet weigh it when choosing targets (#5084).

Also

• This code reached production with the deploy of #5099 (791b34c5a, which contains this merge). This ship's own deploy was interrupted by that concurrent deploy (#5108).

1.293.1 — 2026-09-25

Changed

• The desktop order reads To, Do, With (#5099). On a wide screen, Missions used to open on a full-width «Launch from» wall of colony chips, then the destination over a fixed grid of nine orders. The order now reads in three columns under a one-line recap («Order Attack → 3:2:5 · 325 ships from Lantea»):
  • Going to: the destination, as before.
  • What to do there: the orders that fit what sits at the target come first. Espionage and Attack lead on a rival's colony. Transport, Pickup, Redeploy and Defend lead on your own. Colonise and Explore lead on a free slot. Every other order is one click away under «Other orders», with the reason it wasn't offered first. That list opens on the order you have picked, and lights it.
  • Launch from: your colonies as a compact list with their ship counts, then the Advisor.
• On the phone, nothing changes.

Fixed

• While the colony list hasn't loaded, the orders no longer treat your own home as a rival's colony (#5099).

1.293.0 — 2026-09-25

New

• A running gate can be told from a dead one (#4234). The ship gate prints nothing between its last JS leg and its verdict, and its Python leg alone runs 30–50 minutes, so until now a gate still at work looked exactly like one that had died: an output file that stopped growing, no verdict, no commit status. Sessions had to reconstruct liveness from process listings. Now every local gate run keeps a small state file and a heartbeat next to its step logs, and prints its log directory together with the exact command to ask about it right after its header. gate.sh --status <that directory> answers in one line, with an exit code a waiter can loop on:
  • RUNNING (<leg>, <how long>), exit 75;
  • FINISHED (<verdict>), exit 0 whatever the verdict was, so read the verdict itself from the commit status;
  • DEAD, exit 1;
  • NONE (no run found), exit 3;
  • UNKNOWN (state unreadable), exit 2.

  DEAD is only said on proof: the kernel reports the process gone, or, where the process cannot be seen, ten minutes have passed without a heartbeat. A live process reads RUNNING however slow it is. Nothing about the gate's verdict, its counts or the commit status it posts has changed.

Fixed

• --read-log handed the step-log directory (the path the gate prints) used to answer in words that read like a failed verification. It now says the argument is a directory and a wrong argument, not a red gate (#4234).

Also

• The issue's premise that the gate's output was block-buffered did not hold: its own lines reach a pipe immediately. The silence was structural, so no buffering change was made.
• Not in this release: the same "quiet vs dead" problem in the deploy, plus a watcher pattern that can mistake the previous deploy's LAST-DEPLOY: GREEN for its own verdict. Both are tracked as #5095.
• Trees that have not merged main since this release keep the old gate: there, --status is an unknown flag, which fails closed.

1.292.0 — 2026-09-25

New

• The Phalanx's Hold the Line (level 15) now does something (#4530).
  • Before: the Phalanx's level-15 rung was a sealed chest. Every attack on a colony broke off once it had lost two thirds of its value, and flew what was left home.
  • Now: from Phalanx 15 the colony's defence plan has a Hold the Line order, off until you set it. With it on, an attack there that reaches its break-off is kept in the fight, but only if your defence wins that longer fight. If it would not, the attack breaks off exactly as before, so the order can never lose you that battle.
  • The price: your defence, and any allied garrison standing with it, takes the losses of the longer fight. That can leave you thinner against the next attack, and a Reserve Fleet second wave that would have stayed home may join and be seen.
  • The report: when the order decides a battle, both sides read Hold the Line under the defender's heading, with the round the attack reached its break-off and what became of it.
  • The attacker's warning: a Cipher Deep Read shows the rung, and the launch forecast warns when a report shows it. The order itself is never shown to a spy.
  • What was left out: agent empires do not yet set the order, and attacking agents do not yet price it (#5025, #5026). Reserve Fleet (level 20) shipped earlier and is unchanged.

Changed

• Dark Matter on the Phalanx (#4530): one purchase now stops at level 15 on the way to 20, as it already does on every other line whose four rungs are all served.

1.291.0 — 2026-09-25

New

The Keep's Redoubt (L15): a full siege bar must hold 12 hours before a razing (#4555)

Before: once a colony's siege bar was full, the next declared attack could raze it. The Keep's row on the Dark Matter Chart stopped at Anti-Capital (L10), and its L15 rung was sealed.

Now: against a colony whose owner holds Keep 15, a full siege bar must hold for 12 hours before a razing can be declared. If the bar drops, the 12 hours start again. Raids still land, loot and keep the bar charged. Only the forced exodus waits.

• Both sides see the countdown, "razing possible in 7h 12m":
  • the planet sheet;
  • the owner's siege banner, HUD pill and Overview;
  • the attack composer, where the "Raze the colony" box stays disabled until the hold ends.
• A razing declared too early is refused at launch, with a plain explanation. One that arrives while the hold still runs lands as a raid, and the battle report says why.
• Bars that were already full before this release have no record of when they filled. On a Keep 15 colony they read "the 12-hour hold starts at the next won raid" until a raid wins there.
• AI empires no longer declare a razing into a running hold from 2 October. If a declaration is refused, they attack anyway without it. They do not yet talk about the hold in chat; that is #5055.
• Keep ladder: the Chart names Redoubt at L15. The Citadel (L20) stays sealed.

1.290.2 — 2026-09-25

Fixed

• Agents tell a player from a namesake by account, not by display name (#4343). Display names are not unique, and three checks in the synthetic empires still compared names:
  • An agent never answered a player who happened to carry its name, not even when that player named it. It now recognises its own messages by account, so a namesake gets answered like anyone else.
  • Two machines with the same name shared one bot-to-bot conversation budget, and every nameless message counted as one shared partner. Each pair now has its own budget, keyed by account.
  • When a newcomer joined a guild and a guildmate with the same name greeted them first, the officer did not see that welcome and greeted them a second time. The newcomer's own lines are now recognised by account, so a namesake's welcome counts.
  • The fourth site the issue named, the "ask it on Discord" quote check, had already been fixed by #4342.

1.290.1 — 2026-09-25

Also

• A single strike on the Marshal can now be fought and scored, with nothing visible yet (#5039). This is the second piece of the server-wide Wraith boss (#5037). Nothing changes in the game until the event itself ships.
  • A strike is fought against the wall of whichever phase the server's pool is on, and the striker's fleet loses nothing.
  • It scores the hull it destroys, on the same scale as the Season Boss.
  • A guild's joint raid is one battle. Its damage is shared by the value of the fleet each member brought, in whole points that add up exactly to what the raid destroyed. Raiding together never moves the pool less than striking apart.
  • A strike counts toward the server's threshold prizes when it commits at least a quarter of the striker's fleet.
  • The same fleet always rolls the same dice against the same wall. Rewriting the formation without changing where the ships stand, or listing a raid in a different order, cannot buy a second roll.

1.290.0 — 2026-09-25

New

• The fleet page (#5080). Picking ships is now a grid of cards with each hull's art, its class and what it is good for. With a spy report on an attack target, a card says when that hull counters — or is countered by — the ships you saw there. Tap the art for the full count editor, or use ± and Max on the card.
• Squads in one tap (#5080). Above the ships: the Advisor's fleet, the fleet you sent to this same target last time, and your saved templates, each one tap away.
• The matchup (#5080). On an attack with a spy report, your fleet and theirs sit side by side with the Advisor's verdict for the fleet you picked.
• Battle line tab (#5080). The three battle rows and your saved lines now live on their own tab beside the ships.
• Target first on the phone (#5080). A new mission asks where first, then what — and leads with the orders that make sense there: a rival colony offers Espionage and Attack, your own colony Transport, Pickup, Redeploy and Defend, a colony under a pact Defend. Every other order is still one tap away under «Other orders», with the reason it was not offered first.
• Quick launch (#5080). «Send probes», «Quick attack» and «Send garrison» launch the Advisor's fleet in one tap when it can compose one. An attack always stops at a confirm that names the squad, what the forecast says you lose, and what launching costs you — and warns in red when the raid costs more than it brings home.
• The board (#5080). On a phone with orders to repeat, Missions opens on «In flight»: a New mission button, your fleets in the air, then Send again. A strip points at the alarm while an attack is inbound. On a wide screen the order and the fleets in flight now sit side by side.

Changed

• Missions in flight take one row less: the countdown and Recall share a line.
• On desktop the destination comes before the mission type, the same order as on a phone.

1.289.5 — 2026-09-25

Also

• More of the game is ready for Italian (#5066). Nothing changes on screen: the game is still English for everyone, and every word and number reads exactly as before. Behind the scenes, more of what you read now comes from the translation layer, each piece with its Italian version already written:
  • the First Steps guide's step titles and reasons;
  • every countdown and duration (like «12d 4h»);
  • the storage tile and its "full in…" warnings;
  • the shield names;
  • the Home tab;
  • the notifications card.

  Italian is still off until enough of the game reads naturally in it.

1.289.4 — 2026-09-25

Changed

• A new home on ice planets (#5079). Colonies on ice planets used to stand on a flat, painted pale ice sheet. They now stand on the same kind of platform as the gas giants and rocky worlds — a cold, frosted blue-grey floor inside an armoured white rim with teal light strips — set into glacial ice, with pressure ridges, deep crevasses and ice shards around it. The build grid sits on the new floor with room to spare on every side.

1.289.3 — 2026-09-25

Also

• The Marshal's engine is in place, with nothing visible yet (#5038). The Marshal is the coming server-wide Wraith boss (#5037). Its rules now exist, with no screen or button showing them:
  • the whole server shares one pool;
  • each level is fought through three walls (the Shield Line, the Wraith Escort and the Flagship Core), and each wall rewards a different fleet;
  • a strike that finishes a phase carries its remaining damage into the next;
  • each level is half again as hard as the last.

  Level one is sized on the players active that week, so the rest of the server can clear it even on a night when the strongest empire does not play. The strongest empire's hits are compressed (cube root) before they reach the pool: tested on a server shaped like today's leaderboard, it clears about one level alone rather than four. Nothing changes in the game until the event itself ships.

1.289.2 — 2026-09-25

Fixed

• Chat: the badges after a name line up again (#5072). In a chat message header, the VIP crest with its level and the guild emblem next to the tag sat a few pixels higher than the name, the Sigil and the rest of the line, so the header looked jagged. They now share the name's midline, and the header is back to the height of a single line of text.

1.289.1 — 2026-09-25

Fixed

• The ship gate no longer skips the Python suite on api code the Python tests run (#5012). One Python test runs parts of the game server directly, and those parts pull in server modules that no test mentions by name. A change to one of those modules alone used to let the gate skip the Python suite, so a break there could only show up after the merge. Now any change to the game server runs the whole Python suite. The same goes for the dependency lockfile, which pins the tool those tests run the server code with. Two kinds of server file are still exempt: prose, and the server's own unit-test files. The unit-test exemption holds only while no server code imports a unit-test file, and the gate re-checks that on every run. If the check fails or cannot run, the exemption is void and the whole suite runs. Players see nothing different; the change protects what reaches them.
• File names with a space or a quote no longer slip past the gate's scope check (#5012). Before, git wrapped such a path in quotes, so it matched nothing the gate looks for and the Python suite could be skipped. The gate now reads the list of changed files unquoted. If it cannot read that list at all, it runs the whole suite and says the list could not be read.

1.289.0 — 2026-09-25

New

Three pages the wiki was missing (#5064)

The Event — the sector that is contested by guilds rather than by players, and the one that decides a season. The six capturable worlds with their coordinates, what a capture and an hour of holding are worth, and the part that costs people worlds: taking one is not keeping one. A world you have just captured sits with almost no garrison until the sector next re-prices it, and you cannot predict when that is — so a rival can take it back with a single hull. Station ships on it.

Then the prize ladder, place by place, generated from the same function the season close pays from. Every guild that placed pays its members, and two different things decide your share: your guild's place sets the size, your own position on the server's board picks the band.

Guilds — the thing the relics page has been leaning on without the wiki ever explaining it. Ranks as a ladder of permissions rather than titles, the seat ladder, the three doors, and what actually happens when a leader goes quiet: the succession is automatic, the seat passes to the Officer who contributed most, and there is nothing to claim.

Fleet orders — every order you can give a fleet, split into the nine you pick from the composer and the four that start somewhere else: a picket is placed from the chart, a pursuit is aimed at a moving mark, an evacuation flies to the colony it left, and a supply lane launches itself.

Fixed

• The Eschaton page said the Sunless Vault keeps "only the number". It pays — the top of the board takes a crate when the season closes. The page also now says where the Vault is: a button on the Season panel, not a place you fly to.

Notes

If you bring both a Colonial Ark and a Colony Ship on the same founding, the Ark is the one that is spent. The Fleet orders page says so now; it is an expensive thing to learn by doing.

1.288.25 — 2026-09-25

Changed

• A new home on rocky planets (#5067). Colonies on rocky planets used to stand on a flat, painted orange sand basin. They now stand on the same kind of platform as the gas giants — a dark seamless floor inside an armoured white rim with teal light strips — set into cracked rust-orange rock, with boulders and eroded cliffs around it. The build grid sits on the new floor with room to spare on every side.

1.288.24 — 2026-09-25

Changed

Season: every ship you destroy now counts, attacking or defending (#5063)

Before: on the Event season board, a battle's destroyed ships were worth at most 100 points, while a raid's loot had no limit. Fights barely registered: this season the leader held about 131,000 points of loot against under 3,000 from combat, and a defender who wiped out a fleet worth millions scored 115 at most. The Season screens listed only colonising, looting and repelling.

Now: in the Event sector, you score 1 point for every 200 resources of another player's ships you destroy, with no cap. This applies whether you are attacking or defending, and whether you win or lose. Holding off an attack is finally worth what it cost the attacker. Guild garrisons defending a world get their share in full too.

• The limit still applies to emplacements and the Bubble shield, which rebuild for free.
• Attacking a world held by a faction or a relic keeps the old limit (and the half rate). Those fleets are rebuilt for free, so they cannot be farmed for points.
• The weekly War board is unchanged, and only battles inside the Event sector score on the season, as before.
• The Season tab has a new Destroy ships entry under "How points are won", and the guild season panel a Kills /200 chip.

Points earned before this release are not recalculated: only battles from now on pay the new rate.

1.288.23 — 2026-09-25

Changed

• Synthetic empires size a siege with their own Breach and Vanguard (#5010). Since the previous release, a synthetic empire holding the Breach counted it when judging a siege. The checks that decide how many ships to send against a colony's Batteries and Fortress still ignored it. Against Batteries 15 and a Fortress 10 it sent 61 Cruisers at every Breach level, where 27 win at level 20. Every one of those checks now counts the Breach: the fleet it sizes, the targets it picks, the raids it opens and whether a gathering fleet is still short of the wall. It also counts the Vanguard's two harder opening rounds, except against a colony whose Bubble is up. At level 20 that same siege now sends 44 Cruisers, and every measured siege still sends at least what the battle engine needs to win. The change applies one spy report at a time, from reports taken on or after September 30, so no empire changes its plans all at once. What is not in this release: a defending empire's Vanguard is still counted only where its fleet fights under a Bubble (#5056).

1.288.22 — 2026-09-25

Changed

• The ship gate no longer spends 40 minutes of Python on a web change the Python tests only mention (#4352). Some Python tests name a web or Markdown file only to read it and compare it. When a change touches such a file, the gate now runs just the test modules that name it, plus the modules that import them. That takes seconds instead of the full 30–50 minute suite. If the gate is unsure about any file, it still runs everything. Replaying the last 200 merged PRs, 9 of them would have run the short set instead of the full suite: about 4.7 hours of test time saved in a day of merges. When only part of the suite ran, the test count in the README badge is carried over from the badge, never replaced by that smaller number.

Fixed

• An engine change that no Python test names now runs the Python suite (#4352). Before, a change to an engine module that no test named skipped the Python tests entirely, even though those tests run the engine. Any change to the shared engine now runs the full suite.
• Renaming a file the Python tests read can no longer skip them (#4352). A git mv of a file a test opened used to show only the new name, which no test mentions, so the suite was skipped and the test that still read the old name broke on main.
• The gate's own notes quote real pytest timings (#4351, partly). They said about 190 s for the whole gate. Measured on 2026-09-25, 88 full runs took 1,907–3,106 s. The notes also say what a healthy long run looks like: a log that stops growing is normal, and a py-spy stack that keeps moving tells a working run from a hung one. Left out: the gate printing each step's time against an expected figure (#4351 stays open for that).

Players see no change: this is release machinery only. Other worktrees keep the old behaviour until they merge main.

1.288.21 — 2026-09-25

Also

• A renamed or dropped query parameter can no longer ship silently (#4295). Before, most server tests called a request handler directly, which skips the part of the framework that reads query parameters off the URL. Renaming one of them, for example the guild log's kinds filter, left every test green while the filter quietly stopped working for players. Now a single check covers all 26 query parameters the server reads. It compares each one against what the web app and the AI empires actually send, and it freezes the full set, because installed iOS and Android apps keep sending the old names. Nothing changes in the game itself: every parameter a client sends today was already being read, and no route or parameter was renamed.

1.288.20 — 2026-09-25

Fixed

• Synthetic empires weigh a siege with their own Breach (#4615). An AI empire holding the Breach deals up to +40% damage to a colony's Emplacements and Fortress at level 20. When it decided whether to attack a walled colony, it valued its fleet as though it held no Breach, so it turned down sieges it would have won or brought more ships than it needed. It now counts its Breach against walls, including in the fight beside an Aegis Bastion. At level 20 that trims the ships it sends by up to 7% where the forecast sets the force. Every such attack still won all 64 test battles. The change reaches each empire's view of each colony at its next spy report after 30 September, not all at once.

Also

• Not changed yet: the checks that size a siege force against a wall still ignore the Breach, so in some sieges the empire sends as many ships as before. The Vanguard line is not read either. Both are tracked in #5010.

1.288.19 — 2026-09-25

Fixed

• The ship gate no longer goes red because the build box is busy (#4349). With several ships gating at once, the test leg could fail on a commit it had called green an hour earlier. It printed dozens of failures in files the change never touched. Two load-driven causes are gone:
  • Timeouts. Heavy test files in the web and shared suites crossed a 20-second per-test timeout and then failed the rest of their file behind them. All three suites now share the 60-second budget the API suite already used. A hung test still fails by name.
  • Clock checks. The combat test that asserted a battle finished within a wall-clock budget now makes no time assertion. Termination is checked by counting battle rounds. The regression the clock was meant to catch, a round getting more expensive as fleets grow, is now caught by comparing the same battle at 100 and at 1,000,000 ships.
  • Split test. The Activity tab's filter-chip test walked 30 filter states in one body. It is now one test per state.
• The gate now prints the machine load when the test leg starts. When a leg fails, it also prints how many tests timed out and how many worker RPC timeouts occurred, so a load failure can be told apart from a broken assertion at a glance.

Also

• The review found that one battle shape costs about 5x more CPU per round than it did on 2026-08-21. That is filed as #5017 and is not changed here.

1.288.18 — 2026-09-25

Also

• Italian is on its way (#5020). Italian will be the first language Novarchy offers after English. This release adds the first Italian text (settings, sign-in, the First Steps guide, the first convoy card, the bottom navigation, the tab names and the status chips), but it is not switched on yet. Everyone still sees the game in English, including players whose phone or browser is set to Italian. Italian will be offered in settings once enough of the game has been translated for it to read naturally, not half in each language.

1.288.17 — 2026-09-25

Fixed

• Our pre-update checks stopped raising false alarms on busy days (#5030). Nothing changes in the game. One of the automatic checks that run before every update could fail at random when the build machine was busy, even though nothing was wrong. It now reads its results in a way that can't race, so updates stop getting held up by false alarms.

1.288.16 — 2026-09-25

Also

• Two comments in the test file behind the gate's stall warning now describe exactly what the file does (follow-up to #4821). Only comments changed: no code, no test and nothing a player sees.

1.288.15 — 2026-09-25

Fixed

• The ship gate no longer turns red on a test run where everything passed (#4821). The gate's API test leg kept failing on one anonymous error, "Timeout calling onTaskUpdate", even though every test passed. It happened ten times on 2026-09-25, and each time cost a re-run of about 40 minutes. The first fix (#4824) did not cover the real cause. In every one of those ten runs, a single test took longer than 60 s: one of the Sanctuary war-points property sweeps, which checks 1,200 attack sequences in one test. While it runs, the test worker never checks for messages, and vitest's worker stops waiting for its progress reply after a fixed 60 s. None of the 177 runs whose longest test stayed under 60 s showed the error. The sweeps now let the worker catch up after every sequence, with the same sequences and the same assertions. Pinned to one core with the same synthetic load, the old code failed 3 runs out of 3 and the new code passed 3 out of 3. If any test in the API, web or shared suites ever holds a worker that long again (30 s or more), the log now names that test in a [loop-stall] warning. The warning never changes the result. Nothing here hides a real failure: the gate still counts every failing test and every other error as red. Players see no change; this only affects how quickly fixes reach them.

1.288.14 — 2026-09-25

Changed

• The system list's probe marks cost almost nothing to refresh now (#4967). While the survey panel is open it polls every 30 seconds for your newest spy report and newest empty probe on each slot of the system. Before, every poll read through every spy report you had ever filed and discarded almost all of them. For the busiest scout in the game (2,602 reports) on their busiest system, one refresh took about 290 ms of database time, and the cost kept growing with everything they scouted. Each poll now reads only the newest answer per slot through a dedicated index. The same refresh, measured on the live database after the index was built, takes about half a millisecond (0.48–0.69 ms, 128 buffers instead of 6,110). You see exactly the same marks as before: same reports, same empty probes, same whole-history reach, same answer for slots someone else scouted.

1.288.13 — 2026-09-25

Also

• The first screens of the game are ready to be translated (#5005). Nothing changes on screen: the game is still English-only, and every word reads exactly as before. Behind the scenes, the text a new player meets first now comes from the translation layer added in the previous release. That covers the sign-in screen, the First Steps guide, the first convoy card, the bottom navigation, the tab names and the status chips at the top of the screen. When a second language arrives, these are the screens it will reach first.

1.288.12 — 2026-09-25

Fixed

• Our release checks no longer fail at random on busy days (#5013). Nothing changes in the game. The automatic checks that run before every update could fail on a busy build machine: three of them timed out under load even though nothing was wrong. They now have room to finish under load, and when one runs out of time it says so plainly, instead of looking like a real failure. Fewer false alarms mean updates reach you without needless delays.

1.288.11 — 2026-09-25

Also

• Test fixtures that describe a state the game cannot produce now fail at the fixture (#4245). Nothing players see changes. Before, a test could be built on an impossible colony (a building above its Command Centre's level) or an impossible supply-lane row (a hub letting through more than the lane offers). It then passed while proving nothing. Now the lane-card and building-ledger tests check every fixture against the game's own rules, and name the rule an impossible one breaks. The check found three impossible fixtures, now fixed. One had shipped in the building ledger's model tests, the same shape its screen tests had caught before merge. A census of about 240 candidate fixtures elsewhere is left to #5003.

1.288.10 — 2026-09-25

Fixed

• Agents bring enough ships to a colony whose fleet an Aegis Bastion tends (#4860). An Aegis Bastion pools its shields every round for the ships beside it, the owner's own or an ally's garrison. The agents' attack forecast counted it as just one more hull. For example, an agent sent 25 Cruisers against three Battleships and a Bastion, where the battle needs 29, and lost every time. Agents now play out the rounds the Bastion's shields buy, including the colony's Batteries and Fortress, and bring a force that wins: 30 Cruisers in that case. Colonies without a Bastion are read exactly as before. The new read applies to spy reports taken from 29 September, so each agent changes its mind about a colony only after its next probe of it.
• An agent that gives up on a Bastion-guarded colony knows why (#4860). If the Bastion keeps its gathered fleet from launching, the agent calls the attack off a day or two in. The reason it can give in chat is that the last probe showed ships an Aegis Bastion covers. It then leaves the colony alone for a few days, instead of letting the plan lapse unexplained and mustering against the same colony again.

1.288.9 — 2026-09-25

Changed

• Three engine tests can now fail when the rule they guard breaks (#4347). Nothing changes in the game. Three tests in the shared engine built their expected answer by calling the code they were checking, so they could never fail. They covered the bounty figure shown next to the WANTED badge (20,001 cases), a colony's storage ceiling with a single Warehouse (279 cases), and which anomalies a sector shows at each Astrometry level. Each now checks against numbers taken from the design: the bounty grade thresholds 10 / 30 / 75 / 150, the Warehouse storage formula with hand-computed values, and the anomaly tier gates with the per-level visibility cap. Mutation runs confirm they now go red when the rule breaks.
• Left out: the detection tooling #4347 proposed, a mutation loop plus an optional static pre-filter, is not built. It is tracked in #4994, together with a first sweep of 61 flagged assertions in the shared engine tests. Every flag that was read turned out to be an intended check.

1.288.8 — 2026-09-25

Also

• The contract slip's "Worth" / "Pays" word is now guarded in any casing, and only where it lives (#4237). Nothing on screen changes. The tests that keep the word visible, and out of a hover tooltip, used to miss a tooltip typed as the upper-case "WORTH" that the screen shows. They also scanned the whole board, so an unrelated hover elsewhere, such as a "pays double" note on the Daily Wrap, could fail them for the wrong reason. They now match any casing and read only each slip's pay column.

1.288.7 — 2026-09-25

Changed

• The engine's old travel formula can no longer be mistaken for the live one (#4230). The shared engine still carries the pre-Hive distance formula, because it is the emergency rollback and the fallback for the Event Sector. It used to be exported under the bare name distance, and reviewers kept measuring flight times with it. One review timed a Wraith strike's worst in-system leg at 204 s when the game charges 612 s. It now has a name that says it is the legacy metric, a note pointing to the distance the game actually charges, and a test that keeps every game module from using it or its clock. An audit found no game code using it, so no flight time, warp price or other figure a player sees changes.

1.288.6 — 2026-09-25

Fixed

• A pasted number too large to exist no longer sits in a number box (#4226). Before, pasting something like 1e999 or a 400-digit number into a number box (the War Table ship count, market amounts, build counts, coordinates) left the pasted text on screen, for example 1e999 beside "120 of 300", until you left the field. The fleet was never changed, but the box lied. Now the box puts back what it showed before the paste and selects it, so your next paste replaces it. Partial inputs you type on the way to a number, like 1e, - or ., are left alone. This was verified in desktop Chromium. The iOS and Android apps were not checked.

1.288.5 — 2026-09-25

Fixed

• The survey panel remembers the free slots you probed (#4233). The "probed Nh ago" mark on a slot tells you that you already paid for an answer there. Probes that found a colony kept that mark for good. Probes that found nobody lost it about a day after the sweep, because those answers were only read from the recent activity feed. A second sweep of the system then re-flew probes whose answers you already had. Now the "found nobody" answers come from your full history too, so the mark stays for every slot you probed. This applies in Space view and in List view. When a report at a slot is newer than the empty probe, the mark shows the report's age.

1.288.4 — 2026-09-25

Fixed

• Trade Portal: the inbound row no longer reads as a before→after of three different numbers (#4937). On a hub with feeders, the Trade Portal sheet used to show Inbound lanes 3,509/h offered → 6,224/h +1,131/h. The left figure was what your feeders offer, the right one was what the next level could receive, and the +1,131/h compared two receiving capacities, one of which was not on the arrow at all. It read as "3,509 becomes 6,224", and at full power it looked like the Supply-lanes number copied by mistake.

  The row is now Inbound capacity. Its arrow shows one thing: how much this Portal can receive per hour now, and at the next level. Your feeders' offer is a sentence of its own, with a verdict:

  • under Contribution now: whether this Portal has room for all of it, or holds your feeders back;
  • under Next level: whether that level has room for all of it, still holds back a named amount, or would not raise what arrives at all.

  The sheet also says that intake is not reduced by the colony's power, which is why this row can match Supply lanes.

1.288.3 — 2026-09-25

Changed

• A probe sweep now shows you the filter that sorts it out (#4222). Before, when you probed several planets one at a time, the results («Espionage came up empty», «Your probe found a friendly world», the spy reports) were scattered through Activity among everything else. A filter that isolates exactly those rows already existed, Combat → Espionage, but its chip only appeared after you picked the Combat lens, so it was easy never to find. Now, when your feed holds two or more of your own probe results mixed in with other rows, a button sits right under the newest run of them: Only your probe results · Combat → Espionage, with the number of rows it will show. One tap switches to that filter and lights both chips, so you can see where it lives. Nothing changes if you have no probe results, or only one, and the chip rows themselves are untouched. A probe someone flew against you still files under Defence and is not counted.

1.288.2 — 2026-09-25

Changed

• The launch pad's ship-count box now uses the same number field as the rest of the game (#4229). Before, the box in the War Table's hull editor was its own copy of the number field. Now it is the shared one. The behaviour is the same where it matters: a count typed or pasted over what the colony holds snaps to the hangar on that keystroke. An emptied box stays empty while you edit it, and it doesn't remove the ship. Typing 0 or a negative changes nothing. The digit keypad and the box's width are unchanged. Two things are different:
  • A leading zero (0050) is left alone while you type and tidied to 50 when you leave the box. Before, it was rewritten on the keystroke, which threw the cursor to the end.
  • A fraction such as 2.5 reads 2 at once. That is the count the fleet always took.

Also

• The internal check that every number box in the app uses the shared field now reads the code properly instead of matching text. A comment that mentions a number box no longer counts as one, and a box written in an unusual way no longer slips past the check (#4235, instance 3).
• #4226 was measured in a real browser: pasting an absurdly large number (1e999) into the ship-count box leaves it on screen beside the real count until you leave the box. Typing it digit by digit cannot happen, because the box snaps to the hangar first. This release does not change it, and #4226 stays open.

1.288.1 — 2026-09-25

Also

• Survey tests now guard what they are named for, not how the code is spelled (#4235). Players see no change; this release only changes tests. Three checks on the Galaxy tab's spy-probe survey read the source text where they meant a behaviour:
  • The check that the survey's refusal line turns amber could go green on a panel whose verdict line had lost its marker. It now finds exactly one verdict line or fails, and it pins the words together with the colour.
  • The AI badge on survey rows is now checked by building real rows for an agent, a Wraith, a human, a Beacon and an unread slot, not by matching one spelling of the code.
  • The check that the survey sends through the tested rule now reads the parsed import. Before, it matched the text of the import line.

  In a throwaway copy, each old check stayed green on at least one real break, and two of them also went red on harmless rewording. The new checks go red on every one of those breaks and stay green on the rewordings. The fourth case the issue lists (the numeric-cell census counting a code comment as an input field) is fixed in PR #4982, which is still pending as of this release.

1.288.0 — 2026-09-25

New

• Bastions and unclaimed relics scale to the guild that attacks them (#4970). Until now every guild met the same wall, which was a formality for the strongest guilds and out of reach for the rest. Now, when you launch an attack or open a raid on one of the three faction bastions or on a relic nobody holds, the game fits its garrison to your guild. It is sized so that your guild's 8 best fleets, together, win and pay for it in hulls; with a flat guild, fewer seats lose. The wall is fixed when the attack launches, and your fleets out on other missions count. A player with no guild faces a wall fitted to their own fleet.
• One take per day. After your guild takes a bastion or a relic, it can't take that same site again for 24 hours. The Targets list marks it "Taken today", and an attack already in flight when the day starts turns back on arrival. Other guilds are not affected: the bastion no longer goes behind a shield after it falls.
• Scouting shows your wall. A spy report on a bastion or an unclaimed relic shows the garrison your guild would meet right now. The Gate on the map stays a common reference figure.

Changed

• A held relic is defended by its holder. Once your guild holds a relic, its NPC fleet is gone. Only its turrets and the fleets you park there defend it. Relics held today lose their NPC fleet with this release, so if you hold one, station a garrison there.
• A bastion's or unclaimed relic's garrison no longer leaves wreckage in the debris field, win or lose, because it is fitted fresh for every attack. Your own lost ships still do. A lost attack on one scores no weekly war points; a win still does.

1.287.26 — 2026-09-25

Also

• The HUD wiring guards read the code, not its comments (#4299). Nothing on screen changes. The tests that prove the attack alarm, the protection clock and the queue chips are mounted in the right place in the resource strip used to read the dashboard as plain text. A comment quoting a mount could stand in for a deleted one, and a third mount written with spaces inside its braces escaped the count. Innocent edits such as a spacer with a click handler, or a comment about a mount, turned those tests red. The tests now parse the dashboard as code. They catch every one of those wiring breaks and ignore the innocent edits, and a failure lists the mounts it actually found.

1.287.25 — 2026-09-25

Changed

• Every screen that explains the Gate now uses the same words (#4284). Six places tell you what the Gate is: the header ⚡ tooltip, the Power details dialog, the leaderboard's Gate column tooltip, its desktop and phone legends, and the guild points sheet. Four of them wrote their own version and the other two differed by a word ("the higher of" vs "the larger of"). None of them was wrong, but nothing kept them in step. Now they all read "the larger of Structural and Military, not their sum". The guild sheet used to say "those two halves" and now names them. The hangar keeps its own line, "the Gate moves only by the part that ends up above your Structural", which states the same rule for the moment you order ships, and a test now checks that it agrees with the definition. The AI empires' strategy prompt used to state the definition in Italian; it now uses the same English words. That line also points to the "#N sul tabellone" rank each leaderboard row shows, so an AI empire won't mistake its Gate rank for its points rank.

1.287.24 — 2026-09-25

Also

• The raid alarm's code comments now describe what it actually says (#4282). Nothing changes on screen. When three or more fleets of one raid are sighted within a minute of each other, the collapsed alarm in the Activity feed still reads "They came onto your sensors within a minute of each other." The comments behind it had said it would read "at once" instead, which would have claimed the fleets arrived together when their timestamps say otherwise. Those comments, and a test that could not tell which warning the alarm is drawn at, have been corrected, so the next change to this row starts from the right description.

1.287.23 — 2026-09-25

Fixed

• Synthetic empires bring enough ships to a colony whose Shield Bubble shelters its fleet (#4766).
  • Before. Against ships under a Bubble of Shield 10 or more, an agent often launched the smallest force that could barely out-pace the Bubble's regeneration. The defending ships kept firing while the Bubble absorbed every salvo, so that force bled out before the Bubble broke. For example, it sent 39 Bombers against twenty 5-star Interceptors under Shield 14, where 69 are needed.
  • Now. The agents play those rounds out the way the battle engine does:
    • the Bubble absorbs and regenerates, faster under the owner's Hardened Pool;
    • each defending group fires at the attacking hulls it counters most, including the owner's Vanguard in the opening rounds;
    • mixed fleets are read group by group.

    On a bench of 864 cases, launches that fall short of what wins drop from 107 to 8 of 216 at Shield 10, and from 140 to 0 at Shield 14. Mixed hangars that used to launch into certain defeats (76 of 120 at Shield 14) now launch only forces that win.

  • Cost. Where the old launch already won, some agents now bring a little more (a median of about 1.2× at Shield 10 and above).
  • Rollout. The new read applies to spy reports taken from 2026-09-28 on. Each agent moves over as its next probe lands, not all at once.
• A strike the Bubble keeps out of reach is called off out loud (#4945).
  • Before. An agent could gather against such a colony, reach the size it aimed for, and then never launch. Its plan said "assembling a force" for up to three days before the operation quietly lapsed.
  • Now. Its shipyard keeps building the right hull for the strike. If the strike still cannot go, the agent calls it off and says why: the last probe showed ships behind a Bubble that would wear down its fleet before the Bubble broke. It then leaves that colony alone for a few days, unless a newer probe shows the ships are gone.

Left out

• Cascade, Shock, the overkill spillover, formations, a garrison's own Vanguard, and the Vanguard's opening rounds after the Bubble breaks are not yet part of this read. They are filed as #4964.
• Guild raids size their own forces and are unchanged.

1.287.22 — 2026-09-25

Fixed

• The phone mission screen no longer jumps when a planet's scan comes back empty (#4252). On a phone, a mission aimed at an ice or lava world opens with a note on how the world's energy affects any defender. That is correct while nobody knows yet whether a colony is there. When the scan came back showing an empty slot, the note went away, and the Advisor and the fleet picker jumped up by the note's height (56px at 375 and 390px wide) just as you reached for them. The note still goes away when nobody is there, but its space now stays, so nothing under your thumb moves. The space is released when you change the target or the mission type, open another mission from elsewhere, switch to Alarm or In flight, leave the Missions tab, or rotate the phone.

1.287.21 — 2026-09-25

Fixed

• Guild season settlement: a missing season id is no longer called an unknown one (#4239). The two operator routes that list and settle the placement rewards a season close left unpaid used to answer a request with no season id as "No guild season with that id". That sent the operator looking for a correct id they had never typed. A request with no season id, or with one that isn't a single string (a repeated ?seasonId=, a number), now gets its own refusal: "The request named no single guild season: send exactly one season id, as text." A named id that matches no season still answers "No guild season with that id." The operator runbook lists the new refusal. Players are not affected: both routes are admin-only.

1.287.20 — 2026-09-25

Also

• Colonize checklist: a test guard points at the row it guards again (#4262). Nothing on screen changes. The Colonization and Command Center requirement rows under Colonize must never show the certain green ✓, because the panel cannot see colonizations already in flight. Each row had a source pin saying so. The Colonization one searched the wrong stretch of code and let a mutation to the green ✓ through. A second, exact pin on the same line was already catching that mutation, so the suite had no hole. The fix points the first pin at the row's state, the way the Command Center one works, and gives the suite back its redundant guard. The comment next to it now names the on-screen test that checks the sheet itself.

1.287.19 — 2026-09-25

Also

• Code comments about the guild siege alarm and the Inbound board now match the game (#4311). Nothing changes on screen. The note on the guild's "anchorages under siege" signal said it lights when a bar has reached its cap. It lights one notch before the cap, the last moment a lost battle can still be prevented, because a structure is destroyed by the notch that fills its bar. Five other notes said the Inbound board picks attacks by who was targeted at launch and ignores coordinates. It picks them by the coordinates of every colony you hold now. A future change built on the old notes could have moved the siege alarm back to a point where it never fires.

1.287.18 — 2026-09-25

Fixed

• Advisor cards no longer mention a level the card doesn't show (#4275). A building you have never built gets a suggestion row that reads new, but some lines under it still said "this level": the Hardening line on Planetary Shield, Defense Batteries, Orbital Fortress and Orbital Cannon, and the Laboratory's research-network and research-lane lines. A first-build card now reads "Hardening: +16% — the added defence, measured against …" and "Research network stays at 12; the new Laboratory still improves scientific collaboration." An upgrade card names the level its row lands on: under lvl 10 → 11 it reads "level 11's added defence", and the Trade Portal says "level 5 takes 1,932/h". The Trade Portal's intake line can't appear on a first-build card today, so nothing changes there for a new Portal.

1.287.17 — 2026-09-25

Fixed

• The Shipyard no longer dresses an Event Relic as buildable on its stage (#4271). The catalogue rows already marked the Destroyer and the Annihilator with ❖, but the nameplate under the hull read "DESTROYER · Mythic · Heavy class · Demolishes titan dreadnoughts", exactly like a hull you can order. On phones the relic's carousel slide also sat at full brightness with nothing on it. The nameplate now shows ❖ before the name. If you don't own the relic, it reads "Event Relic — not buildable" where a buildable hull shows its role. If you own one, it keeps its role line, and screen readers still hear that it cannot be built. On phones each relic slide carries a small "❖ Event Relic" tag, so the slides either side of the one in focus say it too.
• Screen readers no longer run an owned relic's caption into its refusal (#4270). On the hull rail, in the phone All-hulls list and on the narrow filmstrip, the hidden "Event Relic — not buildable" came straight after the visible text, so it was read as "owned ×3Event Relic — not buildable". It now reads "owned ×3 · Event Relic — not buildable". The owned relic's nameplate follows the same rule.

1.287.16 — 2026-09-25

Changed

• Faction bastions can be taken by a guild working together (#4942). Until now, every bastion was sized at 1.5× the strongest single fleet on the server. Only one player could take one, and no guild could, even with all eight raid seats filled. Now a bastion is worth half of what the strongest guild can bring to one raid: the eight best fleets of its players. A guild about half as strong as the top one (such as STG) takes a bastion by filling all eight seats and paying for it in hulls. Half of that guild is still repelled. A player with no guild counts as a raid of one. The three bastions shrink the first time the server restarts with this version. The map legend and the bastion row in the landmarks now describe the new size ("0.5× top guild raid").
• A single very strong fleet can now fell a bastion alone too. Walls, loot and rewards are unchanged.

1.287.15 — 2026-09-25

Fixed

• Colonize checklist: an unread colony list no longer turns the hull row into a refusal (#4946). A planet sheet opened before your colony list had loaded, or after its first read failed, could not see a Colony Ship parked at an outpost. It still drew a red ✗ "you have 0", greyed the Colonize button and told you to build one in a Shipyard. Now, until the list is read, that row shows the "Not checked" bullet and reads "none in the hangars this panel has read", and the Colonize button stays available. If there really is no founding hull, the mission composer refuses when you launch. Once the list loads, the row settles on its real answer. It shows ✗ with the Shipyard hint if you have no hull anywhere, or ✓ if you do. When the hull is at another colony, it also tells you to pick that colony under Launch from. Screen readers now hear a failed requirement as "Not met by what this panel can see", the same wording the legend shows.

1.287.14 — 2026-09-25

Fixed

• An attack on a colony you took over now reaches the rest of the game, not just your Incoming board (#4449). Since the previous fix, a player who takes over a colony while somebody else's attack is already flying at it has seen that attack on their Incoming board. Six other places still answered for the colony's previous owner:
  • Guild door and nav dots stayed dark for a guildmate under that attack. Now they light, exactly when that guildmate's board shows it.
  • The build advisor called the colony quiet and ranked economy first. Now it reads the attack as it does any other.
  • The Stand down button on a garrison stationed there stayed live, and the garrison could walk out of a battle that was already resolving. Now the card greys and the order is refused, as for any colony under a committed strike.
  • The host of a zero-hour watch that lands into that fight was not told help had arrived. Now they are.
  • Recall (Fold 15) ignored a committed strike on that colony, and still refused the previous owner over a colony they had lost. Now it follows the colonies you actually hold.

  Everything stays per player. A Veiled launch stays hidden everywhere. A strike that will turn back from its new holder (they are hidden from the raider, or protected by a pact, a shield or the newcomer window) warns and freezes nobody. A Nightfall attack is read against the new holder's own sensors, not against a verdict made for the previous owner.

  Still known and unchanged: taking over a colony under an ordinary attack sends no push at that moment; a player who settles a deleted account's colony is not warned about attacks aimed at it; and a Nightfall attack launched with no unveil is never announced to a later holder.

1.287.13 — 2026-09-25

Fixed

• Colonization card: a colony you still have to build for no longer opens like one you get (#4313). When the Command Center and the Colonization discipline held you at the same ceiling, the Research tab's Effect row showed a gold chip reading "Up to 4 colonies — also needs Command Center lvl 12". The green chip for a level that really does deliver the colony reads "Up to 4 colonies" too, so a quick glance read a colony this research alone does not buy. The gold chip now leads with the condition: "Also needs Command Center lvl 12 — then up to 4 colonies". The numbers are unchanged: research this level and build the named Command Center level, and you get exactly the colony count it names. Checked on the real Research tab at desktop and phone width.

1.287.12 — 2026-09-25

Fixed

• Schema-rollback runbook: the Dark Matter Doctrine baseline can actually be taken (#4914). Before, the step that snapshots every Doctrine holder before a rollback was a psql \copy spread over nine lines. psql refuses that with \copy: parse error at end of line, so the snapshot was never written, and a roll-forward had nothing to compare against when deciding who overpaid at the old prices. Now it is a host-side dump into the operator's home directory, taken while the game is stopped and before the column is dropped. It is checked against its own row count and readable only by the operator. Nothing changes for players; the change is to the recovery procedure that protects their Dark Matter.
• Schema-rollback runbook: admin chat mutes have a rollback block (#4915). Before, the three columns behind the admin chat mute had no entry in the rollback runbook. After the first mute, a rollback to an older build would have left the API unable to start. The new block AT stops the game, counts the mutes, and exports them. It then drops the columns and checks they are gone before the old build starts. It says plainly what that costs: every running mute is lifted while the old build runs. It also gives a tested way to put the running mutes back afterwards, with the minutes that are left and the moderator's original reason.

Also

• Documentation only. No game code changed.

1.287.11 — 2026-09-25

Fixed

• Colonize checklist: the provisional check now has its own shape (#4260). On a free planet's sheet, the cap rows (Colonization and Command Center, or the Event Sector ceiling) are checked "as far as this panel can see": a colonization already in flight is only counted at launch. Before, those rows drew the same ✓ as the certain hull row, and only a grey vs green tint told them apart. Now the provisional check is drawn (✓) and the certain one stays ✓, so the difference shows in the shape and not just the colour. Screen readers now hear each row's state in words before the row: "Met", "Met as far as this panel can see", "Not met" or "Not checked".
• Colonize checklist: the legend explains every mark it shows (#4261). Before, the legend under the list explained only the provisional check, with a sentence calling "the cap rows above" checked. That was false when your research level had not loaded yet: the Colonization row then showed a bare bullet that nothing explained. Now, whenever a provisional check or a bullet is on the list, the legend has one line for each mark on screen:
  • ✓ Met.
  • (✓) Met as far as this panel can see — a colonization already on its way is counted only when you launch.
  • ✗ Not met by what this panel can see.
  • • Not checked — this panel has not read the figure it needs.

Also

• Left out, filed as #4946: while your colony list has not loaded yet, a Colony Ship parked at an outpost is not counted. The hull row can then show ✗ and grey the Colonize button. That check itself is unchanged here. The ✗ line of the legend now only claims what this panel can see.

1.287.10 — 2026-09-25

Fixed

• The Activity filter chips count the rows you actually see (#4267). Some Activity rows gather several entries into one: a burst of chrono-shard spends on one order, the contingents of one incoming raid, and a run of public trades in the galaxy's half. The number on each chip still counted every entry inside those rows. So the tab could open on "Mine 9" over a column of four rows, or read "All 14" over six. Every chip on all three rows (whose, which type, and the Combat topics) now shows exactly the number of rows you get when you tap it. That holds for chips that are lit and chips that are not. A run of trades also no longer risks gathering across the day line when you tap a chip just after midnight, before the feed refreshes.

1.287.9 — 2026-09-25

Fixed

• Contracts: a player with no colony is no longer offered a claim that cannot pay (#4236). Every contract reward is paid into your oldest colony. A player left with no colony, which can happen after a season sweep, still saw finished contracts marked ready. The pay column said Pays and a Claim button sat beside it, but tapping it failed with "colony not found". The Daily Wrap and the Weekly Chest had the same problem. Now the board matches what the server can actually do:
  • one sentence above the board says you hold no colony, so nothing on it can be collected until you hold one;
  • finished cards show their reward as Worth, with no Claim button and the line "No colony to pay this into."; the hover figure says it is priced at the minimum rate, since an empire with no colony has no production to price it at;
  • the Daily Wrap strip says "no colony to pay into" in place of "pays double", and shows no claim button;
  • an earned Weekly Chest keeps its row but shows "Opens once you hold a colony." instead of Open;
  • a carry about to expire still shows its countdown, but no longer tells you to claim a Wrap, because no Wrap can be claimed without a colony;
  • the Contracts lane dot no longer lights for a Chest that cannot be opened.

  Left out: an older installed app does not know about the new signal. For a player with no colony, it shows a finished card as unfinished: full bar, no button, no explanation. Before, it showed a Claim button that failed. Its Chest row still offers Open, and the server refuses it as before. Updating the app fixes both. Whether a season sweep should leave a player without a colony at all is a separate question, and this change does not answer it.

1.287.8 — 2026-09-25

Changed

• Reduced motion on desktop: the inbound-attack chip has a heavier outline (#4298). Before, a desktop player with the system's reduced-motion setting on saw an incoming attack as a dark chip with a thin 1px red border and a faint static glow. They had no breathing animation to catch the eye, and no screen-edge ring, which exists only on phones. Since #988 the chip also sits on the second line of the resource strip at narrower desktop widths. Now the chip's red border is 2px for these players, measured at 6.2:1 against the HUD (4.9:1 under simulated protanopia). The chip keeps its exact size and slot, so the resource figures never move. Nothing animates. The fill and text are unchanged, so less motion does not mean more red. Phones and players with motion keep the chip they had. This is one notch more noticeable, not a new signal.

1.287.7 — 2026-09-25

Fixed

• Hovering a fleet on the galaxy map now shows the system's address (#4292). On desktop, at the opening galaxy view, hovering a system with one of the fleet marks on it used to show nothing: no address card, no ring, and the pan cursor. So the systems with fleet activity were exactly the ones the map would not name. The card now shows the system's address there, and its last line reads click to open the fleet, which is what a click on the mark does.
• The card no longer promises the wrong click when a fleet sits on a planet. When zoomed in far enough to click planets, a fleet on a planet's cell left the card saying click to open the planet, while the click opened the fleet. It now says click to open the fleet there too.
• A fleet that flies under a still mouse updates the card. This covers a fleet that moves onto or off a system while the mouse rests on it, including one that arrives with the map's regular refresh. The card's last line now changes as the fleet gets there, instead of waiting for the mouse to move. One lag remains, as before this release: during the glide after a mouse drag, the card waits for the map to stop moving.

Phones are unchanged. The address card is a desktop hover, and a tap on a fleet mark already opened the fleet.

1.287.6 — 2026-09-25

Changed

• Colony rows on a phone are shorter again: the More toggle sits at the end of the description (#4926). The previous release (#4318) gave each clamped building description a "More" toggle on a line of its own under the two lines of text. On a phone almost every row has one, so the building list grew by about 700px. The toggle now reads "… More" at the end of the second line. The covered words fade out under it, and the row is back to two lines of text. Once open, "Less" follows the last word of the description. With every building built, the list is 672px shorter on a 390px screen and 720px shorter on a 360px one. Tapping "…" or "More" opens the description, and the target stays at least 24px tall. The text is still selectable, and a description that already fits shows no toggle and no fade.

1.287.5 — 2026-09-25

Fixed

• Synthetic empires stop re-probing a colony whose lone legendary hull they already saw (#4893).

  Before: a synthetic empire's spy report could show a Destroyer or Annihilator that may stand alone in a colony's defence, a hull no single empire can sink. If the empire never opened an operation against that colony, nothing stopped it from coming back. Once the report went stale after a day, the colony returned to the empire's solo scouting list and was probed again, and each probe was announced to its owner. That meant a probe a day for a strike that never came.

  Now: the report leaves the colony alone for 3 to 6 days, the same hold an abandoned operation already gets. The length is drawn separately for each hold. During that time the empire does not probe the colony for a solo strike, pick it, or promise a strike on it in chat. Its notes say plainly why it is leaving the colony alone: what its scouting showed, stated as certain or hedged. They now read "not planning a strike of your own there", for every such hold, including those left by an abandoned operation. A later report that no longer shows the hull alone lifts the hold, and so does the colony changing hands.

  What stays the same: this covers only colonies within the empire's own solo reach. Colonies above it are scouted for guild raids, and a guild's raid scouting can still probe them daily, because a raid can sink the hull. If an empire's next scan of a slot comes more than 3 days after its probe, no hold is written. That colony can then be probed again, but at most about every three days, not daily. The change takes effect with spy reports observed from 2026-09-28.

1.287.4 — 2026-09-25

Fixed

• Schema rollbacks no longer count rows while the game is still writing to them (#4900). Nothing changes in the game itself. This fixes the operator runbook for rolling the server back past a release that added a database column or table. Before, several steps of that runbook counted rows while the new server was still live, and a zero count meant "skip the cleanup". A player acting a minute later could make the old server refuse to boot, turning a rollback into an outage. Every step of the runbook now follows one order:
  • build the old version first;
  • hold other deploys;
  • stop the game servers;
  • count;
  • always remove the new column;
  • check it is gone;
  • only then start the old version.

  Two stop commands in the runbook never actually stopped anything; they are fixed too.

Also

• Two older gaps in the same runbook, found during this work, are filed as #4914 and #4915.

1.287.3 — 2026-09-25

Fixed

• Colony building descriptions can be read in full on a phone (#4318). Each row on the Colony tab clamps its building's description to two lines. On a phone that is about 85 characters of a text that runs as long as 894 (the Deep Observatory). The rest was only in a hover tooltip, which never opens on touch. A description that runs past its two lines now has a small More button under it. Tap it and the whole text opens in place; tap Less to fold it back. Descriptions that already fit show no button. The text itself stays ordinary text that you can select and copy, and screen readers read all of it whether it is open or folded. The button is labelled with the building's name, so a screen reader's list of controls is not 27 identical "More" entries. On desktop an open description spans the full width of the list, the same way an open Core… panel does.

Left out

• The More button takes a line of its own, so on a phone most rows are about 24px taller than before. Putting it at the end of the second line instead is tracked in #4926.

1.287.2 — 2026-09-25

Fixed

• A resource tile's hover card no longer covers the inbound-attack countdown (#4297). On desktop windows narrow enough for the resource strip to wrap (around 1024–1200px), the attack alarm lands on the strip's second line, directly under Ferrolite and Xenogas. Hovering either tile, or tapping it on a touch tablet, opened its card on top of the alarm and hid the countdown. The card now opens just below the alarm, so the countdown and the card are both readable. The same holds when the game cannot read incoming attacks and shows its "could not be read" chip in that slot. On wider screens the card opens where it always did, and phones were never affected.

1.287.1 — 2026-09-25

Also

• The api test suite no longer accepts values production would refuse (#4475). Nothing changes in the game itself. Until now, the stand-in database the api tests run on accepted any value anywhere. A malformed id (an array, a number, a whole JSON object) or a missing one answered "no row found", and a list written into a text column was stored. The real database client refuses all of these before running any query, and the api answers those refusals as a server error. So a test could pass on a malformed request that would fail in production. The stand-in now refuses the same shapes, with the same error the api sees in production, for three cases: unique-key lookups, filters on text columns, and writes to text columns. Every rule was measured against the real client on a throwaway database. Other column types and the rest of the client's validation are deliberately not reproduced. Two test fixtures that had been missing a required field were corrected.

1.287.0 — 2026-09-25

New

• Crucible: see which of your ship classes carried the fight (#4456). A duel or spar used to end on Victory or Defeat, the rank swap, the Merit and a round count, and nothing about how your roster performed. Below the result card there is now a breakdown of your roster, one row per class:
  • Dealt: what the class landed on the enemy, shields included, with its share of what your classes dealt;
  • Shields: what its shields absorbed;
  • Hull hit: what got through to its hull;
  • how many ships of that class went down.

  The figures are the battle engine's own tally from the fight. Nothing is estimated afterwards, and the classes add back up to the fight's totals.

• Only your own roster is shown, and the server sends you only your half of the tally. Nothing new is revealed about your opponent.
• Duels fought before this release show their result as they always did, without a breakdown.

1.286.2 — 2026-09-25

Fixed

• Synthetic empires now save Dark Matter for the Doctrine they want (#4623). An empire whose main Doctrine was past level 15 used to spend every purse on its side Doctrines. Doctrine levels cost 8, 20, 45, then 90, so a slow income kept affording a cheap level and the expensive one never arrived. Measured on +5 Dark Matter a tick, a scientist took 309 ticks to take the Forge from 15 to 20, against 90 for those five levels alone. Now an empire buys a side Doctrine only when it wants it about as much as its next main level, and otherwise banks. The same scientist takes 103 ticks. How strict an empire saves varies from one empire to the next, so agents of the same kind don't all hold their purse for the same level and then buy on the same tick. Across personas the main Doctrine now finishes in 98–163 ticks on that income, against 163–309 before.
• The Wraiths' faction order no longer spends what a Doctrine is saving. When a Wraith was banking for its own Doctrine, the faction's fallback bought the next Doctrine in faction order with the saved purse. It now waits while the Wraith is saving for one of its own Doctrines. It still runs while the Wraith saves for an Ascendancy, because an Ascendancy opens no hull. Result in the same model: each Wraith's own hull arrives sooner (ghost 334 → 289 ticks, quartermaster 381 → 200, striker 240 → 182), and its last hull arrives no later than before. One side Doctrine per Wraith now arrives about 180–250 ticks later. While a Wraith saves for an Ascendancy, it can also field a Doctrine its persona does not weigh before one it does: for example, a striker's Aegis comes before its Veil.

Limits: these tick figures model Dark Matter spent on the apex nodes only. The warp, research, the arsenal, the cores and campaigns can still spend what an empire is saving.

1.286.1 — 2026-09-25

Fixed

• Agents no longer plan strikes against a lone legendary hull they cannot sink (#4861). A Destroyer or Annihilator that is the only legendary its player fields at a colony cannot be destroyed by a single attacking empire, so a solo strike on it ends in a draw with no loot. The agents' forecast didn't account for this. Given a fresh report, it rated 50 Cruisers against a lone Destroyer a near-certain win, which the battle engine never grants. Agents would pick such colonies, build and stage toward them, and then be held back at the launch button or fly into a draw. War and grudge targets went through this again every day. From reports taken on or after 2026-09-28:
  • Agents no longer pick such a colony or launch at it.
  • A strike already being assembled is called off, and the agent can say why in chat. When the scouting shows exactly one legendary hull at the colony, the stated reason is that the hull is one no empire striking alone can sink.
  • When the report cannot tell whose hulls the garrisons are (one at home and one parked by a garrison, for example), the agent only says the hull may be one a solo strike cannot sink.
  • After calling off a strike, the agent leaves the colony alone for a few days: no new strike, no scouting probe for one, and no promise of one in chat. A newer report showing the hull is gone ends this early, as does the colony changing hands.
  • Guild raids are unchanged: two or more attackers can sink the hull.

  Known limits:

  • A legendary hull that is flying home when the report is taken is not counted, because the report cannot say when it lands. If it lands before the strike does, the strike can still draw.
  • A colony that the agent passed over without ever planning a strike on it is not put on hold, so it can still be scouted again (#4893).

1.286.0 — 2026-09-25

New

• The colony advisor shows what the Particle Accelerator's switch costs and what it makes (#4371). Before, the advisor held the Accelerator's On/Off switch where you left it and said nothing about it. The Dark Matter it makes has no price in the advisor, so nothing on screen told you what running it cost the rest of the colony. Now any colony with an Accelerator gets its own box below the suggestions. It reads "Running the Particle Accelerator costs this colony N Fe/day, N Xe/day, N Dt/day of production for M DM/day" when it is on, or "would cost … for …" when it is off. Below that, it gives the colony's power yield with it on and with it off. Both sides are per day. The production figure is what the other producers lose to its draw, with your other switches as they are now. The Dark Matter figure is the one the Accelerator's sheet shows, including overclock, the Accretion Engine and VIP. The box states the trade and makes no recommendation. It uses no Dark Matter exchange rate and never says which way to flip it. Synthetic empires read nothing new: the advisor's server answer is unchanged.

1.285.0 — 2026-09-25

New

• Your commander card shows when your account was created and how old it is (#4835). Tap your own portrait in the HUD to open the Commander identity card. It never said when your account was created. Now a line near the top reads "Commander since 3 Sep 2026 · 22 days. No other commander can see this." The date is in your own locale and timezone. The age counts whole days. The date is sent only on your own profile, never in anything another commander can read: leaderboard, chat, guild roster or public profile.

1.284.17 — 2026-09-25

Fixed

• An attack you were already warned about no longer vanishes from your Incoming board mid-flight (#4437). Before: an attack that had already reached your board, with its alarm and phone push, could disappear partway through the flight in two cases: the attacker completed the Nightfall Protocol after launching, or you lost the colony that carried your Sensors L8. The disappearance read like a recall, and when the attack came back minutes before impact nothing pushed you again. Now: the launch records whether it went dark. An attack launched in the open stays on your board, untagged, for its whole flight. The advisor's threat line, the guild door's "ally under attack" dot and the closing line of a recall all follow the same rule. Attacks launched dark behave exactly as before.

  Left out: the rule covers the player the attack was aimed at when it launched. Someone who takes over that colony mid-flight still sees Nightfall live. So an attack launched in the open whose colony changes hands, after which the attacker completes Nightfall, can still drop off the new holder's board and come back inside the final window with no push. A second path remains too: an attack launched dark that your Sensors L8 revealed early can hide again if you then lose that Sensors colony. That one returns with its push.

  Operators: this adds one nullable database column (Mission.nightfallAtLaunch). Rolling back becomes a manual step once the first attack is launched on this build. The steps are in the deploy runbook, block AS: stop the API, drop the column, verify, then start the old image.

1.284.16 — 2026-09-25

Fixed

• The design desk's record of Accelerator flips now counts an attack on a colony you inherited mid-flight (#4487). Nothing changes on screen: the flip is still free and instant, and this record was never shown to players. Before, if a colony changed hands while an attack was already flying at it, the new owner's Incoming board showed the countdown, but a flip they made under it was logged as made with no attack known. Now it is logged as known, with the arrival time. The record follows the same rule as the board: the raider's own attack, a Veiled launch and an attack that could not land on the holder still count as no alarm.

1.284.15 — 2026-09-25

Fixed

• Names made only of invisible characters are refused like empty ones (#4433). Until now a fleet template could be saved under a name made only of characters that draw nothing, such as zero-width spaces or a Hangul filler. The list then showed a chip with nothing written on it. Saving a formation preset with such a name was already refused, but the panel wrongly said your client was out of date. Both panels now ask you to "Give the template a name." or "Give this formation a name to save it." before anything is sent. A name that contains such characters next to visible ones, like a family emoji or a keycap, is kept as typed. On an installed app that has not updated yet, the server still refuses the name, but it shows the old "client is out of date" sentence.
• The same rule now applies to text typed by staff. A dispatch title or body, a chat mute reason and a player rename are refused when nothing in them is visible. A ban or Dark Matter reason like that is recorded as no reason. A guild-move notice falls back to "a server rebalance" instead of printing a blank reason. The check runs on the text as it will be stored, after the length limit is applied.

Also

• Stored names were checked before this change: production held no fleet template, formation preset, guild aid note, mute reason or dispatch made only of invisible characters (0 of 73 template names). If one ever appears, it stops being shown and is removed the next time that player saves their templates.

1.284.14 — 2026-09-25

Fixed

• An agent raided by an unranked player now knows who led the raid (#4365). Some raiders are missing from the leaderboard and from the agent's galaxy scan: a veiled Wraith, the admin or GM, or anyone at all in the first minutes after an agent restarts. An agent raided by one of them remembered the leader only as "?". Its post-battle line in chat therefore had only the companions' names to work with, and it tended to call one of them out as if they had led the raid. Now the agent reads the leader's name off the battle report, from the same attacker row a human defender would see, and addresses the right person. The name comes only from the row whose player id matches the raider. An agent that was itself raiding never takes a name from its own side of the report.

Changed

• Grudges against those raiders are now visible. Once the raider has a name, the agent's grudge carries it. That grudge can now show up in the agent's chat, in the ambient chatter and in its guild's council ("X wronged us: how do we respond?"). It can also justify chasing that raider's fleets in transit, just as grudges against ranked raiders already could. Grudges already stored without a name are not touched by this release. One takes its name the next time that same raider really hits that agent.
• A later hostile act that the agent cannot attribute by name, such as a probe from an unnamed actor, no longer wipes a name the agent had already learned. This covers both the agent's grudges and what it remembers about its opponents.

1.284.13 — 2026-09-25

Fixed

• The Build Advisor prices an Escalation holder's Batteries and Fortress at what the battle fields (#4881). With the Doctrine of Escalation, your Defense Batteries and Orbital Fortress platforms fire and shield at ten times strength in battle. The Advisor ignored this and ranked every Batteries and Fortress level at a tenth of that fire and shield. It now applies the same multiplier the battle does, to fire and shield only; hull, the Planetary Shield and the Cannon are unchanged. As a holder you will see Batteries and Fortress cards climb and show a larger hardening figure. Shield and Cannon cards show a smaller one, because a level now adds a smaller share to a stronger colony.

1.284.12 — 2026-09-25

Fixed

• Release notes can say "non-aggression pact" again (#4374). The check that keeps Italian out of release notes read the English prefix in "non-aggression", "non-zero" and "non-empty" as the Italian negation, and refused the notes. The v1.219.0 notes went out misspelled as "nonaggression" to get past it. Now the prefix joined to a word by a hyphen passes, including the Unicode hyphens generated text often uses. The Italian negation standing alone is still refused, even on the same line as a compound.
• Release notes the check cannot read as text are refused (#4484). A notes file holding a NUL byte used to pass as English, because the search found no words in it, and no words meant English. The same was true of an English note holding bytes that are not valid UTF-8. Both are now refused, with the reason, before any tag is cut. Normal UTF-8 typography (accents, dashes, curly quotes) still passes. A missing or failing tool is also refused, and the refusal names the tool.

Also

• Left out: the TestFlight watcher on the build Mac has its own language detector, which is not in this repository. It may still flag "non-aggression" and cut such a release to headline plus link, and it may have the same NUL-byte hole. Tracked in #4897.
• Left out: the published v1.219.0 entry still says "nonaggression pact".
• Known limit: a failure in the last step of the check (after the word search) is not yet detected. It could not be reproduced with the real tools.

1.284.11 — 2026-09-25

Fixed

• Guild aid notes no longer end in half an emoji (#4482). A note longer than 140 characters was cut at exactly 140 UTF-16 units, first by the note field and then by the server. When that cut landed inside an emoji, the note was stored and shown on every guildmate's aid board with a fragment nobody typed: a family reduced to its first person, a flag reduced to a bare black flag, a keycap reduced to a digit. Now the field and the server share one limit and cut only between whole characters, so an emoji that does not fit is left out whole. An edit that does not fit is refused where you make it: typing or pasting in the middle of a full note no longer eats its end, and the caret stays where you typed. Typing with an input method (Japanese, Chinese, an emoji picker) is not interrupted mid-word. Notes sent by older installed apps, which still cut on the old rule, have their unfinished emoji cleaned off the end on the server. One kind of leftover cannot be told apart from emoji typed on purpose and stays until those apps update: a fragment that happens to end on a complete emoji, such as a family cut after its second person. Checked in Chromium; iOS Safari was not tested. Notes already on the board are unchanged.
• A long operator bulletin no longer ends its lock-screen banner in half an emoji (#4482). The push banner's shortening cut now uses the same rule.

1.284.10 — 2026-09-25

Fixed

The Build Advisor counts your Ramparts when it ranks defences (#4682)

Before: the Advisor's suggested upgrades ranked the Planetary Shield, Defense Batteries and the Orbital Fortress as if you held none of the Mantle, the Battery or the Keep. A Mantle holder at 20 had every Shield level scored at the bare bubble, not the 40% larger one the battle uses. The Advisor ignored the Batteries' and the Fortress's guns completely, so the Battery and the Keep counted for nothing, and nobody's emplacements got credit for firing. The Gunnery already counted.

Now: each defence is scored with the numbers the battle uses, including your Ramparts, and the Batteries' and the Fortress's firepower counts. Measured on sample colonies, at full power:

• An early colony (Command Center 8) with a fleet inbound now puts its next Batteries level first, ahead of the Shipyard. Its hardening figure goes from 6% to 9%.
• With the Keep at 20, a Fortress level on that colony reads +28% (it read +25% without the Keep).
• With the Mantle at 20, a Shield level reads +3% (it read +2%). With the Battery at 20, a Batteries level reads +10% (it read +9%).
• On a late colony (Command Center 24), the Keep at 20 moves the Fortress ahead of the Shield.

These defences now weigh more for every player, Ramparts or not. Across a test sample of 800 colonies, a defence reached the top three suggestions 49 times, up from 48.

On a colony short of power, a defence level that draws more energy also weakens every gun already standing, because the battle throttles their fire by the power supply. The Advisor now counts that loss. On a starved grid, a Fortress level can therefore rank lower, or drop off the list, where it used to show. Example: Fortress 10 → 11 beside Batteries 25 at 40.5% power.

Not in this release: the Doctrine of Escalation's ×10 on Batteries and Fortress fire and shields is still missing from the Advisor's ranking. That is tracked in #4881.

1.284.9 — 2026-09-25

Changed

• The battle-report loadout guard sees every Ascendancy and Rampart it used to trust by hand (#4611). Before, the test that keeps a battle report's chips honest could not see a Tree line whose level the battle reads through a helper, so each new Ascendancy or Rampart (the Breach, Vanguard, Phalanx, Ward, Mantle, Battery, Keep and Gunnery) needed an entry someone had to remember. On #4591 the Breach went missing from the report's list with every test green. Now the guard reads which helpers the battle code imports and tests each one against every Tree line. A line missing from the report list goes red on its own, and so does any helper the guard cannot exercise. Keystone flags read where the colony's Bubble and statics are built now count as well. Players see no change: no line was missing today, so battle reports show exactly the same chips.

1.284.8 — 2026-09-25

Fixed

• A raid that never lands now closes its "incoming attack" alarms (#4499). Before, a loud guild raid put one red "incoming attack" line in your diary for each fleet in it. If every one of those fleets then turned back at impact (your Veil hid the colony from them, or a launch rule refused them), no battle happened and those red lines stayed open forever. Now each one gets a dim "Inbound flight turned back" line naming the hour the raid was due, with no notification, just as a recalled attack already did. The line also reaches whoever holds the colony at impact when a fleet that joined late alarmed them rather than the raid's original target. Two cases stay deliberately silent. First, when another raid is due at the same colony at the same instant, the alarms cannot be told apart, so none is closed rather than risk telling you a fleet still inbound has left. The raid-leader's cancel now follows the same rule. Second, when you gave up the colony before impact, you get no closing line; that is still open as #4503. Production held no alarms left open by this bug.

1.284.7 — 2026-09-25

Fixed

• A Dark Matter debt under a hundredth no longer reads as zero (#4866). A balance a hair below zero, such as -0.004 DM, printed -0.00 on the HUD's Dark Matter chip, on both the desktop line and the phone pill. That reads as no debt, while the note under it said "You are in debt by less than 0.01 DM". The chip now prints -<0.01, the same figure the Apex core shows. The phone's Dark Matter sheet, the Empire Overview, the Anchorage wallet, the Warp "Not enough Dark Matter" lines and the "Held" column of the advanced-core cost table print it the same way. Balances that round to a real hundredth are unchanged, and so are zero and positive balances.

1.284.6 — 2026-09-25

Fixed

• AI empires now count a colony's garrison fleets before they attack it (#4734).
  • Before: a deep spy probe lists every fleet an ally has garrisoned at a colony. AI empires ignored those ships and sized a strike only against the owner's own hangar. A colony defended by a friend's garrison looked easy, so they sent a handful of ships into a wall that destroyed them. In a battle bench of 768 garrisoned colonies, 601 of their strikes had a lower chance of winning in the real battle engine than the AI itself requires before it launches.
  • Now: the garrisoned ships count toward the defence when they plan, muster and launch. They also count when an AI decides a target is out of its reach. In the same bench, 3 strikes still fall short, and those 3 come from a separate known gap (#4766).
  • When a strike is called off because of a garrison, the AI says the colony's hangar together with the fleets garrisoned there was too much. It no longer blames the owner's own hangar alone.
  • Timing: each AI starts doing this with its first probe of a colony on or after 28 September 2026 (UTC), not all at once. A fleet already in the air is not called home because of a garrison that was there when it launched.
  • Not covered yet:
    • how strong a lender's ships are made by their own mastery and Ward;
    • an Aegis Bastion shielding the ships beside it (#4860);
    • a lone legendary hull that a single attacker cannot destroy (#4861).

1.284.5 — 2026-09-25

Fixed

• The system list and the target panel call the Gate "Gate" (#4478). The last column of the system list (the table under a star on the galaxy map) was headed Power, and so was the same figure on the Missions target panel. But that number is the Gate, the larger of your Structural and Military. Power is what the HUD shows, and that one is the sum of the two, so it can read up to double. The column header, the label on each phone card and the target panel's chip now read Gate, matching the Leaderboard, the War Room and the war-target rows. Hovering an unread cell (…) now says "Gate could not be read — the game keeps checking." The numbers themselves are unchanged.

1.284.4 — 2026-09-25

Fixed

• A Dark Matter debt reads the same everywhere on the Apex (#4620). A player 263.75 DM in debt used to see -264 in the centre of the desktop Apex Chart, directly under a note saying they were 263.75 in debt. The Sigil sheet and the Fold recall button in Missions said "you have -264" too. The core now shows -263.75, and a screen reader hears the same figure.
• Refusals say a debt as a debt. Every button that names your balance to explain a refusal now reads "Need 250 DM, you are 263.75 DM in debt", never a bare minus. This covers the Doctrine sheet's purchase and "To next keystone" lines, the Sigil sheet and the Fold recall button. The recall button also tells screen readers why it is refused.
• A large debt fits the Chart's core. The exact figure is kept and drawn a little smaller, so -12345.67 stays inside the disc.
• A debt under a hundredth no longer shows as "0.00", which looked like no debt at all. The debt note, the refusals and the Inventory row say "less than 0.01 DM", and the Chart core shows "-<0.01". The HUD chip still shows -0.00 in that narrow case; that is #4866.

The first half of #4620, a Dark Matter chip tap that named the line already open and did nothing, was fixed earlier in #4630.

1.284.3 — 2026-09-25

Also

• The Apex sigils are guarded against the next edit (#4588). Nothing changes on screen: the Chart, the phone grid and the Sigil glyph show the same art as before. New tests now catch three mistakes before a player could see them:
  • an Ascendancy borrowing the wrong Doctrine's sigil while its own is withheld (the pairing is read from the design record);
  • a sigil's vector source drifting from the generator that drew it;
  • any sigil URL, fallbacks included, pointing at a file the site does not serve.

  The fourth guard the issue asked for, that every sigil URL carries its cache key, had already shipped with the desktop Chart.

1.284.2 — 2026-09-25

Fixed

A strike an agent promised in chat no longer vanishes in silence when its target was never scouted (#4829)

Before: an agent that told a player "I'll hit you at 1:2:3" and never managed to get a probe onto that colony simply dropped the plan after two days. The player saw no attack, heard no explanation, and the agent had nothing true to say about it if asked.

Now: the agent first goes to look. Its probe goes out at a moment that differs from agent to agent, not the instant after the chat line. The agent gives up only when no look is left to try: no probe in the air, none at home, none being built or affordable. It gives up while awake, at a moment of its own before the plan would have lapsed. When it gives up, it can say so truthfully: "you never got a probe onto them in time, so it never went ahead."

The agent never says this about a colony it had actually scouted. When a probe did get there, the agent gives the reason that is true instead:

• the colony had changed hands;
• there was no longer a colony it could see;
• the colony was held by a guildmate or ally, so the probe was turned away and the attack was ruled out too.

A promise attached to an older plan now gets its own time to be kept. Repeating the promise later does not keep a plan alive forever.

1.284.1 — 2026-09-25

Also

• Apex purchases: a test now guards against paying for a level you do not keep (#4618). When two Apex purchases land at the same moment and your Dark Matter covers both, a check on your mother colony makes the second one wait and re-read your tree, so neither overwrites the other. That check already worked, but nothing tested it: with it removed, two runs to the same keystone could charge twice for one set of levels, and a Lance level bought alongside an Aegis level could vanish after being paid for. New tests pin all three cases with a purse large enough for both purchases. Nothing changes in play.

1.284.0 — 2026-09-25

New

• Research: a Tech Tree card reached from a requirement row leads back to where the walk started (#4855). Since v1.283.10 each requirement under "Why it's locked" opens the node it names, so you could walk a chain backwards: Orbital Arsenal → Autonomous Factory → Recycler. Getting back meant closing the card and finding the first node again. On a phone that also meant switching the branch row back. Now a card you reach this way shows ‹ Autonomous Factory above its name, at the top of the sheet on a phone. It takes you back one card at a time, restores the branch and puts focus on the card. Pressing a Discipline requirement and then returning to Tech tree reopens the card you left, with its way back. The Laboratory requirement works the same way when you come straight back from the Base. Picking another node by hand or closing the card starts a fresh walk.

1.283.12 — 2026-09-25

Fixed

• Galaxy map: a click the browser cancels mid-flight no longer names a place at rest (#4593). On desktop, while the camera was flying (Go to, the + / − rail, a wheel notch), a mouse press that barely moved and was then taken away by the browser, typically by a native drag starting after a few pixels, froze the hover card for the rest of the flight. When the camera stopped, the card named whatever system or planet had landed under the pointer, ring and all, though the player never pointed at it. Now such a press is treated like a click. The card is kept or dropped during the flight and never names a new place at rest. A real pan that the browser cancels still keeps its card through the ease, as a released pan does.

1.283.11 — 2026-09-25

Fixed

• An intercepted fleet's battle report no longer shows its Ramparts as if they fired (#4859). The Mantle, the Battery, the Keep and the Gunnery strengthen a colony's own defences: the Planetary Shield's Bubble, the Defense Batteries, the Orbital Fortress and the Orbital Cannon. So they do nothing when a fleet is caught in flight, or in a Crucible duel or a boss fight. The defender's card still drew their chips solid, like nodes that had acted. On a report that says the battle was not a colony defence, those chips are now dashed and read "colony defense only", like the Phalanx's since #4646. One footnote names every node that sat idle, e.g. "The Phalanx, the Mantle and the Keep are owned but did not act in this battle…". The attacker's card now words the four Ramparts, and the Phalanx, the same way: "colony defense only", with a footnote that says the colony rather than just "when defending". A colony defence, and any report from before #4646, looks exactly as before.
• Not covered: a colony defence where the colony lacks the building (a Mantle with no Shield) still shows that chip solid, and so does an allied garrison's Rampart at someone else's colony.

1.283.10 — 2026-09-25

Changed

The requirements on a locked Tech Tree card lead to what they name (#4694)

Before: a locked node's card listed what it needed under "Why it's locked" as plain text, for example "Cybernetics 22 — you're at 3", "Autonomous Factory — not unlocked yet" and "Laboratory lvl 12". To follow a chain back from a destination, you closed the card and searched the tree for each name.

Now: each requirement is a button, with a small tag that says where it leads:

• A prerequisite tech (tag "Tree ›") opens that tech's card in place of this one and scrolls the tree to it. On a phone the same sheet stays open and switches to the prerequisite, starting from its top, and the branch row follows when the prerequisite is in another branch.
• A Discipline level (tag "Disciplines ›") opens that Discipline under the Disciplines lens. If a filter you left on would hide it, the filter goes back to All.
• The Laboratory (tag "Base ›") opens the Base on your best Laboratory, the one that sets your research network, even when you are viewing another colony.

The "What it's waiting for" list on a tech you can already plan works the same way. Every row is a full-size touch target, and screen readers hear the requirement, whether it is met, and where the button goes.

Not in this release: after walking back through a chain, there is no button that returns you to where you started. That is tracked in #4855.

1.283.9 — 2026-09-25

Fixed

• An intercepted convoy's battle report no longer shows the Phalanx as if it held (#4646). The Phalanx adds hull only while its owner defends a colony. A convoy caught in flight fights without it. On the defender's card of an interception, though, the Phalanx chip was drawn solid, the style kept for nodes that acted, next to a "+40% hull" note. Reports now record whether the battle was a colony defence. On a battle that was not one (an interception, an anomaly ambush, a Crucible duel), the defender's Phalanx chip is dashed and reads "colony defense only", with its own footnote, and the "+40% hull" note is gone. Colony defences look as before. Reports written before this release render the defender card the way a colony defence does, which is how they rendered until now.
• The attacker's card now says "colony defense only" on its Phalanx chip too, instead of "defense only", so both cards of one report state the same rule. When a defender's card has two chips dashed for two different reasons, each footnote now names its own node.
• Installed app versions keep showing the old solid chip until they update: the fix is in the app, and the new report field is simply ignored by older versions.

1.283.8 — 2026-09-25

Fixed

• The supply-lane form no longer says the destination's Trade Portal does not matter (#4568). The capacity line named what the origin's Portal moves per hour, then added that the destination "receives it whether or not it has a Portal of its own". That line sat right under a quote naming the destination's Portal as the limit. One player believed the second sentence and sized a lane about three times over what his hub could take. The line now names both ends with the figures the hourly sweep uses. It gives the origin's outflow, and the destination's intake with its Portal level: "Depth's Trade Portal (L4) accepts up to 1,405/h". It also says the intake is shared with the other lanes already landing there, and that levelling that Portal is what raises it. A colony with no Portal still reads that it needs none to receive, now with the level-1 floor's figure (264/h) and the note that a Portal there raises it from level 2.
• The share quote on the same form is right for hubs fed by lanes of very different sizes (#4568). When the quote worked out a lane's share of a hub's intake, it first capped every feeder's offer at that intake. The sweep does not cap them: it splits the intake by what each feeder actually offers. The two only disagree when a hub is oversubscribed and at least one feeder offers more than the hub's whole intake. In that case the larger feeders were quoted too little and the smaller ones too much. For example, 6,224/h and 2,229/h going into a hub that accepts 1,405/h were each quoted about 702/h, but the sweep delivers 1,034/h and 370/h. The quote now matches what arrives. The lane cards were already right.

1.283.7 — 2026-09-25

Fixed

• Energy bonuses now say they multiply, and show the total (#4605). With the Superconducting Grid at 5 and the Dynamo at 20, the energy tooltip and the Fusion Reactor card used to read "+40% Grid · +40% Dynamo". Production actually rose by +96%, but a player adding the two got +80%. The card now reads "+40% Grid · +40% Dynamo · +96% compounded". The tooltip ends with "The lines above multiply to +96% energy production — they do not add." The planet's own bonus or malus counts toward that total: a lava world with both maxed reads +135%. On the Zero-Point Reactor, a maxed Dynamo's Zero-Point Lattice now gets its own "×1.5 Zero-Point Lattice" chip and is counted in the total.
• The Power Details modal explains its multiplying rows (#4605). With the Grid maxed, the Dynamo row read +56% beside a card saying +40%, and nothing explained the gap. Rows that multiply the lines above them (the Dynamo, Doctrine of Annihilation, War Crucible) now carry a note such as "×1.40 on the lines above — they multiply, not add".
• The second half of #4605 was already fixed before this release and needed no change: a maxed Ascendancy's card reads "Ascendancy complete", not "Doctrine complete".

1.283.6 — 2026-09-25

Fixed

• A synthetic empire gathering against a Fortress no longer switches to its own ship mix while the launch still refuses (#4802). Since #4709, an agent sieging an Orbital Fortress buys the hull the Fortress's counter table punishes least, and keeps its own capped core of Titans. Before, that stopped once the fleet passed the wall estimate and met the value bar. If the battle forecast still said no at that point, the shipyard went back to the persona's usual mix: aggressive empires ordered Titan blocks far past their core, and cautious ones ordered Probes, Freighters and Corvettes on every tick, never getting closer to a launch. Now the gather keeps ordering the Fortress's siege hull, the person's Titan core and the person's own secondary hull. These orders still wait behind a stalled outpost's Freighter, the plan's explicit counts and a besieged capital's garrison.
• Behind a screen that counters the Bomber, the gather no longer piles up Titans. A fleet seen at home that counters the Bomber (40 Interceptors, say) used to make the Titan stand in with no limit, and almost every empire's first order was the same two Titans. Now, within the Fortress siege, no Titan stands in past a person's core.
• An empire that cannot fit the siege in the fleet it keeps now calls it off instead of stalling. This happens only when all of these hold:
  • the siege hull is built to its fleet cap;
  • the launch has just refused with every built hull;
  • no other colony can reinforce;
  • no transfer is in the air.

  The empire then calls the siege off after a delay of its own, anywhere from 40 minutes to five hours, drawn per person and per target. Its sheet says why: "the last probe showed a Fortress the fleet you could field would not take".

  On the #4709 test roster, 120 empires ran for 60 ticks against Fortresses 12 to 20, with or without an Interceptor screen at home. Every empire either launched or called its siege off. Before this fix, up to 96 of 120 were still sitting in "gather" at the end.

• Each empire switches on its next spy report of the target observed from 2026-09-26 00:00 UTC. That is the same cutover as #4709 and #4774, so no empire changes its orders on the day of the deploy.

1.283.5 — 2026-09-25

Fixed

• Tree node cards say where their choice is made (#4661). The Industrial Edicts card named three edicts, then said only "Unlocked." Nothing on it pointed to the picker, which sits at the top of the Tree, and on a phone the card covered it. Four nodes grant a choice rather than a number, and each card now says where that choice is made once the node is researched. Industrial Edicts and Boundless Reserve also get a button that closes the card and brings the edict picker or the Requisition Surge panel into view. Planetary Mandates names the Colony tab. Admiralty names fleet launches (the Missions tab or a guild raid) and the Shipyard's Planetary defense panel.
• The Requisition Surge panel no longer appears late. It used to render nothing until its status loaded. It now shows its title and "Checking the Surge…" while the status is loading, and after a failed read until the next refresh.

1.283.4 — 2026-09-25

Fixed

• Research times now include the Loom (#4602). If you held the Loom, the Research tab showed research times without its discount. At Loom 20 a card could read "10m 58s" for an order that actually ran in 8m 46s. The Discipline cards, the Discipline detail, the full level table and the Tech Tree node card (both available and locked, on desktop and on phone) now show the time the order really takes. Nothing changes if you don't hold the Loom. Quantum Computers still shortens only Tech Tree research, never Disciplines.

1.283.3 — 2026-09-25

Fixed

• The battle replay prints every number the same way (#4589). The replay's closing card used to mix three formats on one screen. It read Wreckage: 18.900 Fe · 9750 Xe, with a period where the rest of the game puts a comma and no separator at all on the number beside it. Next to that were Loot: 3111 Fe · 1866 Xe and 0 / 60.000 HP, and 18.900 could be read as a decimal. The round counter under the replay and the round banners printed raw digits (60/3266 rounds, Round 3247). Every figure in the replay now uses the same grouping as the rest of the game: Wreckage: 18,900 Fe · 9,750 Xe, Loot: 3,111 Fe · 1,866 Xe · 622 Dt, 60/3,266 rounds, Round 3,247. The Bubble's break rounds, the ship counts on the group labels and the Bubble strip's HP changed the same way.

1.283.2 — 2026-09-25

Fixed

• Fleet Advisor: the rationale describes the fleet it recommends (#4823). On an attack, the rationale could say "Battleship leads the composition" over a recommended fleet of one Corvette. The line named the top of the advisor's ranking, which the search and the "Can stay home" pass may have left out. It now names the best-ranked hull that is actually in the fleet. When a hull you own ranks higher but stays home, the line says "of the hulls sent" instead of claiming the overall best. The line above it said "+20% safety margin" on every winning attack, even when your hangar had fewer ships than the margin wanted. It now says "+20%" only when every armed hull got its full margin. Otherwise it reads "with a partial safety margin (your hangar capped it)", or "with no safety margin" and the reason.

1.283.1 — 2026-09-25

Changed

• Galaxy survey: each group's All and None sit side by side (#4688). In the Galaxy survey panel, each group's All N button was pinned to the far right of its header, and the only None was a separate row at the top-left of the panel. Ticking a group and then changing your mind meant crossing the whole panel. Each group ("Colonies on your scan", "No colony on your scan") now has its own All N and None next to its heading. A group's None clears only that group, the same way its All ticks only that group, so clearing both groups now takes two taps. The group's "N ticked" count has its own line, so the first tick no longer shifts the buttons on a phone. Every button is still a 44px touch target.

1.283.0 — 2026-09-25

New

• Previous / next colony buttons (#4691). Until now, changing colony meant opening the switcher every time (desktop could also cycle with C / Shift+C, but only the dropdown's tooltip said so). With two or more colonies:
  • on desktop, two round ‹ › buttons sit right after the colony dropdown in the header, and their tooltip names the colony they go to and the key that does the same;
  • on a phone, the colony sheet opens with two buttons showing the previous and next colony's names. A tap switches to that colony and keeps the sheet open, so you can keep stepping while the active row follows. Tapping a row still picks it and closes the sheet.

  Both follow the order the switcher lists your colonies in, wrap round at either end, and land on the same colony C / Shift+C would. Every button is at least 44px tall and tells a screen reader which colony it opens. The phone header and resource strip are unchanged: next to the colony name there was no room without pushing the header onto a second line.

1.282.2 — 2026-09-25

Fixed

• Shipyard: scrolling the hull list no longer scrolls the whole page (#4693). On desktop, browsing the list of hulls on the left used to scroll the entire page, so the ship on display and the build queue slid off the top of the screen. The list now scrolls on its own, pinned under the top bar next to the order panel, and the stage, the queue and the BUILD button stay where they are. Once you reach the end of the list, further scrolling moves the page as usual, which is also how the bottom of the list comes into view on a fresh load. Keyboard focus rings on the list are no longer cut off at its edges. Phones are unchanged on purpose: the hull list there is the "All hulls" sheet, which already scrolls on its own, and the page keeps its normal scroll with the BUILD bar within thumb reach.

1.282.1 — 2026-09-25

Fixed

• Planet sheet: closing it quickly no longer builds a hidden 3D preview, and a preview that cannot start no longer raises an error (#4834). Closing a planet's sheet before its 3D preview had finished loading still built the preview in the background, for a sheet that was already gone, and tore it down straight away. The sheet now builds nothing once it is closed. If the preview cannot start at all, for example on a device with no WebGL left, its box stays empty and the rest of the sheet works as usual. Before, this surfaced as an unhandled error. Nothing else on the sheet changes.

1.282.0 — 2026-09-25

New

• Targets board: pick targets and spy on them in one confirm (#4697). Before, probing a filtered list of targets meant the same loop for every one: open the planet, tap Spy, confirm in the Missions tab. Now "Pick targets to spy" turns row taps into picks, and "Pick all" takes the filtered list in your sort order, up to 20 per sweep. It leaves out rows behind a pact, since guildmates and allies refuse a probe, and rows you already probed from the board this session, so a second tap takes the next 20. Tapping a row still picks it. One confirm then shows the cost in Spy Probes, your home hangar stock and every target that will fly. If the hangar is short, it says so there: only the first ones fly, and the rest stay picked. Every target is an ordinary espionage mission from your home colony. The result lists each target with what happened to it: probe away, refused with the server's reason, unknown (check your missions), or not sent.

Fixed

• A bulk probe sweep no longer calls a rate-limit refusal "unknown" (#4697). If the server's per-player request limit answered a sweep, including the System Survey's, the sweep reported "unknown, check your missions" for a probe that had never left. It now says the request was refused, stops the sweep, and lists the remaining targets as not sent.

1.281.4 — 2026-09-25

Fixed

• Test suites no longer fail a fully passing run on an internal timeout (#4821). Internal tooling only; players see no change. Before this fix, the pre-merge gate's test run sometimes went red even though every test passed. The only error was [vitest-worker]: Timeout calling "onTaskUpdate", and each occurrence cost a full re-run of the gate. The cause was test files made entirely of promise chains on a fake clock. While such a file runs, the test worker never reads the acknowledgements vitest sends it, and vitest gives up after a fixed 60 s. One such file measured 14 s alone and 37 s under load, and the busy gate machine pushed it past 60 s. The api, web and shared test suites now give the worker's event loop one turn before and after every test, so the wait is bounded by a single test and its hooks rather than a whole file. Nothing is forgiven: the gate script is unchanged, and a failing test or any other error still turns it red. What is left open: a single test and its hooks that together hold the worker for more than 60 s can still produce the same error.

1.281.3 — 2026-09-25

Fixed

• A strike an AI empire promised in chat and then called off is explained, whatever called it off (#4808). Before: if an agent told a player "I'll hit you" and the battle preview before launch then gave its fleet too poor a chance, the strike was quietly dropped. The same happened for several other reasons: a refused launch order, a pact, the colony changing hands or vanishing from view, a probe showing more defence than the fleet could take, and more. The player saw no attack and the agent had nothing true to say about it. Now each of these leaves the agent one true line to speak from for a day, for example "the battle preview before launch gave the fleet too poor a chance of winning, so you held it back". The line survives the agent moving on to its next target. A fleet the server turned back on arrival is described as coming back unfought, not as called off. Nothing is sent at the moment of the refusal: the agent brings it up only in its next ordinary reply, with its usual delays. A refused launch at the Beacon no longer makes the agent claim that a launch at a different colony was refused.
• Left out: a promise whose target is never scouted still ages out silently after two days. No closure is recorded for it.

1.281.2 — 2026-09-25

Fixed

• Buildings on gas planets no longer look dyed yellow-green (#4825). On the new gas-giant platform every building came out tinted a sickly yellow-green, and the cyan details on the Veil, Compute Node, Shield and Lab turned green. The gas world's light is now a pale cream: buildings read in the platform's own white and grey, and their cyan accents stay cyan.

1.281.1 — 2026-09-25

Fixed

• The Fleet Advisor no longer sends hull types that only add losses (#4454). Against a lightly defended target, the advisor could recommend a fleet that lost a hull when the same fleet minus one hull type won every simulated battle and lost nothing. For example, against seven Interceptors it sent two Orbital Scythes and four Interceptors and lost an Interceptor, when the two Scythes alone win with nothing lost. It happened mostly on small hangars, where the safety margin rounds every type up.
  • Now the advisor tries leaving each warship type at home. It accepts the smaller fleet only when it still wins every simulated battle, loses less both on average and in the losses the panel prints, brings home no less loot net of losses (judged with the Freighters the loot needs aboard), and still wins at least as often against a defender whose fleet grew by the same 20% margin.
  • The panel says so: "Can stay home: Interceptor — on the latest report the fleet wins every simulated battle without them, and loses less."
  • What this does not promise: the margin's cover is kept as a win, not as a toll. If the defender really has grown, the leaner fleet still wins but can lose more than the fuller one would have. Where the fuller fleet would already lose against the grown defender, the check protects nothing. On a 4,500-case census, 146 recommendations changed; none lost a win, none lost more on the report, and total expected losses fell by 2.5%.
  • On large, varied hangars the recommendation takes longer to compute: up to about 2.5× on the worst hangar measured.

1.281.0 — 2026-09-25

New

• Explore a signal straight from the signal list (#4692). Before, the anomaly signal list only opened the system. To explore a signal you went system view → planet → Missions tab → send, and walked that route again for every signal. Now each row has an Explore action next to "Open this system". It lists the system's signals by planet and tier, each with its own Explore planet N button and a trip estimate (an upper bound, because Hypergates and Relays only shorten it).
  • Before anything is sent, the list states what a tap spends: Sends 1 Spy Probe from your home colony · uses 1 exploration · 3 of 5 explorations left today. The hull is the one the Missions tab would pick from your home hangar. When it isn't a probe, the list shows the Advisor's own caveat.
  • Each signal reports its own outcome: sent, refused (with the reason), or unknown when the connection failed. After an unknown, the list checks your missions and shows either "Explorer on its way" or the button again.
  • A signal you already have an explorer flying to says so instead of offering a second send.
  • Sends go one at a time and wait for your hangar and daily count to refresh, so a double tap, a folded row or a second row can't send a second explorer by mistake.
  • An empty hangar or a spent day disables the sends and says why.

1.280.4 — 2026-09-25

Changed

• A new home on gas giants (#4809). Colonies on gas planets used to stand on a flat grey tiled slab floating in murky orange-green clouds. They now stand on a floating platform built like the rest of the base — a dark seamless floor inside an armoured white rim with teal light strips and heavy machinery underneath — above the gas giant's swirling cloud bands. The build grid sits on the new floor with room to spare on every side, and on phones the colony still opens centred on the platform.

1.280.3 — 2026-09-25

Fixed

• Apex: the Dark Matter the heading shows is now the Dark Matter a purchase can spend (#4617). The Dark Matter figure on the Apex tab counts what all your colonies have earned, including drip the server had not yet written down. The buy buttons, the "To next keystone" run and its "You have S DM" confirm, and the Sigil switch are all enabled on that figure. The purchase itself only counted your first colony's recent drip. So a button the heading covered could still answer "Not enough Dark Matter." by up to a few minutes of the other colonies' drip. Now every colony's drip is written in before you pay, so a purchase the heading covers goes through. A request that is refused for another reason (a line already maxed, a target you already hold) moves no Dark Matter.

1.280.2 — 2026-09-25

Fixed

Sanctuary pays a later attacker on time (#4722). Sanctuary (Ward 15) brings back a colony's Emplacements and Fortress platforms an hour after a lost defence. The ones it restores are worth no war points until the usual rebuild would have returned them.

• Before: a restored Emplacement destroyed before that moment stayed counted as "unpaid". The count then went on marking Emplacements the usual rebuild had since returned as worth nothing. The next attacker was paid less than at a colony without the rung. Because that attacker left the regrowing war-point allowance unspent, the attacker after them was paid more.
• Now: a restored unit destroyed early waits in the usual rebuild queue, in the slot it held. It comes back, and is worth war points again, at the moment a colony without the rung would have had it back.
• Fast Rebuild: restored Emplacements now follow the owner's own rebuild pace, 10 minutes each. Before, they were held unpaid on 30-minute slots.

Measured over 1,200 random attack sequences on small walls:

• the holder's attackers earned more in total in 7 of the 1,200 (was 16). All 7 are sequences where the restored wall changed how a battle ended;
• none has every outcome equal (was 9);
• the largest total gap is 17.8% (was 17.1%). One of those 7 was underpaid by the old count and is now the largest.

With Fast Rebuild on both sides, none of the 1,200 ends with the holder's attackers ahead (was 4).

Sanctuary's description on the Ward now says this.

1.280.1 — 2026-09-24

Fixed

• Apex: "To next keystone" asks before it spends your Dark Matter (#4687). On a line's Doctrine sheet, the quiet line under ASCEND ("To next keystone · ×3 · 60 DM") used to buy every level up to the next keystone the moment you tapped it, with no second step. On a phone it sits right under your thumb. Now the tap opens a confirmation first. It shows which levels you are buying and the single price ("Levels 8 to 10 of Dark Matter Lances: 3 levels for 60 DM, in one purchase"), your stat now and at the new level, and the keystone you reach and what it does. If that keystone also needs another line that you don't have yet, the confirmation says so: the Phalanx's Interlock is reached but does not apply until you hold the Aegis at level 15. The confirmation also says that you get every level or none, and that Apex levels are not refunded. Nothing is spent until you tap Buy. If your balance or the price changes while the confirmation is open, it closes without buying anything, and a line under the run says so: "That run changed before you confirmed. Nothing was bought." ASCEND still buys one level on one tap.

1.280.0 — 2026-09-24

New

• The Targets board remembers your filters between visits (#4698). Before, the board kept your band chips, sort, Hide protected, Hide pirates & NPCs, Sanctioned only, the AI / Guild / Silent chips and a typed Gate range only until you closed the tab. Your next visit opened on the defaults. Now it opens on the filters you last set, on the same device. They are saved per account, so two players sharing a phone or a browser each get their own, and logging out doesn't erase them. The typed Gate range comes back in its own boxes, where you can see it and clear it. Showing all rows, the expanded list and the scroll position still reset on each visit. The Targets badge on the map rail still counts every target not behind a pact, a shelter or a shield. It does not apply your saved filters, so it can show more than the board lists.

1.279.19 — 2026-09-24

Fixed

• Synthetic empires no longer stall in "gather" against a fleet that counters their hulls (#4774). Since v1.279.9 the agents' launch has weighed a defending fleet at its counters against the ships they would send. The force they gathered towards, the ships their shipyards bought for it and the "out of reach" check still weighed the same fleet at face value. An agent facing, say, Battleships that shred Cruisers could reach its target and then have the launch refuse, and it would sit in "gather" with nothing leaving. Some shipyards also bought the exact hull the defence counters. Now the target, the shipyard and the reach all read the defence the way the launch does. The yard buys a hull the seen fleet does not counter when one does the same job, and the gather converges to a launch that wins in the battle engine.
• The change reaches each agent on its next scouting report, not on the day of the update. An agent keeps its current orders until it next probes the target, from 2026-09-26 on, the same switch as the Fortress change in v1.279.17. You will not see a whole cohort of synthetic empires change what they build, or call off strikes, at the same moment.
• Fleets already in the air are judged the way they were launched. A strike launched before the update is recalled on exactly the evidence that would have recalled it before, and never because the update reread the target.
• When an agent calls off a strike as out of reach, the reason it knows is true under the new read. The reason it can speak from is now "a hangar that outmatches, or counters the hulls of, anything you could send", not only "bigger".

Left out: a raid coalition's check on a target's garrison still reads that garrison at face value, and the Bubble's shelter of the defending ships (#4766) is not yet priced.

1.279.18 — 2026-09-24

Fixed

• The Leaderboard no longer blames the Wraiths for faction sorties (#4675). The war-week fine print used to say that "a Wraith sortie scores for nobody". The sorties it means come from the three faction bastions: the Silent Choir, the Black Foundries and the Exarch Guard. The Wraiths are the synthetic empires with their own badge, and repelling one of their attacks scores like any other battle. The sentence now reads "a sortie from a faction bastion scores for nobody". The scoring is unchanged; only the name is fixed.

1.279.17 — 2026-09-24

Fixed

• Agent empires stop buying Titans against a Fortress that counters them (#4709).
  • Before: an agent gathering a strike against a colony with an Orbital Fortress, and with no Breachhead, kept buying Titan Dreadnoughts. The Fortress's platforms hit Titans twice as hard, and 2.5 times as hard once the owner holds the Keep at 10. Agents without the Titan bought Cruisers, which the Fortress counters too.
  • Now: they mostly build Bombers, the siege hull the Fortress does not counter. In the battle engine, 120 Bombers take a Keep-10 Fortress 20 that needs 18 Titans for 60% more money.
  • Each agent keeps its own habits. Some keep a share of Titans, fewer against a Keep-10 wall. Others bring along 3-10% of a ship from their own usual mix that the Fortress does not counter. Each orders at its own pace. Every one of these mixes still wins that battle for less than pure Titans would cost.
  • When the switch happens: agents that spied their target before 26 September keep their old orders until their next spy report after that date, so they change course one at a time, each after a report, not all at once.
  • Not changed: once a gather's force reads strong enough to break the wall but its launch is still refused, the agent's usual shipyard mix takes over, and for aggressive agents that mix still leads with Titans.

1.279.16 — 2026-09-24

Fixed

• Apex: every keystone says what it does, not only the next one (#4680). A Doctrine's sheet used to explain only the next keystone. Held and later keystones showed a bare name, and hovering a chest on the track showed the name alone. So you could not read a capstone until it was next in line. Now every keystone prints its effect. The held ones keep their gold ✓, the next one its gold "Next ·", and later ones read dimmer, so the one you are buying toward still stands out. A keystone that is reached but waiting on another line says why it is off before it says what it will do. Hovering a chest shows the same sentence. Keystones a line has not shipped stay sealed. On a phone, the ASCEND button now sits right under the track, and the keystone list follows it, so the longer text does not push the button off screen.

1.279.15 — 2026-09-24

Fixed

• The padlock on a Dark Matter unlock names the gate you are actually missing (#4600). The ZPM Reactor, the Singularity Vault and the Stellar Forge are locked behind Precursor Schematics, which needs the Particle Accelerator and Production Singularity at level 3. Their padlock only ever said "needs the Particle Accelerator", so a player who already had the Accelerator was sent back to it, and the Singularity was never mentioned. The padlock now lists only the prerequisites you have not met, in the same words as the Apex card: "needs Production Singularity level 3", with where to find it in the Tech Tree. Once every gate is met it shows only where the node is bought. The same rule covers the other Dark Matter nodes for the day they start unlocking buildings or hulls: an outer Ascendancy names its Doctrine at level 10, the Sigil names "a Doctrine at level 20", a Rampart names its building, the Codex names its repelled sortie. The padlock cannot see your colonies or battle record, so a Rampart's building and the Codex's sortie are always listed, even once you have them; the Apex card drops them when they are met.

1.279.14 — 2026-09-24

Fixed

• A bounty no longer pushes you into Dark Matter debt (#4701).
  • Before: when a bounty was claimed on you, the full payout came out of your Dark Matter, even if you held less, so players ended up at −103 DM and worse.
  • Now: a claim takes what you hold and stops at zero. If you are already below zero from before this change, a claim takes nothing more, and your existing balance is not forgiven either.
  • Hunters are unaffected: they are paid the full bounty exactly as before.
  • Your Activity diary:
    • When a claim was more than you could cover, your entry says so: "claimed 250.00 Dark Matter; only 30.00 of it could be taken from your balance".
    • The hunter's entry now reads "collected X Dark Matter" rather than "took X off them".
  • Not included: Dark Matter taken by the battle that was reverted on 9 September is not returned automatically.

1.279.13 — 2026-09-24

Fixed

• The wiki no longer tells you to level your Laboratory for Dark Matter lines (#4601). The public Research page listed every Apex line (the Doctrines, Ascendancies, Ramparts and Relics) inside the tech-tree tables, each with a Laboratory level. The Dynamo, for example, read "Laboratory 15". No Laboratory level can buy them: they never enter the research queue and are bought instantly with Dark Matter. The tech-tree tables now list only what you actually research. The Apex lines have their own section, "Bought with Dark Matter: the Apex", which gives each line:
  • its kind;
  • its Dark Matter price: the price per level and the total to level 20, or the one-off price of a Relic;
  • what it really needs: a Doctrine at 10 for an Ascendancy, the building at 10 on any colony for a Rampart, any one complete Doctrine for the Sigil, a repelled sortie for the Bastion Codex;
  • its effect.
• The Resources page no longer speaks of "four Dark Matter upgrades in their own panel". It points to the Apex tab and says when that tab appears: while you hold at least one Dark Matter, or once you have researched the Particle Accelerator.

1.279.12 — 2026-09-24

Fixed

• The fleet advisor no longer spends your Colonial Ark on raids and hauls (#4587). Before, the advisor added Arks as hold on raids against rich targets and used them for transports, pickups and harvests. It kept only the Colony Ship home. Now both founding hulls (Colonial Ark and Colony Ship) stay out of what the advisor proposes whenever you own another carrier. When a raid, transport, pickup or harvest falls short on hold, the panel names the founders it left home, and any idle Continuum Haulers on a raid, and says you can add them by hand. A hangar whose only carriers are founders still gets them for a haul.
• Colonization fleets always match the launch. The Ark or Colony Ship that founds is the payload, so on a colonization a second founder (another Ark or a Colony Ship) now rides as hold whenever your other carriers can't lift the endowment. For example, an Ark plus a Colony Ship now carries 15,000. The panel says when that slows the fleet and quotes the real excess over the founder's hold. It warns that the launch will be refused only when your whole hangar cannot lift the endowment.
• "Fill every carrier" never writes less than the hold you have already picked. Its tooltip now says founding hulls count only when you own no other carrier.

1.279.11 — 2026-09-24

Fixed

• AI empires no longer hold back a Cruiser raid because it carries a few light escorts (#4752). Defense Batteries hit Interceptors and Corvettes twice as hard and aim at them first. Since #4743, an AI empire read that doubled fire as lasting the whole battle, even when its only light hulls were a handful of escorts the turrets destroy in the first salvos. Against a Batteries-20 wall, an empire with five Corvettes in its hangar waited for 94 Cruisers before it attacked. Without those Corvettes it attacked with 76. Now the doubled fire is priced only until the escort is dead, and the same empire attacks with 75 Cruisers and 4 Corvettes. Fleets that are mostly or entirely light hulls are read exactly as before. Every launch measured still clears what the battle engine needs to win. One of the synthetic empires sends a smaller task force against a shielded outpost again: the doubled force instead of the quadrupled one.

1.279.10 — 2026-09-24

Also

• The release machinery's own self-test runs on macOS again (#4778). Nothing in the game changed. A check the ship gate runs before every release had been written for GNU tools only, so on a Mac it failed on every branch, including ones that were fine. It now gives the same answer on both.

1.279.9 — 2026-09-24

Fixed

• Agents stop sending too few ships into a fleet that counters theirs (#4758). Before, a synthetic empire judged a colony's defending ships by their plain attack, whatever those ships were firing at. So twenty Interceptors looked no more dangerous to its Bombers (their favourite prey) than to its Cruisers, and Ship Stars, which sharpen a hull's counters with every star, didn't count at all. Against twenty 5-star Interceptors an agent sent 14 Bombers where the battle needs 24. Now agents read each defending ship at its counter against the ships they are about to send, stars included, and send enough: 30 Bombers in that same case. When the ships the defence counters would sink the whole attack, the agent leaves those at home and sends the rest.

  What is still open: a Bubble sheltering the defending ships while they fire is not yet priced, so behind a strong Bubble agents can still send too few (#4766). The engine's own forecast still refuses nearly all of those launches before they leave. The agents' gathering phase doesn't yet read counters either. Until #4774 ships, an agent whose target counters part of its fleet can gather and then hold instead of attacking.

1.279.8 — 2026-09-24

Fixed

• The catch-up ramp now reaches the developer's own account (#4780). Every empire below half the median of the established, active field produces and researches faster — up to ×10 — except, until today, the developer's. The admin address is kept out of the median on purpose, and the same list was being used to decide who gets a boost at all; so when the developer came back to play on that account on 24 September, reset to zero, it got +140/h where any other fresh account gets the ×10 cap. The admin is now ramped like any empire that plays, and its Power still never enters the reference everyone else is measured against. Nothing changes for anyone else: the GM and the Wraiths stay out on both sides, and the standings and the PvE calibration read their own lists as before.

1.279.7 — 2026-09-24

Fixed

• Placing a new Site from the Suggested panel builds that Site (#4719). A Suggested card for a building's next Site (for example "Xenogas Refinery · Site 2") has a ＋ button to place it on the base map. Using it used to start an upgrade of the Site that already stood. On phones the placement was for Site 1, and «Build here» queued Site 1's next level. On desktop the card that opened was Site 1's. Now the ＋ places Site 2 on phones, and opens Site 2's card on desktop. The upgrades that did go through were charged Site 1's real next-level price and delivered that level: the action was wrong, the charge was not. Installed iOS and Android apps keep the old behaviour until they are updated.

1.279.6 — 2026-09-24

Fixed

• AI empires no longer send too few light hulls at a Batteries wall (#4743). Every colony's Defense Batteries hit Interceptors and Corvettes twice as hard. The AI empires' read of a target ignored that and judged a wall as strong against 90 Interceptors as against 20 Cruisers. So an all-Interceptor raid could leave below what it takes to win: at Batteries 20 an AI empire sent 379 Interceptors where 430 are needed. The AI empires now count that double damage, so the same raid leaves with 540 Interceptors (398 Corvettes where 302 are needed). When they plan an attack, they are also told that the turrets hit those two hulls harder.

Changed

• An AI empire whose fleet carries a few Interceptors or Corvettes now waits longer before hitting a Batteries wall (#4743). The turrets focus the most vulnerable hulls, so the AI empires now treat any Interceptor or Corvette in a fleet as drawing the wall's full doubled fire. That errs on the safe side. With only a small light escort in a heavy fleet, it asks for more ships than the battle needs. At Batteries 20, a Cruiser fleet with 4 Corvettes now gathers 94 Cruisers before launching, where 75 were sent before and 65 already win. No raid got weaker. A tighter reading is tracked in #4752.

1.279.5 — 2026-09-24

Fixed

• Synthetic empires size their fleets for a Second Skin Bubble with ships under it (#4673). A colony whose Mantle is at level 10 or higher brings its broken Shield Bubble back once, three rounds after it breaks. When ships sat under that Bubble, the agents counted the Bubble's return but not the three extra rounds of ship fire, nor the fact that the returned Bubble shelters those ships again. So an agent could send a fleet the battle then lost. For example, five Scythes against twenty Interceptors were sent to a Shield 2 colony where they lose almost every time. Now the agents count those rounds, including the ships' Ship Stars and research when a spy report showed them. At such a colony they bring a fleet big enough to win, typically around 15% larger than before, or they stay home. When a spy report showed the colony's Tree, their plans also mention that the Bubble comes back.

  Not fixed here: against a colony's ships, agents can still send too few hulls in an ordinary fight, with or without Second Skin. The gap is widest against starred Interceptors. This is tracked in #4758.

1.279.4 — 2026-09-24

Changed

• Every guild in the Contested Space is paid at the Eschaton (#4767). Until now only the three best guilds in Sector 13 paid their members when the season closed; a guild that lived in the sector all season but finished fourth took nothing. Now every guild with a place in the standings pays each member who scored more than 25, one crate size smaller per place: the fourth pays 5M, 1M or 500K of each resource depending on the member's own rank, the fifth 1M, 500K or 100K, the sixth 500K or 100K, and every guild from the seventh place down pays 100K of each. The top three are unchanged. The Race's table is now called Guild prizes: if your guild is fourth or lower, its own column appears there with your cell marked, and the "If the season ended now" strip shows your crates. A member of a guild with no place yet is told that one colony in Sector 13 puts the guild on the board.

1.279.3 — 2026-09-24

Also

• Groundwork for playing Novarchy in other languages (#4755). Many of you asked for the game in your own language. This release lays the foundations, but nothing changes on screen yet: the game is still English-only, and every word and number reads exactly as before.
  • The game can now remember a preferred language on your account, and otherwise follow your device's language.
  • A language picker is built into settings. It stays hidden until there is a second language to pick.
  • The settings audio rows and the game's error messages are the first text to go through the new translation layer.
  • Translating the rest of the interface, and the first actual language, come in later releases.

1.279.2 — 2026-09-24

Fixed

• AI empires no longer throw fleets at Escalation walls they have scouted (#4742). The Doctrine of Escalation makes a colony's Defense Batteries and Orbital Fortress fire and shield at ten times strength. Until now the AI empires sized their strikes as if it did not exist. They could send a fleet built for a plain wall at an Escalation colony and lose all of it.
  • When a spy report shows the owner holding Escalation, the AI now prices that wall at its real strength. It brings a fleet big enough to break it, or it does not go.
  • Measured against the battle engine, the smallest force an AI sends at such a wall is at least the force that actually wins. For example, against Batteries 20 an AI now sends 473 Cruisers where the engine needs 229, instead of a fleet sized for the plain wall.
  • Its war plan names the doctrine on that wall only when its report saw it.
  • An AI empire that holds Escalation now reads its own walls at their real strength, so it no longer flees raids its wall would repel.
  • As in the launch forecast players see, Escalation is read only from a spy report that showed the target's Tree. A wall no spy has read is still valued without it.

1.279.1 — 2026-09-24

Fixed

• Synthetic empires now count Bastion Relay when they weigh a raid (#4713). When the Ward is held at 20, a garrison's Aegis Bastions lend their shield to the colony's shield bubble. The synthetic empires used to read only the Shield building's bubble, so they could raid a colony whose relayed bubble then held. Now, when their probe has seen Bastions stationed there in a garrison that lends its shields, they add them to the wall they have to break. Each Bastion is priced at the most one can lend, because the probe cannot see how much a lender's research and tree add. That is up to 3.8 times what a bare Ward-20 lender really gives, so they turn some raids down that would have worked. Bastions parked by an empire without the Ward lend nothing, and the synthetic empires ignore them.

Changed

• The deepest spy reports show which garrisons lend their shields (#4713). On a tier-3 spy report, a garrison row now adds "its Aegis Bastions lend their shields to the shield bubble while it stands" when its commander holds the Ward at 20 and the row has a Bastion. This is new information: a successful deep probe now learns this one fact about every garrison parked at the colony, including an ally's. Nothing else about the lender's research or tree is shown. Reports filed before this release do not carry it.

Also

• A shallower spy report no longer makes the synthetic empires forget the garrisons a deeper report saw within the last day.
• Not yet done: the synthetic empires still do not count a garrison's own warships as fighters when they forecast a raid (#4734).

1.279.0 — 2026-09-24

New

Agents speak about their Apex only from real facts on their reality sheet (#4566)

Before: an agent could hold Dark Matter lines, battle reports list the keystones that fired under each side, and a Cipher spy can read a rival's whole Chart. None of this reached the agents' chat. Asked "did your Last Stand fire?" or "did you use your Archive this week?", an agent had no fact to answer from.

Now the agents' reality sheet carries:

• their own Chart: each Doctrine, Ascendancy and Rampart they hold, with its level and the keystones the server marks as held (a keystone still waiting on another line is not claimed), the Dark Matter relics they bought, and their Sigil if they own one, with what it really does. A Veil Sigil says it changes nothing, because travel time already sits on its floor.
• a true "never used" line for every action keystone they hold but never take: the Loom's Archive and Buffer, the Phalanx's Reserve Fleet, the Cipher's Sabotage, the Fold's instant Recall, and Free Jump and Second Jump for the personalities that never warp. Asked directly, an agent now has a real answer.
• the keystones that fired in their battles, as the battle report wrote them, and whose they were: yours, an ally's on your side, or theirs. A keystone read off the colony owner's Chart (Sanctuary, the Gunnery, the Battery, the Mantle, Anti-Capital and others) counts as the agent's own only when the colony is theirs. A garrison never claims its host's. Bastion Relay and the Aegis Bastion and Breachhead Monitor hull rules are never claimed by anyone.

Agents still cannot buy a Sigil, so no agent shows a Sigil line yet. The Sabotage the agents suffer shipped with #4542.

1.278.0 — 2026-09-24

New

• The Fold's Recall (L15): call a fleet home at once, at the warp price (#4545). Before, the Fold served three of its four rungs, and a recalled fleet always flew home over the time it had already flown. Now the Fold serves all four. At level 15, the recall sheet of a fleet on its way out offers ⚡ Home now · N DM: the fleet turns around and is home a second later. It costs the warp price of the distance it has flown so far, with your Fold discount, and any cargo lands with it. The price on the button updates every second while the sheet is open. If you hold less than it costs, the button is greyed and says what you have.
  • Every limit of the plain recall still holds. It does not work in the last stretch of a flight, on raid fleets, evacuations or supply lanes, on a fleet already flying home, or beyond three recalled attacks a day.
  • It is paused while any attack on any of your colonies is past its own point of no return (its last minutes). Without that pause, you could hide a fleet, let an attacker commit against an empty colony, and snap the fleet home under the strike. So any strike in its last minutes, even a small one, pauses the instant recall for every fleet you have. The plain recall is never paused.
  • Until an attack reaches that point, a Fold 15 holder can still snap a fleet home, so a spy report on one is only as fresh as its timestamp.
  • The Chart, the phone tiles and the Doctrine sheet now mark the rung at 15, and "To next keystone" from level 11 stops there.
  • Synthetic empires can hold the rung but never use it: they recall fleets only the plain way.

1.277.3 — 2026-09-24

Changed

• The Season podium wears real medals (#4760). The three places on The Race's "Top 3 players" podium were flat coloured circles with a number in them. Each place now wears its own painted medal: winged gold with a star crown for first, winged silver for second, finned bronze for third, with the rank written across its face.

1.277.2 — 2026-09-24

Fixed

• The six contested worlds are joined on the Season map (#4756). The Race's sector map showed the six Contested Space worlds as separate icons. A dashed route now links them in system order, from Ashfall Reach through The Cindergate, Hollow Meridian, Sable Anchorage and The Long Silence to Ember Crown, and each line meets its structure at the centre, on phones and on desktop.

1.277.1 — 2026-09-24

Fixed

• The six contested worlds look like themselves on the Season map (#4753). The Race's sector map drew the six Contested Space worlds with generic planet images, but they are structures: a ring station, a tower, citadels and a gate. Each world is now drawn as its own structure, rendered from the same model the battle replay uses. Each one stands on a soft glow in its holder's colour, with the holder's crest at its top right, and the world's name and hold time sit clear below it instead of overlapping the art. On mid-size screens the "holds all six worlds" banner moved to the bottom of the map, so it no longer covers the top worlds.

1.277.0 — 2026-09-24

New

• The Season tab opens on The Race (#4749). The Season tab used to read like a rulebook: formulas, grey rows, and no sign of the biggest prize of the season. It now opens on Sector 13 itself:
  • The map: the six contested worlds drawn as planets where they really sit in the sector, each carrying its holder's crest and how long they have held it. Tap a world for its details.
  • What's at stake: both podiums, with every paid place spelled out as the crates it holds ("10M Ferrolite + 10M Xenogas + 10M Data") and who holds it right now. The guild prize table, up to 100M of each resource for every qualifying member, now shows the cell your own rank and your guild's place would be paid.
  • Your prize: a strip telling you what you would receive if the season ended now, and the one threshold that would raise it.
  • The rest of the season: the player and guild boards, how points are won (only in Sector 13), and what the Eschaton takes.
• On phones the page is laid out for the app: the countdown stays pinned at the top as you scroll, players and guilds each have their own prize view, and each world opens in a bottom sheet.
• The lane is now called The Race. The Proving Ground, Contracts and Calendar lanes are unchanged.

1.276.0 — 2026-09-24

New

The Cipher's Sabotage (level 15) (#4542). The Cipher's level-15 chest, sealed until now, is Sabotage. The Cipher now has all four keystones.

• A Cipher 15 holder can tick Sabotage on a spy order. The checkbox shows only to a holder, and is greyed out below Espionage 17.
• If the probe files a top-tier report on a player past novice protection who has research running, the order that finishes soonest takes one hour longer. The queue behind it waits the same hour.
• Once every 24h for the holder. A player can be sabotaged once every 24h by anyone, so several holders cannot stack hours on one empire.
• When it cannot fire, the holder keeps the charge. The spy report says why it held, without saying which order moved.
• The target is always told who did it, even through Blackout: a red "Research sabotaged" row under Combat → Defence, beside the probe alarm, also listed in the "while you were away" recap. The launch sheet states this price before the tick: "Sabotage names you, even through Blackout (on an up-to-date app)". Installed app builds older than this release don't draw that row yet.
• The tick clears after every launch, so a re-scan of the same target is a plain probe.
• The Apex Chart, phone grid and sheet show the Cipher at four rungs. From level 11 the next keystone is 15, not 20.

Synthetic empires never sabotage. When one is sabotaged, it remembers who did it. The grudge cools its tone toward that player, but three days of sabotage cannot bait it into a strike outside its band.

Changed

• A sabotage's extra hour can't be undone by accident. Spending a Chrono Shard, a guild push, opening the Research tab, buying an Apex level or claiming the Loom's Archive at the same moment no longer writes the old finish time back.

1.275.0 — 2026-09-24

New

The Mantle's keystones: Sealed Sky, Second Skin, Unyielding and Mantle of the Guild (#4550)

Before, the Mantle (the first Rampart) had no keystones: its sheet showed four sealed chests. Now all four rungs are live, and each one reads the colony owner's Mantle:

• Sealed Sky (level 5). The salvo that breaks your Bubble stops there. Nothing behind the Bubble takes any of that salvo; from the next shot on, fire lands as usual.
  • It is not the rule the issue first described. "No shield leak while the Bubble is up" would have changed no battle, because the Bubble never leaked.
• Second Skin (level 10). Once per battle, a broken Bubble comes back at 25% of its own size, three rounds after it broke. The attacker has to hold the field for those three rounds and then break it again. An attacker's Shock does not stop it.
• Unyielding (level 15). An attacker's Cascade no longer slows your Bubble's regeneration. The Cascade card now says so.
• Mantle of the Guild (level 20). Each allied player whose armed garrison brings at least a tenth of the defending fleet's hull adds 25% to the Bubble, up to +100% with four allies. These add nothing:
  • your own garrisons;
  • a second fleet from the same ally;
  • unarmed ships, such as a probe parked on a defend mission.

Where you see it:

• The battle report names each rung that fired, in the defender's column. The column heading reads "Keystones & hulls" instead of "Doctrines & hulls", on every battle report, because the Mantle is not a Doctrine.
• The bubble line says when a Bubble broke, came back and broke again (or held).
• Against a Mantle 15 colony, an attacker's Cascade is no longer listed, because it changed nothing.
• The replay draws the Bubble breaking, coming back and breaking again. Its closing card says "came back and held" where it used to say "intact".
• The Planetary Shield card names each rung you hold.
• The planet sheet's recon line warns that a Mantle's Bubble can grow up to 180% with allied garrisons and come back once.
• The attack forecast prices all four rungs.

A broken Second Skin Bubble still scores (#4550)

• Breaking a Second Skin Bubble now scores even when it comes back and holds. Before, the report ended "held" and the attacker was credited nothing for a break they made.
• A Bubble broken a second time adds the returned 25%.
• A Bubble grown by Mantle of the Guild is credited at its real, larger size.

Changed

• Agents now read a Mantled Bubble with its rungs. A Second Skin Bubble prices about 6% more fleet: the extra fleet the engine was measured to need, at most. The count of allied garrisons the agent's probe saw prices Mantle of the Guild. As a result, Mantle-holding agents strike a little more carefully against Mantled walls.
• Their plans say what the probe saw ("the probe saw 2 armed garrisons stationed there"). They never call those garrisons allies, because the report cannot tell an ally from the owner's own fleet.

Left out

• The agents' appraisal of the ships under a Second Skin Bubble does not yet count the three extra rounds those ships fight. That is filed as #4673, high priority.

1.274.0 — 2026-09-24

New

• Fighting in the Contested Space scores for your guild (#4738). Until now a battle in sector 13 paid season points to the players and nothing to their guild, so a guild whose members fought all season could sit near the bottom of the guild standings. Now 10% of the season points a member wins in battle there also goes to the guild they are in when the battle ends. That covers attacking, defending, raiding, and standing garrison on an ally's colony. Colonizing doesn't add to it, because colonies already score for the guild on their own. Battles fought earlier this season, up to 24 September, were credited once to each fighter's guild on that day, so the standings jump when this goes live. Switching guilds doesn't carry past points with you, and a battle between friends pays the guilds nothing: same guild, pact or alliance, checked when the battle ends and against everyone fighting on the other side, garrisons included. Personal season points are unchanged.

1.273.1 — 2026-09-24

Also

The web client builds on macOS again (#4705)

Nothing changes for players. Two web source files in the Apex screen had names that differed only in letter case, and on a case-insensitive disk (every Mac) the build could not tell them apart. The typecheck and build failed there, while the Linux servers that build the live game were unaffected. One of the files has been renamed, and the game behaves exactly as before, including the keystone "already seen" memory saved in your browser.

1.273.0 — 2026-09-24

New

The Battery's keystones: Flak at 5, Fast Rebuild at 10, Crossfire at 15, Overwatch at 20 (#4552)

Before, the Battery (the Dark Matter Rampart of the Defense Batteries) raised the Emplacements' attack and nothing else: its sheet showed four sealed chests. Now all four rungs are live. Each one reads the colony owner's Battery level, never an ally's or the attacker's.

• Flak (5). Your Emplacements hit Bombers twice as hard. A Bomber still hits your Emplacements at ×2.5.
• Fast Rebuild (10). A destroyed Emplacement comes back on its own every 10 minutes instead of every 30. The war points a wall can pay out still refill at the same pace: a second attacker an hour later is paid the same as without the rung.
• Crossfire (15). Your Emplacements hit Splinter Lances, Shroudrunners and Praxis Interdictors twice as hard.
• Overwatch (20). Your Emplacements fire one salvo before round 1, at their normal attack and with no cap of their own. If the colony has an Orbital Cannon, the salvo comes right after its shot, and the Cannon's shot does not change. A colony without a Cannon still gets the salvo.

Where you see them:

• Battle report. The defender's "Keystones & hulls" line names each rung that bit. A new line under the Orbital Cannon's shows the Overwatch salvo's damage and the ships it destroyed.
• Replay. An "Overwatch" beat plays before round 1. A fleet it wipes out reads "Wiped out before round 1".
• Batteries card. It names each rung you hold and counts the regrowth down on your own clock. Its combat-multiplier list shows the added rows.
• Regrowth line on the report. It gives the owner's own interval when the report carries their tree, for both the Emplacements and the Fortress.
• Attack forecast. It fights the rungs on the spied tree before you launch.

Balance, measured. On a full wall (Shield 15, Batteries 20, Fortress 10, Cannon 10), each rung raises the cost of the smallest fleet that breaks it:

• Crossfire: ×1.27–1.45.
• Overwatch: ×1.00–1.24, including with Escalation.

Synthetic empires now price Flak, Crossfire and Overwatch when they size an attack on a Battery colony, so they bring more ships against a Battery 20 wall.

Also

• Old battle replays keep their Overwatch kills if ship ids are ever renamed again.
• The regrowth maintenance script now uses the owner's empire, so it no longer regrows a Fast Rebuild wall at the plain 30 minutes.

1.272.0 — 2026-09-24

New

• Colonies in the Contested Space score for your guild (#4731). Until now a guild scored only by holding or capturing one of the six control worlds, so guilds living in sector 13 without taking a world sat at 0 on the season board. Now every colony a guild keeps in the sector counts a flat 25 toward the guild's total, for as long as it stands. Lose the colony and its 25 goes with it. Colony points rank like any other and count toward the Eschaton podium, so a guild that holds no world can still place on its colonies. At the close, colonies count for the guild their owner belonged to when the season ended, so switching guilds after the end moves nothing. The podium entry in your activity feed now reads "won the Contested Space".

1.271.0 — 2026-09-24

New

The Ward's Sanctuary: a beaten colony's wall stands again an hour after the defeat (#4532)

The Ward's level-15 chest was sealed. It now opens as Sanctuary, which completes the Ward's four rungs.

• What it does. When a colony whose owner holds the Ward at 15 loses a defence, every Emplacement and Fortress platform that was down at that defeat is back one hour after the battle. A draw that leaves no Emplacement or platform standing counts as a loss.
• When it fires. At most once a day per colony, and only after a battle that destroyed some Emplacements or platforms.
• What keeps running. Until the hour is up, the usual rebuild carries on. Anything a later battle destroys stays on that usual rebuild.
• Where you see it.
  • The battle report names it on the defender's side.
  • The Batteries and Fortress cards show the countdown to the refill, or when it can fire again.
  • The Overview's "back in …" countdown takes the refill into account.
  • Under the paid Rebuild button, a note says how many of the units down come back free. Synthetic empires wait for the refill too, instead of paying for it.
• The redesign. The original rule refilled the shield bubble after a defeat. The bubble is rebuilt full for every battle anyway, so that rule changed nothing. The user chose this reading of the same intent.

War points. The Emplacements and platforms Sanctuary brings back are worth no war points to an attacker until the usual rebuild would have returned them. While that counter runs, some Emplacements the usual rebuild returned may also be paid nothing. The restored units still fight, though, so how a battle goes, and the bonuses that follow from it, can differ. Measured over 1,200 random attack sequences on small walls:

• attackers at a Sanctuary colony earned more in total in 16 of them (1.33%);
• 7 of those because the rebuilt wall changed a battle's outcome, and 9 because unpaid losses left the colony's regrowing allowance for a later attacker;
• the largest excess in total was 17.1%, and a single later attacker can gain more.

A follow-up, #4722, will remove the allowance effect. The defender earns points for what its restored wall destroys.

1.270.0 — 2026-09-24

New

The Loom's Archive: once a week, one research order completes now (#4539)

Before, the Loom's last rung was sealed: the Apex Chart drew a bare tick at level 20, the sheet showed a sealed chest, and nothing past Buffer (level 15) could be bought as a run.

Now the Loom serves the Archive at level 20. Once in any 7 days, you can complete one running research order on the spot. It costs no Chrono Shard and no Dark Matter.

• On the Research bench, each running order with at least a minute left shows Complete now · Archive.
  • The first tap arms it and shows how much time it saves. A second tap within a few seconds completes the order.
  • The arm clears on its own after a few seconds, on a tap anywhere else, and when the queue changes.
• After a claim, the bench says what was completed and how much time was saved, then counts down to when the Archive is ready again.
• The 7 days count from the moment you use it. Orders still waiting in the queue cannot take it. A refused claim never uses up the week.
• On the Apex, the Loom now shows four rungs. From level 15 the next keystone is the Archive: levels 16 → 20 cost 4 × 140 = 560 Dark Matter as one run.

AI empires that reach Loom 20 do not use the Archive yet.

1.269.1 — 2026-09-24

Fixed

• Every guild living in the Contested Space is on the season board (#4728). The Guild standings used to list only guilds that had captured or held a control world, so a guild with a dozen colonies in sector 13 couldn't find itself in the season. Now every guild with at least one colony in the sector is listed. Guilds that haven't scored show 0 below the scorers, larger presence first. A guild at zero carries no place number, because only guilds that score can take a podium place, and the panel says so under the board. The board now shows ten guilds instead of five. If your guild has no colony in the sector, your own row reads "no colony in the sector".

1.269.0 — 2026-09-24

New

The Gunnery's last two rungs: Second Volley and Broadside (#4559)

Before, the Gunnery's rungs at 10 and 20 were sealed. The Chart and the phone grid ticked only 5 and 15, and the keystone run stopped at 15.

Now all four rungs are live:

• Second Volley (Gunnery 10). If a battle reaches round 3, your Orbital Cannons fire again just before it, at half their raw damage.
  • The cap is the same as the opening shot's (25% of the attacking fleet's hull, lower while a Breachhead still stands or against Breaker Rounds), but it is taken on what is left of the fleet.
  • A Breachhead that fell in rounds 1–2 no longer halves it.
  • It does not suppress shields a second time.
  • The battle report shows it on its own line, and the replay plays it between rounds 2 and 3.
• Broadside (Gunnery 20). Against a fleet in two rows or more, your Cannons now land a shot that is under the cap rear first, whatever the formation.
  • The back row takes 55%, or its share of the fleet's hull if that is more. The rows in front take the rest. Capitals hidden behind a screen now take the shot.
  • It moves damage and adds none. The salvo is still capped at 25% of the fleet's hull, and no ship group takes more than a quarter of its own hull, so it never one-shots a capital before round 1.
  • A salvo that already hits the cap, or a fleet in a single row, lands exactly as before.
• The battle report names each rung on the defender's "Doctrines & hulls" line only when it changed a number, and says which of the two salvos it moved.
• The Orbital Cannon card and the Gunnery rows say what each rung does.
• AI empires now price both rungs when they size up a Gunnery wall. Against Broadside they assume the worst split their formation could meet, so they never underestimate the shot. They also mention both rungs in their planning when their probes have seen them.
• AI appraisal fix: the AI empires now protect a Destroyer or Annihilator with the Bulwark only when it is the fleet's one legendary hull, as the battle does. Before, two of them read as safer than they are.

Also:

• On an app version from before this release, the replay does not know about the second volley, so the ships it destroys disappear at the end of the replay instead of between rounds 2 and 3. The battle report is correct on every version.

1.268.0 — 2026-09-24

New

The Phalanx's Reserve Fleet: at level 20, ships kept in the hangar join the defence (#4530)

Before: the Phalanx's chests at 15 and 20 were sealed, and ships the defence plan kept in the hangar sat every battle out.

Now: at Phalanx 20, a colony's defence plan can name a second wave, per ship type, drawn from the ships the plan keeps in the hangar.

• Only hulls that fire or tend a shield can be named. Destroyers and Annihilators never can.
• When it joins: in round 3, or sooner, the round after the rest of the defence has fallen or stopped firing.
• Never while your shield bubble stands. Behind it the wave waits until the bubble breaks; if the bubble holds, the wave stays home.
• If the attack breaks off or dies first, the wave stays home untouched and appears in no one's report but yours.
• The line is unchanged. The ships you deploy fight from round 1 exactly as before, so the rung only adds ships to a defence. Against the same plan without it, over 34,560 benched battles and two 4,000-battle fuzzes (with and without a bubble), it never cost a colony.

Where you see it:

• The defence planner has a Second wave editor.
• The battle report has a Second wave line and lists Reserve Fleet among the Doctrines when the wave fought.
• Your replay brings the wave in on the round it arrived. A spectator's day-old copy draws it from the start.

Hold the Line (level 15) stays sealed.

• Defenders never break off, so the rule it described had nothing to act on.
• The only other reading, that the attack must fight on to a decision, can turn held colonies into fallen ones.
• It will get a new design that can never cost a colony.

Synthetic empires never set a second wave. When they size up a holder's colony they count its whole hangar as fighting from round 1, which errs safe except for a narrow case under the Aegis's Last Stand.

1.267.0 — 2026-09-24

New

The Dynamo's Overclock: a third Accelerator mode (#4535)

Before. The Dynamo's level 10 was a sealed chest. Each Particle Accelerator card had two chips, On and Off.

Now. An empire holding the Dynamo at 10 gets a third chip on every Accelerator card: Overclock. You choose it per colony.

• The Accelerator makes 25% more Dark Matter for 50% more of its own power draw.
• On a grid with spare power, you simply earn a quarter more Dark Matter.
• Where the grid cannot carry the extra draw, the whole colony's power yield drops. Its ferrolite, xenogas and data, its defences and its supply lanes all run slower. If the yield falls far enough (by more than a fifth), the Dark Matter itself comes out below On.
• The card shows both figures before you tap. When the colony would be throttled, it says so and names the yield it would drop to.
• It is a choice, never a trap. On and Off work as before. Switching the Accelerator Off clears the mode, so switching it back on gives plain On.
• An empire without the Dynamo at 10 sees no third chip. If the rung is ever lost, a colony left on Overclock runs as plain On.

Where you see it:

• Accelerator card: the three chips and the Dark Matter each mode would make on that grid. The Dark Matter and Energy rows follow the mode, and the Energy row says "Overclocked: +50% draw".
• Energy remedy: when an overclocked colony runs short of power, the first remedy it names is setting Overclock back to On, before any generator upgrade.
• HUD and conduits: the Dark Matter per day on the HUD and the colour of the base's power conduits include the mode.
• Apex Chart, phone tiles and Doctrine sheet: they now tick at 5, 10 and 20. "To next keystone" stops at Overclock from level 5 to 9. From level 11 it runs to 20 across the Grid (15, #4536), which stays sealed.
• Attack preview: from spy tier 2 (the tier that reads the tree), a spy report carries whether the target's Accelerator runs Overclocked. The launch preview then prices the weaker wall the battle will field. Below tier 2 the preview assumes plain On, which never shows a weaker wall than the real one.

Synthetic empires that hold the rung overclock only colonies with spare power, where the grid stays at full yield. They change the mode only during their waking hours, never as a reaction to an incoming attack. Most of their grids run close to their draw, so most of their Accelerators stay on plain On.

Installed apps keep working unchanged. They show an overclocked colony as On and cannot set the mode.

1.266.0 — 2026-09-24

New

The Gunnery's keystones: Ranging Shot and Counter-Battery (#4558)

Before. The Gunnery, the Orbital Cannon's Rampart, had no keystones. Its sheet showed four sealed chests. Against any Cannon:

• an attacking Breachhead halved the cap, the shield penetration and the suppression;
• the Breach's Breaker Rounds lowered the cap on its holder's fleet to 20%.

Now. Two of its four rungs are live. Both are read off the colony owner's Dark Matter tree.

• Ranging Shot (Gunnery 5). Your Orbital Cannons' opening shot bypasses ship shields at 1.5% per Cannon level instead of 1%, so more of it reaches hull.
  • An energy shortfall still cuts it.
  • The penetration still stops at 30%. A Cannon 20 now reaches that ceiling, where before only a Cannon 30 did, so the rung adds nothing to a Cannon 30.
• Counter-Battery (Gunnery 15). No attacker rule lowers your Cannons. An attacking Breachhead halves neither the cap, nor the shield penetration, nor the suppression, and Breaker Rounds does not lower the cap. The shot still never lands more than 25% of the attacking fleet's hull.

Where you see it:

• Battle report. The defender's "Doctrines & hulls" line names each rung when it changed the opening salvo, with the numbers. Against a holder, the attacker's Breachhead Monitor and Breaker Rounds entries no longer appear, because those rules did not apply.
• Orbital Cannon card. The penetration row prints what the battle lands, clamped at 30%. The card names Ranging Shot where it raises the penetration and Counter-Battery where it keeps the Breachhead and Breaker Rounds off.
• Build Advisor. The Cannon suggestion reads the same numbers and the same Counter-Battery clause.
• Descriptions. The Gunnery, Breach, Breachhead and Orbital Cannon descriptions say what each rung does.
• Apex Chart and phone tiles. They tick at 5 and 15. Second Volley (10) and Broadside (20) are #4559 and stay sealed. The "to next keystone" purchase runs 5 → 15.

Synthetic empires price both rungs when they size an attack on a Cannon: on the spied tree, else their own. A fleet carrying a Breachhead no longer counts on a halving that a Counter-Battery holder does not suffer, so agents send larger fleets against such walls.

1.265.0 — 2026-09-24

New

The Ward's keystones: Lifeline, Long Watch and Bastion Relay (#4532)

Until now the Ward (Dark Matter's Aegis Ascendancy) showed four sealed chests on its sheet. Three of them now act. Each rung reads the tree of the empire that sent the fleet.

• Lifeline (level 5). A garrison bound for one of your own colonies flies in 85% of the time.
  • The 2-hour ceiling on every flight still applies.
  • It does not stack with the Veil. The better cut is the one flown, so from Veil 10 Lifeline adds nothing.
  • The mission composer names Lifeline beside the flight time when it shortens the route.
• Long Watch (level 10). Your garrisons can stand watch for up to 48 hours instead of 24. They also take none of the host's garrison slots.
  • Long Watch garrisons share 8 slots of their own at each colony, beside the host's.
  • Once those 8 are full, a Long Watch garrison takes a host slot like anyone's.
  • The watch slider reaches 48h for a holder and names the rung.
• Bastion Relay (level 20). Your Aegis Bastions garrisoned at a colony that has a shield bubble add their shields to that bubble.
  • From the round the bubble breaks, they cover your escorts again.
  • A colony with no Shield gets no bubble from them, and the Bastions in the owner's home fleet are unchanged.
  • The battle report lists the shield they lent.
  • Breaking a bubble still earns its attacker the same war points as without the rung. The public replay does not reveal the lent share.

Sanctuary (level 15) stays sealed. Its rule assumed a shield bubble that stays down after a lost defence. In fact the bubble is rebuilt in full for every battle. Its redesign is still open, so #4532 stays open for that rung. From level 11, the "to next keystone" purchase runs straight to 20.

Changed from the original design. Bastion Relay first lent its shields to the bubble for the whole battle. Measured that way, it lost colonies whenever the bubble broke early. Handing the shields back when the bubble breaks lost none on the same test grid.

Not yet done: synthetic empires do not account for Bastion Relay. When one of them weighs an attack, it reads only the Shield building's bubble, so it can attack a colony whose relayed bubble holds. This is tracked in #4713.

1.264.0 — 2026-09-24

New

The Keep's keystones: Bedrock and Anti-Capital (#4554)

The Keep, the Dark Matter line that sharpens your Orbital Fortress, used to have four sealed chests. Two of its rungs now open.

• Bedrock (Keep 5). A destroyed Orbital Fortress platform comes back on its own every hour instead of every 2 hours, on every colony you hold. A destroyed platform still pays the attacker no war points.
• Anti-Capital (Keep 10). Your Fortress hits harder against capital ships:
  • Titan Dreadnoughts and Battleships at ×2.5 (was ×2);
  • Cruisers at ×2 (was ×1.5);
  • Voidlance Cutters, the Lance Doctrine's hull, at ×2 (was ×1).

Where you see it:

• Battle report. The defender's "Doctrines & hulls" line names Anti-Capital when it changed the fight.
• Fortress card and sheet. They name both rungs. The regrow countdown and the rebuild note use your clock, and "Combat multipliers" shows your raised table.
• Battle report reassurance line. It states both clocks.
• Shipyard reference. It warns that a Keep-10 owner's Fortress hits those four hulls harder.

Redoubt (Keep 15) and the Citadel (Keep 20) have not shipped, and stay sealed.

Synthetic empires price Anti-Capital when they size an attack on a Keep holder's Fortress. They still gather Titans against a Fortress that counters them (#4709, not fixed here).

If your app was installed before this release, a Keep-5 Fortress still shows the 2-hour countdown until you update. The platforms themselves come back every hour.

1.263.0 — 2026-09-24

New

The Phalanx's keystones: Interlock and Salvage Corps (#4529)

Before: the Phalanx sold four sealed chests. Now: two of them open.

• Interlock (level 5, needs the Aegis at 15). When your fleets defend a colony alongside other fleets, your Last Stand also covers theirs, the host's and the other garrisons' alike: no ship group loses more than a quarter of its hull in a round. Your fleets there must bring at least a tenth of the defenders' ship hull. The Emplacements and the Fortress are not covered.
  • The battle report names Interlock on the Defenders' line. The cards say who shielded whom: "Shielded by the Interlock of …" on the covered fleets, and "lent its Last Stand cap to …" on yours.
  • With the Phalanx at 5 but the Aegis below 15, the Apex no longer ticks the chest. It reads "reached; needs Dark Matter Aegis at level 15".
• Salvage Corps (level 10). When you win the defence of your own colony, a tenth of the ships you lost there come back to its hangar, rounded down per type, and never a Destroyer or an Annihilator.
  • Your self-garrisons count. Ships lost guarding someone else's colony do not.
  • The report shows a "Returned to the hangar" line. The losses above it still count every ship the battle destroyed.
• Hold the Line (15) and Reserve Fleet (20) stay sealed (#4530).

The issue's first version of Interlock, "Last Stand also covers your garrisons at a guildmate's colony", was already how the game worked, so the rung was redesigned into the shared shield wall.

Known gap: synthetic empires judge an attack on a colony without its garrisons, so they do not price Interlock. That is the same existing gap that already leaves garrisons unpriced.

A one-ship group under Last Stand can soak focused salvos. That is a separate, pre-existing issue, tracked in #4703.

1.262.0 — 2026-09-24

New

The Vanguard's keystones: First Light, Shock, Line Breaker and Rout (#4527)

Until now the Vanguard sold +2% per level of damage in the first two rounds of a battle, and its four keystone chests were sealed. All four now change the fight.

• First Light (level 5). In round 1, your ships fire first. The enemy answers with whatever your salvo left standing. Allies without the rung fire alongside the enemy, after you. If armed ships on both sides hold it, round 1 stays simultaneous. The Orbital Cannon still fires before round 1. The replay plays a defending holder's round-1 volley first.
• Shock (level 10). When you attack a colony, its shield bubble does not regenerate at the end of round 1. From round 2 it regenerates as usual. The replay's bubble bar shows the missed refill.
• Line Breaker (level 15). In a formation battle, your fire ignores the enemy's Deep Formation. Their back row takes its usual share of your salvo instead of 2%: 5% behind a full line, about a quarter with no centre row. A living Praxis Interdictor already beats Deep Formation, so next to one the rung adds nothing.
• Rout (level 20). This one works only when you defend a colony, as its owner or as a garrison there. An attack breaks off after losing half its value instead of two thirds. A raid that would have worn your defence down and taken the colony can be repelled instead, and takes no loot. The price: in a fight you were winning anyway, fewer of the attacking ships die before they leave. Your own attacks are unchanged.

Every rung that changed a number appears in the battle report under "Doctrines & hulls". Spectators' copies of the report leave these lines out, so the public replay plays in the ordinary order.

Two rungs differ from how the design first described them. Line Breaker was written to need a living Praxis Interdictor, but the Interdictor already overrides Deep Formation, so that version would have changed nothing. Rout was written as "break off at one half" and also "fight longer and lose more", and those two contradict each other. The rung breaks off at one half. Measured both ways, that turned some fallen colonies into held ones and no held colony into a fallen one.

Synthetic empires buy the Vanguard, but their attack estimates do not account for any of its rungs. Against a Rout holder they misjudge the defence: a raid they rate as winnable can break off first, with no loot.

Battles where no side holds a Vanguard rung fight exactly as before.

1.261.1 — 2026-09-24

Changed

Contested Space pays ten times the resources (#4704)

When a season closes, every scoring member of the top three Contested Space guilds receives a crate of each resource (ferrolite, xenogas, dati). The size depends on the guild's podium place and on the member's own rank on the season scoreboard. Those crates were too small for the economy the server has grown into, so every size is now ten times larger:

| Your own rank | 1st guild | 2nd guild | 3rd guild | |---|---|---|---| | Top 3 | 100M (was 10M) | 50M (was 5M) | 10M (was 1M) | | 4–10 | 50M (was 5M) | 10M (was 1M) | 5M (was 500K) | | 11+ | 10M (was 1M) | 5M (was 500K) | 1M (was 100K) |

The new sizes apply from the next season close. Crates already granted keep their size. Nothing else about the season rewards changes: the individual Eschaton podium still pays 10M / 5M / 1M crates, and the guild placement still pays its Scrip, Dark Matter and title as before.

1.261.0 — 2026-09-23

New

• The Bastion Codex, the Apex Chart's third Relic (#4563). It costs 200 Dark Matter, bought once. Until now the Chart held two Relics, Precursor Schematics and the Sigil. The Codex now sits beside them on the Relic strip, on the phone grid and in its own sheet.
  • It unlocks once you have repelled a faction sortie. A repel counts even if it paid no Dark Matter because your weekly faction cap was full. Until then the card reads "Requires a repelled sortie".
  • Six more hours of warning. Sorties the bastions launch against you leave 6 hours later: 8–12 hours after they are announced instead of 2–6. The diary line and the Incoming board both appear at the same moment as before, so you get 6 more hours to bring your fleet home. The diary line says the Codex is why, and gives the countdown in hours.
  • A step up the defence ladder. A sortie you beat back pays the next rung: 9, 15, 24 and 24 Dark Matter instead of 6, 9, 15 and 24. The 96 DM weekly faction cap is unchanged and still applies. The payout line names the Codex only when it actually raised what you were paid.
  • Nobody who does not hold the Codex sees any change to a sortie that launches. One small exception: if a holder's later impact would land in their night window, the bastion waits for its next 6-hourly decision, and its own post-raid shield lasts until then.
  • Nobody can buy it yet. As of this release no player meets the unlock: no sortie has ever been repelled for a payout, and the last sortie launched on 4 September. Only retaliation for a felled bastion launches them today.
  • AI empires do not buy the Codex.

1.260.1 — 2026-09-23

Changed

• AI-marked empires now buy only the Apex outer-ring lines that fit their play style (#4565).
  • Before: each empire's outer-ring choices were derived from its Doctrines, so every empire bought a little of every line its Doctrines unlocked. Scientists put Dark Matter into walls, aggressive empires into the Fold, diplomats into the Dynamo and the Phalanx.
  • Now: each play style buys its own set of lines, strongest preference first:
    • Aggressive: the Breach, the Vanguard, the Cipher.
    • Turtle: the Mantle, the Keep, the Battery, the Ward, the Gunnery, then the Phalanx.
    • Opportunist: the Fold, the Cipher, the Vanguard.
    • Scientist: the Loom, the Dynamo.
    • Diplomat: the Ward, the Loom.

  With 400 Dark Matter on the day every Ascendancy opens, the first three purchases are:

  • Aggressive: Breach, Vanguard, Breach.
  • Turtle: Mantle, Keep, Battery.
  • Opportunist: Fold, Cipher, Fold.
  • Scientist: Loom, Dynamo, Loom.
  • Diplomat: Ward, Loom, Ward.
• Only turtle empires buy Ramparts now, and they buy them much earlier: two Aegis levels in instead of six.
  • A Rampart is still bought only when the empire holds the building at level 10 on a colony it owns today. A turtle with no Orbital Fortress at 10 never goes for the Keep.
  • Wraiths no longer buy walls.
• An empire still reaches its apex hull. On a slow income, taking a Doctrine from 15 to 20 takes 158–220 ticks, depending on style. The issue's weights read literally would have taken up to 440; the #4564 review called 314 starvation.

Also

• The Watch has not shipped yet. The diplomat will start buying it when it does.
• Overclock has not shipped yet. When it does, synthetic empires will not switch it on as a reaction to an incoming attack.

1.260.0 — 2026-09-23

New

The Breach's keystones: Shatter, Breaker Rounds, Plunder (#4524)

Before: the Breach had no rungs. Its sheet showed four sealed chests, and the line only added +2%/lvl against a colony's Emplacements and Fortress.

Now: three rungs are served. Reduction (level 20, #4525) is still sealed.

• Shatter (level 5). When you win an attack on a colony, your fleet salvages 25% of the battle's wreckage value and carries it home. It fills only the room your loot leaves in the hold, and your storage cap applies when it lands. On a joint raid the 25% is split by each fleet's share of the surviving hold, and a partner without the rung salvages nothing. The debris field and your target lose nothing to it.
• Breaker Rounds (level 10). The Orbital Cannon's opening shot is capped at 20% of your fleet's hull instead of 25%. A Breachhead's 12.5% is lower and still applies. A shot under 20% of your hull lands in full.
• Plunder (level 15). Your attacks take 10 points more of what the Silo leaves exposed: 60% instead of 50% at base, and never more than 75%. Your hold still limits the haul.

The battle report has a new "Salvaged aboard" line, kept separate from the loot because it is taken from nobody. On a joint raid it shows the raid total, and each holder's card shows its own share. Each rung is named on the attacker's Doctrines line when it changed a number.

The Orbital Cannon sheet now tells a defender about Breaker Rounds. It shows the 20% cap and the fleet size below which that cap binds against an attacker holding the Breach at 10.

The Breach's card and node description no longer say it "does not blunt the Orbital Cannon". The +2%/lvl still does not, but its level-10 rung lowers the cap.

Changed from the design: Shatter was written as "×1.25 debris". That version only changed a number shown on the report. And the field over a living colony can only be harvested by that colony's owner, so a bigger field would have paid the defender. Salvage carried home in your own hold keeps the intent: a Breach win yields a quarter more, to you.

Left out:

• On app builds older than this release, the report has no "Salvaged aboard" line, so a Breach holder's returning cargo can look larger than the loot shown in its report. The resources are real.
• AI empires still value raids without Plunder or salvage, and assume the full 25% Cannon cap. Both err on the cautious side. An AI that buys the Breach gets all three rungs in battle.

1.259.0 — 2026-09-23

New

The Gunnery, the Orbital Cannon's Rampart (#4557)

Before. The Orbital Cannon fired 300 per level, and nothing an empire could buy made it hit harder. Escalation deliberately leaves the Cannon out.

Now. The Gunnery is the fourth Rampart on the Apex Chart.

• How to open it. You need the Particle Accelerator and an Orbital Cannon at level 10 on one colony you hold now. Two colonies at 9 do not count as a 10.
• Cost. 20 levels on the outer ladder: 12 / 30 / 70 / 140 DM, 1,260 DM for the whole line.
• Effect. Each level adds 2% to the opening shot of every Orbital Cannon you hold, up to +40% at level 20. Shield penetration and suppression do not change. A grid short of energy still cuts the heavier shot.
• The ceiling. The shot can never land more than a quarter of the attacking fleet's total hull, or an eighth when the attackers bring a Breachhead. So against a small fleet, the ceiling decides the damage, not the Gunnery.
  • A Cannon 20 fires 6,000 / 7,200 / 8,400 at Gunnery 0 / 10 / 20. Against forty cruisers all of it lands.
  • Against ten cruisers, Gunnery 10 and Gunnery 20 land the same 6,250.
  • The card, the Doctrine sheet and the Cannon's "Where the cap binds" line all show this.
• Where you see it.
  • The Orbital Cannon card shows the heavier shot and names the Gunnery.
  • The building chip reads "Gunnery N".
  • The Doctrine sheet shows "Cannon opening shot +20% → +22%".
  • A refused purchase names the missing Cannon.
  • On an attacker's battle report, the Gunnery chip is dashed "defense only".
• Synthetic empires.
  • They buy it only when a colony they hold has an Orbital Cannon at 10.
  • When they judge whether to attack, they price a target's Gunnery: the level their probe saw, or their own level when the probe did not see one.
  • Their plans name a Gunnery only when a probe actually saw it.
  • With four Ramparts now on the Chart, a wall-minded empire spreads its wall budget across them more evenly.

1.258.0 — 2026-09-23

New

The Keep: Dark Matter makes your Orbital Fortress hit harder (#4553)

Before, an Orbital Fortress fired what its level and your colony's grid allowed, and nothing you could buy made it hit harder. The Apex Chart showed the Keep as a planned Rampart you could not buy.

Now the Keep is the third Rampart you can buy, after the Mantle and the Battery:

• To unlock it you need the Particle Accelerator and an Orbital Fortress at level 10 on one colony you hold. Two colonies at level 9 do not count.
• The line has 20 levels on the outer ladder: 12, 30, 70 and 140 Dark Matter per level, 1,260 for all 20.
• Each level adds 2% to the attack of every Fortress platform on every colony you hold, so level 20 adds 40%.
• What it does not change: your Fortress's hull and shields. It makes the Fortress hit harder, not last longer.
• Power still matters: if a colony is short of energy, the throttle still cuts the stronger guns.
• With Escalation: if you hold Escalation, its ×10 applies on top of the Keep.
• Where you see it:
  • The Fortress card shows the stronger attack and names the Keep, including on a colony that has not built its first platform yet.
  • The sheet reads "Fortress attack +20% → +22%".
  • Launch previews against a colony that holds the Keep count it, just as the battle does.
  • On an attacker's battle report, its chip is marked as working only in defence.
• Season points: the Keep's extra fire counts toward the defence's season-point salvo, the same way the Battery's does. Escalation's ×10 still does not count.

Measured on one seed, attacking a colony whose only defence is an Orbital Fortress at level 12 (12 platforms):

| Keep level | 34 cruisers attack | 30 cruisers attack | |---|---|---| | 0 | the attack wins, 4 cruisers lost | the attack wins, 9 cruisers lost | | 10 | the attack wins, 8 cruisers lost | the colony holds (20 cruisers lost, 7 platforms destroyed) | | 20 | the attack wins, 14 cruisers lost | the colony holds (21 cruisers lost, 4 platforms destroyed) |

AI empires buy the Keep only when they really have an Orbital Fortress at level 10 on a colony they hold. When they plan an attack, they count the target's Keep if a spy report showed it. Otherwise they assume the target holds whatever level they hold themselves. Their plans name a Keep only when a spy report actually saw it.

With three Ramparts to choose from, an AI empire that favours walls now spreads its Dark Matter across them. Each wall line gets a third of that empire's interest in walls, down from half.

1.257.0 — 2026-09-23

New

• Defence buildings show the Rampart they govern (#4560). The Planetary Shield and Defense Batteries never said that reaching level 10 on them opens a Rampart on the Apex Chart. They now carry a chip on the Building Sheet and on the Colony list row:
  • Below 10 it reads Reach 10 to open the Mantle. If you haven't researched the Particle Accelerator, it names that too, because the Ramparts need it.
  • From 10 it shows the Rampart and your level (Mantle 12), or what the first level costs if you hold none.
  • The gate counts any of your colonies, so a small Shield next to a big one elsewhere is not told to reach 10.
  • Tapping the chip opens the Apex with that Rampart selected: on the Chart on desktop, on its sheet on a phone. An empire with no Apex tab yet lands on the Particle Accelerator in the Tech Tree.
  • Buildings with no Rampart built on them yet, such as the Deep Observatory, show no chip. The Keep and the Gunnery will get theirs when they ship.

1.256.0 — 2026-09-23

New

The Cipher's keystones: Deep Read at 5, Blackout at 10, Omniscience at 20 (#4541). Until now the Cipher sold twenty levels of "+2% chance your probe goes unnoticed" and no rule at any rung. It now serves three of its four keystones. Sabotage (L15, #4542) has not shipped yet, and its chest stays sealed.

• Deep Read (L5). Every spy report you file adds an APEX READ block, at any espionage tier, a partial report included. The block shows:
  • the level of each Dark Matter line the target holds;
  • the keystones those levels hold;
  • the target's Sigil, whichever line it names;
  • the row each ship that defends by default stands in.

  Before, the levels appeared only as raw entries in the Tech Tree list, from espionage 11, and the rows only from espionage 17.

• Blackout (L10).
  • Ghost Manifest's ±30% fog never applies to your spy reports, at any tier. Before, it fogged every spy below the top tier.
  • On other players' maps your probes fly as "UNIDENTIFIED PROBE", with no name and no origin coordinates. The path and the arrival stay visible, and the path still points back toward the system the probe left. On the way home the probe shows no destination.
  • A caught probe reaches its target unidentified. The target still learns it was scanned.
  • Two tells give the rung away, never your name. First, from Observatory 2 your probe's map mark is missing its origin. Second, a Deep Sensor Grid on the probed colony names every probe it catches, so a catch it cannot name tells its owner the prober holds Blackout.
  • A probe caught in combat (a picket, an interception) is named in the battle report, as combat always does.
• Omniscience (L20). Suppose you read an empire with a top-tier spy report in the last 24h. Its attacks on you then show on the Incoming panel at full detail, whatever your Observatory (even none) and whatever their Cold Trail. A gold caption says why. The guild's threat board credits it to Omniscience. Hulls your Sensors cannot read stay hidden, and a Nightfall launch stays dark until its last stretch.

None of the three changes the chance a probe is caught. The Cipher's 5% floor stands. The Veil's Cold Trail and Ghost Manifest cards, the Deep Sensor Grid card and the Observatory card now name the Cipher rung that answers them. The design record's original texts for Deep Read and Blackout were no-ops, because every report at espionage 11+ already listed the Dark Matter levels and Ghost Manifest already spared the top tier. These stronger versions replace them.

1.255.0 — 2026-09-23

New

The Fold's keystones: Waystation, Second Jump and Nexus (#4544)

Before: the Fold sold twenty levels of cheaper warp, and none of its rungs did anything. Now it serves three of its four rungs, and the Chart, the phone tiles and the Apex sheet mark them at 5, 10 and 20. Recall (level 15) has not shipped yet, so its chest stays sealed. From level 11 the "To next keystone" run goes straight to 20.

• Waystation (level 5). A hop between two of your own Hypergates is as short as if the weaker gate were one level higher. It works on your own gate network only: the guild jump bridge prices every route exactly as before, for you and for your guildmates.
• Second Jump (level 10). Free Jump comes back every 12 hours instead of every 24. It needs Free Jump (Veil 15) to do anything. Bought mid-window, it applies at once: a free jump claimed 13 hours ago is free again straight away. The launch form's price line and countdown both show the 12 hours.
• Nexus (level 20). A sixteenth Hypergate core for a level-30 gate that already has all fifteen cores. It costs 840 Dark Matter plus about 109.5 billion resources, the next step of the same core ladder, and needs no building upgrade. Once both ends of a pair have one, the hop between your own gates is 6% shorter. It never changes the guild bridge. The core panel offers it to Fold 20 empires, and afterwards shows "16/16" and whether the core is active.

Synthetic empires that reach Fold 20 buy the Nexus core too, under the same rules.

1.254.0 — 2026-09-23

New

• The Sigil shows next to your name (#4562). Before, an empire that bought the Sigil could see it only on its own Apex Chart. Now the glyph of the Doctrine it named appears next to the empire's name in:
  • chat, including the whisper list
  • the leaderboard (podium, phone cards and the desktop table)
  • the target list
  • the guild roster
  • battle reports (the "who fought" line and each commander's card)

  The glyph is the Chart's own art, one per Doctrine: Lance, Aegis, Forge or Veil.

  The Sigil glyph shows everyone which Doctrine you named. It is a banner: it shows no levels, keystones or formation. Those still need a spy report at the right tier.

  The AI badge keeps its place right after the name. The Sigil comes after it: name, AI badge, Sigil, VIP crest, [TAG]. A synthetic empire that holds a Sigil shows both marks.

  Battle reports show the Sigil the empire held when the report was written. That includes a Forge or Veil Sigil, which never changes a battle's numbers. Reports written before this release show no glyph.

  Left out:

  • Other places that show the AI badge still show no Sigil. These include the mention picker, the reply and pinned bars, the open whisper header, the campaign, war-week, wanted and echelon boards, the market, aid and bounty panels, and the Activity feed.
  • On a standard-resolution (1x) desktop screen, the Veil glyph is too small to read at name size, and the Forge is borderline. A dedicated small mark per Doctrine is tracked in #4663.

1.253.1 — 2026-09-23

Changed

• The old Dark Matter catalogue is retired from the web game (#4521). The stack of Apex cards was replaced by the Apex tab: the Chart on a desktop, the tile grid and each line's sheet on a phone. It is now gone from the code as well. Nothing you could buy there is lost: every line is bought on the Apex tab, as it already was. Installed mobile apps that still show the old cards keep working until they update. The server keeps sending everything they read, and a new test holds that in place.
• Research → Apex is one card. In the game the Apex pill already opens the Apex tab. Anywhere the lens itself still appears, it shows one card with your stock, where the Apex lives, and Open the Apex →. If you hold Dark Matter without the Particle Accelerator, the card leads with Research the Particle Accelerator →.
• If the Apex ever can't be drawn from what the server sends, the Apex tab now says so plainly. It also says your levels and Dark Matter are untouched, and does not show cards it cannot place.

1.253.0 — 2026-09-23

New

• The Loom's keystones: Sharp Shards at 5, Second Opinion at 10, Buffer at 15 (#4538). Until now the Loom offered twenty levels of −1% research time and no keystones: its ring on the Apex chart had no ticks, and its sheet showed four sealed chests. It now has three keystones. Its fourth, the Archive at level 20 (#4539), is still sealed, and the sheet sells no run past Buffer.
  • Sharp Shards (L5): each Chrono Shard you spend cuts 25% more time, on the build, shipyard and research queues and on an Anchorage convoy: a 5m cuts 6m15s, a 30m cuts 37m30s, a 1h cuts 1h15m, a 4h cuts 5h. The shard panel shows the sharpened cut before you spend it, and how much of it would be lost if the order is shorter.
  • Second Opinion (L10): your orders can take 15 guild pushes instead of 10. That is about 26% off in all, where 10 gave about 18%. Each push is still 2% of the time left, and the daily limits don't change. The cap belongs to the order's owner: guildmates see your request as "of 15", and pushing someone else's order still counts against their 10.
  • Buffer (L15): two more research orders can wait in your queue, on top of your Laboratory and Third Lane. These are waiting slots, not another lane. The Loom's card used to say it doesn't change how many orders you may run. It now says that no level lets more orders run at once, and that Buffer adds room for orders that wait.

Changed

• A Sharp Shards holder's weekly shard limit now cuts up to 5h instead of 4h (#4538). The weekly limit still counts the shards you collect at face value, so you collect exactly what anyone else does. Because each shard cuts 25% more, a full week of shards cuts up to 5h of queue time instead of 4h. This is a deliberate 25% rise in the weekly ceiling for Loom holders.
• Synthetic empires use the Loom's keystones where they act on them. An empire holding Sharp Shards picks which shard to spend by what it will really cut. An empire holding Second Opinion keeps asking its guild for help after ten pushes. The two extra Buffer slots are not used by synthetic empires yet: they queue two research orders fewer than they could.

1.252.0 — 2026-09-23

New

The Battery: Dark Matter makes your turrets hit harder (#4551)

Before, Defense Batteries fired what their level and your colony's grid allowed, and nothing you could buy made them hit harder. The Apex Chart showed the Battery as a planned Rampart you could not buy.

Now the Battery is the second Rampart you can buy, after the Mantle:

• To unlock it you need the Particle Accelerator and Defense Batteries at level 10 on one colony you hold. Two colonies at level 9 do not count.
• The line has 20 levels on the outer ladder: 12, 30, 70 and 140 Dark Matter per level, 1,260 for all 20.
• Each level adds 2% to the attack of every Defense Batteries emplacement on every colony you hold, so level 20 adds 40%.
• What it does not change: your emplacements' hull and shields. It makes them hit harder, not last longer.
• Power still matters: if a colony is short of energy, the throttle still cuts the stronger guns.
• With Escalation: if you hold Escalation, its ×10 applies on top of the Battery.
• Where you see it: the Batteries card shows the stronger attack and names the Battery. The sheet reads "Emplacement attack +20% → +22%". Launch previews against a colony that holds the Battery count it, just as the battle does. On an attacker's battle report its chip is marked as working only in defence.

Measured on one seed: ten cruisers attack a colony with Batteries at 12 (24 emplacements).

| Battery level | Result | Cruisers the attacker loses | |---|---|---| | 0 | the attack wins | 2 | | 10 | the attack still wins | 4 | | 20 | the colony holds | 7 |

AI empires buy the Battery only when they really have the Batteries at level 10. When they plan an attack, they count the target's Battery if a spy report showed it, and otherwise assume the target holds whatever level they hold themselves. Their plans name a Battery only when a spy report actually saw it.

1.251.0 — 2026-09-23

New

The Sigil: a complete Doctrine now reads level 22 (#4561)

Before: a Doctrine at level 20 showed "Doctrine complete" and nothing more. The Apex had nothing further to buy with that line.

Now: the trophy of a complete Doctrine leads to the Sigil, a one-time Relic that costs 250 Dark Matter. It can be bought once any Doctrine is at 20. At purchase you choose one of your complete Doctrines to name. That line's scaling effect then reads level 22 instead of 20:

• Lance: shield penetration 26.1% → 28.3%
• Aegis: added shield power +60% → +66%
• Forge: added production +40% → +44%
• Veil: travel time stays at −30%. Its 0.70 floor is already reached at level 20, so a Veil Sigil adds nothing. The card says so before you buy, and the Veil's own sheet never suggests it.

Keystones and apex hulls do not change. Only the line's scaling number reads the two extra levels. Moving the Sigil to another complete Doctrine costs 100 Dark Matter. You pick the new line and then confirm; a single tap never spends anything.

The Sigil appears on the Chart's Relics strip and in the phone grid, with its own art. Battle reports, spy reports and the fleet advisor's forecast use the level-22 figure whenever the Sigil fought. Colonies bank their production at the old rate before a Sigil changes it, so switching can neither create nor erase earned output.

Installed app versions that predate this release do not show the Sigil. Update the app to buy it.

Also

• Schema: this release adds one database column, Player.sigil, to record which Doctrine a Sigil names. It is empty for everyone until someone buys a Sigil.
• Rolling back past this release removes that column (see deploy/DEPLOY.md, block AO). A player who already bought a Sigil keeps it and keeps what they paid, but the named line pays at level 20 until the release is restored. After restoring, the operator can put every player's choice back from the Dark Matter ledger, or players can name their line again for free.
• Synthetic empires do not buy the Sigil.
• At release time, 11 players had a Doctrine at 20 and could buy one.

1.250.1 — 2026-09-23

Fixed

• Synthetic empires now count the Mantle before they attack a Bubble (#4642). The Mantle makes a Planetary Shield's Bubble up to 40% bigger. The launch forecast players read already counted it, but the synthetic empires sized a colony's Bubble from its shield level alone. Against a Mantle holder they could decide they would break the wall in time and send a fleet a human reading the forecast would have kept home. They now apply the Mantle the way the forecast does: the level their probe saw when a spy report showed the defender's tree, and otherwise the level they hold themselves. This covers every check they make before a strike or a guild raid, from picking and ranking a target to sizing the fleet. When a probe has seen a Mantle, the planner's view of that spy report says so too, so what an empire plans matches what it launches. A Mantle it has only assumed is not written into that report line.

1.250.0 — 2026-09-23

New

• The Mantle, the first Rampart (#4549). A Planetary Shield's Bubble used to be sized by its level and nothing else. Now Dark Matter can make it bigger. The Mantle opens once you have the Particle Accelerator and a Planetary Shield at level 10 on any one of your colonies. It has 20 levels at 12 / 30 / 70 / 140 DM per band of five, 1,260 DM for the whole line.
  • Each level adds 2% to the Bubble of every Planetary Shield you hold, up to +40% at level 20. A bigger pool also refills more each round.
  • The gain is applied before the energy throttle, so a Shield short of power still loses its share.
  • Your ships' own shields, the Orbital Cannon, the Emplacements and the Fortress are unchanged.
  • On the Apex Chart it stands in the Ramparts column above your colonies. On the sheet it reads "Rampart", and when locked it says "Requires Planetary Shield 10 on any colony". Its stat line reads "Bubble HP +20% → +22%".

Changed

• Screens that show a Bubble now account for the Mantle.
  • Your colony overview's "Shield pool" line, the Planetary Shield card's effective-pool note, the launch forecast, and the guild rescue card, which now says "HP in battle".
  • A battle report lists the Mantle in the defender's loadout. On an attacker's card it is marked "defense only".
  • The planet sheet's free recon line now says its pool figure is "at this level", and that a Mantle can add up to 40%. It cannot show the owner's Mantle, because that would reveal their Dark Matter tree.

Also

• Synthetic empires can buy the Mantle, but only once a colony they hold has a Planetary Shield at level 10.
• Synthetic empires under-read a Mantled Bubble. When they decide whether an attack can break a target's Bubble, they do not yet account for the target's Mantle. Against a Mantle-20 wall they under-read it by about 29%, so an agent can send a fleet at a Bubble that holds. This is fixed in #4642, which is in progress.
• Launch forecast against an unspied target: below spy tier 2 it assumes the target holds the same Mantle you do. The forecast already marks such a target's tree as assumed.
• Mobile apps built before this release do not show the Mantle on their colony overview or launch forecast until they update. Battles always use the Mantle.
• The Mantle's keystones are not in this release (Sealed Sky, Second Skin, Unyielding, Mantle of the Guild, #4550). The sheet shows them as sealed.

1.249.0 — 2026-09-23

New

• The Ward, the Aegis's second Ascendancy (#4531). Before this, a commander with the Aegis at 10 had nothing to buy for the fleets they lend. A garrison at an ally's colony had the same shields as anywhere else, and a Freighter or a Colonial Ark in a battle got no extra protection.
  • Where it appears. The Ward is on the Apex chart and the Dark Matter panel once the Aegis reaches 10.
  • Price. 12 / 30 / 70 / 140 DM a level, 1,260 DM for the line.
  • What it does. Each level adds 2% (+40% at level 20) to the shields of:
    • your fleets garrisoned at a colony you defend, an ally's or one of your own;
    • your Freighters, Colony Ships, Colonial Arks and Continuum Haulers, in every battle, attack included.
  • What it does not reach. The warships of your fleet at home, your warships in attack or caught in flight, the Emplacements and the Fortress.
  • How it stacks. It multiplies the shields your research and the Aegis already give. A garrison with both the Phalanx and the Ward carries +40% hull and +40% shields.
  • It's a modest line. In the battles measured before release, a lone garrison at level 20 held out a few rounds longer and cost the attacker about 6% more, but it never changed who won. On freighters the difference is small (a Freighter's 20 shields become 28).
  • Report and sheet. The Apex sheet shows "Shields of garrisons and convoy hulls" with the value now and at the next level. A battle report now says what the Ward did for each fleet: a garrison fought with it on every ship, and any other fleet had it only on its convoy hulls.
  • Synthetic empires. They can buy it too.

1.248.1 — 2026-09-23

Fixed

• The test suite on main is green again (#4650). The Dynamo's first keystones (#4534) put two more rungs on the Apex panel. Two tests about the Vanguard and the Phalanx assumed the panel always showed exactly sixteen rungs, so they failed, and every change waiting to ship inherited the failure. Now each test checks that its own line adds no rungs, whatever other lines serve. Players see no difference: only tests changed.

1.248.0 — 2026-09-23

New

• The Dynamo's first keystones: Brownout Grace and Zero-Point Lattice (#4534). Before, the Dynamo sold twenty levels of +2% energy and no keystones: its ring on the Apex Chart was bare, and its sheet showed four sealed chests. It now serves two of its four:
  • Brownout Grace (level 5): for the first hour of a power deficit your colony produces at full rate. This happens at most once per colony per day. A power deficit never stopped a colony; it cut ferrolite, xenogas and data by the share of the draw your plants cover. Brownout Grace holds that cut off for an hour. The Particle Accelerator's Dark Matter and your defences still feel the shortfall at once.
    • The Colony tab, the building card and the energy tooltip say how long the grace has left.
    • A deficit that never ends is graced once, not once a day. Switching the Accelerator off and on again does not buy a second hour.
  • Zero-Point Lattice (level 20): your Zero-Point Reactors produce 50% more energy, and the Dynamo's own +40% applies on top. On a Precursor colony with a level 10 reactor, energy production goes from 92,807 at Dynamo 19 to 139,761 at 20. The building sheet's energy-per-resource figure includes it.
  • Levels 10 and 15 (Overclock and the Grid) have not shipped and stay sealed. From level 6 the next keystone is 20. "To next keystone" sells that whole run in one purchase: 1,170 DM from level 6.

Changed

• The endgame plateau price is 1.2% higher. Each plateau level of the four uncapped tier-7 lines (Doctrine of Annihilation, Beyond the Horizon, Production Singularity, Abyss) now costs 108.60G ferrolite, 66.83G xenogas and 22.62G data, up from 107.29G / 66.03G / 22.35G. The plateau is priced as 60 days of a maxed empire's output, and Zero-Point Lattice raised that output. The Dark Matter step and the level where each line flattens are unchanged. No empire is on a plateau yet.

Also

• Synthetic empires count Zero-Point Lattice in their power planning. They do not plan around the Brownout Grace hour.

1.247.0 — 2026-09-23

New

• The Phalanx, the Aegis's Ascendancy (#4528). Once the Dark Matter Aegis reaches level 10, you could not buy anything further down that line. The Phalanx now opens there. It is priced on the outer-ring ladder, 12 / 30 / 70 / 140 DM a level and 1,260 DM for all twenty levels. Each level adds 2% hull to your ships while they defend a colony, so +40% at level 20. That covers your own fleet at any colony you hold, and your fleets garrisoned at a colony you defend, an ally's included. It never applies in attack or to a fleet caught in flight (an interception, escort included), and never to the Emplacements or the Fortress. In a joint defence each empire's contingent uses its own level. A thicker line also draws fire first: an attacker shoots the Phalanx holder's ships before an identical line without it. Synthetic empires can buy it too.
• On the Apex sheet the line shows "Hull when defending a colony" as it is now and at the next level. The attack forecast measures a defender's Phalanx the same way the battle does. A battle report lists it in the loadout. On an attacker's card it is marked as defence-only, and a defender's card notes that it sits outside the printed Hull multiplier.

Fixed

• Being refused a Dark Matter line gated on the Forge or the Veil (the Dynamo, the Loom, the Cipher, the Fold) showed a raw code such as prereq_missing_dm_forge. It now reads "Requires Dark Matter Forge at level 10 first." (and the same for the Veil). A test now checks every such gate against the Tech Tree, so a new one cannot ship without its sentence.

1.246.0 — 2026-09-23

New

• The Vanguard: a Lance Ascendancy that hits hardest in the first two rounds (#4526). After maxing Dark Matter Lances, the only other thing that spoke sold was the Breach, and it only pays against a colony's walls. Nothing Dark Matter bought rewarded a fleet for deciding a battle quickly. Once your Lances reach level 10, the Vanguard is on sale on the Dark Matter panel for 1,260 DM for the whole line. Each level adds 2% to the damage your ships deal in rounds 1 and 2 of a battle, whether you are attacking or defending, and against ships and a colony's defences alike. That is +40% at level 20. From round 3 on, your fleet fights exactly as it did before.
  • It does not change what your ships aim at.
  • It adds nothing to the Orbital Cannon, your own Emplacements or your Fortress.
  • It stacks with the Breach: both at level 20 is +96% against a wall in the opening two rounds.
  • In a raid, each empire opens with its own Vanguard level.
  • Battle reports list it among your loadout.
  • The Apex sheet shows its total now and at the next level, labelled with the two-round window.
  • Synthetic empires can buy it like any other Ascendancy, but they do not yet factor it into how they judge a raid.
  • Its four keystones are not in this release (#4527).

1.245.3 — 2026-09-23

Fixed

• The Cipher's Apex sheet now has a stat line (#4540, #4543). Selecting The Cipher on the Apex sheet showed no now → next figure. It now reads "Chance to go unnoticed ×1.20 → ×1.22" at level 10, a multiplier on your probe's odds that uses the same words as the card.
• The Fold's stat line is shorter (#4543). It now reads "Warp cost −20% → −22%". Hovering it explains the rounding: the cut comes off the jump's distance, and the price is then rounded up to whole DM, never below 1. The card above it gives the measured average saving.

Also

• Every Dark Matter line must now have a stat line on the Apex sheet, or a stated reason for having none. The test that enforces this found the Cipher's missing line, and it will catch the next one.

1.245.2 — 2026-09-23

Fixed

• The Apex sheet now shows The Fold's stat (#4543). Before, selecting The Fold on the Apex sheet showed no now → next figure at all. Now the sheet prints the cut and what it applies to: "Warp cost, on the distance before whole-DM rounding −20% → −22%" at level 10, and "−40%" at level 20. The price is still rounded up to whole DM afterwards, so the saving on a single jump is often smaller than this figure, and a 1-DM hop never gets cheaper.

1.245.1 — 2026-09-23

Fixed

• The Breach's card no longer promises that your ships turn on the wall (#4626).
  • Before: since v1.236.0, the Breach's card on the Dark Matter panel said it could make your ships "turn on the wall first and leave the enemy fleet for later". The v1.236.0 changelog entry said so too, including that interceptors facing battleships switch at level 1. That claim was wrong in live battles. It was measured on a targeting path no colony battle uses. Every attack on a colony is fought in formation rows. Your salvo splits across the defender's rows the same way at every Breach level. In that test, an interceptor salvo put the same damage on a battleship in the enemy's rear at level 0, 1 and 20; only the wall's share grew.
  • Now: the card says what actually happens. The Breach adds no damage against ships and does not change how your salvo splits across their rows. The wall stands in their front row, and only an enemy ship in that row can lose your fire to it. Bringing the wall down sooner is what sends your fire on to their fleet sooner. The note older app versions read from the server says the same.
  • Nothing about the Breach's effect changed: it is still +2% per level against Emplacements and the Fortress, +40% at level 20. Whether it should pull fire across rows is an open design question, not a change in this release.
  • Installed apps from v1.236.0 onward keep the old card text until they update.

1.245.0 — 2026-09-23

New

• The Apex on a phone is a tile grid (#4519). On a phone the Apex tab used to show the stacked catalogue: every line as a card of prose, one after another. It now shows two columns of tiles in the desktop Chart's order: the four Doctrines as big square tiles, then each Doctrine's Ascendancies under its name, then the Ramparts outside-in over "Your colonies", then the Relics. A tile carries the line's sigil in its level ring (keystone ticks only where a rung is served), its name and level, one state line ("Next 45 DM", "Keystone in 2 levels", "Locked: Orbital Fortress 10", "Complete") and the keystone dot. Tap a tile to open that line's sheet from the bottom of the screen, with its stat, track, buy button and run to the next keystone. Above the tiles are your empire's one stock, what all your colonies add in a day, and the countdown to the next keystone. It fits from 320px wide with no sideways scrolling.

Changed

• On a phone, the keystone dot on the Dark Matter chip now clears one line at a time (#4519). It clears when you open that line's sheet. Opening the Apex tab no longer clears every lit line at once, because the tab no longer shows every line's card. Tapping the chip opens the lit line's sheet directly, and tapping it again for a new keystone on the same line opens it again.
• On desktop, tapping the Dark Matter chip now always goes to the line it lights (#4620, first item). If a door had already opened the Apex on a line and you then picked another line on the Chart, a chip tap naming the first line again did nothing, and its dot stayed lit. The tap now brings the Chart back to that line and clears the dot. The other item of #4620 (the Chart core printing a debt as −264 instead of −263.75) is not in this release.
• Without the Particle Accelerator, the phone grid is drawn dark, with the one "Research the Particle Accelerator" button above it. A player with no Dark Matter at all is still sent to the Accelerator, as before.
• An app installed before this release keeps the old stacked catalogue on a phone until it updates.

1.244.0 — 2026-09-23

New

• The Fold: Dark Matter buys a cheaper warp (#4543). Until now a warp jump cost one Dark Matter per 15,000 units of distance, and nothing an empire could buy made it cheaper. The Fold is the Veil's Ascendancy. It opens once the Dark Matter Veil reaches level 10 and has 20 levels priced 12 / 30 / 70 / 140 DM (1,260 for the whole line).
  • How it works: each level takes 2% off the distance a jump is priced on, up to −40% at level 20, and the price is then rounded up to whole DM, never below 1. It applies to raid warps and bought return legs too, and the launch form quotes the discounted price.
  • On today's map: every level makes some routes cheaper, from 8% of routes at level 1 to 88% at level 20. A maxed Fold saves about a third on an average jump. A 1-DM hop stays 1 DM.
  • The Free Jump (Veil level 15) still waives one whole jump, now at the discounted price.
  • It never makes a jump faster: the 600-second arrival floor is untouched.
  • Synthetic empires can buy the Fold, but they still plan their warps at the full price.

1.243.0 — 2026-09-23

New

• Synthetic empires can now buy Ascendancies (#4564). Until now, only human players could buy the outer ring of the Apex Chart. The AI-marked empires and the Wraiths stopped at the four Doctrines, and their Dark Matter went nowhere else. They now buy the four Ascendancies that have shipped, each once its Doctrine reaches level 10 like everyone else:
  • the Breach, on the Lance;
  • the Dynamo and the Loom, on the Forge;
  • the Cipher, on the Veil.

  Each empire favours the spokes of the Doctrines its style already leans on.

• They also save up for a Doctrine instead of dribbling Dark Matter into a cheaper Ascendancy level.
  • Measured on a slow income: an aggressive empire takes its Lance from 15 to 20, which is what opens the Cutter, in 142 ticks, against 90 before Ascendancies existed. Without the saving rule it took 314.
  • The extra time is the few levels a newly opened Ascendancy gets while it really is that empire's best buy.
• When the Ramparts arrive, synthetic empires will only buy one once they actually hold its building at level 10 on a colony they own.

Also

• A synthetic empire's siege planning does not yet count the Breach it holds. It may pass on a siege it would have won, but it will not launch one it should not (#4615). Synthetic empires can still spend on minor Doctrines while saving for their main one. That was already true before this release and is tracked in #4623.

1.242.0 — 2026-09-23

New

• The Doctrine sheet (#4517). On the desktop Apex Chart, selecting a line used to show the old catalogue card beside it. That card had a per-level rate, a thin ladder, a buy button that turned grey without saying why, and a disabled "Doctrine complete" button once the line was finished. The sheet that replaces it shows:
  • the line's sigil, name, kind and branch;
  • the stat as two totals, e.g. "Shield penetration 14.0% → 15.3%". The Lance's compounding is counted, so ten levels are 14.0%, not 15%. The Aegis and the Forge say their share is "Added to" the total;
  • twenty steps in four bands, with the Dark Matter price per level under each band;
  • a chest at levels 5, 10, 15 and 20. A held chest is ticked. The next chest is named along with what it does. The level-20 chest shows the hull it unlocks. A rung that has not shipped yet (the Dynamo's, for now) is a sealed chest with no words;
  • one ASCEND button whose refusal says why, in its text and its accessible name: "Need 45 DM, you have 12", "Requires Dark Matter Forge level 10", "Requires Orbital Fortress 10 on any colony", or in debt "you are 263.75 DM in debt";
  • the run to the next keystone under the button, "To next keystone · ×5 · 225 DM", which adds "· you have N" when the stock doesn't cover it;
  • "Answered by", and "Feeds" with how many of the lines a hull needs are already at its gate ("Breachhead Monitor (1 of 2 at 15)");
  • a trophy instead of a button once the line is complete: "Doctrine complete · Voidlance Cutter unlocked", or "Ascendancy complete".
• A purchase in flight refuses another on the same line. Two quick taps, or leaving a line mid-purchase and coming back, buy one level, not two.

Phones still show the stacked catalogue until the tile grid (#4519) opens this sheet in a bottom sheet. Not included: a per-hull build count on the trophy (the server does not send one), and the link to the Sigil (#4561 has not shipped).

1.241.0 — 2026-09-23

New

The Cipher: the Veil's Ascendancy (#4540)

Before, the Veil line stopped at its Doctrine. Once your spy probes had all the Espionage they could get, nothing else made them harder to spot.

Now The Cipher appears on the Dark Matter panel once you hold the Veil at level 10. It costs the outer-ring price: 12, 30, 70 or 140 Dark Matter per level, 1,260 DM for all 20 levels. Each level raises the chance that a spy probe goes unnoticed by 2%, up to 40% better at level 20.

What that looks like in practice: a probe sent with Espionage 10 against a commander with Astrometry 8, a level-5 Sensor Grid and a level-4 Observatory slips through 15% of the time without the Cipher, 18% at level 10 and 21% at level 20.

What it does not do:

• It never makes a probe invisible. The game still catches every probe at least 5% of the time, and the card says so. A spy whose Espionage has already brought the odds down to that 5% gains nothing from it.
• It changes nothing about what the spy report contains.
• The Cipher has no keystone rungs yet. They come with #4541 and #4542.

Only players buy the Cipher for now. Synthetic empires do not buy it, and do not talk about it.

1.240.0 — 2026-09-23

New

• The Dark Matter chip tells you when a keystone is in reach (#4520). Before, the only way to learn that your Dark Matter covered the run to a line's next keystone was to open the Apex and read every card. Now a gold dot sits on the HUD's Dark Matter chip, on both the desktop line and the phone pill, whenever some line's stock covers every level up to its next keystone. On the desktop Chart, the line that lit the dot carries the same dot on its ring. Tapping the chip opens the Apex on that line, and opening its sheet clears the dot. The same keystone does not light it again on later refreshes; the line's next keystone does, once you can afford it. Your cleared dots are remembered in this browser. A second device shows the dot until you open the sheet there too. Lines whose keystones have not shipped yet never light it, and neither does a debt.
• The Apex core counts down to your next keystone (#4520): "Next keystone in 2d 9h" at today's income, or "A keystone is in reach". It is left out when your income is zero.

Changed

• Whole Dark Matter on the Apex (#4520). The Apex catalogue printed its stock to the hundredth ("Stock: 44.90 DM"), although nothing is priced in fractions. It now reads "Stock: 44 DM". A debt is still printed to the hundredth, matching the debt note above it. A price is enabled exactly when that printed figure covers it. The HUD chip keeps its hundredths.

Fixed

• The HUD's Dark Matter per day now matches the Apex core (#4520). The HUD chip and the Empire Overview estimated the daily rate from your colony list, while the Apex core showed the server's figure, so one screen could read "+0.00/day" in the HUD and "+3.6/day" on the core. The HUD and the Overview now show the server's rate, the same one that credits your stock.

Known issues

• Tapping the chip does nothing in one narrow case: the Apex was opened earlier on the same line, and the player has since selected a different line on the Chart (#4620). The desktop Chart core still prints a debt rounded down to a whole number (−264 for 263.75), as it did before this release (#4620).

1.239.0 — 2026-09-23

New

• The Loom — the Forge's research Ascendancy (#4537). The Dark Matter panel has a new outer-ring line behind the Dark Matter Forge at level 10. Each level takes 1% off research time, up to −20% at level 20, on Disciplines and Tech Tree nodes alike, and on orders that wait in the buffer as well as those that start at once. It costs 12 / 30 / 70 / 140 Dark Matter per level in bands of five, 1,260 DM for the whole line. It stacks multiplicatively with Quantum Computers under the same existing floor: with both maxed, an order runs at 60% of its base time. Research time never goes below 15 seconds.
  • What it does not do: it shortens the order you are running, not the number of orders you can run at once. Research lanes and buffer slots are unchanged.
  • Known gap: the duration previews on the Research tab and the Tech Tree still show the time without the Loom, so a Loom holder sees a longer estimate than the server actually runs (#4602).
  • Its keystones (L5 / L10 / L15 / L20) are not in this release (#4538, #4539), and the card lists none.
  • Synthetic empires do not buy this line yet (#4564).

1.238.0 — 2026-09-23

New

• Buy to the next keystone in one tap (#4518). Before, a Doctrine card sold one level per tap: getting from Lances 7 to the Sundering keystone at 10 took three taps and three round trips. Now a quiet line under the card's buy button reads "To next keystone · ×3 · 60 DM". One tap buys every level up to the next keystone, in a single purchase at the summed price. Either every level lands or nothing is spent. When your stock can't cover the whole run, the line stays visible, greyed, and says why ("· you have 150"). It is hidden when the next level already is the keystone, and on lines whose keystones haven't shipped yet (the Breach and the Dynamo), which still sell one level at a time. The installed iOS and Android apps don't show the line, and their one-level purchase works exactly as before.

1.237.0 — 2026-09-23

New

• The Apex is a star chart on desktop (#4516). The Apex tab used to be a column of long prose cards. On a desktop it is now a chart:
  • Core: the Accelerator, with your Dark Matter stock as a whole number and what you earn per day.
  • Inner ring: the four Doctrines.
  • Outer ring: each Ascendancy, fanned out beyond its Doctrine.
  • Ramparts: a column standing on your colonies, in the order an attacker meets them (Watch, Mantle, Battery, Keep, Gunnery).
  • Relics: a strip along the bottom.

  Each line is its sigil inside a level ring, with a gold tick for every keystone you hold, its name and level, and one line of state: Next 45 DM, Keystone in 2 levels, Locked: Forge 10 or Complete. Lines that counter each other are joined by an arc, which lights up when you point at either end. The Breachhead and the Interdictor hang between the Doctrines, with a thread to every line that pays towards them; a thread turns gold once that line's share is met. Click a line (or press Enter on it) and its card, with the buy button, appears beside the chart. The page's old explanation is now the tooltip on the core, not a paragraph.

  • Without the Particle Accelerator every line is drawn dark. The one "Research the Particle Accelerator" button sits above the chart, and the core reads "Dark Matter" rather than "Accelerator".
  • Lines that land later appear in place on their own: the chart is drawn from where the server says each line hangs.
  • On a phone the Apex keeps the stacked list of cards until the phone layout (#4519) ships. On narrow desktop windows and tablets (roughly under 1,100 px wide), the card sits under the chart, and tapping a line scrolls it into view.

Fixed

• A maxed Ascendancy's card now says "Ascendancy complete" (a Rampart's will say "Rampart complete"), not "Doctrine complete" or the generic "Line complete". Its tooltip names the level above its own cap, not "21". This is part 2 of #4605; part 1 (the energy surfaces) is still open.
• The Apex page no longer tells a player who is already on the Apex tab to spend Dark Matter "on the Apex tab". It now says "Spend it on the lines here, or on Warp Drive jumps (Missions)."

1.236.0 — 2026-09-23

New

• The Breach: Dark Matter that brings walls down (#4523).
  • Before: the only Ascendancy on the Apex Chart's outer ring was the Dynamo, which is an economy line. Dark Matter bought nothing that helped a fleet against a defended colony: the Emplacements and the Fortress took the same damage from you whatever you had spent.
  • Now: a new line on the Dark Matter panel, the Lance spoke's Ascendancy. It opens once your Dark Matter Lances are at level 10 and runs to 20 levels on the outer ring's ladder: 12 / 30 / 70 / 140 Dark Matter per level in bands of five, 1,260 for the whole line. Every level adds +2% to the damage your ships deal to a colony's Emplacements and its Fortress, so +40% at the cap.
  • What it does not do: it adds no damage against ships, it does not blunt the Orbital Cannon's opening shot, and it does not help a wall's own guns.
  • It does change where your fleet shoots. A Breached wall is a more attractive target, so from the first level some of your ships turn on the wall first and leave the enemy fleet for later. For example, interceptors facing battleships switch at level 1, while the heaviest hulls take four levels or more. The card on the panel says so.
  • In a coalition raid each empire brings its own Breach: your ships hit at your level, your ally's at theirs.
  • The battle report now shows a Breach chip on the side that fought with it.
  • The four keystones are not in yet. The Breach's rungs at levels 5, 10, 15 and 20 arrive with their own issue (#4524). Until then the card shows no rungs, rather than four you cannot buy.

Also

• Synthetic empires do not buy the Breach (#4564). As with the Dynamo, the agents' Dark Matter spending still knows only the four Doctrines. No synthetic empire holds the Breach or will bring one to your walls until #4564 lands.

1.235.0 — 2026-09-22

New

• The Apex server now sends everything the upcoming star chart will draw (#4513). The Dark Matter screen now gets, for every Apex line:
  • its ring: Doctrine, Ascendancy, Rampart or Relic;
  • the arm it hangs from: the Doctrine an Ascendancy grows out of, or the building a Rampart guards;
  • its four keystones, in the player's own words, each marked held or not;
  • the lines that counter it;
  • the cross-Doctrine hulls it is paying towards, with the level needed and the level held;
  • the next rung it is climbing to (none for a line whose keystones have not shipped yet, such as the Dynamo), and its price ladder.

  The response also carries the empire's Dark Matter income per day and its stock as a whole number. The income is the same arithmetic that credits the stock, so a freshly upgraded Accelerator is counted at its new level in both.

• The keystone rungs on the Dark Matter cards now come from the server, so the card and the chart can never describe the same rule two different ways. An app from before this change still shows its built-in text.

Nothing a player sees changes on the current Dark Matter cards apart from where the rung text comes from. The star chart itself arrives with #4516–#4518.

1.234.0 — 2026-09-22

New

• The Dynamo: Dark Matter can finally buy a bigger grid (#4533).
  • Before: the Apex Chart's outer ring was empty. Dark Matter bought one of the four Doctrines and, once, the Precursor district — and nothing an empire could buy made its colonies produce more power. If you wanted a bigger grid you built more reactors, and that was the whole menu.
  • Now: a new line on the Dark Matter panel, the Forge spoke's Ascendancy. It opens once your Dark Matter Forge is at level 10, runs to 20 levels, and every level adds +2% to the energy your colonies' plants produce — +40% at the cap, on every colony you hold. It is the first node priced on the outer ring's dearer ladder: 12 / 30 / 70 / 140 Dark Matter per level in bands of five, 1,260 for the whole line, against 815 for a Doctrine.
  • What it deliberately does not do: the draw is untouched. The walls, the Shield and the Particle Accelerator cost exactly what they cost before, so the choice between minting Dark Matter and powering the guns stays as sharp as it was. On a colony already in surplus, power yield stops at 100% and stays there — the extra production is room to build, not more yield. The one exception is named on the card: a Zero-Point Reactor converts spare power off the raw balance rather than the capped yield, so on a Precursor colony at ZPM 10 the conversion really does move, x1.0796 to x1.0854.
  • The Energy row on a plant's card and the energy tooltip in the header now name the Dynamo on its own line, beside the Superconducting Grid's, so the percentage you see is fully explained by the lines under it.
  • Nothing changes for an empire that does not buy it. At Dynamo 0 every energy number is exactly what it was before this release.

Also

• Synthetic empires do not buy Ascendancies yet (#4564). In this release the Dynamo is a purchase for human players only: the agents' Dark Matter spending still knows only the four Doctrines, so no synthetic empire will compete with you for the Dynamo — or for any other outer-ring line — until #4564 lands.
• Known and not fixed here (#4601): the public wiki still lists the Ascendancies under a Laboratory column, so it states the Dynamo's requirements as "Forge Tier 8 / Laboratory 15". The card in the game is the correct one: the Dark Matter Forge at level 10.

1.233.1 — 2026-09-22

Changed

• Apex lines can now be locked behind a building, not only behind research (#4511). Nothing on your screen changes yet: no line in the game asks for a building today. When the five Ramparts arrive, each one will ask for its own defence building at level 10 on any colony you hold — one colony at 10 is enough, two colonies at 9 are not, and a colony that changes hands stops counting the moment you lose it. The locked card names what is missing in words ("Orbital Fortress 10 on any colony"), and a purchase attempted without it is refused instead of taking your Dark Matter.
• A building you have just finished counts straight away, without having to touch that colony again first, and a second Site of the same building counts as itself rather than as nothing.
• A line you already own stays on the Apex panel with the level you paid for, even if you later lose the colony that opened it.

Also

• Groundwork for the Ramparts of the Apex Chart (epic #4508). The agents do not know about building-gated lines yet: that is #4564, and it lands before the first Rampart.

1.233.0 — 2026-09-22

New

• The battle report tells you which Doctrine rules fired (#4512). A defender is usually offline when their walls matter, and until now nothing in the report said what their Dark Matter had bought. Each side of a battle report now carries a Doctrines & hulls line naming the rules that actually changed the fight, with what each one did and the round it first did it — for example Hardened Pool (shield bubble regeneration rate raised by 50%, first in round 1) · Last Stand (hull losses capped at 25% of a group per round, first in round 3). It covers Overcharge, Sundering, Cascade, Hardened Pool, Deep Formation and Last Stand, plus the Aegis Bastion's shield pool and the Breachhead Monitor's halving of an orbital Cannon.
• The replay's finale card names them too, one line per side, names only — the card stays a verdict you can read at a glance.

Changed

• Holding a rule is not the same as using it (#4512). A keystone is listed only where it moved a number in that battle: Overcharge against shields a salvo swamps anyway, Cascade on a bubble that broke before it ever regenerated, Deep Formation while the bubble swallowed the whole salvo, and Sundering on a counter already at its ceiling are all held, all silent. A line that appears has earned itself.
• Reports written before this change are unaffected and read exactly as they did. Watching someone else's battle as a spectator does not show the line: which Doctrines an empire holds stays something you have to find out by other means.

1.232.1 — 2026-09-22

Fixed

• Galaxy map: a click during a camera flight no longer leaves the hover card behind (#4572). While the camera was flying — the + / − buttons, Go to, a wheel notch, a tap to zoom — one more click that changed nothing about where the camera was going was treated as the end of a mouse pan: a click that opened a planet, a click on empty space, or a second click on the star the map was already flying to. From that moment the card stopped following the move, and when the camera came to rest it named whatever system or planet had landed under the pointer — a place the player had never pointed at. Now only a real drag counts as a pan: through the rest of the flight the card the pointer named is kept, or dropped, and never replaced by another address.

1.232.0 — 2026-09-22

New

• The Apex is its own page, and the Dark Matter chip opens it (#4515). The Doctrines and the Precursor Schematics used to be the third lens of the Research tab: you opened Research, found the "Apex" pill, and read them under a bar about something else. Now tapping your Dark Matter — the figure under your commander's name on a desktop, the purple pill on a phone — takes you straight there. On a phone the Apex is also a door in the Menu, next to Research. Research keeps its "Apex →" entry, and every link that promised "Research → Apex", like the shipyard's VIEW UNLOCK on a locked apex hull, now lands on the page itself.
• The page shows both doors only to empires the Apex exists for: holding Dark Matter, or with the Particle Accelerator researched. Everyone else sees exactly what they saw before.

Changed

• On a phone, the lines that used to live in the Dark Matter sheet — a debt, the factions' haul over the last seven days, the VIP stipend — now print on the Apex page, above the Doctrines, so a gold pill still explains itself in one tap.
• The stock in the Apex heading turns gold when you owe Dark Matter, like every other place that prints it.
• While the Apex is open, the Dark Matter figure in the header is underlined, so the band says where you are.

Also

• The desktop tab rail is unchanged: an Apex entry there pushed Guild onto a second row on a 1920-wide screen.

1.231.3 — 2026-09-22

Also

• Sigil art for the Apex Chart (#4522). Every node of the coming Apex Chart now has a sigil drawn and served from novarchy.com: the four Doctrines, the eight Ascendancies and the five Ramparts (seventeen lines in all), plus the Sigil and the Wraith Codex. Precursor Schematics reuses its district building. You won't see them in the game yet: nothing renders them until the Chart itself ships (#4516 on desktop, #4519 on phone). Until then the Dark Matter panel looks exactly as it did.

1.231.2 — 2026-09-22

Fixed

• Colony advisor: the build suggestions agree with the switch line above them (#4370). Since #4366 and #4369 the advisor has printed "Worth switching back on" or "Worth switching off" above its build suggestions, but the suggestions were still priced as if the switch had not been made.
  • Before: with a second Refinery powered down and the line saying to run it again, no power plant was suggested. With a full Xenogas warehouse and the line saying to cut a Refinery, the top suggestions were power plants saying "every plant here runs short".
  • Now: the suggestions are priced with the switch the line advises, and a short note under the line says so ("Every build suggested here is priced with the Xenogas Refinery (Site 2) switched back on."). A power plant is credited only with what it adds on top of the free switch, never with the switch itself. When a second switch would follow the first, the note says the line will suggest it next.
  • The suggestions don't reorder on their own. When a warehouse fills while you are looking and the switch advice changes, the cards stay where they are. Until then they can still describe the state before the change. A button, "The switch advice changed · update the suggestions", re-prices them when you tap it. They also update when you build, queue or switch something, or open the panel again.

Also

• The panel was rendered headless (SwiftShader) at phone and desktop widths and reviewed by eye; it was not checked on a real device.

1.231.1 — 2026-09-22

Changed

• Apex Chart foundation: the outer ring's price ladder (#4510). Nothing changes in game yet. Every Dark Matter line you can buy today costs exactly what it did (8 / 20 / 45 / 90 DM per level, 815 to max a Doctrine). Underneath, the engine can now price the Chart's coming outer ring, the Ascendancies and the Ramparts, on its own ladder: 12 / 30 / 70 / 140 DM per level in bands of five, which comes to 1,260 DM for a full line. An outer line opens only when its Doctrine reaches level 10. While it is locked, its card says so ("Requires: Dark Matter Lances level 10"), and a purchase attempt is refused. No outer line ships in this release: the first is The Breach (#4523). Part of the Apex Chart epic (#4508).

1.231.0 — 2026-09-22

New

• The Apex opens to anyone holding Dark Matter (#4514). Before, the Research tab's Apex lens simply did not exist until you had researched the Particle Accelerator. If you were holding Dark Matter without it (125 empires on the day this shipped), nothing on screen said what it was for, and a Doctrine padlock's "View unlock" sent you to the Tech Tree instead. Now the Apex tab appears once you hold at least 1 Dark Matter. Without the Accelerator it is drawn dark: every Doctrine is listed, readable but not for sale, and one button, Research the Particle Accelerator →, opens the Tech Tree on that node with its requirements listed. Nothing on the dark Apex can be bought: purchases still need the Accelerator, exactly as before. With the Accelerator researched, the Apex is unchanged.

Changed

• The shipyard's padlock on an apex hull now says "The Doctrines open once the Particle Accelerator is researched". It used to say the Apex tab appears then, which is no longer true for anyone holding Dark Matter.
• If you arrive from a padlock holding no Dark Matter at all, the Research tab still opens on the Accelerator in the Tech Tree and says why.
• If you are looking at the dark Apex and spend your Dark Matter below 1 (on warp jumps, say), you land back on Disciplines. You are not told you asked for something you didn't.

Not in this release: the full Apex Chart view (#4516) and the Apex's own screen outside the Research tab (#4515). This is the existing panel, opened to more players and drawn dark.

1.230.16 — 2026-09-22

Also

The Apex Chart has a design record (#4509). This is a design document only: nothing changes in the game yet.

Before. Dark Matter had nowhere to go past the four Doctrines. Four empires hold all four lines at level 20 and see only grey buttons. The richest of them holds more than 13,500 Dark Matter it cannot spend on the Apex tier. Eight players hold 100 or more Dark Matter but have no Particle Accelerator, so the Apex view does not appear for them at all.

Now. DESIGN-DARK-MATTER-APEX-V3.md sets out the Apex Chart that the rest of epic #4508 will build.

• The four Doctrines stay unchanged, as the inner ring.
• Eight Ascendancies form an outer ring, two per Doctrine. Each opens once its Doctrine reaches level 10 and costs 1,260 Dark Matter to complete.
• Five Ramparts, one per planetary defence, each open once that building reaches level 10 on any colony.
• Three one-shot Relics: Precursor Schematics, the Sigil and the Wraith Codex.
• The whole Chart costs 20,240 Dark Matter, more than any empire holds today.

Every production number in the document was re-measured for it. Where a measurement disagrees with the issue text, the document says so. It also flags five decisions still open before their parts ship: Reduction, Redoubt, Grid, which building gates the Watch, and the id of the Citadel.

1.230.15 — 2026-09-22

Fixed

• The galaxy map's tooltip no longer walks through addresses while the camera moves (#4290). On a desktop, with the mouse resting anywhere but the middle of the map, a wheel notch used to blink the tooltip off and bring it back naming a different system, and the white hover ring hopped from star to star. Five notches could name three other systems in one move, a Go to named planets in two sectors on the way, and when the camera stopped, a card sat on a system you had never pointed at. The ± buttons, tap-to-zoom and ⤢ Fit did the same. Fit was the worst: it jumps the camera, so the old card simply stayed put. Now a camera move can keep the card of the system under your mouse, or take it away once, but it never names a new one. The next system is named when you move the mouse. Two cases keep the card on purpose. Zooming across the point where planets become too small to aim at changes the card between "Planet 9:14:9" and "System 9:14" instead of dropping it. After a mouse drag, the chart easing to a stop no longer blinks the card of the star you dragged.

Also

• Still open, tracked in #4572: a click during a camera flight that leaves its destination unchanged can make the rest of that flight behave like the end of a drag, so the card is named when the camera stops. After a drop the cursor also returns to the drag hand even where a click would still open a planet. The card comes back on the next mouse move.
• Checked by frame sequences of a wheel notch, five notches, the zoom across the aiming threshold and a fast drag, before and after, rendered headless (SwiftShader) and reviewed by eye, with no real device. The tests record every hover report frame by frame across each kind of camera move: 13 of their 21 cases fail on the previous build.

1.230.14 — 2026-09-22

Fixed

• Ship tooling: test placeholder tokens no longer read as secrets (#4361). Nothing changes in the game. This fixes the release pipeline's own secret check.
  • Before: the secret check flagged any token=. Web tests mount components with token="t" (282 times in 191 test files), so nearly every web change was reported as "possible secret in added lines" and sent to the full review lane. Across the last 400 commits on main, 74 were flagged and none held a secret.
  • Now: a line whose only match is one of six fixed placeholder values (t, test, tok, token, test-player, a.b.c) is counted separately and shown as "placeholder literal(s) not counted". Anything else on the line is still checked. Over the same 400 commits, 25 were flagged.
  • Still flagged: the remaining hits are type annotations and variable pass-throughs such as token: string; they are outside this fix.

Changed

• Ship tooling: the secret check now catches keys it used to miss (#4361). Before, it only saw a secret sitting after a key named like one, such as API_KEY= or password:. It now also recognises values that identify themselves:
  • AWS access key ids, and AWS_SECRET_ACCESS_KEY= in both spellings;
  • GitHub ghp_/github_pat_ tokens, including inside an Authorization: header;
  • sk- API keys and Slack xox tokens;
  • bearer JWTs;
  • a password written inside a connection URL.

  Of a 31-shape test set of realistic secrets, 9 went unflagged before; all 31 are flagged now, in source files and in test files.

• Trade-off: blanking a placeholder also stops the check from incidentally catching other keys on the same line that aren't named like secrets (for example pw: "…" next to token: "t").
• Rollout: worktrees that have not pulled main keep the old check until they do.

1.230.13 — 2026-09-22

Fixed

• Agents state the Exodus bounty band on the Gate, and past the limit (#4477).
  • Before: an AI empire whose colony was razed and relocated out of its band was told that anyone hitting it paid "no bounty at all under three times my Power" in the Frontier, and a doubled bounty "from twice my Power" in the starting sectors. Both were wrong. The bounty compares Gates (the larger of Structural and Military), not the Power headline, which can be up to twice as large. And the limit itself is free, so the price starts only once an attacker is more than that multiple.
  • Now: the agent says what the web has said since #4427. In the Frontier, no bounty unless the attacker's Gate, or a raid's Gates added together, is more than 3 times its own. In the starting sectors, the bounty is doubled and starts once that figure is more than 2 times its own. The multiples are read from the server's own values, and a test holds them equal.
  • Who sees it: agent chat is currently switched off in production, so players won't see these lines until it is turned back on. The change does not alter what the agents plan or do: the lines only reach the agents' chat, and their plan prompt already stated the band on the Gate.

1.230.12 — 2026-09-22

Fixed

• Decoys at your other colonies can no longer silence the attack that lands first (#4493).
  • Before: the game sent at most four notifications a day, shared by every kind. An attacker could aim four cheap fleets at four of your colonies, use up those four in seconds, and the real attack launched a minute later never reached your phone. The last thing it said was about a decoy.
  • Now: the fleet that lands first on any of your colonies still rings when the day's four are spent, as does a fleet landing sooner than a warning already on your phone about the same colony. The day still has a hard ceiling of twelve notifications. Only those two kinds of alarm can go past four; everything else still stops there.
• A decoy the attacker recalls no longer counts (#4493). Before, a recalled fleet could keep standing between you and the real attack behind it on the same colony, and the warning kept its "about 10 minutes". Now the game checks which fleets are really still coming.
• The daily ceiling now holds even when several notifications leave at once (#4493). Before, notifications sent in the same moment could all get through past the limit.

Also

• What it costs you: on a heavy day you can now get up to twelve notifications instead of four. An attacker who launches many fleets at one colony, each landing sooner than the last, can make several of them arrive within a minute. Opening the game silences the rest for five minutes.
  • On past days' attacks, replayed read-only, players with notifications on would have received about 17% more, and 1 of 116 player-days would have reached twelve.
  • A day with many corrected warnings can leave room for one battle report fewer than before: 2 over those 45 days.
• Not changed: nothing is sent while you are in the game or during your night hours, and the in-game feed still lists every fleet. A strike landing after a fleet you were already warned about can still go unannounced on a day whose four are spent. The warning already on your phone brings you in first, and the game shows every fleet coming.
• iPhone notifications are still off on the live server: the Apple credential is not installed, so only browser notifications reach players today.

1.230.11 — 2026-09-22

Fixed

• An incoming-attack alarm is closed when the strike turns back at the colony (#4485). A Nightfall alarm reaches whoever holds the colony once their Incoming board would list the attack. If the reason it waited for came back before impact, the strike turned back at the colony without a word. That reason can be a guildmate of the raider who left and rejoined, a pact, or a Veil raised past the raider's Sensors. The red "incoming attack" line, and its push, stayed open for a fleet that never landed.
  • Now a player who was alarmed about that attack gets the closing line a recalled attack already writes: "Inbound flight turned back: the hostile fleet that was due at 21:14 turned around — that one is not coming." Because it is written at the moment the fleet was due, it names the clock time the alarm counted down to rather than "any moment now".
  • A raid whose every contingent turns back closes each leg's alarm in the same way.
  • A raid that partly lands writes no such line: its battle report closes it.
  • A player who raises a Veil against a loud attack aimed at them now gets their alarm closed when the fleet turns back blind.
  • A recall, or a raid cancel, after the reason came back now closes the alarm too. Before, it wrote nothing, because the Incoming board had already dropped the row.
  • A player who was never alarmed still hears nothing, one alarm gets one closing line, and the closing line sends no push.
  • AI empires no longer read these lines as a feint: an agent that rejoined the raider's guild does not hold a grudge against a guildmate for a strike that was aimed at the colony's previous owner.
  • Still open: a raid's loud launch alarm when every contingent turns back blind (#4499), and a player who leaves the colony before impact (#4503).

1.230.10 — 2026-09-22

Changed

A schema change can no longer ship without a rollback note (#4336). This is ship tooling: players see nothing.

Before. A change to the database schema could reach novarchy.com with no rollback block in the deploy runbook, and nothing in the ship machinery noticed. SQL migration files were already gated, but a schema change applied at boot had no guard. The deploy only printed that the schema had changed, and review was the one thing that ever caught a missing block. That gap matters within minutes of a deploy. Once a player writes to a new column, rolling the server back crash-loops it, and the one command that unblocks it is written nowhere.

Now.

• The deploy refuses to ship a schema change that the runbook does not name.
  • A new or changed column must be named together with its table in the runbook text the same range adds.
  • A new table must be named.
  • A block written later in the same range counts, so a refused deploy is healed by writing the block.
• The PR preflight asks the same question before the merge, while the fix is still one commit in the same PR.
• A true revert of a schema change passes, because it removes the column and deletes its block together.
• Both checks fail closed: an unreadable production version or a git error refuses.
• A human can override one deploy with an explicit flag that does not carry over to the next.
• The check confirms that a note was written and names the change. It never judges whether the note is right. That stays with review.

Replayed over the history of main: of 130 past schema commits, 44 named their changes, 24 changed nothing a rollback touches, and 62 would have been refused. 53 of those had no runbook edit at all, and 9 edited the runbook without naming the change.

Left out. The runbook's duplicate ladder letters are not fixed, and the blocks that past changes never got are not backfilled. Trees shipping from an older checkout keep the old, unguarded scripts until they pull.

1.230.9 — 2026-09-22

Fixed

• The AI Advisor's commentary knows the switch line (#4370). The colony advisor prints "Worth switching back on" or "Worth switching off" above its build suggestions (since #4366 and #4369), but the AI Advisor that writes the commentary never saw that line.
  • Before: with a Refinery powered down the grid reads 100% power, so the Advisor's headline could call the grid healthy right above a line saying that Refinery was worth running again.
  • Now: the Advisor is told which producer the line names, which way to switch it, and where (the Energy choice in the building sheet), and it may not contradict that.
  • It is not told the line's countdown or figures, which change every few minutes near a full warehouse, so a commentary written earlier can't quote a number the line has moved past.
• The Accelerator card's energy figure (#4370). On a colony with a producer powered down, the Accelerator card's "Energy yield X → Y" counted that producer as drawing power after the upgrade but not before.
  • Before: the card could show a drop in power the upgrade doesn't cause, or not appear at all.
  • Now: both figures describe the same colony.

Also

• The build suggestions themselves are unchanged in this release, apart from that Accelerator card.
• Coming in a following release: the suggestions priced with the switch the line advises. With a Refinery powered down and the line saying to run it again, power plants will be ranked, credited only with their own share of the gain, not with the free switch. With a full warehouse and the line saying to cut a producer, cards will no longer say "every plant here runs short". The engine for this is in place but not yet turned on for the panel.

1.230.8 — 2026-09-22

Fixed

• An agent no longer brings up a deleted player's words when somebody new takes that name (#4469). Agents remember who they talked with and what was said. Some lines name a person without replying to them: an idle-channel opener, a guild council line, a league taunt, a raid call. On those lines, if a player had been deleted and a newcomer later picked the same display name, the agent could be handed the deleted player's old words next to a name that now belonged to the newcomer. Now the agent only remembers that it once spoke with a different commander who used that name, and quotes nothing. Someone who renames keeps their history under the new name, and a person talking under their own name is remembered exactly as before.
• Agents forget the conversations of a player who has been gone for two days, the same way they already let go of that player's grudges. One missed check is never enough. Anybody seen on the leaderboard or the galaxy map, heard speaking in galaxy chat, or known as an agent, a Wraith or the Observatory keeps their history. What is still forgotten is a player who stays off the leaderboard, hidden from or absent on the map, and silent in galaxy chat for 48 hours in a row.

1.230.7 — 2026-09-22

Fixed

• A sooner attack on your colony no longer stays silent behind an earlier alarm (#4486). An "Fleet inbound" push used to hold its colony for a flat 15 minutes. A second, faster attack on the same colony in that window never rang, so the lock screen kept saying "It arrives in about 2 hours" while a fleet landed in minutes, and the next thing you saw was a battle report. Now the sooner attack rings through, and its banner names both fleets: "A fleet is heading for Lantea. It arrives in about 3 minutes. Another fleet follows in about an hour."
• A fleet launched after the one you were warned about had already landed is announced. It used to be muted by the banner about a battle that was over, and since an attack alarm is only sent once, it was never announced at all.
• A later fleet no longer pushes a sooner one off the lock screen. An alarm about a slow fleet could replace the banner of a faster one still on its way. Every attack banner now leads with the fleet that lands first.
• In the browser, a replacing alarm now makes a sound. A notification that replaced an earlier one for the same colony used to swap the words silently. Chromium browsers (Chrome, Edge) now alert again. Firefox and Safari ignore the setting, so there the replacement may still be silent.

Changed

• Up to 12 attack alarms a day instead of 4. The daily limit of 4 notifications stays, shared by every kind. On top of it, each attack alarm that reached you can be followed by up to 2 corrections that day: a sooner fleet on the same colony, or one launched after the fleets you were told about had all landed. Corrections do not use up the day's 4, so an attacker can no longer silence you by sending decoy fleets that each land a little sooner. Every one of those alarms is about a real fleet. A raid is still one alarm, and a slower fleet is still not announced again inside the 15 minutes.
• Still open: decoys sent at four different colonies can still use up the day's 4 before the real strike (#4493).

1.230.6 — 2026-09-22

Fixed

• The bounty band is stated on your Gate, the number the server actually compares (#4427). The bounty starts when an attacker's Gate (the larger of Structural and Military) is more than 3× the defender's, or more than 2× on colonies in sectors 1–4. For a raid, the members' Gates are added together.

  Before, about thirty lines said "Power", the header figure that adds Structural and Military together and can be nearly double the Gate. Examples: "none at all under three times your Power", "starts from twice your Power", "Within your Power band (⅓–3×)", and "Far below your Power".

  They now name the Gate:

  • the Galaxy tab's sector notes, the Exodus screen and the diary's Exodus rows;
  • the Leaderboard row tooltips and the planet sheet band notes, including the Beacon/Bastion/Relic note, which printed its Gate as "Power N";
  • the Attack button tooltip, the launch warnings and the mission composer's target line;
  • the War Room ("your Gate N") and the war target rows;
  • the raid door and its rule tile, and the Target Board header;
  • the Wanted pill and the phone Wanted sheet, and the Wanted board.

  The sector and Exodus lines also say "more than" where they said "under", because a Gate at exactly 3× is still free. They now say a raid's Gates are added together too: four raiders at 0.9× your Gate each do pay a bounty, even in the Frontier.

• Synthetic empires plan against the same rule (#4427). Their strategy briefing said the band "looks at buildings" and called it the "Power Band". It now says the band compares Gates while fleets decide the battle, and that interception follows the same Gate band.

Left out: the agents' chat lines about the Exodus still say "Power" (#4477). The system list and target panel still head their Gate column "Power" (#4478).

1.230.5 — 2026-09-22

Fixed

• A Nightfall attack on a colony that changed hands now alarms its new holder when the strike can reach them, even if it could not at first (#4467). A Nightfall attack's alarm (the diary line and the push) is sent at the start of its last stretch. If you had taken over the colony mid-flight and the attack could not land on you at that moment, or the attacker could not see you, you were never alarmed: for example you were in the attacker's guild, still a newcomer, under a shield, or behind a Veil. That held even if it changed before impact, which it can: you leave the guild or are kicked, you launch your own first attack, you move your night window, the attacker's spy probe reaches your colony, or his Sensors are completed. In any of those cases the fleet struck with no warning but the countdown on your board. The alarm is now checked again every minute until impact, and sent as soon as your board would show the attack. A colony that was empty and gets settled in that stretch is covered too. It is never sent with less than a minute to go, when the countdown on your board is the warning. This applies to alarms due after this release; it is a rare case, and production had no such attacks when it shipped.
• The push no longer rounds the last minutes up (#4467). With 90 seconds left, an incoming-attack notification said "in about 2 minutes". Below three minutes it now rounds down, as it already did for hours: "in about a minute".

Changed

• An alarm that reaches you late says so, instead of blaming the Nightfall Protocol (#4467). A Nightfall alarm sent after that stretch began reads "This attack became visible to you only now." (for a raid: "They became visible to you only now.") instead of saying the fleet ran dark and that a Deep Sensor Grid at level 8 would have seen it: that was not why it was late. Older app versions keep the old sentence until they update.

1.230.4 — 2026-09-22

Fixed

• Agent lines cut for length keep their last emoji and their numbers whole, and the Kepler Observatory never posts an empty bulletin (#4425). Players will not see any of this yet. Agent chat has been off since 1 September, and the Observatory's bulletin, the aid-board notes and Wraith chat are all under that same switch. The change is installed for the day chat comes back on.
  • Before: when an agent's line hit a length limit, the cut could land in the middle of an emoji, so the line ended on half a glyph (a family emoji reduced to one person, a flag reduced to a bare black flag). It could also land in the middle of a number, so "we lost 10 ships" went out as "we lost 1", "1.2M points" as "1.2", and a 3-2 score as "3". The Observatory's bulletin could end on a decimal point ("leads with 1."), and a bulletin with emoji in it could be refused as too long. A bulletin made only of invisible characters was sent and refused by the server.
  • Now: a cut never splits an emoji. A figure at the cut is either whole or left out entirely, including its suffix (k, M, %), a score, a date or a minus sign. The bulletin ends on a real sentence and counts its length the way the server does. A reply with nothing visible in it is not sent anywhere, and a bulletin like that is logged. Agent aid-board notes always fit the server's own limit whole.
  • Left out: a human player's aid-board note can still have an emoji split by the server's own cut (#4482).

1.230.3 — 2026-09-22

Fixed

• On/off switches take only on or off (#4459). A malformed request could flip these switches the wrong way:
  • the Accelerator and per-building power switches;
  • warp on raids and missions;
  • razing and dock raids on missions;
  • pin, push notification and Discord post on admin dispatches.

  Before, a text "false", an empty list or an empty object all counted as yes. So "warp":"false" asked for a paid Dark Matter warp, and "on":"false" switched the Accelerator on. Now only a real true or false counts. A missing value still means off, and anything else is refused with a clear error. The game, the mobile apps and the AI empires already send real true/false, so nothing players do changes.

• Guild actions with no target answer with an error instead of crashing (#4459). Seven guild actions failed with a server error when a request named no player or guild: change rank, kick, hand over leadership, invite, answer an invite, decide an application, and grant from the Fund. They now return the same error each action already gives for an unknown player or guild.

Also

• The in-memory test database still accepts some values the real database rejects. That is tracked in #4475.

1.230.2 — 2026-09-22

Fixed

• Agents remember a conversation by who they had it with, not by the name that person carried (#4341). Before, an agent's memory of who it had talked with, and what was said, was filed under the other side's display name. Display names are not unique, so two commanders with the same name shared one history. An agent could greet a player with an argument it had had with somebody else of that name, and since v1.229.2 that player actually got an answer. Renaming started a fresh history, and a player who shared the answering agent's own name was never remembered at all.

  Now each conversation belongs to the player, whatever they are called:

  • a rename carries the history along;
  • two commanders with the same name keep separate histories;
  • on any line that names someone who shares a name with somebody else the agent has talked to, including the comment after a battle, the agent never borrows the other one's words.

  The memories agents already had were carried over when this release started. 515 conversations were kept, with every word and every count intact. 55 could not be tied to a single player now in the game: whoever held that name has left, or someone else carries it. They were set aside rather than risk showing one person someone else's conversation. If that player turns up again within 48 hours, the conversation comes back. Memories of agents no longer in the roster were left as they were.

  Not covered yet: a conversation with a player who was deleted can still surface on a line that names nobody directly, if a newcomer has taken the same name (#4469).

1.230.1 — 2026-09-21

Fixed

• Research, shipyard, Dark Matter and guild diplomacy refuse a request field of the wrong type (#4459). These are the cases #4430 left over that could change game state. Before, a request that sent an id wrapped in a list slipped past the catalogue checks:
  • Research (["logistica"] instead of "logistica") started a second, fully charged order on a Discipline or Tree node that was already researching, at the same level.
  • Shipyard queued and charged the order and stored the list in the colony's ship queue.
  • Dark Matter upgrade wrote the list into the Dark Matter receipt.

  Proposing a pact of a kind the game does not offer answered a server error (500). A reparations demand with an amount that was not a number was filed as a bill whose resource was stored as empty. Each of these is now refused with a clear error, and nothing is charged or written. Requests sent by the game's own clients (web, iOS, Android, AI empires) behave exactly as before. A read-only check of production found no queue, order, receipt or demand affected by any of these.

  • Two request shapes that were accepted before are now refused. No shipped client sends either:
    • an old Italian ship name wrapped in a list (["intercettore"]). The same name sent as plain text is still translated and built;
    • a research request that carries both keys with a Tree node of false or 0.
  • Still open in #4459: requests where a true/false field is read loosely ("false" counts as true), and a missing player or guild id on seven guild-member and treasury routes, which still answers a server error.

1.230.0 — 2026-09-21

New

• How much of a resource you are missing, on the Base (#1664). Since #1630 an upgrade's price turned red on exactly the resource you were short of, but nothing on the Base said by how much. Now every Base surface that prices an upgrade prints, under the price, «Still needed −3,120 Xe · −9 Dt», in exact figures: the map's hover card (mouse), and the Build-menu row and the Building Sheet (what a phone reaches, since the hover card never opens on touch). The figure is your colony's current stock against the price, rounded up, which is the same comparison the Upgrade button and the server make, so banking what it says always clears the shortfall. It stays silent while that level is already being built (its price is paid), on locked buildings, on levels needing an Advanced core, and on levels the Command Center caps. On an empty treasury it says «the full price» instead of repeating it.

Fixed

• The Base map's hover card is no longer cut off at the map's edges (#1664). It used to grow upward from the pointer with no limit and reserve a fixed width, so near the top of the map its top was cut, and a long price ran off the left or right edge. It now sits above the pointer when it fits and flips below it near the top. It stays inside the map on every side, wraps a long price between resources instead of mid-figure, and paints above the map's Operations panel and camera buttons. Checked by rendering it headless (SwiftShader) and reviewing the frames by eye, not on a real device.

Also

• Not changed: the Building Sheet still colours affordable figures white, where the hover card and Build menu use green.

1.229.5 — 2026-09-21

Fixed

• A guild aid note made only of invisible characters is now no note (#4416). A request's note could be made entirely of zero-width spaces, joiners, word joiners, soft hyphens or braille blanks. It was stored as text, and the aid board showed a note with nothing in it. It is now stored as no note, the same as a note of plain spaces. A note with anything visible in it is kept as typed, emoji included. The 140-character cut on the note no longer leaves half an emoji at the end.
• Formation preset names that draw nothing are refused (#4416). Saving a preset whose name is only invisible characters now fails, the same way a blank name always did. A preset saved earlier with such a name is no longer listed, so it can't block saving your other presets.

Also

• The refusal for an invisible-only preset name still shows the generic "your client is out of date" text, which is wrong for this cause. Honest wording is tracked in #4433, together with fleet template names, which still accept invisible-only names.
• No one has checked how many notes or preset names like these are already stored in production. Existing notes stay until their request closes.

1.229.4 — 2026-09-21

Fixed

• An attack on a colony that changed hands no longer strikes a holder the launch would have refused (#4450). An attack lands on whoever holds the target when it arrives, but the checks that refuse an attack are made at launch, about whoever held the colony then: you cannot attack yourself, a guildmate or a guild you have a pact with, a newcomer under novice protection, a shielded empire, or someone inside their night window. If the colony emptied and was refilled during the flight, nobody asked those questions again. A raider who colonised the freed slot himself had his own fleet fight and loot his own colony. A guildmate, a newcomer born on that slot, or a shielded or sleeping settler was struck and looted as if they had been the target.
  • Now the arrival asks the same checks of whoever holds the colony, when the fleet arrives. The player the attack was launched at is not asked again: their launch-time answer stands, as before. If the new holder would have been refused, there is no battle. The whole fleet turns back with an empty hold, and nothing is written to the holder.
  • The attacker's diary says why: "Attack called off on arrival". The line names the protection (the colony was yours; it belonged to a guildmate or to a guild yours had a pact with; to a newcomer under novice protection; it was under a shield; the fleet arrived in its owner's night window). It never names the owner. An empty slot, or one hidden by a Veil, still reads "Target not found". Installed app versions that predate this release show that older line for a refused strike too, until they update.
  • Raids are judged contingent by contingent. A contingent that may not strike sits the battle out and comes home whole; if none may, the raid turns back with no battle.
  • The inbound warning agrees with the impact. Since v1.229.1, the new holder of a colony sees an attack inherited with it. They are now shown it only when it can actually land on them. One limit remains: the late Nightfall alarm is sent once. If the protection that silenced it ends before impact (the holder or the raider leaves the guild, or a newcomer makes their own first attack), the countdown appears on the board, but no alarm is re-sent. That follow-up is #4467.
  • How often: this is a correctness fix for a rare case. Measured read-only on 2026-09-21 over every attack since 2026-07-24: no attack launched at a player had ever struck a different player, and no battle had ever pitted an empire against itself.

1.229.3 — 2026-09-21

Fixed

• A request with a text field sent as a list or an object no longer crashes the server (#4430). Some requests carried a field that should be text as a list or an object instead, for example a chat channel sent as ["galaxy"]. Before, the server answered them with an internal error. That covered sending a chat message or marking it read, pinning, unpinning or reporting a message, the guild diplomacy actions, guild membership actions, creating a guild, naming a colony, placing or moving a building, speeding up a queue, repairing a legendary hull, opening a crate, granting from the guild treasury, and the Google, Apple and developer sign-in routes. The sign-in routes need no account, so anyone could trigger it. Each now refuses the request cleanly with the error it already uses for a bad value of that field, and nothing is written. The game and the agents always send these fields as text, so nothing changes for anyone playing. The first name Apple shares on the first sign-in is optional: if it arrives in an unusable form, the empire gets its generated name, as it always did, and sign-in is not refused.

Also

• Fields that are not text but still misbehave when sent in the wrong shape are tracked in #4459. They include the diplomacy proposal type, research orders, ship orders, a few on/off switches, and a missing required id on the guild-member routes. They are not fixed here.

1.229.2 — 2026-09-21

Fixed

• A player who shares an agent's name is treated as a person again (#4342). The AI empires used to decide whether someone was a person by looking up their display name among the agents' names. Display names are not unique: a Google sign-in keeps your first name as-is, and many agents carry a bare first name. A player whose name matched an agent's therefore counted as a machine to all of them:
  • naming an agent in galaxy chat got no answer;
  • whispers went unanswered;
  • threats left no grudge;
  • rules questions were never picked up;
  • joining an agent-led guild brought no welcome.

  The agents now use the account-level marker the game already shows as the AI badge, so a person is a person whatever they are called. A machine missing from the agents' own lists is now also treated as a machine. No player on the live server was affected at release time. One known gap remains: an agent's memory of recent conversations is still keyed by display name (#4341), so it can mix up a person and an agent who share a name.

1.229.1 — 2026-09-21

Fixed

• An attack on a colony that changed hands mid-flight now warns whoever holds it (#4438).
  • Before: if a colony changed hands while an attack was flying at it (the old owner folded away and someone colonised the empty slot, for example), the attack disappeared from every inbound board. The old owner no longer held the colony, and the new owner never saw the attack because it had been aimed at someone else. It still landed, on the new owner, with no countdown, no HUD alert and no option to evacuate.
  • Now: the attack is on the inbound board of whoever holds the colony, graded by that colony's Observatory. The Dashboard alert, the all-clear banner, the evacuate option and the guild threat board all follow.
  • Recalls and Nightfall follow the holder too. If the attacker recalls (or cancels a raid), the "attack withdrawn" line goes to the holder who was watching the countdown. A Nightfall attack's late alarm and push go to whoever holds the colony when it unveils, but only when the attack was launched at a player without Sensors L8. An attack launched at the Pirate Beacon, or at a player with Sensors L8, has no scheduled unveil, so a later holder sees it on the board in its final minutes without an alarm or push.
  • Stealth is unchanged: an attack launched against a colony the attacker could not see through its Veil stays invisible to everyone. A new holder is warned only if the attacker can see their colony, which is the same check that decides whether the strike lands at all. Your own Sensors still pierce Nightfall as before.
  • What the old owner sees: once you no longer hold the colony, you get no alarm or withdrawal line about it.
  • Left out: taking over a colony with an ordinary attack already inbound does not send a push; you see the attack on the board from that moment. The guild door indicator, the advisor's threat line and the garrison stand-down lock still read the original target, and a colony taken over from a deleted account is not warned. These are tracked in #4449. What happens at impact for the new holder is tracked in #4450.

1.229.0 — 2026-09-21

New

The colony advisor says when a running producer is worth switching off (#4369)

Before: on a colony short of power, a producer whose warehouse was full kept running. Everything it made overflowed, and its power draw still slowed every other producer on the colony. The advisor said nothing. Since v1.223.0 it could say "switch this back on", but never "switch this off".

Now: the advisor prints its own line above the suggested builds, in the same box as the switch-on line:

Worth switching off: Xenogas Refinery (Site 2). Your Xenogas storage is full, so what it makes is lost. Set its Energy choice to Off in the building sheet. +724 Fe/h, +110 Dt/h reaching storage, colony-wide. The colony then runs at 95% power.

When the warehouse is not full yet, the line says when it will be: "Your Xenogas storage fills in about 5h". If some of that producer's output would still have reached storage, the line says how much it gives up.

When the line appears:

• only when that producer's own resource is full or fills within 8 hours, and the colony is short of power;
• only when switching it off gains at least 1% of the colony's output, by both of the ways the advisor weighs resources;
• it names one producer, the one that gains most.

It goes away when the warehouse drains, and it follows a mandate change.

It never contradicts the switch-on line:

• It stays silent while the switch-on line has something to say.
• Following one line never brings up the other naming the same building.

Measured, on 1,500 random colonies short of power per row:

• One resource at the cap: it appears on 1,463, median +20% of output.
• No resource forced to the cap: it appears on 153, median +6.4%. These are colonies whose stock happened to sit near a cap.
• Without the full-warehouse condition: the same arithmetic would have spoken on 798, at a median +1.2%.

Not included:

• the Accelerator's own switch (#4371);
• crediting power plants for a switch they make possible (#4370).

The line is shown in the colony panel only; the AI Advisor is not told about it yet.

1.228.2 — 2026-09-21

Fixed

• The catch-up ramp says which figure it is measured on (#4411). While the ramp was boosting you, the Overview said "boosted until your Power reaches half the median established empire", and the research Bench said the discount lasts "while your Power sits under half the median established empire". The game has always measured the ramp on your Gate, the larger of your Structural and Military. The Power figure at the top of the screen is the two added together, so it can read up to double. A balanced empire could see its Power past the line and still be on the ramp. Both lines now say "until your Gate (the larger of Structural and Military) reaches half the median of established, active empires", which matches the note on the Leaderboard.
• Chrono shard entries in the diary name the right ratio (#4411). A shard earned by striking up used to read "torn from an empire 2.13× your power". The ratio has always been between Gates, so it now reads "torn from an empire with 2.13× your Gate (the larger of Structural and Military)". A raid's entry reads "…× your raid's combined Gate (your Gates added together, …)". Older entries in your diary read the same way, and the figures they print are unchanged.
• AI empires talk about the Gate the way players see it (#4411). When an AI empire mentions the catch-up ramp, or brags about a shard it tore from a bigger empire, it now works from the same "Gate" figure the HUD and the Leaderboard show. The size gap in a brag is the one measured when the attack hit, never recalculated from the empire's current size.

1.228.1 — 2026-09-21

Fixed

• Crucible Vault: a dormant player's own line reads as one phrase, and never tells them to wait (#4444). In the Monday Vault, a dormant player's "You" line used to show two unlabelled phrases side by side: "0 pts none while dormant nothing while dormant". A dormant player whose banked points would still reach a paid place got the same "nothing while dormant" there, while the stage chip told them to "duel before Monday to keep it". On a phone the Vault sits alone on its own segment, so the two contradicted each other. The Vault line now says "dormant" once: "dormant: no place, no prize", or, when the banked points are above place 10's, "dormant: enough for a paid place right now — duel to bank again".
• The stage chip gives the same hint in the same words, in the present tense. A dormant empire banks nothing while place 10 keeps banking, so a lead over place 10 shrinks by the hour, and "before Monday" invited the player to wait when the move is to duel now. At an exact tie with place 10 the hint is not shown, because the game cannot tell from whole points which way the tie breaks.
• On the Ladder tab, "Your standing" now reads "If it ended now: none while dormant" for every dormant player, in reach or not. A player still dormant at Monday's roll is not paid, so no paid place is claimed under that label.

Also

• The frames of the three dormant cases (nothing banked; enough for a paid place; too few) were rendered headless (SwiftShader) at 390 and 1440 px and reviewed by eye, with no check on a real device.

1.228.0 — 2026-09-21

New

• Crucible: the Monday prize goes to whoever held the ladder longest (#4440, a player's suggestion)
  • Before: the ten ranks standing on the board at the instant of the Monday roll were paid, so a week at #1 lost to a knock-out in the final hour.
  • Now: every hour at a rank banks points: 500 an hour at #1, 499 at #2, one less per rank, down to 1 at #500 (nothing below). At Monday 00:00 UTC the ten with the most points that week are paid the same prize table as before, by place, and the count starts again.
  • Time is counted exactly, to the second, from the moment a rank changes hands. Nothing depends on when the server ticks.
  • Dormant hours bank nothing. An empire still dormant at Monday 00:00 is not paid, and its place goes to the next by points. One duel before then (fought or defended, won or lost) ends the dormancy, and the points banked before it count again.
  • This week (the week of 21 September) counts held time from this release onward, for everyone alike. Last week was already paid by the old rule.
  • The Crucible tab now shows:
    • your points this week, what your rank banks an hour, your place in the week, and how far you are behind place 10;
    • the prize your place would pay if the week ended now;
    • in the Vault, who holds each paid place right now, with their points and current rank;
    • week places read "place N", board ranks "#N".

Also

• Installed apps keep showing the old rule text ("Top ten hold their rank, get paid") until they update, and light the prize row at your board rank. The prize figure they quote follows the new rule: it is the prize for your place in held points if the week ended now.
• The new Crucible screens were rendered headless (SwiftShader) at phone and desktop widths and reviewed by eye, not on a real device.

1.227.0 — 2026-09-21

New

The relics have a page: novarchy.com/wiki/relics (#4442).

Every relic in the galaxy has carried a line of its own story since the day it shipped — "A ring the width of a moon, still folding space for traffic that stopped coming", "It measures a drift no chart admits to, and broadcasts the correction every hour, to nobody" — and until now the only way to read one was to be standing on it. All forty are on the page, grouped by Travel, Production, Shipyard and War, each with what it is, which sector it sits in, and what holding it is worth to everyone in your guild.

The rules are there too, including the one that costs a fleet if you do not know it: a guild already at the cap can win the battle and still not take the relic. It is left unclaimed for whoever comes next.

The lore pages have art. New key images on the relics page, the Eschaton page and the factions page, plus portraits of the Lantern Road, the Deep Orchard and the Suspended Yard.

The Eschaton page now says what the previous cycle left behind, and what The Sunless Vault actually is: a wall you throw your fleet at to measure how hard it hits. Nothing is spent and nothing is lost — it is a ruler, not a raid.

Fixed

• The relic list used to be a wide table, which on a phone meant swiping sideways to read a sentence whose relic name had already scrolled away. It is a list now, and it wraps.
• Page images declare their size, so the text no longer jumps as a picture loads, and the first image on a page loads first instead of last.
• The lore art is less than half its former weight.

1.226.1 — 2026-09-21

Fixed

A screen reader now says when the last inbound attack leaves the board (#4333)

Before: with a screen reader, you heard «Under attack on …» when an attack appeared, on any tab, but nothing when the last one left the board. The alarm chip simply disappeared. The only way to learn the board was clear was to open the Missions tab and read it.

Now: on any tab, when your inbound board goes from carrying attacks to carrying none, you hear, once and politely:

No inbound attacks detected. Your Inbound board covers every colony you hold, not one in the air but off the board, not ships in dock, not a neighbour's strength.

It says what was detected, never that you are safe, and it carries the same scope note sighted players see under the board: an empty board can still be missing an attack. The sentence is the same whatever emptied the board (an impact, a recall, or an attack that went dark), so it tells you nothing the board itself doesn't show.

Nothing is announced when the game loads, when you switch tabs or rotate your phone, or when a poll fails and then recovers onto the same empty board.

Not verified on a device: this was tested in a simulated browser, not with VoiceOver or TalkBack. On the Missions tab, the green banner can be up to one 30-second poll behind or ahead of the spoken sentence.

1.226.0 — 2026-09-21

New

Novarchy has a manual, at novarchy.com/wiki (#4420). Ten pages, readable without logging in, on a phone, and linked from anywhere — no account, no bundle, no waiting for the game to load.

Three pages are written for people:

• Getting started walks your first ten minutes in the order they actually happen, with screenshots of the real game. It leads with the one thing nobody tells you: build the Fusion Reactor first. A colony with no power runs at −100%, so an extractor built before a reactor produces nothing at all and never says why.
• The Eschaton is the world — the end-of-cycle event the galaxy is named for, the Veil, and a straight account of who else is out there: the empires run by language models, and the Wraiths. Both carry a mark beside their name, and neither will ever claim to be a person.
• Protection answers the question every persistent strategy game gets asked first — can I be wiped out while I sleep? — with the engine's own numbers rather than reassurance.

Seven more are generated from the game itself at every release: buildings, ships, research, planets, resources and the three factions. Not one number on them is typed by hand. They call the same functions the server calls, so a balance change reaches the wiki with the change, and a table can never quietly drift out of date. Costs are the real ladder, ship stats are the real stats, and the counters that decide a battle are the ones your shipyard is using right now.

Fixed

• A mistyped or retired /wiki/… address now returns an honest 404. Every other address on the site answers 200 with the game shell, which is how a dead link could look like a working page.
• The sitemap listed two URLs and was maintained by hand. It is now built from the page list, so it cannot lose a page — and search engines can find the manual.
• /changelog and the wiki now link to each other.

Notes

The wiki's numbers are the engine's, but its prose is written by people and was reviewed hard before shipping: several first-draft sentences described rules the game does not have, and were cut rather than reworded. If you find one that is still wrong, it is a bug — say so and it gets fixed like any other.

1.225.3 — 2026-09-21

Fixed

• Agent chat lines no longer break emoji apart, or come out blank (#4417). Most agents never use emoji, and the rest use one now and then. When a model wrote an emoji an agent was not allowed, the cleanup cut it one piece at a time. That broke the emoji built from several pieces:
  • a family emoji came out as a blank line with the agent's name and no words;
  • the England flag also came out blank;
  • with one emoji allowed, the family came out as just the man;
  • a keycap like 3️⃣ left a stray box behind.

  Since v1.224.1 the server refuses a blank chat line. In a whisper, that meant the agent asked its model again on the next round for a reply that would end up blank again. Now:

  • An emoji over the limit goes whole, and one that stays is the one that was written.
  • A keycap keeps its number, so "we have 3️⃣ fleets" becomes "we have 3 fleets", not "we have fleets".
  • A line with nothing visible left is silence. In a whisper, the agent leaves the message on read, the way someone who never types emoji does when their only answer would have been a thumbs-up.

  How often each agent uses emoji is unchanged, and no timing changed. Players will not see this yet: agent chat has been switched off in production since 1 September. The fix is live in the agents service and applies to every line once agent chat is turned back on.

1.225.2 — 2026-09-21

Fixed

• A chat link with a key written twice no longer breaks the chat (#4422). A chat URL that carried channel twice made the chat and its pinned bar answer with a server error. A mention search with q twice did the same, and asking for facts only (scope=facts) twice still listed every commander. Each of these now reads the last value given, the same rule the guild log follows since v1.225.1. Two different after or before cursors are still refused, and two page sizes still give the default page. No app ever sent these URLs: the error needed a hand-typed address, and nothing was ever written or lost.

1.225.1 — 2026-09-21

Fixed

• A repeated take no longer breaks the guild registry (#4288). Opening the registry with a repeated or garbled page size in the address (?take=25&take=30, ?take=abc) used to fail with a server error. It now opens: the last value given is the one used, and a value that isn't a number gets the normal page. The game's own screens always send a single value, so no one hit this in normal play. Measured against a real database, the old failure was a loud error and never a registry that looked empty.
• The colony advisor's "no narration" switch holds when repeated (#4288). Sending narrative=0 twice used to turn the narration back on. It now stays off.
• Admin player search accepts a repeated q (#4288). The search no longer errors on it and uses the last term.

Also

• A repeated query key now follows one rule across these routes, the same one the guild registry's kinds filter already used (#4276): it is normalised, never a server error. The same defect on three chat routes (channel, and q/scope on mentions) is left out of this release on purpose and tracked in #4422.
• A page size of Infinity (or a number too large to represent) now gets the normal page of 40 entries instead of the maximum of 100.

1.225.0 — 2026-09-21

New

• «Send from» on the Overview's colony cards (#1138). On a phone, each colony card on the Overview ended in Base, Colony and Yard. To send a fleet from a colony you picked it there, then opened Missions. Every card now has a fourth button, Send from: one tap selects that colony and opens Missions with the launch pad already on it. It appears on every card, including a colony with no ships yet, so the row never changes shape between updates. The same cards show on tablets and narrow windows; the wide desktop table is unchanged.

Changed

• A half-built mission keeps its launch pad (#1138). If you have a mission in progress, «Send from» (or picking another colony in the header) no longer moves it. "In progress" means a mission with ships you picked yourself, or cargo you typed for a Transport, Pickup, Colonise or Redeploy. Picked ships already worked this way. Typed cargo did not: before, a transport's cargo quietly moved to whichever colony you switched to, and was drawn from that colony's warehouse instead of the one you typed it for. Now the pad stays where you built the mission. The «Launch from» section says which colony it launches from ("You are viewing Vega, but this mission launches from Hoth"), and one tap on another colony's chip moves it. Cargo left over from a cargo mission doesn't count once you switch to a verb that carries none, such as Attack: that pad follows as before.

1.224.1 — 2026-09-21

Fixed

• A chat message with nothing visible in it no longer posts (#4241). Before, a message made only of invisible characters (zero-width joiners or spaces, word joiners, a soft hyphen, the Hangul filler, braille blank and the like) went through. It showed up in galaxy chat, guild chat or a whisper as a row with a name and no words. Now it is refused with the same "empty message" error as a blank one, and your text stays in the composer. Messages that merely contain such characters post exactly as before, including family emoji, keycaps and Persian words that need a ZWNJ between letters.

1.224.0 — 2026-09-21

New

• The Leaderboard explains how the game helps whoever joined late (#1601). A newcomer who opened the Leaderboard used to see Gates hundreds of times their own, and nothing on that screen said a late start is not a lost one. The mechanics that help existed, but each was explained only where it applies, if at all. A collapsible "Joined late? What works in your favour" note now sits between Your Echelon and All empires. It lists five rules, each quantified on the Gate the board ranks by:
  • Catch-up ramp: while your Gate is under half the median of established empires, production and research run faster, up to ×10.
  • Underdog bonus: the value you destroy counts up to ×3 in war points against a bigger target. In a raid, the comparison uses the raid's combined Gate.
  • Chrono shards: winning an attack on an empire with at least 1.5× your Gate can drop a shard, within the weekly pool of 4 hours.
  • Bounties: an empire more than 3× your Gate (2× on colonies in sectors 1–4) that attacks you usually takes a bounty.
  • Weekly races: the war board and your Echelon race start over every week.

  The note opens by itself only for a player the catch-up ramp is boosting, and shows their live multiplier. It stays collapsed on later visits once they close it. Every number is read from the rule it describes, and tests fail if either one changes without the other.

Also

• Filed #4411: the Overview line, the research Bench and the agents' own status sheet state the catch-up threshold in "Power", but the rule measures the Gate. It is not fixed in this release.

1.223.12 — 2026-09-21

Fixed

• Galaxy map, desktop: the camera buttons draw their icons, centred and matched (#4407). Before, the column at the top right of the map (⌖ home, ⊞ system view, + and − zoom, ⤢ show the whole galaxy, ↩ back) was six typed characters. Each one's size and position depended on whatever font your computer had for it. In our test render they came from five different fonts, so a tiny minus sat beside a large square, some glyphs sat off-centre, and the fit button turned from a small arrow into a wide, flat hook when it became ↩. Now all six are drawn icons, the same set the phone map has used since #4399, with the same stroke as the map's left-hand rail. Each one is centred in its button, + and − are a matched pair, and ⤢ and ↩ are the same size, so the button no longer changes shape when it swaps. Buttons, tooltips and keyboard behaviour are unchanged.
• Galaxy map, phone: ↩ is now exactly the size of ⤢ (#4407). The phone's ↩ icon was one unit shorter than ⤢, so the fit button still changed height slightly when it turned into ↩. Its loop is now one size, and the other phone icons are unchanged.

The frames behind this change were rendered headless (SwiftShader) and reviewed by eye, desktop and phone, in both states of the fit button. There was no render on a real iPhone or Android device, and none in a real desktop browser on macOS or Windows.

1.223.11 — 2026-09-21

Changed

• Galaxy map on phones: the camera buttons are larger, match, and sit centred (#4399). The round buttons at the bottom right of the map (⌖ your home, ⊞ system view, ⤢ show the whole galaxy, and ↩ back to your previous view after a fit) showed typed characters whose size and position came from whatever font the phone had, so they did not match each other. In our headless render ⌖ was a small mark, and ↩ a flat hook about 7 px tall, slightly off-centre. They are now drawn icons in the same line style as the map's left rail. They are larger and sized as one set, each is centred in its circle, and ↩ is the same width as ⤢, so the button no longer jumps when it swaps. The buttons, their labels and their tap targets are unchanged. The change was rendered headless and reviewed by eye. It was not rendered on an iPhone or an Android device: the shapes no longer depend on the phone's fonts, but no real device has been looked at yet. The desktop rail still shows the typed characters. That is filed as #4407.

1.223.10 — 2026-09-21

Changed

• Phone Overview: one fleet section instead of two (#1713). On a phone, the Overview used to show the STATIONED FLEET chips above your colonies and the FLEET LEDGER cards below them. The numbers were nearly the same, drawn two different ways. When you have two or more colonies, they are now one section, in the old strip's place above the colonies:
  • The FLEET LEDGER header is the tap to open and close it.
  • A line under the header gives the split: how many ships are docked and across how many colonies, how many are in flight, and how many are garrisoned.
  • While closed, it shows one chip per ship type with the icon and every ship of that type you own.
  • When open, each ship type's card is headed by that same chip, with the total up front instead of a trailing TOTAL. The card still shows your per-colony counts, the in-flight cell and the garrison cell.

  The Empire header on the phone now says «hulls docked», so its count cannot be mistaken for the chips' owned total. The section keeps its place while your fleets are still being counted, so the colony cards no longer jump down when the count arrives. One trade-off: if you have two or more colonies and no ships at all, the section briefly shows «still counting» and then closes, which moves the page up once. With one colony, and on desktop, the Overview is unchanged.

1.223.9 — 2026-09-21

Fixed

• Pickup: the number under each cargo cell is now the number the fill buttons write (#4354). On a Pickup, the faint «/ N» under each cell showed what the destination holds, minus what your other fleets have already booked there. The fill buttons also stop at the free room in the launch colony's Warehouse, where the fleet unloads, and the caption left that out. With your warehouse 5,000 short of full, a destination holding 80,000 and a 10,000 hold, the cell read «/ 80,000» and max wrote 5,000, with nothing on screen to say why. The cell now reads «/ 5,000 fits in the launch colony». The suffix appears only when that warehouse is what set the number, the same way a Redeploy says «fits there» about its destination. When the destination's stock or another fleet's booking is the limit, the caption stays a bare «/ N». A full launch warehouse reads «/ 0 fits in the launch colony» under every cell.
• Pickup: the panel no longer blames your other fleet for a full warehouse (#4354). When another of your fleets was already booked at the destination, the panel said the offered amounts were what that fleet would leave. It said so even when your own warehouse was the real limit, and even when the manifest was empty because the warehouse was full. It now mentions the other fleet only when its booking actually lowered what is offered. When your warehouse is the limit, the empty-manifest message says the amounts stop at what the launch colony's Warehouse can still store.

1.223.8 — 2026-09-21

Fixed

• Relic and bastion signs no longer cover whole systems on the zoomed-out galaxy map (#3547).

  Before: a relic's diamond (26 px) and a bastion's chevron (22 px) kept their close-up size at every zoom, and hung above their star. At the zoom a phone opens the map at, one sign was wider than the system it named and reached toward the next star. A sector's relics stacked into a column that read as one shape.

  Now:

  • At galaxy zoom, a relic alone on its star is a 10 px diamond and a bastion a 9 px chevron, both sitting on the star.
  • From sector zoom in they return to full size, with their name and effect line.
  • A sign whose star also carries another mark keeps its full size at every zoom. The other mark is your home, your colony's gold mark, a pirate beacon, an anomaly signal, or a second relic or bastion. Shrinking a sign there would have merged it with that mark.
  • Hiding that mark's layer lets the sign shrink with the rest.

  The pirate beacon got the same treatment in v1.223.4 (#1605).

1.223.7 — 2026-09-21

Fixed

• Battle replay: the Fortress no longer hides the Emplacements beside it (#4289). In a replay of a colony that had built both static defences, the Fortress stood at the end of the defender's front rank nearest the camera. It is by far the widest thing on the line, so the Emplacements next to it, and any light ships the defender had placed in that rank, were drawn behind it. The Fortress now stands at the far end of its rank, and its rank-mates are drawn in front of it. The Fortress keeps its size and the camera frames the battle exactly as before. The only other change is on the defender's side: its rank-mates each shift one place along the rank. The same holds for old reports from before formations, where the defence is drawn as one flat line.
  • Measured on the replay's occlusion bench against its 40% ceiling: the worst defence shape fell from 2.34 to 0.33 with ranks, and from 3.66 to 0.35 on old flat reports.
  • The other half of #4289, a capital ship covering a light ship beside it in the same rank, was closed by v1.223.5 (#4380). Re-measured here, it stands at 23% coverage.
  • Not covered: the ordering holds for the replay's own camera. A player who drags the camera round to look at the defence from behind, and some director close-ups, can still see the Fortress in front of its rank.
  • Rendered headless (SwiftShader) and reviewed by eye. Limits: software rendering, a handful of frames, one scenario. Those frames happened to catch milder angles than the bench's worst case.

1.223.6 — 2026-09-21

Changed

• Galaxy map: the fit button now goes both ways (#1722). ⤢ used to frame the whole galaxy and forget what you had been reading, and getting back meant pinching in and finding the place again. Now, after a fit, the same button turns into ↩ "Back to your previous view" and flies you back there. It goes back to ⤢ as soon as you pan, pinch, zoom or fly somewhere. It keeps keyboard and screen-reader focus while it switches, so pressing it twice takes you there and back. The button stays on phones, where it is the only one-tap way back to the whole galaxy. Desktop and phone now call it by the same name, "Show the whole galaxy". Left out: the glyph itself is unchanged in size (20 px in the 48 px phone button). Enlarging it is filed as #4399.

Fixed

• A fit is remembered across the system view (#1722). After pressing ⤢, opening a system and coming back could restore the zoomed-in view saved on an earlier visit to the map. It now lands on the whole galaxy you left.

1.223.5 — 2026-09-21

Fixed

• Capital ships no longer tower over the support hulls that outrank them in the battle replay (#4380). Capitals used to render on a larger ruler of their own. The Colonial Ark (catalogue 3.0) read shorter than a Battleship (2.9), and the Hauler (3.2) shorter than the Battleship, the Destroyer and the Voidlance Cutter. Now all six capitals are drawn to the same proportion of their art as every other class, about 0.87. A class the catalogue ranks higher is never drawn shorter, capitals included.
  • What changes on screen: capitals are about 29% shorter in the scene. The Titan and Annihilator are still the longest hulls in the fleet, now about 1.4 times a cruiser instead of 2 times. The replay frames the front by its width, so a battle that holds capitals is framed tighter, and escorts, support hulls and the enemy line all read larger than before.
  • A side effect: a capital sharing a front row with a light ship no longer draws over it. The overlap drops from about 52% of the smaller hull to 23%. The Fortress still draws over the Emplacement beside it; that is tracked separately in #4289.
  • Not changed: the shipyard cards, which frame each ship by its own size.
  • How it was checked: by measurement, and on replay frames rendered headless (SwiftShader) and reviewed by eye, before and after, on the same battle. The limits are software rendering rather than a GPU, and a handful of frames. The Ark and the Hauler beside capitals were seen only in an edited test fleet, because no stock scenario fields them.

1.223.4 — 2026-09-21

Fixed

• Pirate beacons no longer cover whole systems when the galaxy map is zoomed out (#1605).
  • Before: every beacon was drawn at 18 px, the size it has close up. At the zoom a phone opens the map at, that was as wide as the whole system it marks (1.03× a system's width), and twice that at full pinch-out.
  • Now: at galaxy zoom, a beacon alone on its star is 11 px (0.63× a system at a phone's opening zoom). It returns to 18 px when the map switches to sector zoom.
  • Exception: a beacon that shares its star with your colony, your home or a visible relic or bastion stays 18 px, so its orange does not blend into the gold mark there.
  • Position: the "far from their sector" half of the report was already fixed before this release. A beacon sits on its own star at every zoom.
  • Left out: relic and bastion icons keep their size at galaxy zoom; that is #3547, still open.
  • Verified by measurement only. The size and position of the mark were measured in the running map code on a phone and a desktop viewport. The final look (including the smoother texture of the 11 px ring) has not yet been checked in a screenshot.

1.223.3 — 2026-09-21

Fixed

• «fill the hold» no longer looks live when it cannot do anything (#4355). On a cargo mission with no ships picked yet, the button used to look ordinary and do nothing when tapped: no numbers, no message. That was every first run, and on a redeploy it was the default, because a redeploy opens with no fleet picked. The button is now dimmed whenever a tap would change nothing, following the same rule as the per-resource «max» buttons next to it. When the reason is the missing fleet, a line under the cargo cells says so: "No carrier picked yet — pick one under Fleet to fill its hold, or use fill every carrier." It is text on screen, so it works on phones, which have no tooltips. The line appears only when those two fixes would actually load something, not when the pad is empty or the destination warehouse is full. Numbers you typed are never touched. Not changed: after a fill, picking more carriers leaves «fill the hold» dimmed, because every cell is already written. Tap «clear», then fill again.

1.223.2 — 2026-09-21

Fixed

• Galaxy map: tapping a system no longer opens the sheet of a fleet flying beside it (#1642). At the opening zoom, a tap on a system is meant to step the camera in. Before, if any fleet in transit was drawn within a fingertip of that star (24 px, which on a phone is wider than the gap between two neighbouring stars), the tap opened that fleet's intercept sheet instead, and tapping again opened it again. Now a fleet mark takes the tap only when the tap lands on the mark itself: within a small circle around it, 10–15 px depending on the mark's size. Every other tap steps the camera in. The circle is a little larger than the drawn triangle, so a star sitting right against a mark can still yield to it. That trade is deliberate, so the mark stays hittable with a finger.
  • On desktop, the hover card over such a star now names the system, because clicking there really does zoom in.
  • When two fleet marks overlap, a tap opens the one the finger is on, not whichever sits nearer.
  • On a very crowded chart, a hollow fleet mark that could not be drawn no longer takes taps.
  • Hiding Relics or Bastions from the Layers sheet does not change any of this. Those places were never tap targets, and fleet marks are not behind any layer switch.

1.223.1 — 2026-09-21

Fixed

• Support ships in the battle replay are the size their class says (#1287). The Colony Ship, Freighter, Colonial Ark, Hauler and Aegis Bastion came out too small in the replay, at about 0.6 of the art they replaced. A Colony Ship looked shorter than a Bomber flying beside it, and a Freighter shorter than most fighters. All five are now about 1.4 times larger, so they sit in the same size band as the gunships and cruisers. Every support ship now renders at least as long as any escort or cruiser its class outranks. The Spy Probe is unchanged; it was already the right size.

Also

• Checked by measurement only: every hull's in-scene length was measured through the game's own sizing code, and a new test holds the whole fleet to the ship-class ranking. Nobody has watched the change in a live replay yet.
• One known gap remains: the Colonial Ark and the Hauler still render shorter than the capital ships they outrank, because capitals sit on a larger size scale of their own. Tracked in #4380.

1.223.0 — 2026-09-21

New

The colony advisor says when a powered-down producer is worth switching back on (#4366)

Before: since v1.217.0 you can power a producer down to free its energy. Once you did, the advisor treated that as settled. You saw a healthy grid and a list of builds, and nothing told you the switch was costing you output. Across 1,500 test colonies with a producer powered down, switching it back on was already a gain 88% of the time, with a median of +14.4% of output.

Now: when running a powered-down Extractor, Refinery, Compute Node or Stellar Forge again would raise what reaches your storage by at least 1%, the advisor shows its own line above the suggested builds:

• which producer to switch back on;
• what it adds per hour;
• what its power draw takes from the rest of the colony;
• the power the colony would then run at.

A full warehouse is the clearest reason to power a producer down, and the line respects it. When the producer's resource is at the cap, switching it on would only throw its output away, so no line appears. The line follows your stock: it goes away as the warehouse fills and comes back when it empties.

Not in this release:

• advice to switch a running producer off (#4369);
• the suggested builds still being priced with the producer off, and the written advisor summary not yet knowing about the switch (#4370);
• the Particle Accelerator's own on/off switch (#4371).

1.222.0 — 2026-09-21

New

Relics and Bastions can be switched off on the galaxy map (#1720)

Before: with every switch in the map's Layers sheet turned off, the relic diamonds and bastion chevrons stayed on the chart. No switch reached them, and on the current map they hid the sector layout underneath.

Now:

• The Layers sheet has two new switches, Relics and Bastions, both on when the map opens. Turning one off removes that family's marks and names from the chart.
• A hidden relic no longer crowds a bastion that shares its star: the bastion sits centred with its effect line again. A pirate beacon no longer steps aside for a mark you can't see.
• The Places board still lists every relic and bastion. Tapping one to fly there switches that family back on, and only that family, so you never land on an unmarked star.
• A switch flipped while the map is still loading now takes effect when the map appears.

The switches reset when you leave the map, like the other layers, and nothing is saved between sessions or accounts.

1.221.1 — 2026-09-21

Fixed

• The weekly Echelon results no longer arrive hours late on a busy Monday (#4348). On Monday 21 September the Echelon week should have closed at 00:02 UTC. It closed at 02:05 instead, because it had to credit 119 winners' home colonies in one go, and any one of those players doing something in their colony meanwhile cancelled the whole attempt. For those two hours:
  • the Echelon board already showed the new week, but still ranked everyone by last week's progress;
  • no promotion, relegation or reward had arrived;
  • progress made after midnight counted toward the week that had just ended, not the new one.

  Now each winner's home colony is credited on its own, right after it is read, so players can keep playing through the settle. Measured on a copy of that week's 119 colonies: under traffic heavy enough to make the old settle fail 20 times out of 20, the new one failed none. If a winner's home colony is destroyed at that moment, the reward goes to their next-oldest colony instead of being skipped.

1.221.0 — 2026-09-21

New

Plan a tech-tree node behind the research that unlocks it (#1555)

Before: a node whose prerequisite was already being researched looked exactly like one nobody had started. It sat behind a padlock, its card listed "Why it's locked" with a ✗ beside research you were paying for at that moment, and it offered no control. Research that has started can't be cancelled, so the only way to queue the node was to watch the lane, wait for it to finish, and come back.

Now the node can be planned: placed in the queue behind the order that unlocks it.

• The card offers Plan unlock (waits for its prerequisite). Under What it's waiting for, a requirement already in progress shows an hourglass and reads in the queue. It no longer shows a cross.
• In the tree the node is marked Can plan on phones and shows an hourglass on desktop. A new legend entry explains the mark.
• On the research bench the order reads "Not paid yet: waiting for the research it needs to complete." It never claims to be waiting for a lane when a lane is standing free.
• Nothing is paid when you place it. The cost is charged when it actually starts. It's the price of that level, the same whether you plan early or late. Once the research it needs completes, it becomes an ordinary queued order and starts when a lane and your treasury allow it.
• A planned order never holds up the rest of your queue. Other research behind it keeps starting as lanes free up.
• If you remove the research a planned order was waiting on, the order isn't silently dropped. The bench marks it Stuck and gives you both ways out: queue that research again (the order then carries on by itself), or remove the order, which costs nothing.
• A node that also needs Dark Matter you don't have yet can't be planned. Parked like that, it would have stopped your whole queue the moment its prerequisite finished.

Planning needs a free queue slot, which comes with a Laboratory at level 12 or above, a Science-10 guild, or the Third Lane doctrine. Without one, the node stays locked until its prerequisite completes, as before.

1.220.0 — 2026-09-21

New

Evacuate — the inbound alarm can be answered by running (#741)

An attack is coming and the alarm counts it down. Until now the only thing you could do from that card was Send a garrison — and a garrison has to come from somewhere, so it needs a second colony. Most players have exactly one. For them the alarm was a countdown with no action on it at all: they watched it arrive and lost the fleet they had been warned about.

The card now offers Evacuate the hangar. It lifts every hull off the colony under attack — probes and colony ships included — and lands them back there a few minutes after the raid is over. No form, no coordinates to retype, no second colony needed.

What running costs, so it is a decision and not a reflex:

• your fleet is away for as long as the raid is due, plus a few minutes — the card tells you exactly how long before you commit;
• it cannot be called back once it has left;
• hulls only. Your warehouse stays behind, and the raid still takes it. The colony defends without a fleet.
• it is only offered while an attack you can see is actually inbound on that colony.

The control asks twice — the first press states the price, the second sends it, and you can cancel in between. Once the fleet is up, the alarm card says so and counts down to its return instead of going blank, and the fleet list shows it standing off your colony.

The garrison is still there, below it, whenever you have somewhere to launch one from.

When your fleet comes home, Activity says Fleet back in the hangar. In the rare case where the colony it left is gone by the time it returns, the ships are not lost — they are rerouted, and the next line says where they landed.

1.219.1 — 2026-09-21

Fixed

• An agent could say that a raider "came in with" himself (#4339). When an empire is raided, it remembers who came in alongside the raider and says so in chat — "Ivan Keller raided you, together with noemi80". It worked out who to leave off that list, the raider himself and the empire itself, by display name. Two things went wrong with that. Against a raider who wasn't on the leaderboard, the agent couldn't put a name to him, so nobody was left off, and it would say "Ivan Keller raided you, together with Ivan Keller and noemi80". And display names are not unique in Novarchy, so a real co-raider who happened to share the raider's name, or the agent's, silently disappeared from the story. Agents now tell raiders apart by account, read from the battle report itself. On top of that, no raid line prints a companion spelled like the raider it is about, or like the agent speaking. That is what fixes raids remembered before this release: the lists stored for them are not rewritten, but they no longer print the raider twice.

  What is not fixed, so nobody reads more into this entry than it delivers: a companion is still named in chat by display name alone. If two players share a name, a reader cannot tell which one was meant. And a genuine namesake of the raider is now remembered but deliberately left out of what the agent says, because printed next to the raider it reads as the same man counted twice.

1.219.0 — 2026-09-21

New

• The Wanted board hides the commanders you cannot attack (#1547). The Wanted board at the top of the Leaderboard listed every commander with a price on their head and said nothing about whether you could go after any of them. With most human players in two allied guilds, much of it was people the server would turn away on arrival — and the only way to find out was to launch a fleet and read the refusal. Now the board opens with Hide protected switched on, and holds back every commander an attack on whom would be refused right now: your own guild, an ally, a nonaggression pact or the withdrawal notice of one just cancelled, a newcomer still under the shelter, or a post-raid shield — on the whole empire, or on every one of their colonies you could aim at. It always says how many it is holding back ("3 hidden — behind a pact, a shelter or a shield"); switch it off and they come back marked ✕ Pact, ✕ Novice or ✕ Shielded. If it holds back everyone, it says so and names the switch. The header counts the commanders actually listed, and the switch keeps its setting when you tap a row, go to the Galaxy and come back.

  The answer comes from the same rules that refuse the launch, in the same order, so the board and the attack agree — including during a pact's withdrawal notice, when the treaty already reads as dissolved but attacks are still refused. It is a list of who is held back, not a promise about who is left: an attack can still be refused for night protection, which depends on when your fleet lands and cannot be known from a list. Your own row, if you are wanted, is never hidden. Suggested by jaybee/Usherin.

  Not changed: the ranking below the Wanted board is not filtered — it is a table of where everyone stands, and hiding rows from it would leave gaps in the ranks.

1.218.1 — 2026-09-21

Fixed

• A flaky Apple sign-in test no longer turns the release gate red when the build machine is busy (#4353). Nothing changed in the game: Sign in with Apple behaves exactly as before, and its once-a-minute limit on asking Apple for fresh keys is untouched. What changed is the test that checks key rotation. It read the time before its first sign-in and then pretended 61 seconds had passed, so it had one undeclared second of slack against a 60-second limit. On a loaded machine that second ran out, and a valid token was reported as invalid. The test now stops the clock before the first sign-in, so how busy the machine is no longer matters. It also checks the limit on both sides, one millisecond short and exactly on time. A neighbouring test that could pass without checking anything under load now counts Apple requests, so that shortcut is closed.

1.218.0 — 2026-09-21

New

• A «max» beside each cargo cell (#1142). Sending a lot of one material meant doing the arithmetic by hand: «fill the hold» spreads the load across Ferrolite, Xenogas and Data in the ratio of your stock, so to send only one you had to read the hold capacity off the header, subtract whatever was already typed in the other two cells, and key in the result. Every cargo cell now has a small max that loads that one resource with everything that still fits. It counts what you typed in the other two cells against the hold and leaves them as they are, and it never writes more than the launch would accept: no more than the fleet can carry, no more than you actually have, and always whole units. It appears on Transport, Colonize, Redeploy and Pickup, and each keeps its own limit: a pickup is also capped by what fits back home, and a redeploy by what the destination warehouse can still take in. Once a cell already holds its maximum, or there is no hold to fill yet, the button greys out instead of doing nothing. It works together with «fill the hold»: max one resource, then fill the rest, or the other way round. Suggested by Remy.

1.217.0 — 2026-09-21

New

Power down a producer to get its energy back (#1275). A colony that had outgrown its reactors had one switch and one only — the Particle Accelerator's. Everything else drew energy forever, and buildings cannot be demolished, so an Extractor browning out the grid had no off position at all.

The four resource producers — Ferrolite Extractor, Xenogas Refinery, Compute Node and Stellar Forge — now carry an Energy choice box on their building sheet, with On and Off. Powered down, a building draws nothing, which raises the colony's energy factor for everything else, and produces nothing while it is off. Its level is untouched, so the empire score it counts towards does not move: you save energy, never weight.

The off state is honest everywhere the colony is shown — the energy bar and the conduits, the production chips and the "Full in…" clock, the ⚡ delta on each build row (marked as not being billed), the building sheet's own projections, and the Suggested Upgrades ranking. Switching a building back on resumes it immediately; nothing is lost in between.

Also

Defence structures and the Particle Accelerator are deliberately not switchable. The wall a colony fields would have honoured the switch, but the wall the game advertises to scanners would not have, so powering a Shield down would have freed its whole draw and kept its deterrent — better than leaving it on, which is not a choice worth offering. The Accelerator keeps its own existing switch rather than gaining a second one. Both may follow later.

Synthetic empires read a powered-down colony correctly, so their economic and defensive estimates match what the server will actually do.

1.216.1 — 2026-09-21

Fixed

• An agent could mistake another player's chat line for one of its own, and its own for another player's (#1688). Agents are allowed their own turns of phrase: before the anti-repetition gate asks "has someone else already opened a line this way?", it sets the speaker's own lines aside. It worked out which lines were the speaker's own by matching on the display name — and display names are not unique in Novarchy, by design, with older accounts and agents already sharing them. So a player using the same name as an agent had their lines set aside as if the agent had written them: the agent was then free to open a line exactly the way that player just had, and the list of phrasings it was told to steer away from quietly lost one. Agents now tell each other apart by account, not by the name on the badge, so a shared name no longer lets a repeated opening through.

1.216.0 — 2026-09-21

New

The Research tab shows the whole ladder, not just the next level (#1534)

Opening a Discipline used to price exactly one level: the one you were about to buy. At level 17 the panel told you what level 18 costs and nothing else — so planning three levels ahead meant buying your way there to find out.

There is now a Show all 24 levels button under the Research action. It opens a table with one row per level, all the way to the ceiling, showing for each:

• what it costs in Ferrolite, Xenogas and Data;
• how long it takes for your empire — the figures already include your Laboratory network and your current research modifiers, so they are the durations you will actually wait;
• what you end up with at that level — the total you reach, not the step.

The row the Research button would buy is highlighted, levels you have already paid for are dimmed, and the table opens scrolled to where you are. It stays collapsed until you ask for it, so the panel still opens on the decision rather than on a wall of numbers.

Three players asked for this independently over twelve days, each comparing it to the buildings' progression table, which has worked this way for a while.

Details the table gets right that a simpler one would not:

• Colonization stops selling colonies partway up. Once your Command Center can hold no more, further levels shave time off every building you queue instead — and the table says so, level by level, rather than repeating "Up to 10 colonies" for fourteen rows. It also respects your actual Command Center, so it never offers a colony slot the game would refuse.
• Espionage moves at every level, even where the report tier does not, so consecutive rows are never identical over a purchase costing millions.
• Big costs stay readable and stay exact. Past a million the figures abbreviate to fit; an Exact figures switch above the table turns every digit back on, which works on a phone where hovering does not.
• The table is legible with a screen reader. Each row is headed by its level, so the costs are never announced without saying which level they belong to, and the level you can research next is marked as current rather than only tinted.

Also

• Corrected two source comments that described the Discipline ceiling as 25; it has been 24. (The 30-level ladder is the one buildings have.)

1.215.0 — 2026-09-21

New

Raiders choose what fills the holds (#795)

A raid that beat its target but could not carry everything it was owed used to scale all three resources by one factor. The hold came home as a mirror of whatever the target was sitting on — and since colonies overwhelmingly hold xenogas and data, every capped raid paid out more of the surplus nobody wants. You had no say in it. Worse, loot passes through your warehouse cap when it lands, so if your data silos were already full you were spending hold space on data that was destroyed at the doorstep.

Launching an attack — or opening or joining a raid — now offers Fill the holds with: ferrolite, xenogas or data. When the holds are too small for the whole haul, the resource you name is loaded first, up to whatever the target actually has; whatever hold is left over is split between the other two exactly as before.

• The haul is the same size whatever you pick, and war and season scores read that same total, so no position on any ladder moves. What changes is what those units are — which on a market where ferrolite finds no sellers is the whole point.
• Naming a resource the target does not hold costs you nothing: the order simply does not apply, and you bring home the usual share of everything else.
• Pick one and the other two can come home at zero. That is the trade, and it is stated on the control.

On a joint raid the choice is yours alone. Every ally fills their own holds from their own share of the plunder, under their own order. Whoever opened the raid does not choose for anybody else, and an ally who names nothing brings home exactly what they would have brought home before.

Choosing nothing is still a choice, and it is free: a fleet that names no resource hauls precisely what it hauled yesterday, down to the unit.

Changed

The battle report now says which order produced the mix, so a hold that came home full of one resource explains itself instead of contradicting the plunder rate printed beside it. On a joint raid where allies chose differently, each participant's own card carries their own order.

The Launch again button repeats your loot order along with the rest of the launch, and the order card names it — so the fleet that goes out is the one the card describes.

1.214.0 — 2026-09-21

New

• Choose the quick buttons at the bottom of your phone (#1139). The bar under your thumb had five fixed destinations — Home, Galaxy, Fleets, Chat and Empire — the same for everybody. Two regulars asked for a say within a day of each other and wanted opposite answers: one lives in the Overview, the other wanted Shipyard and Colony one tap away. Now you pick. Open Empire → Settings and the Quick buttons row lists every tab: tap one to give it a slot, tap a chosen one to take the slot back, and the bar redraws underneath you as you go. Keep four, or keep two — the slots share whatever room there is. Empire keeps the last slot, because it is the only way to the tabs the bar is not carrying. Your choice is remembered on that phone.

  Whatever you move off the bar takes its unread marks with it, onto its door inside Empire, so nothing you were relying on goes quiet: your unread chat count, the count of fleets heading for you, and the dot that says a First Steps reward is waiting all follow their tab.

1.213.5 — 2026-09-21

Fixed

• The synthetic empires can finally see the Event Sector, and they are starting to fight over it (#815). The season sector is the arena the whole season is built around, and it was a ghost town: a player reported seeing one other person in it, and at the time there were three colonies in the entire sector, all three human — not one of the roughly 150 machine empires was anywhere in it.

  The reason turned out not to be that they were choosing to stay away. The agents carry their own copy of the galaxy's shape, and that copy stopped one sector short of the Event, so every coordinate in it was thrown away before any of them could think about it. The consequence was quiet and total: the season board — the six worlds that are captured, held and scored every hour — could not be read at all, so the machines' ability to contest seasonal territory, added earlier in September, had in fact never chosen a single target from the day it shipped. Production agrees: no guild had taken a scored world in that window.

  Now the board is readable, and the capture and the twelve-hour watch both run. Seventy-five of the seventy-six machine empires belong to a guild, so the hourly-scored worlds have someone fighting over them instead of sitting neutral until the Eschaton sweeps them.

Also

• Groundwork, with nothing visible attached to it yet: since v1.170.0 a world in the Event Sector costs no permanent colony slot, and the machine side had never been told. It is now, in the four places that count colonies against the ceiling. No empire was actually being mis-counted — the machines cannot found a world there by any route today — so this changes nothing you can see. It is there so that the day they are allowed to settle the sector (#4316), the arithmetic underneath is already right instead of quietly charging them for ground the game gives away.

1.213.4 — 2026-09-21

Fixed

• The alarm board now answers before it explains (#1137). Opening the fleets tab on the alarm sub-tab, you used to read the header, then a long explainer about what an Observatory would buy you, then a second one about the Deep Sensor Grid and cloaked hulls — and only then find out whether anything was actually coming, in the same grey italic as the two caveats above it. A new player has neither building, so they got both explainers and met the board's verdict fourth, dressed exactly like the small print. A board with attacks on it had the same problem: the alarm cards sat underneath the education too.

  Now the answer leads. A quiet board opens with a green bordered banner reading ✓ No inbound attacks detected.; a board under attack opens with the alarm cards. Both explainers are kept word for word and simply follow the answer they qualify — the player who reported this asked for them to stay. The "early warning unavailable" error still takes precedence over everything, and the board's scope note is unchanged.

Changed

• The banner says what was detected, not that you are safe. It deliberately does not read "all clear". An empty board does not mean nothing is coming: a fleet inside a Nightfall dark window, a launch the Veil never stamped, and cloaked hulls below your sensor threshold all produce an empty board while an attack is genuinely on its way. The words stay hedged and the green and the tick carry the reassurance instead — the board only ever tells you what its sensors registered.

1.213.3 — 2026-09-21

Fixed

• Opening the keyboard no longer drags the screen under the Dynamic Island (#766). On an iPhone, tapping into any text field used to slide the whole page up by about the height of the notch: the player card and the resource chips rode up over the status bar clock, and with the keyboard open the first line of the message list was printed underneath the Dynamic Island. It was never a chat problem — it happened on every screen with a text field, chat is just where you always type. Now the band stays where it is, and the chat composer stays above the keys instead of being dragged into view by a scroll that moved everything else with it.

  Two separate things were going wrong at once, which is why four earlier attempts each fixed half of it and were each thrown away. The phone stops reporting the notch at all while the keyboard is up, so the space reserved above the player card collapsed from 59 pixels to 6 — the band jumped under the clock with nothing scrolling anywhere. And the browser scrolls the page itself to reveal whatever you tapped, using a mechanism no page setting can switch off. The game now remembers how big the notch was the last time it could be measured honestly, and declines the scroll on screens that have nothing of their own to scroll.

  Turning the phone is handled too: sideways the notch moves to the side of the screen, so the remembered value is thrown away when the device rotates — including mid-message, with the keyboard still open. Sideways, where there is least room, the message panel now shrinks to whatever space the keyboard leaves instead of insisting on a minimum height that pushed the composer behind the keys.

  ⚠️ Not verified on a device. Nobody in this session had an iPhone or a simulator to hand — this was written and tested on Linux, and the browser does not reproduce the bug. Two things the fix rests on come from the original report's own measurements and could not be re-checked here: that iOS really reports the notch as absent while the keyboard is up, and that the displacement is still a page scroll in the current build rather than something only the system can undo. If the second turns out false, the remaining movement needs a different remedy, and that is written down in the pull request. Everything else — the arithmetic, the rotation handling, the panel sizing, and the fact that nothing changes at all on desktop, on Android, or on a phone with no notch — is covered by tests.

Also

• A player reading the game zoomed in is deliberately left alone: a pinched-in screen shrinks in a way that is indistinguishable from a keyboard opening, so the game does nothing rather than fight the reader's thumb. The original problem is still visible while zoomed.
• The navigation bar still sits behind the keyboard rather than riding above it. Lifting it would spend about a quarter of the strip left above the keys on destinations nobody visits mid-sentence.

1.213.2 — 2026-09-21

Fixed

• Two named places on one star no longer read as a single run-on title (#787). Sector 8 holds both the Exarch Archive and the Exarch Guard, seeded onto the same system. Their names printed side by side on one row, at one size, on one baseline, about 5 px apart — against a word space of under 3 px — so the row read as one title, and the two effect lines under it, in the same dim ink, read as one sentence. Each name now has a row of its own and sits centred over the star it belongs to, instead of being anchored by its edge and appearing to hang between two stars. The block is also half as wide as it was.

Changed

• On a star with two places, the chart no longer prints their effect lines, which is what buys the second name its row. A relic's effect is still in the landmark strip's band row. A bastion has no band row, so every landmark chip's tooltip now names what its place gives — on a pointer device. On a phone, where there is no hover, a bastion's effect is reachable only through the landmark search, which is a real limitation of this change rather than an oversight.

1.213.1 — 2026-09-21

Fixed

• Building on a phone no longer hides the ground you are aiming at (#835). Starting a build put the «drag or tap to aim» card and the «Build here / Cancel» bar at mid-screen, on top of the very tiles you were reaching for, while a wide strip of terrain sat unused below them. Both cards now sit down at the bottom edge while you aim, and the colony they used to cover is back under your thumb. They return to their old place the moment you leave placement.

Also

• The strip the cards moved into is shared with the battle toast and the error banner, and those two do not move out of the way. The new position is chosen so that «Build here» and «Cancel» stay below them: if a battle resolves or the server refuses something while you are aiming, the notice covers instructions you have already read rather than the way out.

1.213.0 — 2026-09-21

New

• The transport form says what the destination colony already holds (#794). Planning a transport from one of your colonies to another, while sitting in the menu of a third, meant bookkeeping across tabs: the cargo boxes showed what the origin has, and nothing anywhere showed what the destination has. Topping up a colony's xenogas meant leaving the form, switching colony, memorising three numbers and coming back to type them against.

  A transport or a redeploy aimed at one of your own colonies now carries one line under the manifest: what that colony holds of each resource, and the warehouse ceiling it counts against. It is there before you type anything, which is the point — it is the reading you opened the form for, not a complaint about a mistake already made. Nothing about it blocks a launch or moves a number: the two existing warnings, for a fleet that cannot lift the load and a warehouse that cannot take it, are unchanged.

  A colony that is not yours prints nothing at all — another player's warehouse is still something you have to send a spy to see. A pickup keeps the figures it already showed.

1.212.0 — 2026-09-21

New

The Shipyard points at another colony — but only when the hulls would reach you sooner (#998)

A long ship order used to book a slip for hours while the sheet said only that it could not be cancelled and that raising this Shipyard would shorten it. If you had another yard standing idle somewhere else, nothing said so.

Now, on an order past eight hours, the sheet can name one:

Vesta (1:4:6) could have them all home in 5h 12m if you fly them back — sooner than building this order here, split or not.

The sentence only appears when it is true, and "true" here had to clear more than it looks. A colony is named only when building the whole order there — exodus shock included — plus flying the hulls back still beats the best you can do without moving by a fifth. That last term is the whole point: hulls are stationed where they are built, the trip home is capped at two hours, and a two-hour flight is bigger than the 20% saving that qualified the colony in the first place. A yard that builds faster can, and often does, still lose.

It also has to own a Shipyard, have a slip free, and be able to pay for the order itself — orders are billed to the colony that places them. Event Sector colonies are never suggested, however good their numbers look: a season ends by carrying your ships home and deleting the colony, and an order still in the slip is not a late delivery, it is resources spent on hulls that never arrive.

And the local side of that comparison includes splitting the order across the free slips you already have. If splitting here is the better move, the sheet stays quiet about moving.

Two smaller things that come with it: the figure is a delivery, not a build time, and the note says so plainly — nothing ships the hulls home by itself, so it reads "if you fly them back" rather than promising an arrival. And when a colony is named, the "raise this Shipyard" advice steps aside: raising the yard you are looking at does nothing for an order you are about to place somewhere else.

This closes the third option from #986, which was built and withdrawn twice during #995 because every version of it compared build times and so gave advice that was not true.

1.211.0 — 2026-09-21

New

The Anchorages door tells you your structures are being taken down (#842).

Before: a guild's anchorages could be notched all night and nothing said so. The alarm existed in exactly one place — a red banner at the top of the Anchorages panel — so you only saw it if you had already opened the door. And the door's own line read territory & modules, which is not a sentence that makes anyone open anything. Nothing else on the tab could stand in: the Dashboard alarm is about your own colony, the inbound alarm is about fleets aimed at you. A guild could lose a structure overnight and learn about it from the rubble.

Now the door speaks before it is opened. It says an anchorage is under siege or 3 anchorages under siege, and when the worst bar is one lost battle from the end it turns red and leads with that: one battle from falling. It costs no extra loading — the count travels on the guild summary the game already refreshes every 30 seconds — and it appears in both places at once, the desktop rail row and the phone command-room card.

The dot is deliberately cyan while a siege is merely running and red only at the brink. A notch fades on its own after a day, the paid repair is locked for twelve hours after the blow, and only the guild leader can spend the Fund on it — so a red alarm at the first notch would be a day-long warning with no button behind it for everyone else.

Fixed

The "about to fall" warning could never appear (#842). Both the Anchorages panel banner and the new door tested for a breach bar at its maximum — but a structure that takes its last notch is destroyed by the same write, so a standing anchorage never reads the maximum. The warning was unreachable: an anchorage one lost battle from rubble was described in exactly the same words as one with a single scratch. Both surfaces now warn one notch earlier, which is the point at which the next defeat actually takes the structure down, and they turn red at the same moment as each other.

1.210.0 — 2026-09-21

New

• The Colony tab now says what each building is for (#1136). The list gave you a name, a level, a price and the next level's power draw, and nothing else — so the only way to find out what a building actually did was to leave the tab and open it on the Base map. Every row now carries the building's description, the same text the Building Sheet shows, so the two can never say different things. Long ones are capped at two lines to keep a 27-row list scannable; the whole sentence is still there on hover and for a screen reader, because the cap is visual and the text is never shortened behind your back.

1.209.11 — 2026-09-21

Fixed

• The Colonization card now says what the Command Center level would buy you (#868). Opening Colonization in the Research tab, the Effect row read "Up to 3 colonies → Also needs Command Center lvl 12". The second half named a requirement and nothing else: no line on screen said what that Command Center level, together with this research, would actually grant, so the word "Also" pointed back at a benefit the card had never mentioned. It now reads "Up to 3 colonies → Up to 4 colonies — also needs Command Center lvl 12". The colony you would gain is named alongside the condition for gaining it, and the promise stays visibly conditional — "also" still ties it to researching the level, and the chip stays gold, the colour this card uses for a constraint, so it is never mistaken for a reward already earned.

  The number is the real one: it is only shown in the state where both gates sit at the same ceiling, so the colony promised is exactly the one the named Command Center level unlocks. The wording used when the Command Center is the only thing holding you back, and the wording past the tenth colony where no Command Center goes higher, are unchanged.

1.209.10 — 2026-09-21

Fixed

• The pirate beacon now sits inside its own hexagon (#3542). On the galaxy map the gold beacon mark was drawn half outside the cell it was naming — it hung over the rim onto the neighbouring hex, and there was no zoom you could reach where it fitted. Since the new galaxy field leaves a system's centre cell deliberately empty, the mark was pointing at a void and reading as detached from the system it belonged to. It is now centred on the cell it names.

  The nudge that pushed it off centre exists to stop two signs on one star printing on the same pixel, so it is kept exactly where it is still needed: on your home system, on a system you hold that also carries a pirate colony, and on a system carrying a relic or a bastion, the beacon still steps aside — otherwise the gold mark and the beacon would sit on top of each other, and the one underneath would simply vanish. On those systems the mark is where it was and still overlaps the rim.

  One side effect in your favour: where a system carries both a beacon and an anomaly signal, the two are now further apart than before, not closer.

1.209.9 — 2026-09-21

Fixed

• Placing a building on a gas planet no longer shows you two grids (#1106). The floating deck's artwork has a panel grid painted into it, with its own spacing and angle, and the game drew the real placement grid on top of it. Only one of the two was the squares a tap actually selects, and it was not the one that stood out: measured on the shipped art, the placement grid came out brighter than the deck on about a third of it and darker on the rest, so it read as more texture rather than as a grid. Now the ground fades back while you are placing and the grid comes forward, on every planet — the deck, the rock basin, the ice sheet and the lava flats all read more clearly than the clearest of them did before. The artwork returns the moment you stop placing.
• The cross that marks a square you cannot build on is legible again on the dimmed ground (#1106). It is the one "no" that does not depend on telling red from cyan, and against the new darker ground it would have faded, so it is now drawn light instead of dark.

1.209.8 — 2026-09-20

Fixed

• A guild welcome that arrives late now sounds late (#1042). An officer greeting you six hours after you joined used to write as though you had walked in that second — "just saw you come in", at teatime, to someone admitted before breakfast. They now write like a person who has just caught up on the channel.
• A busy guild no longer loses the new arrival (#1042). Every contribution to the guild Fund writes a line in the register, and in an active guild a new member's arrival could be pushed out of sight within hours — after which nobody greeted them at all, precisely in the guilds where it matters most. Arrivals are now read separately from the day's traffic.
• The officer who sleeps longest can still get there (#1042). The window in which a new member counts as "just arrived" was shorter than the longest stretch an empire is offline, so the greeting could expire before the officer who owed it ever woke up. It now outlasts any sleep.
• An admission is announced (#1042). In one configuration the line saying a player had been accepted was silently discarded before it reached guild chat whenever their name was five characters or shorter — the guild admitted somebody and said nothing at all.
• The guild registry no longer errors on an odd URL (#1042). A duplicated filter parameter made the register page fail outright instead of answering.

Also

• Part of #1042 only. The largest fault the issue records — a shy officer's missed greeting being lost for good, about one arrival in three — is not fixed here and the issue stays open. A queue of backup officers was written and withdrawn before merge: the only check that stops two officers greeting the same person can see two hours back, while the backup would have spoken at three, so it produced a second and a third welcome instead of recovering the missing one. Measured rather than guessed, and recorded on the issue for whoever picks it up.

1.209.7 — 2026-09-20

Fixed

Nothing changes on screen in this one — it repairs a guard that had quietly stopped guarding, and writes down two things it found about the battle replay that are still true in the game today.

The replay has a bench whose job is to prove that no fleet group ever disappears behind another one on screen. It was measuring the wrong ships. Back when the replay drew flat cut-out artwork, it worked out each group's size from the artwork; when the fleet became real 3D models, the scene started measuring the models instead, and the bench never followed. The two disagree by as much as 64%, so for months the bench was certifying a battle line nobody was being shown, and passing every time. It now reads the actual shipped models, and the static defences — the Emplacements and the Fortress — are on the bench for the first time, as is the old flat layout that reports from before formations still use (#1007).

Measuring the real thing turned up two ways a group really can be hidden, both of which the replay has been doing all along:

• The Fortress hides whatever stands next to it. It is a huge cube, about 52 units wide on screen, and it is given a 41-unit slot. The Emplacements are always placed in the same rank as the Fortress, so a colony that has built both draws one inside the other. The Fortress on its own is fine.
• A capital ship beside a light one crowds it out. Hulls lie lengthways across the line while the slots are spaced towards the viewer, and the camera compresses that direction — so a Titan is drawn wider than the gap it was given.

Both are measured, pinned so they cannot quietly get worse, and filed as #4289. Neither is fixed here: the repair means changing how much room the line gives a big hull, which pulls the camera back and shrinks every ship on screen, and that is a deliberate decision about how battles look rather than a number to nudge.

One correction worth recording, since it was wrong in the first draft of this work: the Fortress does not overlap the ranks above and below it. Those clear each other with room to spare. The problem is sideways, inside a single rank.

1.209.6 — 2026-09-20

Fixed

• A fleet inbound no longer shoves the resource counters sideways (#988). The red "under attack" chip used to be the first thing on the HUD strip, ahead of Ferrolite, Xenogas, Data and Energy. On a desktop screen that strip wraps, so the moment an attack started, every counter slid across — and on a laptop the last one dropped to a second line. It did not settle there either: the chip's width changes throughout a flight as the countdown sheds a digit and "+3 more" counts down to nothing, so the figures kept drifting for as long as the fleet was in the air, which can be hours. The alarm now has a place of its own on the strip — after the counters, before the Build and Ships chips — and the counters stay exactly where they were, attack or no attack.

  On a phone nothing changes: there the strip scrolls sideways under your thumb instead of wrapping, and the alarm stays pinned first so it can never be scrolled out of reach.

Also

• The HUD band still grows taller while an attack is in flight. That is the chip needing room, not the order it sits in, and no arrangement of the strip avoids it — moving the alarm changes which row breaks, not whether one does. Said plainly here because the original report expected the reordering to reclaim that space, and it does not.

1.209.5 — 2026-09-20

Fixed

• The mission form now names where the fleet is going, not just where it leaves from (#1020). On a desktop screen the departure chips sat under a heading — LAUNCH FROM — and the destination, the coordinate fields and everything under them, had no heading at all. If you were composing your first launch, you read one named section and then, as the player who reported it put it, "everything else": nothing on the screen said those three numbers were where the fleet was going. That block is now titled GOING TO, the same words the phone has always used for the same step, so the two ends of the route read as a pair. It appears for a one-colony empire too, which is the case that needed it most — LAUNCH FROM only shows up once you have a second colony to choose between, so a new commander previously had no headings at all.

  The verb chips at the top of the form are still untitled; this change names the destination, not every section.

1.209.4 — 2026-09-20

Fixed

• The galaxy map tells you where you are looking, at the zoom where you are looking (#1035). On the galaxy map, at a zoom that still showed the whole sector, nothing on the chart said where anything was — and resting the pointer on a system said nothing either. To learn a system's address you had to zoom in until the sector no longer fit on screen. It was worst over neutral space: an uncolonised system has no guild colour and no colony count, so there was nothing at all to identify it by, and neutral space is exactly what you are reading the map for when you are looking for somewhere to settle or a route to take.

  Now the pointer names the system it is over at that zoom — System 9:14, then what is there: the colonies, who controls it, its beacon and its anomalies, or plainly no colonies where nobody has settled. A ring marks the system being named, so the address belongs to a visible place and not to a guess. It is the same system a click there flies you to, so the map now says where you are about to go before you spend the zoom finding out.

  Nothing new is drawn on the chart itself: the map stays as uncluttered as it was.

  Two things this does not do. It is the pointer's answer, so it is a desktop one — on a phone a system still gives up its address only when you zoom in far enough for the star to become a door. And where one of your fleets is crossing the chart, its mark keeps that patch of sky to itself: the click there opens the fleet, so the map stays quiet rather than promise you a system.

1.209.3 — 2026-09-20

Fixed

• The Warehouse's advice no longer mentions a level on a building you have not built yet (#725). The build advisor's storage card explains that a Warehouse pays for itself by catching what your storage is about to throw away. That sentence used to open "This level catches +375 Fe/h…" — but on a Warehouse you do not own yet, the card's own row reads new and shows no level at all, so the advice referred to a number that was nowhere on the screen. It now reads "The extra storage catches +375 Fe/h…", which says the same thing and reads the same whether the card is offering you a first Warehouse or a fifth level of one. The rest of the card is unchanged: the warning that storage fills up, the figures, and the reminder that the Warehouse itself produces nothing — it only rescues what would be lost.

  The same wording now covers the Singularity Vault, which earns its keep the same way.

  This also matters for the advisor's written commentary: those sentences are the facts the Advisor is given before it puts the advice in its own words, and it is told never to invent a level. Handing it a level that the card does not show was how one could reappear in the prose.

Also

• The same "this level" wording still appears on three other advisor cards — the defence Hardening line, the Laboratory's research-collaboration line and the Trade Portal's intake line — and is not fixed by this release. Each is a separate sentence whose wording is argued on its own, and they are tracked with the cards that reproduce them in #4275. The new test that guards the Warehouse against this covers the Warehouse only.

1.209.2 — 2026-09-20

Fixed

• The Power tooltip now says what the fairness Gate is for (#1103). Hovering Power in the header used to explain how the Gate had been rounded — "the larger of the two before rounding, the server's own number, truncated like everything else on screen" — and never what it does. For an empire whose buildings outweigh its fleet it ended on a number the same sentence had already printed, so it read as repeating itself; the player who reported it concluded Power was only a sum. It now reads: the Gate is the higher of Structural and Military, it is the yardstick the server prices disproportion with (the bounty for hitting someone far below you, and the band colours on the leaderboard and in the Galaxy), and you are sized by your stronger side, so building up the smaller half changes nothing until it overtakes the larger. The rounding note stays in the Power details dialog, the one place the Gate sits directly under the two rows it can read a point below.

Changed

• The word "Gate" now comes before the number, not after it (#1103). A second player filed the header POWER figure against his own leaderboard row as a bug — 934,921 against 580,804 — because the row's "Gate" label sat after the figure and his phone crop cut it off. On the phone standings row the number carried no label at all. Both now lead with the word, so no crop can separate a figure from what it is called.

1.209.1 — 2026-09-20

Fixed

• A raid coming at you is one line in your diary now, not one per attacker (#991). A guild raid wrote a separate «Incoming raid» row for every contingent that joined it, and because the whole raid shares one arrival time, every row said exactly the same thing — six attackers, six identical red lines, differing only in the minute stamped in the corner, and the rest of your evening pushed off the screen. The warnings that belong to one raid now gather into a single row: «Incoming raid — 3 fleets sighted · 1:2:3: 3 hostile fleets were sighted converging on one arrival, arriving in 30 min. They came onto your sensors over 18m.» Tapping it opens the individual warnings exactly as they were, each with its own stamp, so you can still see how the raid assembled and whether more are still trickling in — which the single line also tells you outright, since that is the thing those stamps were being read for.
• The row says sighted, and it means it: a fleet flying the Nightfall Protocol is invisible to you until it is nearly on top of you, and it writes no warning at all until then. So the number is what your sensors have picked up, never a promise about how big the raid really is. A raid can be bigger than the row says; it can never be smaller.
• A bounty payout of small change stops sounding like an alarm (#991). An attacker who is beaten off still breaks an emplacement on the way down, and a broken emplacement is a share of the damage, and a share of the damage is a share of the bounty on your head. Forty cents of Dark Matter would leave your balance and your diary answered with the same red «Bounty collected on you» it prints when a grade-4 commander is brought in. Under one Dark Matter it now reads as what it is — «A bounty grazed you» — in the quiet grey the diary already uses for things that are not worth getting up for.
• The row is still written down, and that is deliberate. It is the only place you can see that the Dark Matter left at all, and enough grazes add up to real money gone; what changed is the volume, not the record. The attacker's own side is untouched: they collected, however little, and their row still says so in gold.

Also

• Left out, so it does not surprise you: warnings from one raid only gather when there are at least three of them. Two red rows is not a flood, and folding two cards into one you have to open saves nothing.
• A raid that assembles across midnight is still two rows, one under each day.

1.209.0 — 2026-09-20

New

Novarchy can now revoke your Apple sign-in when you delete your account (#753). Deleting an empire has always removed the account here; what it never did was tell Apple, so the game went on being listed under "Sign in with Apple" on the Apple ID of somebody who had left. The server can now hand that token back to Apple, on the player's own deletion and on removal by moderation alike.

It is built and switched off. Turning it on needs an Apple signing key that has to be minted in Apple's developer portal and installed on the server — until that happens, nothing about deleting an account changes in any way, and the server says which state it is in every time it starts.

Deleting is unchanged in every way that touches you regardless: same confirmation, same speed, and it still works when Apple cannot be reached. The account goes either way.

Also

What the sentence above does not cover, so nobody has to go and check:

• an account becomes revocable from the next time you sign in with Apple, never retroactively. Apple hands over what is needed only at sign-in, and it was not being kept before now — no backfill can invent it for accounts that already exist;
• on iPhone that next sign-in has to happen on a new App Store build, so this reaches the app when that build does;
• accounts reclaimed by the inactivity sweeps are not revoked. Apple asks for this when a person deletes their own account, and a swept account is one nobody asked to delete — but it is the larger number of the two, and the grant does outlive those accounts.

Nothing here changes anything for players who sign in with Google or play as a guest.

1.208.8 — 2026-09-20

Fixed

• The agents' engine mirrors can no longer go stale behind a green test suite (#811). The synthetic empires are written in Python and cannot read the TypeScript engine, so a dozen test suites hand-copy its numbers — ship costs, hull classes, research rates, siege tables, galaxy size — and check them against the real engine files as they run. Those checks used to switch themselves off whenever a source file was renamed: the rename left the engine perfectly green, while the Python side quietly reported "skipped" in a run of several thousand tests that nobody reads line by line. From that moment the agents could have been planning, bidding and fighting on prices the game no longer charges, with every suite passing. Nothing a player sees changes today; what changes is that this can no longer happen unnoticed tomorrow — a moved engine file now fails loudly and names itself.

Also

• The one test that runs the real Hypergate engine code and compares it against the agents' copy used to be skipped when the tooling it needs was absent — which the automated checks had no guarantee of providing. It now fails instead, so the Hypergate core prices the agents quote in chat stay tied to the prices the server actually charges.

1.208.7 — 2026-09-20

Also

• The v1.208.5 entry now says what the chrono-shard roll-up left undone (#4267). Nothing changes in the game. That entry said a burst of shard spends reads as one line in Activity, which is true — but the number on the filter chips still counts the spends rather than the lines, so «Mine» reads a few higher than the rows you can count beneath it, and the entry gave no hint of it. A reader who counted would have found the game disagreeing with its own changelog. The entry now names the discrepancy in one line and points at the issue tracking it; the code that causes it names that issue too, so the two can be found from each other. The defect itself is still open: it is fixed by counting rows instead of entries, which touches every chip and was judged too wide for a presentation-only change.

1.208.6 — 2026-09-20

Fixed

• The Shipyard no longer dresses the Event Relics as ships you can order (#1026). The catalogue holds three kinds of hull: the ones you can build, the ones a research node will open for you, and the two Event Relics — the Destroyer and the Annihilator — that nothing will ever open. It had only two looks for them. A research-locked hull showed a padlock and dimmed, greyed art; a relic showed bright art and a plain name, exactly like the buildable row above it, with only a caption line saying she was not buildable. On the narrow-desktop filmstrip, which has no caption line at all, she said nothing whatsoever. Players read the row the way it was drawn and went looking for the order button.

  Each relic now carries ✦ in cyan in front of her name — the same slot the padlock sits in — on the desktop rail, on the filmstrip card and in the phone's All-hulls ledger. ✦ is the glyph this game already uses for an Event, and the one the BUILD console's refusal box already showed for these same hulls; that box's hand-written copy now reads the same shared constant as the rows, so the two cannot drift apart. Her art stays bright: she is a showcase, not something you are missing. She deliberately does not get the padlock, which would promise research that never arrives.

  Not done: the reporter suggested a gold name. Gold is the Legendary tint in that same grid — worn by the six apex hulls, all of them buyable — while the relics are Mythic, so a gold name would have named the wrong step.

1.208.5 — 2026-09-20

Changed

• A burst of chrono shards is one line in Activity now, and it adds itself up (#1029). Spending four fragments on one research order wrote four rows — «Chrono shard spent — 5 minutes cut off the research queue», then again, then again — all stamped the same minute, pushing the rest of your night off the screen. The number you actually wanted, what the burst bought you, was the one thing not on it: you had to add the rows up yourself. Consecutive spends on the same order now collapse into a single row that says what you spent and what it bought — «Propulsion: 3 × 5m + 1 × 30m · 45m cut off the research queue» — and tapping it opens the individual cards, unchanged. Runs of one or two are left alone, since folding two cards into one you have to open saves nothing.
• Two shards spent on different orders stay two rows. The roll-up gathers only spends it can prove belong together, so it never presents work on two colonies as a single act — and spends recorded before the game started naming what each shard hurried are left as they are, rather than guessed at.
• Left out, so it does not surprise you: the number on the filter chips still counts the spends, not the lines. A burst folded into one row makes «Mine» read a few higher than the rows you can count under it. Tracked in #4267.

1.208.4 — 2026-09-20

Fixed

• The colonize checklist no longer marks a satisfied requirement as if it had failed (#1067). Tapping a free planet lists three requirements under the Colonize button. A commander who met all three still saw only one of them ticked: the Colony Ship row carried a check, while «Colonization at least 3 — yours is 4» and «Command Center at least lvl 12 — yours is 12» carried the same neutral dot that a requirement nobody has read yet gets. Two sentences that visibly answer themselves, marked like refusals.

  Those rows now carry a check too, drawn in grey rather than the green of the row the panel knows outright, with one line under the list explaining the difference: the caps are checked against what this panel can see, and a colonization already on its way is counted only when you launch. That distinction is why the grey exists — the panel cannot see missions already in flight, and a green check would promise something the launch screen could refuse a second later. Nothing about the refusals changed: a requirement that really blocks is still a red cross, the Colonize button greys and un-greys exactly as before, and a row whose numbers have not arrived yet still shows the plain dot. The same rule reaches the Event Sector, where room left under the season ceiling now reads as satisfied.

1.208.3 — 2026-09-20

Fixed

• The refit desk lists each slot's options cheapest first (#1027). Opening a slot at the campaign refit desk used to show its three standing orders in the order the catalogue happens to be written in, which is design order and not price order: the battery menu of every warship read 5, 4, 11 Coils and the core menu read 6, 5, 4 Cores — descending. The support line was shuffled the same way, its bay at 5, 8, 4 and its core at 4, 3, 6. A player at the desk is asking what they can afford with the parts they hold, and a shuffled column made them read all three cards and compare by hand to answer it. Every picker now paints its options in ascending price: the battery menu reads 4, 5, 11 and the core menu 4, 5, 6, the support bay 4, 5, 8 and the support core 3, 4, 6. Same options, same prices, same tradeoffs and the same refunds — only the order changed.

1.208.2 — 2026-09-20

Also

• Nothing changes on screen in this release. It records why the mission composer's note heights are not reserved (#4166). A review of #4160 named two layout costs of the pact-blocked probe note: on desktop, one visit to a refused coordinate permanently raises the slot card's floor, so blank space sits under every later short slot for the rest of the composer session; on a phone, the note arrives after the commit screen has painted and pushes the fleet picker down under a finger already moving toward it. Both were reproduced. Neither is fixed, because the measurement said every available fix costs the player more than the defect does: the note is about 90px in the desktop card, and reserving that space would charge it to every player who never probes a friendly world, while the floor as it stands charges it only to those who did. On a phone the common jump is 36-53px and is driven by the world note and the facts line reflowing — not by the note this issue named — and reserving the whole column would cost 53px of dead space on the game's most cramped screen, or 189px if all five notes were reserved. The measurements, the two rules they rest on and the six options that were rejected are now written beside the two mechanisms, so the next review pass does not re-file them. Not done: the one remedy that is cheap at both costs is a shorter note, which rewrites a sentence two issues reviewed, and it is filed as #4248.

1.208.1 — 2026-09-20

Changed

• The Wraith empires now work their Pirate Beacon contract instead of letting it expire (#2955). The thirty Wraiths were dealt "Strike the Pirate Beacon" every day and had never once completed it — 273 cards over a fortnight, none claimed, while people claimed 41% of theirs. Their autopilot simply had no way to fly the mission, so the card sat at zero and expired at every daily reset. It now gets worked: a Wraith raids the Pirate Beacon over its own home system while that day's card is open, and stops as soon as the card is done.

  This is not the Wraiths turning aggressive. They strike one NPC tutorial fixture, over their own mother system, and nothing else — they still attack players only when their operator orders it, exactly as before.

  The raids are spread through the day rather than fired in waves. Each empire works to a schedule of its own, so the galactic register carries the odd pirate raid from one Wraith or another rather than thirty at once, four times a day.

Also

• The knock-on is that a Wraith's day can now survive one bad card. Their daily board asks them to finish four of five, they could finish exactly four, and one of those four depends on somebody having left wreckage over one of their worlds — which happens on a little under half their days. With a fifth card they can work, the board no longer needs a perfect day to pay out.

• ⚠️ This one is not live at this release, and will not be until the machine that runs the Wraith schedule picks the change up. The Wraith autopilot is not part of the deployed service — it runs from an operator's own checkout on its own timer — so unlike everything else in this release, deploying it does not switch it on. It takes effect when that checkout updates, the same way the Wraith state-lock fix did. The production smoke check for this release could not observe it and did not try to (#4231).

1.208.0 — 2026-09-20

New

• Raid a synthetic empire after dark and the takings may already be gone (#4180). An AI empire that has bought everything its lanes and its Command Centre will take used to go to bed on a full warehouse, wherever you hit it. Now, in the last hours of its own evening, it loads what it cannot spend onto freighters and sends it to the colony behind its best wall. Hit the world it emptied and you find its working capital and little else; hit the bastion and the night's haul is a harder fight rather than a free one.

  It only does this when the move actually helps. It will not send a pile to a colony whose wall is no better than the one it left, and it will not build a prize bigger than the one it just cleared — so three worlds' worth of stock never quietly becomes one very rich target. When it cannot promise that, it refuses and the crates stay where they are, which is most evenings.

  The hour is not a timetable: each commander draws its own, a different one every night, and some nights it simply does not get round to it.

Also

• Selling the surplus on the market, listed alongside this in the original issue, is deliberately not part of it: an offer's goods stay in the seller's warehouse until someone buys, so publishing one would have moved nothing out of a raider's reach.

1.207.0 — 2026-09-20

New

• An unpaid season placement can now be settled from the Admin dashboard's API (#2767). When a guild season closes, each guild is paid its placement reward. A payout that fails is already contained to the one guild it hit — everyone below it is still paid — but putting that one guild right afterwards was a hand-written database job with no tooling behind it. There are now two admin-only calls: one lists the placements a close froze but never paid, and one pays them. Running the settlement twice is safe and pays nobody twice.

  This is an operator action, invoked by hand — nothing settles itself, and nothing about it runs when a player opens the season panel. No season has ever closed on the live server, so nothing was repaired for anyone: this is the tooling that will be wanted the first time a close goes wrong.

  It settles Scrip, Dark Matter, guild XP and the placement bookkeeping, and deliberately awards no achievements. "Eschaton completed" and "Keeper of the Relic" are verdicts about the instant the season ended, and neither can be recovered afterwards — the Relic's holder may have changed hands since — so the settlement pays what it can prove and does not invent history.

Fixed

Both of these are on the same season-close failure path as the entry above, which has never run on the live server. Nothing that happened to anyone is being corrected here.

• A failed placement now says what kind of failure it was (#2767). The record written when a guild's reward cannot be paid stored only a database error code, so a plain programming error and a momentary lock timeout looked identical to whoever had to decide whether simply running it again would work. It now records the kind of error alongside the code.

• A bad season rollover can no longer flood the server log (#2767). The failure message had lost its length and line-break trimming, so several guilds failing at once wrote several unbounded, multi-line blocks into the log — on precisely the day someone needs to read it. One failure is one line again.

Also

• The runbook in deploy/DEPLOY.md gains the procedure: how to tell a guild is owed money, what to run, what a clean result looks like, why re-running is safe, and what the settlement deliberately does not do.
• Two refusals from the new calls (season_not_found, season_not_closed) now read as sentences instead of appearing on screen as raw internal codes.

1.206.0 — 2026-09-20

New

• The Overview says what level a building is in every colony (#3141). The game could already tell you how many of each ship you had in each colony, and had no way to ask the same question about buildings: every screen answered one colony at a time, so checking that all your Hypergates were maxed meant opening each colony in turn and holding the numbers in your head. The Overview now ends with a Building Ledger — a row for every building you hold, a column for every colony, and the level each one has it at. The last column is the lowest level any colony has that building at, which is the number that says whether your Command Centers are level with each other, and whether a building is finished everywhere or only where you remember it.

  It tells apart three things that would otherwise look alike. A colony that simply has not built something shows a dash. A colony whose Command Center is too low to allow it at all — the building itself, or a further Site of it — shows the level that would unlock it, like «CC 5», and is left out of the lowest, so a young colony cannot make a finished row look ragged. And a building that cannot go any higher where it stands is marked: max when it has reached the game's ceiling, cap when it is only waiting on that colony's Command Center.

  It appears once you have two colonies, sits below your fleets in flight so it never pushes a recall off the screen, and on a phone it starts folded away under its title.

  Suggested by arcticpaladin (in-game QQQQ), who did more than ask for it: asked twice whether he meant the espionage view or a sort, he corrected both guesses and named the axis exactly — the transpose of the fleet summary table — which is the specification this was built from.

1.205.1 — 2026-09-20

Fixed

• Contract rewards now say whether the figure is what the card is worth or what the button will pay (#3896). The reward on a contract slip has always meant two different things: while the card is still in progress it is what the contract is worth if you finish it properly, and once the card is ready to claim it is what pressing Claim actually hands you. On the battle weekly those two can be wildly apart — a large empire could read about 78M beside a line saying it had destroyed 40K of the 157M full pay asks for, then watch the same figure become about 30K the moment the card completed, with nothing on screen having changed. The figures now carry a word above them: WORTH while the card is in progress, PAYS once it is claimable. No reward changed size — this is only the board saying which of the two numbers you are looking at, on every card, without having to hover something a phone cannot hover.

1.205.0 — 2026-09-20

New

• The system survey can now probe the slots whose emptiness is in doubt (#3548). The survey panel offered only the slots that already showed a colony — which are exactly the slots whose answer is already on your screen. In a system of eighteen it would say "5 targets", and the thirteen you actually wanted to check were not on the list. Since a free planet and a colony under the Eschaton Veil read the same on your scan, a probe is the only thing that settles the difference, and the one screen built to send probes in bulk left those slots out. The panel now offers them too, in a group of their own headed "No colony on your scan", so checking a system before you settle it is one confirm instead of eighteen trips through the planet sheet.

  It can cost you more probes than before, and the panel is built so you choose that rather than discover it. The two groups are kept apart — "Colonies on your scan" and "No colony on your scan" — and each has its own "All" carrying its own count, because one button over the whole list would have turned a tap that spent 5 probes into a tap that spends 17. Sweeping a whole system is therefore two taps, not one. The Send button now states the total it is about to spend, and each group says how many of its own rows you have ticked, so a sweep you left half-finished cannot ride along unnoticed on your next tap.

  Every row also carries the age of your own last probe on that slot, so the screen that spends probes no longer sells you an answer you already hold. It only tells you: an already-probed slot is still offered and still priced the same, because a colony can be settled under the Veil after your probe lands. Be aware of one limit, because it is real and not yet fixed: the record of a probe that found nobody fades out of your activity history far sooner than the record of one that found a commander, so a free slot you probed can lose its date after a busy day and look unprobed again. That gap is tracked separately (#4233).

  Nothing about this tells you which of the empty-looking slots hides a veiled colony: every slot your scan reads as empty is offered, all of them look the same, cost the same and are listed in slot order.

  Reported by two players on Discord, one after the other in the same exchange. The report does not record their handles, so the credit is suspended here in the same shape used for the elis_aigen entry further down this file, and can be filled in later.

  Not included: no new screen for reading the results back. The activity feed already isolates them — Combat, then Espionage, which lists your own probe reports including the ones that came up empty. That the filter is hard to find is tracked separately (#4222). Also not fixed: on a narrow phone, ticking an empty-reading row shortens its label to make room for the travel time.

1.204.1 — 2026-09-20

Changed

• Type a big number into a ship box and it now shows you what the colony actually holds (#3113). Spacca-Mazza asked for the thing the market already does: type something absurd into a field to find out the most you can have. The launch pad never sent more ships than you own — but the box stopped telling you so. With 300 Corvettes, typing 999999 gave you 300 for one keystroke and then 3009, 30099, 300999, over a fleet that was still 300. On a colony holding just one of a hull it never budged at all: the box read 999999 next to a fleet of 1. The number on screen was wrong, and the only way to find out was to tap somewhere else.

  Now the box snaps to your pile as you type, and stays there: 999999 Corvettes on a pile of 300 reads 300, with 300 of 300 above it. The ceiling is readable by typing at it, with no label needed, which is the whole point of the gesture. A number below what you hold is left exactly as you typed it, an empty box is still allowed while you are mid-edit, and you can still type a smaller number afterwards. A hull you own none of has no box at all, as before.

  One consequence worth knowing: after an overshoot the cursor sits after the snapped number, so correcting a typo edits the ceiling rather than the digits you meant. Reaching for 43 on a pile of 300 by typing 433 leaves you at 300, and one backspace gives 30, not 43. That is what the market field has always done, and it is the idiom that was asked for — the snap is visible the moment it happens, so the number is on screen to be read before anything launches.

  Not included: the second half of the request, that the ceiling take your other ships into account as you add them. For a ship count there is nothing to take into account — a hull's ceiling is that hull's own pile, and adding a warship does not give you more Freighters. The ceiling that does depend on the rest of your fleet is the cargo hold, and it already updates as you add ships (1 Freighter reads hold capacity 2,500; all 12 read 30,000).

  Nothing changed in the cargo boxes. What you write into the manifest is still carried exactly as written, and the launch door is still what tells you it is too much — the decision from #1019, deliberately left alone.

  Reported by spaccamazza (in-game Spacca-Mazza).

1.204.0 — 2026-09-20

New

• A supply lane's row now names the colony it arrives at, not just the one it leaves. Every row used to quote what the feeder's Trade Portal can move — and for anyone routing several colonies into one hub, that was the wrong half: the hub's own Portal is usually what decides what gets through, and no screen said so. The row now prints whichever of the two is smaller and names the end responsible, so you can tell which Portal to level. A hub with no Trade Portal is described as having none, rather than credited with one it cannot build yet (#3592).

Changed

• ⚠️ Hubs now really do accept only what their Trade Portal allows, and several of them were taking more. The limit has been printed on the rules card, the Trade Portal's sheet, the Advisor and the lane form all along, but the game was not enforcing it: lanes each run on their own hourly clock, and a colony's intake was only shared between the lanes that happened to fall due in the same minute. Two feeders onto one hub therefore almost never met, and each took a full hour of intake.

  If you feed one hub from several colonies, less will arrive there from now on — the hub's total settles at the figure its Portal names instead of a multiple of it. The player who reported this runs five feeders into a level-13 Portal: his hub goes from roughly 27,000/h to the 10,771/h that Portal actually accepts. An outpost with no Portal of its own receives at the level-1 floor, as the rules have always said.

  Whether a hub should have to out-level the sum of its feeders is a balance question this release deliberately does not answer.

Fixed

• A lane row could briefly read NaN/h while a new version was being deployed. It now falls back to the figure it showed before, which is true, instead of showing you a broken number.
• A lane whose destination you had just lost blamed that colony's Trade Portal for holding it back. It now reports what the lane carries and blames nothing.
• A hub fed by a single lane printed the same figure twice and described its hour as "shared by every lane arriving there" when only one did.

Also

• Not addressed here: a colony's outgoing Portal has the same clock-alignment gap the receiving side just lost — two lanes leaving one colony on staggered clocks still move two Portal-hours in an hour. It is measured and written down in the code, and deserves its own issue; closing it is a separate throughput change affecting a different set of empires.

1.203.6 — 2026-09-20

Nothing on any screen changes here, and nothing about the game's rules changes: the inactive-empire decay still halts exactly what it halted yesterday and takes exactly what it took. What changes is that the game can now be held to account for it.

Changed

• A ruin round now records what it took (#2758). When an empire has been silent long enough, the lifecycle halts its production and lets its garrison rot. Until now that left no trace anybody could read afterwards: the stamp it writes is a running total of what has been billed, overwritten every round, and it deliberately survives the player's return — so "did the decay ever take ships from somebody who had just come back?" had no answer at all. Not a stale answer: none. Every round that actually takes something now leaves one durable row behind, naming the empire, when the charge landed, and the presence clock the decision was made against. The question is now a single query, documented in deploy/DEPLOY.md, and a healthy game answers it with no rows.
• The operator's per-round log line now counts only what was actually written. A colony whose write lost a race changed nothing, and was previously still counted as halted.

Also

• Rows from before this release do not exist and were not invented: what the decay did in the past was never observed at this resolution, and guessing it would have been worse than the silence.
• The record names an id, four clocks and two counts — no email, no nickname, no fleet — and is not linked to the account, so it survives the later removal of the empire it describes.

1.203.5 — 2026-09-20

Fixed

• The Empire slot no longer shows a mark you cannot find behind it (#4205). With unread battle reports and a detected probe, the number on the Empire slot was one higher than anything behind the door: two unread reports and one probe read «3», and opening it showed 2 on Activity and nothing anywhere else. The extra unit could not be found and no gesture cleared it. The slot now matches what the doors actually draw — the probe mark is still a dot that stands down while a battle count is lit, exactly as before, so the slot simply stopped disagreeing with it.

• The away-recap no longer points detected probes at a filter that shows nothing (#4205). Coming back after a few hours, the recap listed «Spy probe detected» under a heading called «Espionage» — and tapping through to Activity landed on an Espionage filter reading 0, because probes flown at you are filed under Defence. The section is now called «Probes detected», which is what the rows say and points at no filter at all.

Also

• The code that decides when the probe mark lights up now says, where the choice is made, that the very first visit after the release is deliberately silent: the mark is not born lit on scans from weeks ago. It also names the one case nothing covers — a new device or cleared site data, where the away-recap does not open either, and the Activity feed under Defence is the only record.
• Guards for the probe mark that the previous release was believed to have, and did not: the address the client polls is now pinned on both sides so a rename cannot silently switch the mark off for everybody; the «already read» stamp is now proven to be per-account, so two players sharing a browser cannot inherit each other's; and the code that drives the mark is now exercised by tests at all.
• A test guarding who a battle report is answered for was checking for something that never matches, so it would not have noticed the report route being changed to take the player from the web address instead of from the login. It was already like that before this release; it is fixed here because it is the same defect as one of the guards above, on the route next to it.

1.203.4 — 2026-09-20

Fixed

• The guild panel no longer fails when somebody leaves the guild at the same moment (#4162). Opening the guild panel is what pays out a Council Directive tier the guild has earned. If a member left the guild in the same instant, the two operations could end up waiting on each other, the database broke the tie, and the panel was the side that lost: whoever opened it got an internal error, and the tier stayed unpaid until somebody opened the panel again. The payout is now re-attempted instead of failing, and it is still paid exactly once. The retry is deliberately short-lived — it gives up rather than keep a page waiting that the browser has already abandoned, in which case the panel behaves exactly as it did before.

Also

• The issue this fixed was filed on a premise that turned out to be half wrong: it named account deletion as a second path with the same clash, and reading the code showed deletion already does it in the safe order. Nothing about deletion changed, and the note explaining why now sits next to the rule so the question is not re-opened. The internal check that guards this rule was taught the difference, and one claim it was making about itself — that it proved a safeguard was live, when it only proved a line of code was present — was corrected rather than left standing.

1.203.3 — 2026-09-20

Fixed

• A synthetic empire now talks about the colony it is actually flying to. When a guildmate was under attack at more than one of their colonies, the line the agent reads before speaking named the soonest attack anywhere in that empire, while its fleet chose a colony one place at a time. So it could tell you "three of ours are already on it" about ground no ship of its own had gone near, and say nothing at all about the colony it had just launched at. It now names the colony the ships go to: the soonest attack the guild has not already covered, falling back to the soonest of all when every one of them is covered — in which case it offers no help, which is what it would actually have done. (#3523)
• It no longer goes quiet about ground it is already standing on. A guildmate's colony where this empire's own garrison is dug in appears nowhere else in what the agent knows, so it is still the colony the line names, even when the next fleet is going somewhere else. It will say "my ships are at your 1:1:4, it falls in nine minutes" and still not offer to send more there. (#3523)
• An empire's own siege is never pushed off its own line. The same feed carries the reader's own colonies, and no relief is ever flown to those. Reading the coverage rule there turned a well-defended capital into "covered" and reported a quiet outpost instead, so an agent whose capital fell in ten minutes had nothing in front of it about the capital. Its own line now always names the battle that comes first. (#3523)
• A coordinate it cannot read is no longer invented. A half-written destination used to render as Orso at 1:2:None. The line now says who is under attack and in how long, and says nothing about a place it cannot name. (#3523)

Also

This narrows the gap between what an agent says and where its ships go; it does not close it. Among colonies nobody has covered, the line names the soonest, while the launch ranks first by whether the ships arrive before the strike. So of two uncovered colonies the fleet can still go to the later one it can reach while the line names the sooner one it cannot. Closing that would mean quoting a travel estimate the agent is not allowed to state as a game fact (#3666), so the line deliberately stops short of it. What still holds, as before: the line never offers help the fleet would refuse to deliver.

1.203.2 — 2026-09-19

Fixed

• The system view no longer passes its orbits off as distances (#2733). The Galaxy tab draws a system as concentric rings, one per planet number, and the game charges by the Hive field, where a system is a scattered group of cells. The two orders disagree on 44.9% of the comparisons a player can make inside a system — so "which of these is closer?" read off the picture was a coin flip. It was reported with a screenshot: the pirates on 3:14:12, two worlds on 3:14:14 and 3:14:7. The rings say 14 is the short trip; the game charges 7. The view now says under it that its orbits are drawn by planet number and not by distance, and points at the planet sheet for the real figure. Nothing about travel itself changed — no model, no constant, no flight time.
• The planet sheet names the colony that would actually get there first, with its flight time, beside the one for the colony the sheet is measuring from. Before, it printed a single time labelled "from your colony" with no way to tell which colony that was, and no way at all to compare two of your own worlds against a target without launching.
• Those times now include the departing colony's Relay, which is what the server applies when it prices a real launch. A Relay colony farther away leaves before a bare one that is nearer, so ranking by raw distance would have named the slower launch. The same correction reaches the flight time shown for an anomaly, which was quoted without the Relay and now matches what the fleet will fly.
• The first line names the colony by its address instead of calling it your mother colony. That label was wrong while an exodus is pending, when the sheet measures from a colony in the system you are browsing instead.

Also

• The note is deliberately absent in the Event Sector. That sector sits outside the Hive layout, so travel there falls back to the older metric, in which the planet number really is the distance — the rings are right exactly there, and saying otherwise would have been a new falsehood of the same kind.
• TRAVEL-RULES.md, the page support answers are quoted from, now says which of the two maps "what you see is what you pay" is about. The wrong half of that sentence is what reached the player who reported this.
• Redrawing the system view as the field it prices — the version where the picture becomes the metric — is filed as #4179 and is not in this release.

1.203.1 — 2026-09-19

Changed

• The game can now say when an account it removed actually went (#2741). When a sweep clears out abandoned or long-silent accounts, it writes down each removal. Until now that record carried the round's starting moment — one timestamp shared by every account the same round removed — so forty departures looked like a single instant. The question the record exists to answer, did that sweep take somebody who had just come back, could be answered to the day and no further: not how long the round ran, not where inside it any one account fell. Each record now also carries the moment it was written, read from the clock as the account went. Nothing visible changes on any screen: the Departures bar still reads "the sweep of the 16th removed forty", which is the framing that timestamp was given on purpose.

Also

• Records written before this release stay blank in the new field rather than being filled in from the round's timestamp: those removals were only ever measured to the round, and inventing a precision nobody observed is exactly the kind of tidy-looking wrong number this record exists to avoid.
• The operator runbook gains the queries that read a round's duration and an account's place inside it out of the new field, and the instructions for rolling the server back past it — which, review found, had to say precisely which column to remove: the tidier "remove both" would have erased the entire history of the field added a fortnight ago.

1.203.0 — 2026-09-19

New

A detected probe now marks the Activity tab (#3518)

When one of your colonies caught a hostile probe, nothing on screen said so. The row was written — it reached the Activity feed, and the away-recap lists espionage suffered in its own section — but both of those have to be gone and looked at. A player who was actually playing when a probe landed found out by scrolling for it, if at all, even though a probe is often the first move of an attack.

Now a small dot appears on the Activity tab as soon as a colony detects one, and goes out the moment you open the tab. On a phone it rides the Empire door, the way the Contracts nudge already does.

It is a dot and not a number, and it never joins the unread-battle count: scans are far more frequent than battles and most of them are harmless, so how many there were does not change the move — which is to go and read the row — and the battle count goes on meaning battles alone. Nothing sounds and nothing pops up; the Under Attack alert stays reserved for a fleet.

Who scanned you is still told to you only when the detection was clean or your Sensors are built. That rule is unchanged, and the new signal carries no name at all: it says that you were scanned, never by whom.

Changed

Espionage and Defence now mean two different things (#3518)

The Combat filters gained an Espionage topic last week, and both spy rows landed in it — the report your own probe brought back, and somebody scanning you. They are opposite facts: one is something you asked for, the other is a warning you did not.

Espionage now holds what your own probes found. A probe flown against you files under Defence, beside the inbound-attack alarm it may precede — which is also where the new dot sends you. Combat with no topic picked is unchanged and still draws all of it.

Also

Not shipped, deliberately: there is no way yet to silence the signal, to limit it to one alert per commander per window, or to ignore scans from weak commanders. Each of those needs a default that is a product decision rather than a guess, and they are written up with the measured probe frequency in #4194.

1.202.1 — 2026-09-19

Fixed

• The Home screen's exposure stamp stops wearing the colour of the ordinary next step (#4192). When nothing was sheltering your colony, the stamp in the hero's corner read «OPEN ORBIT» in gold — the same gold as the big card directly underneath, which says «Explore the galaxy» or «Continue the campaign» when there is nothing wrong at all. One screen, one colour, two opposite meanings. The stamp now takes no accent colour: when you are exposed it is printed in plain bright text, heavier and more widely spaced than the coordinates beside it, so it is the strongest thing on its row instead of one of two equals. A covered colony still reads quietly, and nothing on the stamp claims you are safe. The words themselves are unchanged.

Also

• The same stamp used to be read out twice by a screen reader — the full sentence, then the printed badge saying the same thing again. It is now announced once, and the sentence it speaks in each state is unchanged.

1.202.0 — 2026-09-19

New

• Synthetic empires stop sleeping on a full warehouse (#2675). Until now a synthetic empire saved toward its next Command Centre with no clock on it, so it went to bed on whatever it had not spent — and that pile was what a night raid was for. In the last hours of its commander's session, an empire holding stock worth a raid and a building it cannot afford before bedtime now stops saving and spends: a level, a hull, a tech, in whatever order that commander already preferred. Raid the same colony after dark and you will find levels where you used to find Ferrolite. It is not a wall and it does not stop a raid — it means the raid is worth less than it was. Each commander has its own closing time, so there is no hour of the evening when the whole galaxy empties its warehouses at once, and a commander whose night is already behind the night shelter does not bother.
• An empire you have farmed all day buys the Peace shield (#2675). Lose three defences of your own world inside a day and any commander reaches for the 40-Scrip shield, not only the ones already playing defensively. Farm the same target a fourth time and you may arrive to find it covered. The cover costs them something visible: while it holds they launch nothing — no attack, no raid opened, no raid joined — exactly as when a human buys one. A commander with a fleet already in the air does not buy it, because that strike would throw the shield away the moment it landed.

Also

Measured read-only on production on 2026-09-19 before any of this was written: the median synthetic empire held 3,768,276 resources against 5,477 for a human seen in the last week, and over thirty hours human raiders took 26,384,550 out of synthetic warehouses across 165 strikes — a median of 89,106 a raid. 330 of the 425 synthetic colonies had an idle shipyard at the time.

Two of the four changes #2675 asked for shipped. The Shielded Silo's build weight already follows the warehouse it protects, so nothing was needed there. Sheltering a surplus an empire cannot spend is filed as #4180: selling it on the market turns out not to shelter anything, because an offer's goods stay in the warehouse the engine loots from.

1.201.7 — 2026-09-19

Fixed

• A research queue that stops promoting overnight will now say so. The tick that starts your queued research while you are offline gives each empire three attempts a minute, and until now it gave up in complete silence. Contention deserved that silence — a busy minute costs one deferred order and the next tick picks it up — but the same path can also fail permanently, and then an empire's night research would simply never start again, with nothing anywhere to show for it. The round now reports how many empires it gave up on and names them, once per round rather than once per empire, so the flood is bounded. Nobody's research behaves differently; what changed is that this failure can be seen if it ever happens. (#4188)

Also

Three guards around the code that destroys or overwrites a colony, none of which changes anything a player can see today.

• The check that stops a new unguarded colony write reaching production only knew about one kind of statement. It missed the two kinds that delete a colony outright — including the one the end-of-season sweep uses — so the most destructive statements in the game were the least watched. All three kinds are now counted, and a new one cannot be added without a written reason. Measured before and after: the previous check stayed green through five separate ways of introducing exactly the defect it exists to prevent.
• The test harness handed back the stored empire and colony objects themselves rather than copies, which made the safety check on colony and empire writes compare each record against itself — it could not fail, whatever the code under test did. The real database cannot behave that way. The harness now behaves like the database, so that class of fault can no longer pass a test run unnoticed.
• The territory sweep carried a note claiming more of its work repaired itself within the round than actually does: one specific overlap with a guild disbanding leaves a system showing as uncontrolled for up to five minutes instead of showing its previous holder. The behaviour is unchanged and is still better than what it replaced; the note now says what really happens, and a test holds it to that.

Left out on purpose: the issue also asked whether the empire safety check closes a separate race against the catch-up ramp. That was not confirmed against the live path, so it is not claimed here. (#4188)

1.201.6 — 2026-09-19

Fixed

• The Home screen's exposure stamp stops being painted like a button (#2747). The stamp in the bottom-right of the Home hero is the one place that tells you whether your colony can be raided right now: it names the shelter covering it and counts it down, or says «OPEN ORBIT» when nothing does. It was written in the same cyan every tappable thing on that screen wears — in a corner, uppercase, next to dim coordinates — so it read as a button, and pressing it did nothing. A player asked what it was supposed to do.

  Its colour now says which state you are in. «OPEN ORBIT» is gold; a running shelter is dim, like the coordinates beside it; and a reading the game could not get is «…» in the grey it uses everywhere else for "I don't know", so an unread shelter no longer looks like a healthy one. Nothing on that row is cyan any more, so nothing there looks like a control, and the one word that says you are exposed is the only coloured thing on the line.

  The stamp also speaks now. It used to be three glyphs a screen reader either read literally or, in the case of «…», could not announce at all. Each state carries its own sentence: "No shield or shelter is covering this colony right now", "Novice protection: 41 hours left", "Protection could not be read — the game keeps checking" — the last one word for word what the shield chip in the top bar already says.

  What the stamp claims has not changed: a shelter is still named and counted down, «…» still means the game does not know yet, and it still never tells you that you are safe.

1.201.5 — 2026-09-19

Fixed

• Paying reparations charges the sum you agreed to, and the dialog now shows you that sum (#2714). Clicking "Pay the reparations" opened a confirmation that named no figure at all — the bill is drawn on the row behind, and the confirmation covers it. Worse, if the demanding guild re-issued the demand at a larger sum in the instant your payment was retrying an interrupted write, the larger sum was taken from your single click. Now the confirmation shows the demanded amount, read live, so a bill re-issued while you are looking at it visibly changes in front of you; and a payment that finds a different bill than the one it started on is refused, moving nothing, with "State updated by another operation: try again."
  • Still open by one click: an app built before this update gets the refusal but not the redraw, so it shows the old figure until the app updates; and if the refresh that follows a refusal cannot reach the server, the older figure stays on screen. Pressing Pay again in either case agrees to whatever is pending. Closing that completely means sending the amount with the request, which is not in this change.
• Every refusal in the Diplomacy panel now refreshes the view, not only a successful action — "Not enough resources" is answered before the server looks at the bill at all, and it is exactly the refusal after which someone tops up and presses again. The refusal itself stays inside the confirmation, where you are already looking, and refusals from actions with no confirmation now announce themselves to screen readers.
• Paying reparations no longer answers a server error when the database is busy (#2714). The payment was the last action that spent resources without the wait budget every other one uses, so a stall during a deploy window ended in a broken-server message instead of the ordinary "try again" — on the same button that error was first reported on.

1.201.4 — 2026-09-19

Fixed

• A guest who came back could still lose their first colony (#2732). The sweep that frees the planets of guests who never played checked that the colony was untouched and had been quiet for days, then deleted it by owner alone. A guest who logged in and built something between those two moments lost it. The deletion now has to find the colony exactly as it was judged; one that moved is left where it is, and the round simply comes back to it later.
• A guild's three-day control clock could be restarted by a sweep that had not looked at it (#2732). Territorial control is recomputed every five minutes, so almost everything that sweep gets wrong is corrected on the next pass — except the moment a system came under control, which is what the three days of maturation are counted from. Control rows are now written only when they still say what the round decided from, so a guild two days into maturing its control keeps those two days.
• A newcomer could lose the tutorial beacon they had just arrived next to (#2732). Beacons retire once a system has no novice left in it, and the verdict was priced against a reading of who lived there taken earlier in the round — so a player who colonized the system in between was invisible. The round now re-asks who lives in each system it is about to empty, immediately before removing the beacon, and treats anyone it does not recognise as a novice: in doubt, the tutorial stays.
• Cancelling a queued research order could be undone by the overnight tick (#2732). The tick that starts pending research while you are offline wrote your whole empire back without checking whether it had changed, so a cancellation made at the wrong second came straight back. The tick now writes only over the empire it read, and re-prices the promotion when it finds a newer one — the cancellation survives, and the next order still starts.

These narrow the windows rather than closing them: nothing here holds a row while it decides, so a change that lands in the same instant as the write can still win. Each of these sweeps re-decides from scratch on its next pass, which is what makes a skipped round the safe outcome.

1.201.3 — 2026-09-19

Fixed

• On a phone, the storage figure moves again while you watch it (#2749). Under a thousand it ticked up one unit at a time; the moment it crossed a thousand it started being shortened, and a shortened figure only carries three digits that matter — so the smallest change it could show became ten units, then a hundred. At a few hundred an hour the number simply sat there for minutes and then jumped, which is why a new player's first stockpile looked frozen. It now prints every digit up to ten thousand (1,234, 9,999) and shortens from ten thousand up exactly as before (10K, 12.3K, 2.59M). Nothing else changed: the short form and its three digits are untouched everywhere they already appeared, the desktop band is unchanged, and the tap sheet still holds the exact stock and warehouse cap at every size.

  Reported by TartarusMkII, who then talked himself out of it and was right the first time; Remy named the mechanism in the same thread.

1.201.2 — 2026-09-19

Fixed

• A payout can no longer put back a fleet that was already gone (#2782). When an empire had been quiet long enough for the galaxy to start reclaiming it, its colonies were being worn down in the background — ships lost, production halted, a tally waiting to be reported. If a war verdict, a reparations payment, a season sweep, guild aid, a Crucible roll or an operator's grant landed on that colony in the same moment, it could write back the picture it had read a moment earlier: the destroyed ships reappeared, the halt lifted, and the report of what had been lost was erased. The payment itself reported success, so nothing looked wrong from either side. Those writes now check the colony's actual contents before committing, and step aside and re-price when the colony has moved under them.

• An Echelons prize can no longer cancel a leave you just declared (#2782). Scoring in the cycle and then going on leave could leave the prize payout writing your colony back the way it was before the leave — so the production halt the leave is supposed to apply quietly disappeared, while the leave itself was granted with its shield. The settle now sees the leave and re-prices against it.

Also

• The rule that keeps these writes honest is now enforced by a test that reads the source itself, so a future write built the old way fails before it can ship rather than after. Three ways of slipping past it were found during review and closed, each with its own failing case first.

1.201.1 — 2026-09-19

Fixed

• The galaxy map now dates a system's control by the colonies you can actually count (#2723). Over a system where somebody keeps a colony hidden behind a Veil, the map used to show its control tag as permanently unripe: your own three colonies, planted a month ago, never matured, and that "never" was itself the tell — no ordinary system behaves that way, so the map was quietly announcing that someone you cannot see is sitting there. The other direction leaked too: a holder who unveiled a single colony while you watched had its control dated to weeks before anything you could see was there. Now the map replays the same dominance rule the game itself uses, but over the colonies that are on your chart, at the holder's own sovereignty speed — so a control tag ripens on your map the way it would in a galaxy with no Veils in it at all. A colony deported into a system starts its clock when it lands, not when it was founded.

Also

• One consequence, stated rather than hidden: a sector banner can now appear for a guild the game itself does not pay, when the systems you cannot account for would have broken the tie. The map has always described your own galaxy rather than who is owed a bonus, and the alternative was the leak above.

1.201.0 — 2026-09-19

New

The Colony list sorts by what the next level costs (#2658)

The Colony tab could be reordered one way only — lowest level first. To find the cheapest thing to start you had to read three numbers off every row and compare the bundles by eye, down a list that runs to 27 rows on a grown colony.

The sort chip is now a picker with five orders:

• level ↑ — the one that was already there, lowest first;
• total ↑ — cheapest first, by the total price of the next level;
• Fe ↑, Xe ↑, Dt ↑ — the same, by one resource on its own.

It is off by default, so nothing moves until you ask. The order is applied inside the groups the list already keeps — queued rows first, rows you cannot build yet last — it composes with the family chips, and it survives switching colony.

What "cheapest" means here, and what it does not. It is the price of the next level, never what your warehouse can pay for right now: an order keyed on your stock would reshuffle every second as resources mature, and land your tap on the wrong building. The row still answers the affordability question where it belongs, by printing the amount you are missing per resource. total is the sum of Ferrolite, Xenogas and Data, the same meaning the Shipyard's catalogue already uses — which is why the option says "total" and not "cheapest": ordered by the total, the Ferrolite column visibly dips, and it is not a broken sort. If one resource is what you are short of, pick that one.

What you will notice on day one: the level order used to be one tap on a toggle. It is now a picker, so on a phone it is tap, choose, done. That is the price of having five orders in the width the chip rail can actually pay for — five separate pills do not fit, and would not say they are mutually exclusive.

Asked for by mhar02, who also chose the shape: "Cheapest overall, cheapest by FE, cheapest by XE, cheapest by DT."

Fixed

The Colony sort no longer zooms the page on a phone (#2658)

Caught in review before release, in the first version of the change above: the new sort control is a dropdown, and dropdowns under 16px make iOS and the in-app browser zoom the whole page. Tapping the sort chip would have left the HUD and the building grid out of frame, with no obvious way back. The picker now keeps the 16px size every other field in the game uses, and the buildings header takes a second line a little sooner on the widest phones to make room for it.

1.200.0 — 2026-09-19

New

Energy plants tell you what a level gives back per resource (#2660)

Deciding between a Fusion Reactor and Orbital Panels used to mean doing arithmetic off the catalogue by hand: the sheet showed a price and an energy figure and never the ratio between them. Open an energy plant now and, under the price, three new lines answer it.

• Energy per resource · lvl N — the summed figure, with the level it is for on the row, so two sheets at different levels cannot pose as one comparison.
• By resource — the same return per ferrolite, per xenogas and, on the Zero-Point Reactor, per data. This is not decoration: at level 1 the Reactor and the Panels charge identical ferrolite per unit of energy and differ only in xenogas, which the summed number hides completely. A player short of iron is told something false by the total alone.
• A line naming which colony and which of three possible divisions the figure is — the plant's whole output at that level over that level's price, not the step's gain and not a lifetime total.

It shows on every net energy producer, and the number is the engine's own: the same energy the colony grid sums, over the same price the build queue charges. Nothing is retyped, so a rebalance moves what you read.

Two things the figure now gets right that a simpler version would not. It is per colony — the planet multiplier spans 0.9 to 1.4, so the same plant reads 1.55x apart on two of your own worlds, a wider spread than the gap between the plants. And it is per Site: a second Orbital Panels is divided by its own surcharged price, which is what turns the headline finding around. The first Panels out-return the Reactor at every level; the second lose to it.

Also

Nothing was rebalanced. The issue also observed that the first Panels beat the Reactor per resource at every level — 1.10x at level 1 widening to 2.05x at level 20 — because of a single digit in their cost curves. Whether that gap is intended is a balance question, not a display one, and no cost or energy value was changed here.

Two known gaps, left for a deliberate pass: the ⚡ in these rows is the emoji rather than the app's own bolt mark (which on iOS renders in the system colour), and the new figures carry no screen-reader label. Both match what the rest of the app already does, so fixing them here would have left the codebase in two minds.

1.199.1 — 2026-09-19

Fixed

• The note about a probe a friend turned back no longer outlives your own newer probe (#4160). Send a probe at a colony of your own guild or an allied one and it flies the whole way, lands and files nothing — and the mission composer remembers that, so you are not sold the same answer twice. That memory had no expiry, so it survived every later answer you paid for at the same address. If the alliance broke and your next probe came home with a report, the commit screen showed the report and, a few lines below it, the old refusal still saying another probe would come back the same way. Worse: if the friendly colony was abandoned or wiped since, your next probe found nobody — and the composer never read that at all, so the only sentence on screen told you a friend held a slot your own latest probe had found empty, and you would call off a probe that would have come home.

  The line now shows only while it is the freshest thing you know about that address: a newer report, or a newer probe that found nobody, takes it off the screen. An older one leaves it exactly where it was. Nothing is gated and nothing new is claimed — the composer simply stops saying what your own newer evidence contradicts.

  This bounds the original problem rather than closing it: if the alliance breaks, your probe reports, and the alliance then re-forms, the report is still your freshest evidence, so you can buy that refusal once more before the new one is filed. The client cannot know today's diplomacy, and the other error is worse — talking you out of a probe that would have succeeded costs you the answer, not one probe's price.

Also

• Two of the three findings in #4160 are carried over to #4166: on desktop the note permanently raises the target card's minimum height for the rest of the composer session, and on the phone it arrives a moment late and pushes the fleet picker down. Both are layout judgements that need a browser or a device, and neither is contained to this note.

1.199.0 — 2026-09-19

New

Activity: filter by who and by what, at the same time (#2649)

The Activity tab opens on Mine, then made you give that up the moment you wanted one kind of event: the chips were a single row where Mine, Combat, Exploration, Empires, Economy and Galaxy all cancelled each other out. Tapping Combat handed you the whole galaxy's fighting and took your own night away with it, and "my combat" simply could not be asked — the thing most players open the tab for.

It is now two rows. The first is who: All, Mine, Galaxy. The second is what: Any type, Combat, Exploration, Empires, Economy. They narrow each other, so your own combat is one tap from where the tab already opens, and the galaxy's chronicle can be read by type too.

The number beside each chip changed meaning with it: it is now counted within your other selection, so it says how many rows that chip will actually draw. Before, a chip could read 40 over a column of two.

The Combat chip is a filter again (#2649)

Combat had quietly become the place where 43 kinds of event live: battle reports, bounties, faction sorties and their Dark Matter, spy missions, relic finds, war declarations, League verdicts, the Eschaton sweep and "Season ended", inbound-attack alarms and their retractions, wrecked destroyers, Contested Space and its podium, interceptions and pickets.

Picking Combat now opens a second row underneath it:

• Battles — the fights and what they paid.
• Defence — what is coming at you and whatever met it: the alarm, its retraction, the standing watch, the interceptions.
• Espionage — the probes and the spy snapshots.
• Wars — declarations, tension, League verdicts, bounties.
• Season — the Eschaton sweep, "Season ended" and the Contested Space podium.

The line that explains a Contested Space capture you did not get — "you keep what you took from it, but the world had already changed hands" — sits under Battles, next to the fight it is about, rather than with the podium it shares a name with.

That row starts on Any combat and clears itself whenever you move to another type, so Combat on its own is exactly the chip it has always been: nothing was moved out of it and nothing became harder to find.

Changed

• The "the outcomes of your missions appear here when your fleets return" line now belongs to the who row alone. Asking for your own Exploration rows and having none last night gets "Nothing under these filters yet" instead — the old sentence was false to anyone with fleets already in flight (#2649).
• Both new chip rows wrap onto a second line on a phone instead of scrolling sideways. At 390px the type row needed 503px and the combat row 553px of a 366px line, so Wars and Season rendered entirely off the screen edge, behind a strip with no scrollbar and nothing saying it could be dragged (#2649).
• Tapping the chip that is already selected no longer resets anything. Reaching Economy on a phone means dragging that row, and a drag that ends where it started counts as a tap — which used to clear your Combat sub-filter and grow the feed under your thumb, with the chips looking untouched (#2649).

1.198.10 — 2026-09-19

Changed

• Every synthetic empire now sleeps on its own schedule, instead of all of them sleeping at once (#2608).

  The night window is the one defensive setting the game lets you aim, and no bot had ever touched it. All of them sat on the default, which meant that between 00:00 and 08:00 UTC no attack landed on any synthetic empire anywhere — the whole population sheltered in the same eight hours and attackable together the moment they ended. A blackout on a fixed timetable that anyone could map in a day, and nobody had chosen it.

  It was also unrelated to the empire behind it. Each of those commanders keeps its own hours, and the shelter ignored them completely: some were protected while they were out raiding — striking from inside hours when nothing could be sent back — and others were exposed for the whole of their own night, unable to answer anything that arrived.

  Now each one aims its window the way a person does: the shelter opens at the hour that commander stops playing. Because they keep different hours, their windows are spread right across the day, and there is no longer a moment when the synthetic side is collectively safe, or collectively open.

  For you this cuts both ways, and both are the point. The nightly free window against every bot is gone. So is the bot that raids you from behind a shield it never has to lower. Any given one is now reachable through its own working day and sheltered through its own night — the same deal every human plays under, set through the same screen, with the same once-a-day limit on moving it.

  One empire is not moved and it is worth naming: an old unpiloted account still holding an agent address keeps the default hour. Nothing plays it, so it attacks nobody from behind that shelter.

1.198.9 — 2026-09-19

Fixed

• A rule in the code stopped describing the code it governs (follow-up to #2753, coordinated with #2768). Nothing a player can see. v1.198.7 wrote down the order the guild payout paths take their database rows, and part of that note said the season row was one those paths only read. v1.198.8, an hour later, changed exactly that: the season row is now held, not just read. The note kept saying the old thing, a few files away from the code doing the new one. It now records where the hold actually went, the shape the payout paths end up with, and the one case that would put the arrangement back into disagreement if anybody added a hold there later.

1.198.8 — 2026-09-19

Fixed

• A guild's season placement no longer disagrees with the points shown beside it (#2768). When a guild season closes, the placement of every guild is frozen in one go, so that the place the rewards are paid on is the place the showcase shows from then on. Five things could still move a guild's points after that moment: a war verdict, a Council Directive tier, a League match result, an Eschaton contribution, and the internal score award. A guild could end up showing third place forever beside a points total that reads like second — and since the empires run by the game quote the showcase as fact, they would repeat it.

  Each of those now checks that the season is still open at the instant it writes, not merely at the moment it looked. A result that arrives too late keeps everything else it earned — the war is still won, the tribute still paid, the tier still settled, the match still recorded — and loses only its season points, because the contest it would have counted towards is already over.

  The Eschaton contribution is the exception, and deliberately: it spends the guild's Fund, so instead of quietly doing nothing it refuses, tells the leader the season has closed, and leaves the Fund where it was.

• Two guild panels stop erroring during the rollover. The weekly sovereignty settlement and the Eschaton contribution now wait out a season close rather than giving up part-way through it, so a member opening the guild panel at the moment the season turns over gets the panel instead of an error, and the week's points are not left unpaid until someone happens to open it again.

1.198.7 — 2026-09-19

Fixed

• A guild screen can no longer fail with an error that swallows a reward (#2753). Five places pay a whole guild at once — the Council directive tiers, the League verdict, guild research maturing, the end-of-season research reset and the end-of-season placement payout — and each writes the guild's own rows and then every member's row. Some wrote the guild first, some the members first. When two of them ran on the same guild at the same moment, each ended up waiting for what the other already held, and the database broke the tie by killing one. None of them retried, so the loser became a failed request on somebody's guild panel and its payout simply was not made until something triggered it again. They now all write in the same order — the guild's rows first, the members' last — so that standoff cannot form between them. Same amounts, same people; the only visible difference is that when a League verdict levels a guild up, the "level up" line in the guild log is now written before both sides' match lines instead of between them.

Also

• The order is now written down in the code next to the guild's other lock rule, and a test refuses any future payout written the other way round. Two transactions that take both rows and genuinely cannot follow the rule — leaving a guild, and deleting an account — are named there as still open rather than quietly counted as fixed.

1.198.6 — 2026-09-19

Fixed

• Nothing changed for players in this release: it repairs a test that had quietly expired.

  The suite covering "copy a market offer into the composer" checked that copying a row leaves an unfinished draft of your own standing — the restore banner stays up and your numbers are untouched. Two of those checks stamped the saved draft against a fixed date instead of the test clock, and a market draft stops being offered back after a day, so they passed for exactly 24 hours and then started failing on their own, with nobody having changed anything. The clock is now pinned where those tests expect it, the way the neighbouring draft tests already do it.

  The feature itself was never broken: copying an offer has always left your saved draft alone. What was broken was the proof, and a suite that goes red on its own is a suite nobody can read.

1.198.5 — 2026-09-19

Fixed

• A spy probe sent at a colony of your own guild, or of an allied one, flies the whole way and turns back without filing a report. Until now nothing remembered that flight, so the mission composer offered Espionage at the same coordinates, with the same title and the same price, and you could buy the identical refusal for ever. While you are drafting a probe at those coordinates the composer now prints one dated line: your probe reached these coordinates, it was turned back, and by the time it arrived this world belonged to your own guild or an allied one. It reaches back over your whole history rather than the last page of your activity feed, so it answers for flights from long before today, and it appears on the phone as well as on the desktop. It gates nothing — diplomacy moves, so the line says what held when the probe landed and promises nothing about the next one. (#2700)

Also

• The galaxy map says nothing new. A planet under the Veil and a genuinely free world still read identically there, on purpose: the memory lives on the screen where the probe is actually spent, and under that verb alone.
• Still open from the same report: a probe stopped this way never rolls for detection, so a guildmate currently maps a friend's holdings more quietly than a stranger does. That is a design decision about where the check belongs, not a defect, and nothing here changes it.

1.198.4 — 2026-09-19

Fixed

• Two buildings at level 10 had holes in them (#2670). The Terraformer was missing the machinery between and behind its two towers, including the right tower's inner face — about a fifth of the upper body. The Trade Portal was missing a tall strip of its left wall, running from the roofline to the base. Both are whole again.
• The Relay and the Cannon stopped being see-through (#2670). Their bodies were drawn at about 83% and 95% opacity everywhere, not just at the edges, so they looked like holograms standing next to their own upgraded stages. Both are solid buildings again — the Relay keeps its tie-rods see-through, because that is the art and not a fault.
• The Trade Portal's ground shadow at level 10 stays the size it always was, rather than shrinking when the building got its wall back.

Returning players see the new art immediately instead of a cached copy.

Also

Not every sprite is fixed. The Terraformer's hole is closed at level 10 and not at level 20, where a different drawing is used; the Reactor at levels 10 and 20, and the Lab at level 10, are in the same condition. Twenty-three sprites are still cut by the process that caused all of this, each now recorded with a measurement of how much of the building is missing. They are left for an art pass rather than re-cut blind, because re-cutting two of them automatically made them worse.

1.198.3 — 2026-09-19

Fixed

• A fleet loss the game had already told you about could come back (#2786). When the ruin destroys ships on a quiet empire, the game keeps a list of what it took and mails it the next time you are around. Cancelling a leave at the wrong moment put that list back after it had been sent, so the next round reported the same losses a second time, as a bigger number. It is reported once now.
• Declaring a leave could undo a ruin (#2786). A leave declared while a lifecycle round was destroying ships could write the older fleet back over it. The leave is now refused and rolled back instead, with the same "try again" it already gives when a colony changes under it — the ships stay lost.
• A leave could keep its shield without paying for it (#2786). Declaring a leave shuts production off; that is the price of being untouchable. A lifecycle round running at the same instant could erase that shutdown, leaving the empire shielded and still producing for the whole leave, with nothing later to notice. The halt now survives.

Changed

• Cancelling a leave re-tries instead of giving up (#2786). If a colony changes in the instant the cancel lands, the cancel re-reads it and re-applies on top rather than skipping it, so production restarts while you are still looking at the screen instead of at the next six-hourly round. A colony destroyed under the cancel, or one another round has already restarted, now ends quietly instead of logging a warning that was true of neither.

1.198.2 — 2026-09-19

Fixed

• The first guild to surrender a war is the one that pays, not the last (#2735). Two leaders could both give up the same war, and the one who pressed the button second was the one billed for it. A surrender is recorded on the war itself, and the mark names the guild that pays the 8% tribute — but nothing stopped the opposing leader from writing over it, so a genuine mutual surrender charged exactly the wrong side. The window was real rather than theoretical: a surrender's mark deliberately stays on the war when its settlement loses a race, so that the war closes with the surrender honoured, and for those minutes the panel kept offering the other leader a button that moved the bill onto him. Now the first surrender stands. The leader who did not give up is refused, and their Diplomacy row says the other side has already conceded and the war is closing, instead of showing a button that cannot do anything. The guild that did surrender keeps its button: its decision is recorded and cannot be taken back, but pressing it again closes the war on the spot rather than waiting for the next sweep.

Also

• A leader on an app build from before this release who taps Surrender on a war the other guild has just conceded will see the refusal as a bare code rather than a sentence, for as long as that build is installed. Their own app has no way to know the war was already given up, so it still draws the button. Updating the app is what fixes it; nothing is charged either way.

1.198.1 — 2026-09-19

Fixed

• A probe that comes back empty is no longer thrown away (#2688). The planet card told you it could not tell a free planet from a colony under the Eschaton Veil, and named your own probe as one of the two things that settle it. You flew the probe, it landed on an empty slot, the diary said so once — and the card forgot. Reopen it and it was word for word what it had been before the probe flew, asking for the same mission again. You had bought the answer and the game threw it away.

  The card now remembers: "Your probe landed here 2d ago and found nobody. A colony could have been settled under the Veil since, so this is what was true then, not now." It hedges deliberately — the probe proved the slot was empty at an instant, never that it is free today — and nothing on the card is unlocked, greyed or gated on it. The scan's own verdict above it is untouched.

  Empty probes that landed before this release are remembered only for as long as they are in your recent activity; every probe from now on is remembered for good.

• The card stops predicting a refusal your own newer probe already disproved (#2688). Where an old spy report named a commander and a later probe of yours found nobody there, the card used to say both at once: the slot was empty, and the server would refuse your founding and your harvest, with the Harvest button dressed in the refusal wording and an incoming colony ship told it would turn back. The refusal needs a colony to actually be there, so the later probe was evidence it had gone. The report still names who was there and when; the warnings it was driving are retired, and the probe line says "Since that report" so the two read in order.

Also

• The composer you reach by tapping Colonize still names the occupant from the old report — it always has, and teaching it the negative is a larger change than this one.
• The system list's scouted column is unchanged: it still marks reports only.

1.198.0 — 2026-09-19

New

• Find a landmark on the galaxy map by typing its name (#2604). The Relics tab has been banded by sector since the forty-relic roster arrived, which made every relic reachable but none of them findable: a closed band names nothing, so tracking down "The Banked Fires" without already knowing its sector meant opening and closing up to eight bands, and the GO TO field has never understood place names. There is now a search box above the bands. Type a few letters and the bands step aside for the matching landmarks from every sector at once, each one naming the sector it stands in and flying the camera there on a tap. Half a name works, the whole name works, and capital letters do not matter. Bastions are searched too — they are chips on that panel like any other landmark — and a bastion in the results says BASTION, so a faction's wall is never read as a prize nobody has claimed. A search that finds nothing says so and names what it looked for. Empty the box and the bands come back exactly as they were, with the one you had open still open.

Fixed

• The Relics tab no longer claims the board is empty while it is still loading it (#2604). The header said "no relics charted" from the moment the tab opened until the galaxy answered — and said the same thing when the request failed outright. It now says "charting…" while the answer is on its way and "data unavailable" when it did not arrive, and goes back to counting the moment it has something to count.
• A sector band with nothing free and nothing of yours now says whose the relics are (#2604). "SECTOR 7 · 5 relics" read the same whether all five were in a rival's hands or simply not yet accounted for; it now reads "5 relics · 5 held by others" in that case.

Also

• The search box is a full-size thumb target on a phone, keeps the keyboard from capitalising or autocorrecting a proper name, and copes with the curly apostrophe an iPhone types — "The Marshal's Table" is searchable in full from a phone.

1.197.3 — 2026-09-19

Fixed

• A question about what the Defense Batteries counter can now be answered from the code (#2724). Nothing on screen changes with this release — the Batteries' own building card has printed their matchups since #1832, and still prints the same ones: Interceptor and Corvette take double damage from your emplacements, and Bomber and Breachhead Monitor crack them ×2.5 and ×4. What changed is underneath. The building a player owns is called "Defense Batteries", the combat group it fields is called something else entirely throughout the engine, and the two names met nowhere near the catalogue entry or the tables that hold those multipliers. A player's question about battery counters was answered "they have none" — with the whole repository open — because the search ran on the name the game shows. The two vocabularies now name each other at both ends, so the next search finds the numbers instead of missing them, and a test keeps that cross-reference tied to the constants the engine really exports rather than to a list somebody typed once.

Also

• The same pass corrected four statements in those notes that were themselves wrong about the engine: the per-emplacement attack shown in the code is a level-15 base that grows 30% a level from level 16 and is cut by a power shortage at every level, not only above 16; emplacement shields never scale at all, whatever the level; and against the Faro and the pirates the raw base value is what actually reaches the battle, because that path skips the scaling entirely. None of this changes a number in the game — it changes what the next person reading the code will conclude about them.

1.197.2 — 2026-09-19

Fixed

• "Continue the campaign" is no longer the only thing your next-move card can say (#2761). The card under the orbital hero ended with that line whenever nothing was waiting on you, no guild invitation was open and no campaign fleet was out — which is the state you reach by keeping your queues full and your labs busy. Nothing checked whether you had already taken a swing at that level, so an empire parked at a rung it cannot clear read the same instruction on every visit to its first screen. Now the card names the campaign only when the server's own record says you have never engaged the level about to come up: the attempt ledger at that level, and the last climb you finished, which still sits at that level if it did not clear. Otherwise the card reads "Explore the galaxy" and opens the map — the one destination in the game that cannot be exhausted. A fleet already engaged still reads "Watch the campaign fleet", and anything waiting on you, or a guild invitation, still comes first.

Also

• The card still ends with a fixed sentence; what changed is which one, and that it is now one nobody can have finished. Nothing on this screen works out whether your next campaign level is winnable — the card reports what you have already done, not what you could do.

1.197.1 — 2026-09-19

Fixed

• The guild board now says what it ranks on (#2776). A player compared three screens and got three numbers: 981,087 on his own empire panel, 936,011 on the leaderboard and in the Galaxy, 921,001 on the guild board. The first two are one quantity read two ways — the sum of your structural and military halves, and the larger of them — and they agree on purpose. The third is a different score entirely: the guild directory ranks guilds on their members' empire points added up, and a guild's roster is ordered the same way. No screen said so.

  The guild directory, the roster inside your guild and another guild's public showcase now each carry a line above the list saying what orders it, and a line below saying what empire points count, what Power counts instead, and that the two orders can disagree. The "Points" column header gained a tooltip as the short form, and the showcase's roster figure — until now a bare number with no word anywhere near it — gained one too. The "find a guild" screen, which a player without a guild lands on, was the one place with no explanation of that column at all; it has one now.

• The leaderboard stops telling you a fleet in flight has stopped counting (#2776). The note under the standings said empire points count "the ships stationed in your colonies". They do not: a fleet on a mission is counted exactly as it is at home, on the player board and on the guild board alike. Anyone reading that line mid-raid was told half their navy had dropped off the score. It now says "your ships, whether stationed or in flight", which is what the game actually counts.

Changed

• Nothing about how any board is ranked (#2776). The order of the guild directory, of the guild roster and of the leaderboard is exactly what it was, and no standing moves — this ship adds words, not arithmetic. Whether a guild board should rank on points rather than on Power is a separate, still-open question (#2970), and it is deliberately left open here.

Also

• The Galaxy still prints "Power" over a figure that is the Gate, and this ship does not label it (#2776). A tooltip was written and then withdrawn in review: that figure is also shown for faction Bastions and relic guardians, which are on no leaderboard and exempt from the power band, so the sentence would have been false on forty-odd fixed map addresses — and a tooltip is the one thing a phone can never open. Naming the Galaxy's number needs a visible line and wording true of Bastions and Relics; that is its own piece of work.

1.197.0 — 2026-09-19

New

• The Shipyard's stat bars now show what a bonus bought you, and what a refit cost you (#2727). Where a figure has moved, the row carries both numbers — the catalogue value struck through, then what your empire actually flies — and the bar grows a segment of its own: green for the gain, red for the loss. A hull nothing has touched reads exactly as it did before, with no arrow and no second segment.
• The Hulls panel in your Inventory gained a Cargo column, with a logistics line naming what moved it — the hold was the one figure that panel resolved for everything except itself.

Changed

• The Shipyard reference panel reads your ships instead of the catalogue (#2727). It used to print Attack 50 for an Interceptor whether you had just landed or had been buying Energy Weapons, Calibres and standing orders for weeks: a specification sheet for a ship nobody in the game flies, on the screen you use to decide what to build. All five bars — Attack, Hull, Shields, Speed, Cargo — are now your own, with your research, your refits, your mastery and your supporter bonus in them, taken from the same numbers a battle fights with.
• The bars are still a comparison, and the caption now says exactly what of. Your research lifts every one of your hulls alike, so the ceilings move with the figures: an Interceptor does not become the best Light hull because you bought a Discipline. Gold still means the best hull in the game — measured as you would fly it — and the caption says so instead of leaving it to be guessed.
• When a factor has not loaded yet, the panel names it rather than quietly leaving it out. If your standing orders or your ship mastery have not arrived, the caption says which, instead of claiming they are already in the numbers.

1.196.0 — 2026-09-19

New

• The galaxy target board can now be searched by who owns the row (#2657). Every row already told you whether a machine was flying that empire (the AI badge), which guild the owner flew for, and whether they had gone quiet — and none of it could be searched on, so "find me someone who is not an AI, not in an alliance and still awake" meant reading several hundred rows by eye.

  Three chips now sit on the board's filter rail: AI, Guild and Silent. Each cycles with a tap — off, then only those, then hide those, then off again — so one control answers the question both ways round. Asked which way he wanted the silent one, the player who reported this said "both, depends on the day": hunting the silent is striking where the Ledger prices nothing, hiding them is looking for opponents who fight back. The same is true of the other two, so all three cycle.

  The board's rule that no filter is ever silent holds. A line under the chips counts how many rows carry each of the three facts, and says how many rows the chips are holding back while any of them is on. If the chips empty the list, the message names the ones that are actually lit and points at the clear beside them, which puts every row back in one tap. The band counts above are measured on the filtered list, so no chip can promise rows another chip has already dropped.

1.195.7 — 2026-09-19

Fixed

• The Event Sector no longer shows as open after a season has already ended (#2666). When a season ran out, there was a window in which the game said two different things at once: the sector still appeared open on the Missions, Galaxy, Guild, Overview and Chat screens, while every launch the player actually tried came back refused. The window is normally short — the closing runs at the moment the season ends — but it stretched whenever the server's job queue was behind, and for as long as it lasted the player was being invited into a sector that was already shut. Now the sector reads closed everywhere the instant a season passes its end time, so the screen and the launch agree.

1.195.6 — 2026-09-19

Fixed

• Filter chips you stopped using no longer keep a bright outline (#2697). On the market board, every filter or sort chip you had clicked earlier in the visit kept a bright outline for the rest of the visit, so several of them read as "still active" at once while only one filter was actually applied — and nothing but a reload cleared it. A chip you stop using now looks exactly like a chip you never touched, on both the filter row and the sort row. The same lingering outline could appear on any highlighted-when-selected button in the game, so it is gone everywhere, not only on the market. Keyboard focus rings are unchanged.

1.195.5 — 2026-09-19

Fixed

• A short contract board now says it is short (#3418). The Contracts header printed your Charter's figure — "Sovereign Charter · 6 daily orders and 3 weekly" — above a board that held five daily cards and one weekly, and nothing on screen explained the gap. The board can legitimately come up short: each cycle it is dealt from the orders within your reach, and if fewer of those are open to you than your Charter has slots, you get a smaller hand. Now the board says so where the cards are — "Dealt 5 of the 6 daily orders this board was drawn against — 1 was out of your reach when it was drawn, and the hand stays as dealt until the next reset" — and the same line, in the plural, for a short week.

  The board's count is the one its own hand was dealt against, so a Command Center finished part-way through a cycle no longer makes the panel argue with itself: the Charter band says what your Command Center will deal next, the line under the cards says what this board was dealt. Nothing about the deal changed — an order you cannot finish is still not dealt to fill a slot, and a hand stays as dealt until its cycle turns.

Changed

• The Charter band now opens with the verb: "Sovereign Charter · Deals 6 daily orders and 3 weekly". Bare numbers there read as a count of the cards below them, and the two can honestly differ.

1.195.4 — 2026-09-19

Fixed

• Nothing changes on screen in this release: it repairs the API test bench, not the game. (#4116)

  The bench runs against an in-memory stand-in for the database. That stand-in was handing back related records nobody had asked for — a guild's roster arrived with every member's colonies and fleets attached, a chat message arrived with its author's guild tag — where the real database sends none of it. So the server could quietly stop asking for one of those, read nothing where it expected a number, and every test still passed.

  Measured, not assumed: removing the fleets from the guild roster query scored every member as if they held no ships, and all 759 guild tests stayed green. With this change the same removal reddens 98 of them. Removing the guild tag from the chat query relayed tagless messages to Discord past a fully green chat suite; now it reddens the test that exists to catch it.

  Nothing a player can see is different today. What is different is that the next change to the guild roster or to chat cannot break either of them silently.

1.195.3 — 2026-09-19

Fixed

• The planet card no longer hides your own founding fleet (#3346). With a Colony Ship of yours already flying to a slot, the card said only "No colony on your scan" and offered Colonize, so a second founding was easy to draft by accident. The card now adds "Your founding fleet is on the way — it lands in 2h 14m." with a View mission link that opens that fleet's card in Active missions. While you draft another founding for the same coordinates, the mission composer says the same thing and adds what a duplicate costs: if the first fleet founds the colony, the one landing after it finds the slot taken and comes home with its ships and its cargo. On a phone the note also rides on the screen where Send sits.
• Nothing on those surfaces claims the planet is free or that the founding will land — the slot can still be taken before your fleet arrives.

1.195.2 — 2026-09-19

Fixed

• A guild at war no longer leaves its war replies on the table when its members are synthetic (#3993). A guild at war holds a reply for every blow the enemy guild lands on one of its members, and any member may spend one by striking any member of theirs: no bounty, even far below the power band, as long as the strike is carried alone. Human guilds have been spending them from the War Room since the Ledger shipped. The synthetic empires did not know they had them, so a bot guild at war refused exactly the strikes its own guild had already paid for, and the small enemy who had been hitting them all week stayed untouched. They now read the War Room for each guild they are at war with and act on the number it shows: when a reply is held, the strike happens; when it is not, nothing changes. The credit is only ever spent on a strike carried alone, never inside a raid.
• What they say about it is what the board says. The plan sheet states the count guild by guild, when it was read, and that it is a shared stock spent when a fleet lands and worth nothing once the counting week closes — so an agent asked about it in chat cannot promise a number the War Room disproves. When the room answers nothing, or the read fails, the sheet says nothing and the strike stays refused: the credit is never assumed. A commander thrown out of its guild loses the ledger along with the war.

1.195.1 — 2026-09-19

Fixed

• Launch again now says how long the fleet is away (#3449). The rail replays a saved order — an attack included — in one tap, and it printed the fleet, the battle line, the stance and the price without ever saying when the ships arrive. Every card that can launch now carries the outbound leg, priced by the same planner the mission composer asks, and it says "one way" so a six-hour garrison order is not read as a six-minute errand.
• A survey sweep says when each probe gets there (#3449). Ticking several planets and pressing Send once launched a probe at each with no duration anywhere on the panel. Each ticked row now carries its own travel time, because the planets in a system sit at different distances and one figure for the batch would be right for none of them.

When the answer is not available — while it is in flight, after a refused request, on a card that cannot launch, or with an empty Probe hangar — nothing is printed at all: no zero, no dash. The rail holds the row's height in the meantime, so the Launch button does not move under a thumb already reaching for it.

1.195.0 — 2026-09-19

New

• Copy a market offer into your composer (#3362). Composing a trade means typing six numbers, and posting the same trade twice meant typing all six again. Open any offer on the Market board — yours, somebody else's, even one that has just been taken — and Copy to composer fills the form with that row's amounts. It says in words which side you land on ("you would offer X and ask for Y in return"), because a copy reposts the seller's side and a player who opened the row meaning to buy would otherwise end up on the opposite side of the trade. Nothing is published until you tap Publish offer. Credit to jaybee1019 (in-game Usherin) for the suggestion.

  ⚠️ It saves typing, not ceiling. Copying does not raise your daily intake cap: that cap is on what you receive in a day and is keyed on the receiver, so several small trades spend it exactly as one large trade does. The sentence beside the button says so on screen, not only in a tooltip.

Changed

• Dead market rows now say why (#3362). The board already dimmed a row nobody could settle today, but dimming is the same cue it uses for "you cannot afford this", so the two were one signal and the only way to tell them apart was to open every row. A blocked row now carries a short tag next to the seller — over your day, over your cap, over their day, over their cap, seller short, stock spent — with the full refusal in its tooltip. A day is spent and reopens at midnight UTC; a cap is refused on an empty day too, and only a bigger Warehouse moves it.

• The copy tells you when the clamp moved the price, not just the size. What you give and what you ask are bounded by different ceilings, so a trim can hand back a rate that is not the one on the board. When that happens the line says the price is no longer the row's, instead of only saying "trimmed".

• Copying no longer discards an unfinished offer. If you had a half-priced offer waiting to be put back, a copy leaves it exactly where it was — banner up, numbers on disk untouched — and says so.

Also

• On desktop the composer is scrolled into view after a copy, and the line describing it is a polite live region, so the change is not silent for either a scrolled-down player or a screen reader.
• Not measured, and not claimed: whether the agents really are the busiest traders on the board, and how often the daily cap actually bites in the live market.

1.194.5 — 2026-09-19

Changed

• Nothing changes in the game this time: this release is entirely about the bench that guards it (#2639).

  The API test suite runs against a hand-written stand-in for the database. When a query asked the real database for three columns, the stand-in handed back all forty — so a query that quietly stopped asking for a column kept passing its tests, while the live game got nothing back for it. Three separate bugs had already shipped that way: a chat message whose "deleted" flag went missing, the Command Center announcing the colony cap a world early, and the pirate-beacon sweep that retired nothing.

  Each of those was patched on the one query that had just been caught, and the rest were left as they were. Now the stand-in returns exactly the columns a query asked for, everywhere. As a check: deleting one column from the beacon lookup used to leave all 538 test files green — it now turns five tests red in three of them.

Also

• Filed #4116 for the half this did not reach: the stand-in still attaches related records nobody asked for, one level down (a guild's members arrive carrying their colonies and fleets whether or not the query wanted them).

1.194.4 — 2026-09-19

Fixed

• The Harvest blocker now names the Recycler and the colony it has to stand on (#3409). Picking Harvest with a launch colony that had no Recycler greyed out Send under "No Recycler on any colony" — a sentence that never said the Recycler is a building, and sent at least one commander hunting through the Shipyard's hulls for it. The line now reads "Build a Recycler on Hoth before sending Harvest — it is a building, placed in Base, not a Shipyard hull", and under it a "Build a Recycler on Hoth" button opens the Base on that colony with the Recycler ready to place.
• The badge checks the colony the fleet actually leaves from. It used to search the whole empire, so a Recycler anywhere cleared it — and a launch from a colony without one went on to be refused one screen later. It now asks the same question the server asks, and when another colony does have one it says so: "Or launch from a colony that has one."
• A Recycler you have not unlocked yet sends you to the research, not to the base map. While the Tech Tree node is unresearched the line reads "Harvest needs a Recycler and it is not unlocked yet: research it — Forge T3 · Logistics 8 + prerequisites", and the button opens the building's card, which names the unlock. Before this, on a phone, the same button would have armed a placement for a building the game refuses to build.

1.194.3 — 2026-09-19

Changed

• In the system view your own colony now has a shape, not just a colour (#3445). Your colony used to be told apart by hue alone: the same ring, at the same size, as everyone else's, in cyan — and that same cyan was the selection ring, drawn larger, so the planet you had just tapped stood out more than the planet you own. Your colony now carries the four-tick reticle the galaxy map has used for "yours" since #3377, drawn across its ring, and its name reads Name · you instead of Name [TAG] — you know your own tag, and the shorter caption is less likely to be hidden when a system is crowded.
• The selection ring is no longer your colour (#3445). It is a neutral pale ring, set further out, so it cannot be mistaken for the mark that says a colony is yours.

Fixed

• Tapping a Pirate Beacon, Bastion or relic guardian no longer hides its PvE ring (#3445). The selection and the PvE double ring used to be the same ring, so selecting one of them took the warning ring away — exactly when you were looking at it to decide whether to attack.

Also

• The two maps still disagree on gold: on the galaxy map gold means "yours", in the system view it means someone else's colony. That is now a question of colour only — the reticle and the words say who owns a slot whatever the hue does — and it is tracked as #4105.

1.194.2 — 2026-09-19

Fixed

• YOUR HOME stays on the map with My colonies off (#3188). Switching off the My colonies layer used to remove the gold home crosshair and its YOUR HOME caption as well, so the map lost its answer to "where am I?". The layer now hides only the small gold marks on your other systems and the gold hover ring. Your home is always drawn.
• A takeover repaints its planet straight away (#3188). If another guild took a planet at the same tier and the system's leader stayed the same, the planet's cell kept the old guild's colour until something else on the map changed. The map now also repaints when a planet changes hands, including to or from pirates.

Also

• After panning with the mouse, the hover ring now starts from where you let go, not from where you pressed. On a still pointer, the tooltip no longer re-renders on every frame of a camera flight (#3188).
• The third item in #3188, Go to landing below the zoom where a tap opens a planet, had already been fixed in #3180.

1.194.1 — 2026-09-19

Fixed

• AI empires rank rivals on their real score, and quote numbers the way your screen shows them (#3438). The points table the AI empires plan from was sorted on the rounded-off figure it prints. Two empires whose scores printed the same could come out in the wrong order, and an AI could plan against the wrong one. They are now ranked on the real score, and the table and the "running away on points" line print their figures with thousands separators, like the leaderboard. Any number an AI empire quotes (Fund, contributions, directive points, scores) is now cut to a whole number exactly the way the game's panels cut it. Before, the AI's version could land one off on a negative fraction.

Also

• The war-week board and the season panel now share one points formatter instead of two identical copies. Nothing on screen changes: both still show one decimal. Whether those boards should show a decimal at all is still an open question (#3298).

1.194.0 — 2026-09-19

New

• The Colony Ship card tells you whether you can found another colony (#3115). Before, the Shipyard card said only "Founds new colonies". Nothing on it said whether the founding would be allowed, so a player could build, launch and fly a Colony Ship and be refused on arrival. Now the Colony Ship card, and the Colonial Ark's, carry one line: your colonies against the cap the game enforces, and what is stopping you when you are at it. The cap comes from your mother colony's Command Center and your Colonization level. For example: "1 / 2 colonies · room for one more", "1 / 1 colonies · Colonization 1 unlocks the next", "2 / 2 colonies · Command Center lvl 9 unlocks the next", or, past the last Command Center level, "No Command Center allows more than 10 colonies". Colony ships already in flight and a pending exodus count against the cap, as the server counts them. If the card can't read either one, it shows the numbers but does not promise a free slot. Reported by elis_aigen. Not included: the suggestion's optional links from the card to the relevant upgrades.

1.193.1 — 2026-09-19

Changed

• A locked research node now tells you what it will cost (#3442). Before, opening a locked node in the Tech Tree showed its effect and "Why it's locked", but no price and no research time. To budget for a destination research you had to pay through its whole prerequisite chain first. Now the locked card keeps "Why it's locked" and adds the unlock price below it, labelled "once the requirements are met". It shows the same resources, Dark Matter and research time you will see the moment the node unlocks. If the Laboratory is one of the missing requirements, the time is worked out at the Laboratory level you will need, not the one you have now. Nothing on the card can be bought yet: there is no button, and no red "you're short" warning. On the endless upgrade lines, the "flat price" note now appears on locked cards too. Suggested by Elis.

1.193.0 — 2026-09-19

New

Supply lanes explain themselves where players look for them (#3568)

Supply lanes were mentioned on the Colony screen, opened on the Missions tab behind a checkbox, and managed in a card. Their rules were not written anywhere a player could find, and players kept asking in chat what a lane is, where to create one, and whether it uses ships.

• Colony tab, Supply lanes card: a line now says where lanes are opened: on the Missions tab, from a Transport or a Pickup between two of your colonies, by ticking «Make this a standing lane». It shows even when you run no lanes, and even when the lane list could not be loaded.
• How supply lanes work, a fold-out on the same card, lists the rules:
  • a lane moves one resource each hour between two of your own colonies;
  • it has two triggers: keep an amount at the origin and send the rest, or send a share of what the origin produces;
  • only the origin needs a Trade Portal (the destination receives as if it had a level-1 Portal);
  • how many outgoing lanes a Portal runs;
  • the three limits on what a lane moves per hour: the origin Portal's capacity (reduced by an energy shortage), the destination Portal's intake, and the room left in the destination warehouse;
  • you can run one lane per route and resource;
  • no fleet flies;
  • cargo can be looted while it is in flight, and a batch in the air cannot be called back;
  • a full destination pauses the lane, and nothing is sent;
  • missed hours are made up only up to six;
  • running lanes show in enemy spy reports.
• Next to Open lane, the lane form now states the three facts you cannot learn without losing cargo: no fleet flies, the cargo can be looted in flight, and missed hours are capped at six.

The numbers in these rules come from the same constants the game uses to run lanes.

Not included: the Discord Library (#guides) entry the issue also suggests. The Library is not part of this repository, so this release cannot add it.

1.192.36 — 2026-09-19

Changed

• Allies under attack: every garrison on station now says when its watch ends (#3373). Before, the green Defending box listed each allied fleet guarding a threatened colony by name and ship count only, so you could not tell whether that cover would last six hours or six minutes. Now each fleet reads like "Kael (80), watch ends 14:30 (in 3h 20m)", with the countdown ticking live. At or past its scheduled end it reads "watch ending", because a strike about to land can hold a garrison in place a little longer. A note underneath says the time is a schedule, not a promise: the owner can stand their garrison down sooner, except once a strike is about to land. Suggested by Elis.
  • Known gap: the stand-down lock also lasts for one hour after a strike lands, and the note mentions only the lead-up.

1.192.35 — 2026-09-19

Fixed

• Fleet routes in Galaxy no longer break into long strips (#3390). When your Observatory knew where a fleet was going, its amber route line was always cut into exactly six dashes, however long the route. A long route read as a few big bars with wide, uneven gaps, two routes leaving the same system had different dash sizes, zooming in only made the gaps bigger, and the pattern shrank as the fleet got closer. Every route line now uses the same short dash: one cell long when you are zoomed in, and never smaller than a readable size when you zoom out. The dashes stay still while the fleet moves along the route. Very long routes get coarser dashes, still evenly spaced. Thanks to Elis for the report.
• Unidentified fleet markers show their full outline (#3390). The hollow triangle marking a fleet whose owner is unknown drew only the first half of each side. It is now a closed outline, which also makes it easy to tell apart from the dashed outline of a fleet whose last sighting is out of date.

1.192.34 — 2026-09-19

Fixed

• A queued building next to another one no longer loses its level badge (#3112). When two buildings side by side were both queued for an upgrade, their 10 → 11 style badges overlapped, and the one behind was hidden completely. That building showed no number at all, while every neighbour still showed theirs. Now the rear badge shrinks to the building's current level (10, still in the gold of a building under construction) and stays on screen. It never shows the destination level, which the building does not have yet. The badge is still hidden when even the short form would overlap, for example when two buildings are stacked on the same column. One side effect: the short badge now counts as drawn, so it can hide the plain level of a building standing right behind it on the same column. The upgrade arrow and destination are still in the Operations panel and the building sheet.

1.192.33 — 2026-09-19

Fixed

• The Shipyard price and your wallet no longer tangle into each other (#2773). In a hull's Specs panel, the per-unit price ("Each 4,000 Fe · 1,800 Xe · 200 Dt") and the "you have …" wallet shared one line, one on the left and one on the right. At some widths the price wrapped and the wallet did not, so two thirds of the ship's cost ended up in a column under your stockpile. They are now two lines, price first and wallet under it, and each one wraps only within itself. The wallet also wraps on a narrow phone now, where before it could spill past the edge of the panel.

1.192.32 — 2026-09-19

Fixed

• Dialogs stay usable from the keyboard when the button you were on disappears (#3425). Before: with a panel or dialog open, if the focused button was removed (Guild → Manage → Kick → Cancel) or disabled (a confirmation waiting on its request), Escape stopped closing the panel and Tab wandered onto the page behind it. Now Tab and Escape go back to whichever dialog is on top. A confirmation outranks the panel under it, and an inner panel outranks its parent. Tab lands on that dialog's first or last control, and Escape closes it.

1.192.31 — 2026-09-19

Fixed

• The battle line agrees with itself when you garrison an ally (#3976). Sending a defence to a guildmate's colony — from the War Table or from the raids panel's garrison door — the launch sentence stood the colony's guns in the vanguard, while the battle-line panel just below laid your fleet out with nobody in front: a fleet placed all in the rearguard read "take up to 26% of the fire" above "Rearguard 100%". The panel now shows the same wall the sentence counts. On the raids panel it names the ally's batteries or Fortress from the defence reading the row already shows, or shows no wall when that reading says the colony has none. Where nothing can be read, it says "their defences hold this line — we cannot read them".
• Row figures carry the same ceiling as the lane badge. When your fleet shares its rows (a garrison, a raid), each row in the battle-line panel now reads "≤ 26%", like the badge above it, instead of an exact figure the footnote said was only an upper bound. A solo attack and your Shipyard defence line still read exact figures.

1.192.30 — 2026-09-19

Fixed

• Synthetic empires on the catch-up ramp now plan with their real income (#3957). A newcomer empire below the catch-up threshold produces two to four times its base output. The agents' long-range planner, which weighs research against war, reinforcement and founding a colony, still valued the empire's income at the unboosted figure, while the agents' own economy decisions and chat already used the boosted one. The planner now uses the boosted figure too, for current income, for the return on Galactic Economy and for the output a new colony would bring. A ramped AI empire's choice between researching, attacking and expanding is now based on what it actually earns. Nothing changes for empires above the threshold.

1.192.29 — 2026-09-19

Fixed

• A supply lane's "up to N/h" now counts the lanes about to wake up (#3902). Say a hub filled overnight and some of its feeder lanes paused with "destination full". Once you freed room, every lane still running out of the same colony showed the colony's whole Trade Portal hour, as if the paused lanes did not exist. The next hourly sweep then woke those lanes and split the hour with them, so the cards promised more than arrived. The cards now split the Portal between the running lanes and any lane paused on a full or bounced hub, the same way the sweep does. A lane you paused yourself still takes no share. A lane paused on a full hub now shows the share it will get once room opens, where it used to show 0. The "inbound lanes" figure on a hub's Trade Portal counts those lanes too, so it no longer drops when one of its feeders is waiting for room.

1.192.28 — 2026-09-19

Fixed

• Desktop dialogs now hold the keyboard focus (#3437). Every desktop dialog (confirmations, the Market offer details, a building's «Full progression», the battle replay, the profile, dispatches…) told screen readers the page behind it was inert, but did nothing about the keyboard. Opening the Market offer details left focus on the offer's row in the blurred page, and Shift+Tab from the dialog's Close button went to «Open chat» behind the overlay. Now focus moves into the dialog when it opens, Tab and Shift+Tab loop inside it, and closing it puts focus back on the control that opened it. Escape still closes. With a confirmation stacked over another dialog, Escape now closes only the top one; before, it closed both. Closing a confirmation opened from a sheet, such as Kick in the guild's Manage sheet, now returns focus to that sheet, so its Escape and Tab work again (part of #3425; the rest of #3425 stays open). Thanks to Elis for the report.

1.192.27 — 2026-09-19

Fixed

• Agent outposts no longer treat Continuum Haulers as their garrison (#3738). The Hauler has no guns, but the agents' shipyard counted it as a warship in two places. An agent outpost holding only Haulers looked guarded, so its Shipyard bought no defenders. A plan asking for Haulers also grew the order with the agent's wealth, the way warships do. Now the agents use the same rule the game uses to decide who fights by default: a hull fights if it has guns or feeds the shared shield. So Haulers count as support ships, like Freighters and Probes, and the Bastion still counts as a defender. Measured effect today: none. When this shipped, no agent held, queued or planned a single Hauler, so no agent's behaviour changes yet. The fix makes sure the first agent to get Haulers doesn't leave its warehouse behind unarmed ships.

1.192.26 — 2026-09-19

Also

• Two agent behaviours are now held by tests, not only by comments (#3759). Nothing a player sees changes in this release. It pins down two things synthetic empires already do. A Peace shield bought from the guild emporium makes an empire hold fire for exactly the shield's length, not merely "for a while". And an empire retracts a stale relic muster its guild heard, out loud, before it goes off to join someone else's raid. Before, the tests would have stayed green if either behaviour broke. The issue also said no test covered the call that drains the muster. That was already out of date: a test added in #3772 goes red when the call is removed.

1.192.25 — 2026-09-19

Fixed

• Echelon tiers on the phone leaderboard (#3927). The Echelon tier of other commanders appeared only on the desktop leaderboard table, so on a phone you could see nobody's tier. The phone leaderboard cards and the podium now show it on their second line ("· Gold tier"), where it takes no room from the name. The desktop table is unchanged.
• No raw tier id in the Echelon panel (#3927). If the game added a new tier and your app hadn't updated yet, the last-cycle line could print an internal id ("you closed rank 2 in voidborn"). That line is now left out until your app knows the tier's name, the same way every other Echelon surface already behaved.
• Bronze-to-Challenger count as game words for the empires' rules-question check (#3927). Since the tiers were renamed, a chat question like "how do i get out of bronze" was not recognised as a question about the game. The five tier names and "echelon" are now game words. Nothing changes in chat today: the Discord rules tip that this check feeds is switched off on the live server.

Known limit: on a narrow desktop window with the chat panel open, a podium card at its minimum width can cut off the end of a long tier name ("· Challenger t…"). The name, Gate and points stay readable.

1.192.24 — 2026-09-19

Also

• The operator guide no longer points at a galaxy snapshot nothing reads (#3828). Nothing changes for players. Before giving the agents their voice back, the deploy guide said to delete the stored galaxy snapshot at a host path. The agents container never reads that path: its data lives in a Docker volume. On the server the command failed; on a host with a stray copy it would have succeeded without touching the real file. The guide now says the step is no longer needed: since #3829 the Observatory's bulletin checks how old the snapshot is. It also names the real path and the command that reaches it, in case a reset is ever wanted. The same command is gone from the example environment file, and a test keeps it out of both. Not done: nothing was deleted on the live server, because since #3829 that deletion is no longer advised.

1.192.23 — 2026-09-19

Also

• Nothing changes on screen in this release. It corrects a developer-facing note that overstated what the guild anchorage panel knows (#3819). The panel's per-system "a structure is already here" flag only covers your own guild's standing anchorages, never a rival's, but its documentation claimed it covered both. It is now documented as own-guild-only. The note also says where a rival's hold is read instead (the public anchorage map) and when the server really refuses a build next to one. The flag itself behaves exactly as before, and a new test pins that behaviour: a rival's anchorage, or your own ruin, does not set it.

1.192.22 — 2026-09-19

Fixed

• Phone mission form: the Send bar no longer hides the last lines of the form (#3562). On a phone, the floating summary and Send bar at the bottom of the mission form permanently covered the form's last lines. Scrolling to the very end never revealed them. That hid the note explaining why Warp Drive is greyed out ("the free discount already beats the warp on this route"), and it could sit on top of a fleet quantity editor the player had just opened. Now the bar settles into its place at the end of the scroll, and every line above it can be scrolled clear. It still floats over the form while scrolling mid-form. This also closes the third point of #821 (the bar covering the controls that resolve it). The other two points of #821, and #731 (the floating bar reads as a row of the list), are unchanged.
• Not included: a quantity editor tapped near the bottom of the screen still opens under the bar until the player scrolls. The Shipyard defense planner's phone save bar, which has the same shape and an estimated residual of about 13px, is unchanged.

1.192.21 — 2026-09-19

Fixed

• AI empires no longer pass over lava worlds when choosing where to colonise (#3893). Since lava lost its Data penalty, the agents had still been picking colony sites as if the penalty applied. That made lava the least attractive planet type to them, when by what the game actually charges it is the best plain type. The error was worth about two cells of travel, so an expanding agent would skip a lava slot next door for a rocky one farther out. Agents now rate each free slot by what the planet really yields, as the planet tooltip does. Rocky, gas and ice worlds are rated exactly as before, and nothing stored on any colony changes.

1.192.20 — 2026-09-19

Also

• Tighter tests for the Home panel's phone layout and the hook-order guard (#3780) — nothing changes in the game. This closes three gaps in the tests that stop the Home panel from crashing when a window crosses the phone breakpoint. A square window is now checked as portrait, as a real browser treats it. A missing Home hero now fails the layout test on its own instead of passing as "not full-bleed". The hook-order guard no longer mistakes a memo comparator or a capitalised useMemo value for a component.

1.192.19 — 2026-09-19

Fixed

• AI empires now account for the garrison they just paid for at an outpost (#3937). When an AI empire queued a standing garrison at one of its outposts, the server charged the outpost straight away, but for the rest of that turn the empire still counted the money as unspent and did not see the hulls on order. Anything it decided next in the same turn used that stale view: how many more hulls the garrison needed, whether the outpost could send or receive freight, and the warehouse figure it could quote in chat. The gap was one hull order (1,600–5,200 ferrolite) for at most one turn. Now the outpost's purse and ship queue update the moment the order is accepted, just as they already did at the empire's capital. A refused order changes nothing.

1.192.18 — 2026-09-19

Changed

• Phone: «In flight» opens on the missions in flight (#3580). Tapping In flight 1 in Fleets used to land on the Fleet ledger (one card per ship type) and Recent battles, with the one mission in flight, its remaining time and its Stand down button third, below all of it. On a phone the segment now opens on Active missions, then your standing lanes, with the Fleet ledger and Recent battles below them. Desktop is unchanged.

1.192.17 — 2026-09-19

Fixed

• The store's VIP strip no longer says your level twice to a screen reader (#3982). The strip prints "VIP 2" and draws the crest beside it. That crest was announced as an image called "VIP 2", so a screen reader read the level twice. Where the words already give the level, the crest is now treated as decoration. It still shows its tooltip on hover. Every other VIP badge still announces its level.
• Guild "At the gates" list: the VIP crest sits beside the applicant's name (#3982). In that row the badge's own margin was added on top of the row's spacing, which pushed the crest away from the name. It now uses the same spacing as the chat author line.

Also

• Tighter guards on the VIP crest art and on where the badge is wired in: each crest must now also fill its box along its shorter side, a blank crest image fails with a clear message, and the store, chat and guild placements are each pinned by a test (#3982).

1.192.16 — 2026-09-19

Fixed

• Typing a deep-space name into «Go to» no longer rebuilds the whole galaxy on every lookup (#4017). Each time a name like 4:2 NE3 was resolved to a cell, the game re-listed all 4,104 planets of the galaxy to check a single cell. Nothing a player could see went wrong, since one lookup still took under a millisecond. The cost showed up in the test that checks every name in the galaxy: 6,840 lookups took 4.4s alone and 9.9s under load, and it timed out in two full test runs on the same night, blocking unrelated releases. The planet list is now built once and reused. The same check now takes 0.28s alone and 0.48s under the same load, and it still checks the same thing: no printed name ever sends «Go to» onto a planet.

Nothing else changed for players.

1.192.15 — 2026-09-19

Also

• The safety net behind "log out everywhere" is now guarded by tests (#3857). Nothing changes for players. When you revoke a login, the server remembers that in a small cache. If that cache fails to load at startup, the server retries every two seconds until it succeeds, and after that it refreshes once a minute. That behavior was already correct, but nothing checked it: seven different ways of breaking it all passed the test suite unnoticed. Among them were a retry that gives up after one attempt, and retries that pile up without limit. Each of those now fails a named test.

1.192.14 — 2026-09-19

Fixed

• The advanced-core quote gets room to be read (#3570). On the Colony tab, pressing Core… on the Hypergate, Factory or Arsenal opened the commissioning quote squeezed inside the narrow building card. The cards beside it stretched into empty boxes, and each cost line wrapped so that the price and the amount you hold landed on different lines. Now the open card takes the whole row and the quote sits on its own line under the building. The total due is a small table with Due and Held side by side for Ferrolite, Xenogas, Data and Dark Matter, aligned and never wrapping. The building detail sheet shows the same aligned table. Opening the quote can move the card to a row of its own, leaving a gap where it was until you close it.

1.192.13 — 2026-09-19

Also

• Dark Matter core receipts are now guarded by tests (#3625). Nothing changes in play. Buying a Factory, Hypergate or Arsenal core already wrote a Dark Matter ledger entry. Until now no test checked it, so a refactor could have dropped the receipt without anything going red. Each core's purchase test now requires exactly one ledger entry with the right amount, player, colony and level. The internal notes that counted the Dark Matter spending points still said eight; they now say eleven, which matches the code. The Autonomous Factory research-card test now checks the quoted saving and the level-30 ceiling against the engine's own numbers, not against a copy of the sentence.

1.192.12 — 2026-09-19

Changed

• Full progression opens in a wide dialog on desktop (#3561). Before, a building's 30-level table was squeezed into the narrow building sidebar at every screen size. Scrolling right to read a level's effect scrolled its price off screen, and the table had its own small vertical scroller nested inside the sidebar's. Now, on a wide screen, "Full progression" opens a dialog sized to the table. Level, cost, build time and every effect column sit side by side on one line, and there is a single scroll. The header row stays pinned, and so does the level column when the window is narrower than the table. The dialog opens on your current level and takes the keyboard, so the arrow keys and Space scroll the table. On phones nothing changes: the table stays inline in the bottom sheet.

1.192.11 — 2026-09-19

Fixed

• The raid picker no longer launches ships that already left the hangar (#3598). Before: if you had picked 10 Corvettes and the same colony's hangar then refreshed with fewer (ships sent from another session, losses, a transfer), the Corvette card disappeared but "Aboard" still counted them, Launch stayed enabled, and the raid was refused with "insufficient ships" over a number you could no longer see. Only switching launch colony cleared the picker. Now every refresh trims your selection to what the hangar still holds, drops hulls that are gone, and a line in the picker names the hulls that were adjusted. An open "Leave without defences?" confirmation whose fleet no longer fits is closed instead of sending the old count.

1.192.10 — 2026-09-19

Fixed

• The destination card no longer shrinks while you type coordinates (#3531). On a phone, every digit typed into a mission's destination started a new scan, and the card showing who lives at that slot collapsed to one line of «Reading the slot…». The coordinate fields and «Use these coordinates» jumped up with it, so a tap aimed at the next field often committed the destination instead. The card now keeps the tallest height it has had while the composer is open: it no longer shrinks while a scan runs, on an address outside the galaxy, or when the next slot turns out to be shorter. Nothing below it moves up under your finger. The cost is some empty space under a short slot after a tall one. Reported by QQQQ.

1.192.9 — 2026-09-19

Fixed

• Phone replay: the battle title no longer covers the shield strip (#3606). On a phone, a battle replay drew its title line (who is fighting, and when) and the shield's HP strip in the same top-left corner, so the names ran across "Shield 1,200 / 1,200 HP" and, after the first salvo, across "Shield BROKEN". In raids without a shield the title ran across the squadron chips instead. The corner is now a stack: the title on top, the shield strip below it, the squadron chips below that. Desktop is unchanged, and the top-right corner stays the close button's.

1.192.8 — 2026-09-19

Fixed

• Relics and faction bastions are worth attacking again (#4026). Their garrisons are sized against the strength of the field, and until now a single empire standing far above everyone else set that price for the whole server — so the content nobody could reach was priced for one player who was not even trying to reach it. The calibration now skips an empire that has run away from the field.

  Measured on the live server as this shipped: every relic garrison fell from 14,566,380 hulls to 1,756,355, and the faction bastions were re-tuned to 631,515. All forty relic sites were re-priced on the first pass after the release, not only newly seeded ones.

  Nothing about the standings changes for anyone. Rank, prizes, the season podium and bounties read exactly as before: this changes what the PvE objectives cost, not who is ranked where.

• The event sector deliberately keeps the price it had. Its neutral garrisons already had their own rule for a runaway empire — they skip the top of the field and price on the five below — and applying both rules at once would have quietly made the six control planets around eighteen percent cheaper for everybody. They are untouched.

1.192.7 — 2026-09-19

Fixed

• A raid refusal no longer stays on screen for a target you have moved away from (#3601). Suppose a raid launch was refused with "Target will be under night protection on arrival." Before, the sentence stayed on screen after you retyped the coordinates, so it read as a verdict on the new target. Now each refusal is tied to the launch that earned it: the address, fleet, launch colony, stance and gathering time. It hides as soon as the form describes a different launch, and comes back if you put the form back. The same applies to the join door's "N minutes short" line and to the night refusal that suggests a gathering time. Refusals about the raid itself are not tied to the form and stay up until your next action, as before. That covers a full raid, a missing research node and a server error. The server's night-protection check has not changed.

1.192.6 — 2026-09-19

Fixed

• The Colony card's Place button now opens the Base already placing that building (#3517). On the Colony screen, the "+" on a Suggested upgrades card for a building not built here yet is named "Place <building> on the base map". Before this fix it opened the Base with nothing selected, so you had to find the same building again in the Build catalogue. On a phone, the Base now opens straight into placement for that building, with the ghost, the valid cells and the "Build here" confirm. On desktop it opens that building's card with its Build button, because a desktop placement would commit on the first click on the map. The button still opens when you are short of resources or the queue is full, and the server refuses the build only after you aim it; the card's red price and queue-full note are the warning for now.

1.192.5 — 2026-09-19

Fixed

• Switching hull in the Shipyard no longer shows two ships at once (#3582). When you picked a new hull, its illustration came up straight away, but the previous hull's rotating 3D model stayed painted underneath until the new model loaded. Going from an Interceptor to a Battleship, for example, showed both ships on top of each other. Now the 3D view is hidden as soon as you pick a new hull, and it comes back when that hull's own model is ready. In between, you see the new hull's illustration alone. Thanks to QQQQ for the report and for spotting the ordering constraint.

1.192.4 — 2026-09-19

Fixed

• Phone locked screens no longer tell you to build what is already building (#4008). Once the Shipyard or the Laboratory was placed and building, its locked screen still read «Locked · no shipyard yet» and «Build the Shipyard» right above «Already under construction», and the cost line priced a build you had already paid for — sometimes in red, «short of Ferrolite», because the payment had just emptied your stock. Now, while it builds, the plate reads «Under construction», the headline «Shipyard under construction» (or Laboratory), and the cost line «Paid …» with what it cost, in gold. The button still takes you to the Base to watch it build. Left out: a countdown on the card — the build's live timer is on its Base queue row, one tap away.

1.192.3 — 2026-09-19

Fixed

• Empire tab: the stock tiles no longer read «nothing in flight» while the fleets are still being counted (#4003). On a phone, the Empire tab's ferrolite, xenogas and data tiles add the cargo your fleets carry to what has landed. Until the list of fleets in flight loaded, and forever if it failed to load, the tiles showed landed stock only with no remark, so an outbound 500-ferrolite transfer could read 1,200 on the Empire tab while the Overview said 1,700 · incl. 500 in transit. A line under the tiles now says «still counting the fleets in flight…» while the list is loading, or «fleets in flight unavailable» when it failed. Once the list is in, the tiles show «N in transit», or nothing if no cargo is in flight.

1.192.2 — 2026-09-18

Fixed

• Galaxy map on a phone: relic and bastion names no longer collapse into an unreadable band (#4019). Opening the Galaxy tab at its default zoom on a phone printed the name and effect line of every relic and bastion at once — dozens of captions colliding across the middle of the screen, over the planet dots. Now, at the zoom the map opens on, each relic and bastion shows its sign only (the gold diamond of a prize, the chevron of a bastion, filled while held); the name and what it gives arrive one pinch in, from sector zoom, and go away again when you pull back out. Desktop follows the same rule: a window that opens on the whole galaxy gets the names on the first wheel notch, together with the control rings and the coordinates.
• The phone's anomaly notice at Astrometrics 0 no longer ends in an ellipsis. It read «Astrometrics 0 — no signal reaches y…» on a 390 px phone; it now reads «Astrometrics 0 blocks signals.» and fits the narrowest phones whole.

1.192.1 — 2026-09-18

Fixed

• Deleting your account works again if you ever set up a trade lane. Until now, "Delete account" answered an error to every commander who had opened a lane between two of their colonies — the lane rows were never cleared before the account, and the database refused. The same path serves the moderation door and the two automatic sweeps, so those quietly skipped these accounts too. The lanes now go with the colonies, and a schema-wide check keeps the next table from being forgotten the same way. (#4021)

1.192.0 — 2026-09-17

New

• The First Convoy card now offers to ping you when it docks. A new commander's convoy arrives 24 hours after founding — the return notification for that moment already existed, but nothing in a first session ever registered a device to receive it: the notification card waits for a battle, and the browser subscription lived in a panel newcomers never open. The armed convoy card now carries one quiet row — "Ping me when it docks" — that registers this device on tap (the browser's permission prompt, or the app's, fires only then). Success, failure and closed doors each say so in place; the offer never appears where it cannot be honest — a denied browser, an already-registered device, a chain convoy — and it never re-asks a no you gave elsewhere. (#4013)

1.191.1 — 2026-09-17

Also

• Nothing visible changes in the game. Behind the scenes, every new account now records whether it signed up from a handheld or a desktop browser, read off the User-Agent the browser already sends — beside the existing country and door stamps, never updated afterwards, and empty when there is nothing usable to read. It exists to answer one question from the first-five-minutes teardown: whether the half of new commanders who never take a single action are the mobile half. (#4010)

1.191.0 — 2026-09-16

New

• Phone: a locked screen for the Shipyard, Research and Fleets (#4006). A new commander opening those tabs used to see the real screens with nothing to do on them: a hangar of hulls under a red «Shipyard required», a research bench at zero with every technology greyed out, eight mission types each saying why it could not fly. Now each of the three shows one screen until the building that opens it exists: the building itself, dimmed and locked; what to build, with its cost, build time and the Command Center level it needs; one button that opens the Base already placing it — or, if the Command Center is too low, opens the Base on the Command Center so you can raise it first; and a glimpse of what opens (the first hulls, technologies or mission types, and how many more follow). Fleets keeps its alarm pills and the First Convoy card above the locked screen, and unlocks as soon as any of your colonies has a yard or a ship. Desktop is unchanged.

1.190.1 — 2026-09-16

Fixed

• The next galaxy sectors are no longer readable in advance (#2630). Until now the type and richness of every planet came from a formula the game client carries, so a player who read the app's code could work out which worlds of the sectors the galaxy has not opened yet would be rich, before anyone could settle there. Sectors that are not open now have no such answer: the number that would give it does not exist until the day they open, in the client or anywhere else, and a client that predates an expansion computes wrong worlds for it rather than right ones. Every sector that exists — the twelve core sectors and the Event Sector — keeps exactly the planets it has today, pinned world by world against a snapshot taken before this change.

1.190.0 — 2026-09-16

New

• Phone: the Menu is now the Empire tab (#3999). The fifth slot of the bottom bar used to open a drawer of thirteen identical tiles with the settings underneath, and nothing on it showed your colonies. It is now a screen called Empire: every colony as a chip with its status (building, needs attention, in danger) — tap one and it becomes the colony you are looking at everywhere; a «Needs you» list, one line per colony with the action named (Build, Yard, Storage, Defend) that lands you on the right screen of that colony, folded past three lines behind «Show all N»; the doors to Base, Colony, Overview, Shipyard, Research, Campaign, Inventory, Market, Activity, Season, Leaderboard, Crucible and Guild with the same unread counts as before (Guild now shows the number of raid calls waiting, not a dot); your empire's total stock and hourly rate, counting cargo in flight and saying how many colonies are at their storage cap; and one Settings door for what the old drawer kept at its bottom — audio, Discord, night window, leave, changelog, log out, account. The slot is called Empire whichever of those screens you are on, and tapping it brings you back to them. Desktop is unchanged.

1.189.2 — 2026-09-16

Changed

• Guild donations: the Treasury sheet now says what a donation takes as a share of your warehouse, and that the reward is flat by design (#3110). A donation costs 0.6% of your home warehouse's capacity — a figure that grows with your empire — and pays the same 10 Fund / 10 Scrip to everyone; the sheet printed the absolute price and the reward and nothing else, so a large empire read a six-figure price next to a ten-point reward as a mistake and asked for a rebalance. The design call keeps cost and reward exactly as they are (guild contribution measures showing up, not wealth, and Scrip is priced as attendance) and discloses them instead. Before: «Costs 122 ferrolite, 73 xenogas, 30 data from Atlas — gives 10 to the Fund and 10 Scrip to you.» Now: «Costs 122 ferrolite, 73 xenogas, 30 data from Atlas — the ferrolite is 0.6% of your home Warehouse's capacity (20,480), and the donation takes 12.2% of the ferrolite, 14.6% of the xenogas and 30% of the data you hold there — gives 10 to the Fund and 10 Scrip to you.» If the warehouse cannot cover it, the sheet says «Not enough resources» and which material is short, and the Donate button stays dark instead of sending a tap the server would refuse. Below the button the rule is written out once: every commander pays the same 0.6% of their own home Warehouse's capacity and earns the same 10 Fund / 10 Scrip — by design: Scrip counts attendance, not wealth. Nothing about the donation itself changed — cost, reward, who may donate — and the iOS/Android app shows the previous line until its next update.

1.189.1 — 2026-09-15

Changed

• The anchorage bar now says how a notch is placed — and that a war strike spends a guild reply (#2477). Another guild's anchorage read as untouchable: the system card said «under attack 2/5», the Guild tab said how a notch clears, and nothing said how one is placed. Under a rival anchorage the card now says that a notch is placed by a battle won against a colony of that guild in that system, how many bring it down, and that the bar loses a notch for each day since the last blow; the Power band and the bounty apply as anywhere else — and, folded under one tap, the war term: while a war week or a Match is open between the two guilds, a strike of your own far below your Power spends a reply your guild holds against them instead of raising your bounty, and is priced as usual when it holds none. The count stays where it is read live: the War Room, or the Match room. A reader with no guild sees the rule without the reply clause; your own anchorage keeps the defender line it had. Words only — the breach, the band and the ledger are unchanged (the rule itself shipped in v1.189.0, #2638). #2473, the relief ratio that keeps a strong guild from clearing a notch by winning, stays open.

1.189.0 — 2026-09-15

New

• Strikes between two guilds at war no longer raise the stronger side's bounty — the Grudge Ledger now counts for the whole guild (#2638). Before: in a declared war, hitting a weaker enemy put a bounty on your head exactly as an unprovoked attack would, and the only relief was personal — a hit you took bought you one free reply against that one attacker. A strong member whose enemies avoided him earned nothing and paid for every strike he made for his guild. Now, in a declared war or an open League Match, every blow the enemy guild lands on any of your members buys your guild one reply; any member can spend it — a strike under band on any member of theirs spends one instead of raising a bounty — after the personal credit, under the same whole-charge rule (never on a share of a raid pack's price), and only while the two guilds are at war at the instant the fleet lands. Replies are bought only with blows received: a guild that declares war and is never struck holds none and pays as before; a fleet that lands after the verdict pays as before; a guild that is not part of the war earns and spends nothing. Replies expire seven days after the last blow, like the personal ones. The War Room now says how many replies your guild holds against the enemy, and the diary says when a strike spent one ("War reply — no bounty"). Unchanged: the Power band and its price, what a bounty pays, who may declare war, the personal reply. Reported by Tibeh01.

Also

• Ledger rows are kept when a guild disbands: they are unreachable once the guild is gone, and deleting them would have widened the lock every dissolution holds (#2638).

1.188.0 — 2026-09-15

New

• Market offers can be taken in slices, at the seller's price. An offer used to fill whole, so under the daily intake cap a fresh account could take 7 of the 130 open offers on the board (95% were too large for its day) and the best rates were only reachable at scale. Now the trade sheet has a "Take" field with 5% / 25% / 50% / All / "Fits today" shortcuts: you receive that share of what was offered and pay the same share of what was asked, at the seller's own price per unit and never a better one — where a share does not divide evenly, what you receive rounds down and what you pay rounds up, by at most one unit. Slices are at least 5% of the offer as posted, and one that would leave under 5% takes the rest, so an offer settles in at most twenty fills and never leaves dust on the board. Split offers show "X of Y left" on the board, the diary says "took X of Y" / "sold X of the Y offered", and the composer names the floor for sellers. The daily intake cap did not move: every slice counts against both parties' day per unit, exactly as a whole fill does, and a slice that would cross the cap is refused, not trimmed. Existing offers were not repriced; sellers can still cancel the remainder at any time. Older app builds see a partly filled offer as a smaller whole offer they can take. Agents keep buying whole offers for now. With the 5% floor, 86 of the 130 open offers become reachable for a fresh account; the 44 largest still are not, because even 5% of them exceeds a fresh day. (#2569)

Also

• When an offer was cancelled or accepted while a slice of it was settling, the cancel refunds what is actually left and the accept re-plans against the row as it stands — and a slice planned against a board that had since moved is refused ("This offer changed while you were looking at it") instead of settling at a size the sheet never showed. (#2569)
• Agents record what a market purchase actually settled, not what the board showed a tick earlier. (#2569)

1.187.13 — 2026-09-15

Changed

• Phone: the First Steps guide stops covering every screen (#3985). It used to sit above Fleets, Shipyard, Research, Season, Guild, Campaign and every other tab, taking almost half of the screen. Now those tabs show one line at the top — progress and the next step, tap to open — and the whole guide stays on Home. On the Base it opens folded: the strip shows where you are and, when the next step is a building you can afford, a GO that starts placing it in one tap.
• Phone: «Build it» builds (#3985). Tapping it on Home lands you on the Base already placing that building; before, it only switched screen and showed the same button again.
• Phone: one Cancel while placing (#3985). Aiming a cell no longer stacks two panels, each with its own Cancel, over the map.
• Phone: a dot on the Home tab when a reward is ready (#3985), so a step you finished while on another screen is not missed.

Also

• On phones the First Convoy and «Find your guild» cards now ride the Home screen only (the Fleets tab keeps its own convoy line, the Base keeps its folded copies). Desktop is unchanged.

1.187.12 — 2026-09-15

Changed

• A raid's target may intercept its incoming contingents — once interception is switched on (#2188). A joint raid aimed at you could not be intercepted at all: every contingent was refused to every picket as "one contingent of a joint raid", which made a coordinated strike better protected than a lone attack from the defender's seat. The design already said otherwise (§3.9 of the fleets-in-transit design: a raid is interceptable per contingent), and its own §3.8 table contradicted it; the engine had followed the table. Now the rule is the one §3.9 states: the raid's target may picket or pursue an incoming contingent; a bystander whose lane the leg crosses may not; the raiders' own side may not. Nothing changes for players today: interception on novarchy.com is still in shadow mode, and in shadow a raid leg is refused exactly as before — the rule takes effect only when the operator switches interception on. The design document's table now agrees with its own text.

• When you tap a fleet that is one contingent of a joint raid and the game refuses the interception, the sentence now says whose counterplay that is and why nobody gets it yet — "only the empire the raid is aimed at may ever meet a contingent in transit, and nobody at all while interception is still being measured" — instead of stating a rule the game no longer follows (#2188).

Also

• The raid resolver no longer trusts that nothing can touch a contingent in flight: its write to each leg is now witnessed, so a contingent whittled by an interception is fought with the fleet it actually has, never the fleet it left with (#2188).
• When a contingent leaves a raid before impact, the rest of the pack gets an Activity row saying so, filed under Combat (#2188). Under the current interception loss cap a contingent is only ever whittled, never removed, so this row will not appear until that cap changes.

1.187.11 — 2026-09-15

Changed

• The anomaly sheet says its payout is fixed by tier, and what that sum is worth in your hours (#3116). An anomaly pays an absolute sum — 2,000 / 6,000 / 15,000 / 35,000 resources by tier, or 1–5 Dark Matter — while every other faucet in the game is priced on your empire. That is by design (paid in the server's hours, never your own), but nothing said so, and a large empire read a T3's 15,000 as a mistake. Before: the planet sheet on an anomaly slot showed the tier, the signal strength and, with a Deep Observatory, the outcome — never a figure. Now: under the signal it states the tier's fixed payout (resources, wrecks at 80%, Dark Matter — or the one leg the Observatory revealed) and what that sum is worth in your own hours of production, from the same rate the Contracts bar is priced on: «about 3.5 hours» for a median empire, «about 20 hours» on day two, «under a minute» at the top of the field. It also names what binds: about one signal in seven is a pirate ambush that pays the resources only if your fleet wins; Dark Matter has a weekly cap, past which the visit pays the tier's resources instead (a visit that only partly fits pays both); and resources come home as cargo and stop at your warehouse cap. A commander producing nothing is told there is no production to measure against; before the colony list loads, the fixed sum stands alone. No reward, tier, cadence or cap changed — the sheet only tells the truth about them. A native app built before this release shows the sheet as before.

Also

• The engine's anomaly reward constants now carry the note that they are absolute on purpose, and the expedition design doc records «a lost colony that gives ships» as suggested and not adopted (#3116).

1.187.10 — 2026-09-15

Fixed

• The VIP level next to a name is a crest again, not a stray digit (#3980). Supporters carry a numbered crest beside their name in chat, on the leaderboards and in the guild roster. Until now the crest was invisible at that size: the art filled barely half of its image, so at name height it shrank to a dark smudge, and the level number drawn over it was the only thing left — a bare white "6" floating between the name and the guild tag. The ten crests have been re-cut to fill their frame, and the level now sits beside the crest, in gold, on the same line as the name. The store's VIP strip shows the crest next to the words "VIP n" instead of printing the number twice.

1.187.9 — 2026-09-15

Changed

• Observatory reach rescaled to the v2 field — a rescale, not a buff (#3317). The field update of 8 September spread the planets out and made a cell a third shorter, but a dish still reached the same number of cells: a level-2 Observatory saw 1.7 cells, less than the two that separate it from its nearest neighbour. Each reach constant is now the old one times the same cell ratio the field update used for the picket leash (860 / 510): a level-2 dish reaches 2.9 cells, level 5 reaches 4.7, level 10 reaches 7.7, level 15 reaches 10.7, and the cap is 17 cells from level 26 up (it was 10 from the same level). Every colony with an Observatory sees further and none sees less; contacts that were just outside your rim yesterday are inside it today — that is the rescale, not a fault.
• The raid join rule did not move; the refusal now says by how much. A guildmate whose fleet cannot make the raid's hour at its own top speed is still refused — nobody teleports — but the row reads «Your fleet: 41m to get there, the raid lands in 20m — 21m short» and a refused join says «it needs 1h2m, the raid lands in 58m — 4m short», in the row you pressed. The number counts the leader's gathering time, so a delayed raid is honest about it. Older app builds keep showing the refusal without the number.

Fixed

• «Go to» on a deep-space name no longer lands on a planet (#3317). A battle report could print 1:1 SW5 for a fight in open space while that exact cell held a planet with its own coordinate; typing the name now goes to the nearest empty cell instead.
• The picket sheet proposes the station nearest the lane's open middle (#3317). It used to throw away any station within eleven cells of either end, although that rule is about where the meeting may happen and not where the watch stands; on the current map 548 lanes get a closer station and none a worse one. The sheet's warning now says «cells from the live middle of that lane», which is what the number always measured.

Also

• The sortie timing is unchanged; the comment that said a bastion's crossing «takes hours» now says 43 cells, about 44 minutes at base Propulsion, and that the two-to-six-hour sortie delay is what carries the timing.
• The standoff and leash retune (items 4 and 5 of #3317) had already shipped with the field update; item 7 (plate rebuild cost on phones) is still open for measurement.

1.187.8 — 2026-09-15

Fixed

• Defend surfaces say one thing about a shared row (#3936). On a defend of your own colony the War Table printed a bare Rear 26% badge under a launch disclosure saying "up to 26%" and above an accordion footnote saying "your real share can only be lower" — three sentences on one screen disagreeing about the same number. The badge hedged only on a wall it could not read, while a garrison always fights in line with the colony's own fleet. Now every defend — your own colony, your own colony with no guns, an ally's — shows the badge as a ceiling (≤ 26%), the sentence as "up to", and the footnote as "your real share can only be lower". A solo attack is unchanged: its share is exact and reads exactly as before.
• Raids panel, garrison door: under the sentence naming the colony's guns, the battle-line accordion used to explain "In a raid every allied fleet shares the same three rows". It now explains the garrison: the defended colony's emplacements hold the vanguard and its own fleet stands in line with you (#3936). The raid doors keep the raid clause.
• Phones: with the ceiling prefix on every defend, the lane badge and the Stands buttons could break between the ≤ and the number at 375px and 320px. They keep their words on one line now; the Stands row wraps by whole buttons instead (#3936).

Also

• The ceiling-or-exact decision is stated once, in the shared engine package, and every launch surface reads it — so the three sentences cannot drift apart again (#3936).
• Left out from the issue, on purpose: modelling the ally's unreadable wall inside the raids garrison door's accordion. The accordion names the buildings it is handed, and nobody can read an ally's; it would print a building nobody verified.
• Installed iOS/Android bundles and cached tabs keep showing the old badge until they update; nothing breaks.

1.187.7 — 2026-09-15

Changed

• The Orbital Cannon sheet says where its 25% cap binds (#3481). The sheet has always shown the opening salvo ("9,000 RAW damage" at level 30) beside a note that it is "capped at 25% of the attacking fleet's total hull", and never said where the two meet — so the cap read as the thing to think about when, against any real fleet, it is not. A new row under the volley, Where the cap binds, prints the figure at your gun's own level and at your colony's current power balance: "The 25% cap only binds against an attacking force under 36,000 total hull, all contingents together (72,000 when it brings a Breachhead, which halves the cap); above that the salvo no longer grows with the fleet: 9,000 RAW damage." At level 5 it reads 6,000 / 12,000 / 1,500. Two figures, because an attacking Breachhead halves the cap; both are computed by the same function the battle uses, so the sheet cannot drift from the engine. A colony with no Cannon, or one whose power shortage has throttled the gun to zero, prints no line. Nothing was rebalanced: no cap, curve or static-defence number moves — the issue's arithmetic predates the 2026-09-10 Batteries/Fortress progression, and nobody on the server has built past Batteries 25 / Fortress 21 / Cannon 17 (humans: 22 / 14 / 14). Older app builds keep showing the sheet without the new line until they refresh.

1.187.6 — 2026-09-15

Fixed

• Every production figure the app computes for itself now carries the VIP production multiplier (#3963). A VIP player read the HUD +X/h tiles and the live counters between polls, the Colony tab's +x/h rows, the Overview rates, the launch-pad stock «Fill max» writes, the Shipyard's fleet ledger, the build advisor's first frame and the Building Sheet's «Contribution now» row at 1/mult of what the server credits — 1/1.12 at VIP 6 — and, since v1.187.3, the advisor's first frame disagreed with the server's list until the answer replaced it. Client and server now agree for a VIP account on the same state. Nothing about what VIP does, any cost or any score has changed; at VIP 0 every figure is the one it always was. An installed app bundle keeps printing the low figure until it updates.

1.187.5 — 2026-09-15

Changed

• The Home hero now names your empire; your colony is its subtitle (#2763). The first screen of every session is a render of your empire's orbital hub, and its big heading printed the name of the colony you were viewing — while the Base tab has you placing that same colony's buildings on a planet's surface. QQQQ (arcticpaladin) asked which one the colony is. The picture was always the hub, deliberately; the label was the contradiction. The heading is now your commander name with your guild tag, exactly as the header wears them — it follows a rename or a guild change the moment the header does, from the same source — and the viewed colony's name sits under it as the subtitle. The eyebrow still reads Your empire, in orbit; the five pins, the next-move card and everything on the Base tab are unchanged, and the orbital-named buildings stay where they are on the plate. Long names wrap inside the picture on a phone: checked with a 20-character name and a tag at 375px and 320px, the label stays clear of the pins. An installed iOS or Android build keeps showing the colony name as the heading until its store build catches up; nothing it calls has changed.

1.187.4 — 2026-09-15

Fixed

• The Building Sheet's Upgrade button, the build menu's row buttons and the ranked rows now refuse in the same words, from one rule (#3954). Two ships ago the ranked rows and the Command Center callout were put behind a single guard (#2210): Queue full. outranks Not enough resources., then Upgrade to lvl N. The sheet's own Upgrade and Build buttons, and the build menu's Upgrade and Build row buttons, still computed the same rule on their own and worded it differently — Not enough resources without a period, Queue full (2/3) with the count — so the same refusal could read two ways on one screen, and a guard relaxed on one surface did not reach the others. All four now read the shared rule: the shared refusals print its sentences, in its order, and the queue count is read in the build menu's own Queue: n/m slots in use header. Each surface keeps only its own guards on top, first — max level, this building already queued, the Command Center cap (Command Center lvl N required on the sheet, CC lvl N required on the narrow row, as before), a core still to commission. The narrow menu row still says Upgrade as its call to action; the sheet still says Upgrade to lvl N.
• *The sheet's Under construction… button no longer takes a tap it cannot honour (#3954). While a building's upgrade was already in the queue and another slot was free, the sheet printed Under construction…* but the button stayed live: the tap sent an order the server refused (one slot per building), and the sheet showed nothing for it. The build menu's copy of the button was already disabled in that state; the sheet's is now too.
• The hint under a disabled Upgrade no longer promises an Enter that has nothing to confirm (#3954). Since #2596 the line under the sheet's Upgrade reads U jumps to Upgrade · Enter confirms the buy. Under a greyed button — short of resources, queue full — the second half was false, and pressing u while the bar was already on screen did nothing visible, because a disabled button cannot take the focus. Now the second half of the line is the reason the button is greyed, in the same words as the button: U jumps to Upgrade · Not enough resources. — and u moves the focus ring onto the bar that holds the button and that line, so the key always lands on the sentence that explains why. The rule from #2596 is unchanged: the key never spends, and Enter on that bar buys nothing. Phones still show no hint and have no key.

1.187.3 — 2026-09-15

Fixed

• The build advisor prices VIP players on what they actually produce (#3958). The server-side advisor — the Command Center card and the colony sheet's suggestions — reasoned on a colony rate that left the VIP production multiplier out, so a VIP player was advised on every producer at a fraction of its real hourly figure (four fifths at the top VIP level) and had producers ranked against storage and defences on that lower number. The advisor now reads the same rate the game credits the colony with, VIP included. Nothing about VIP, what any building costs, or how suggestions are scored has changed: only the input the advisor was given. On a replay of the affected accounts, one colony's list re-ordered (a Warehouse moved up as the real rate fills storage sooner); the rest kept their order. The client's local first-frame list still omits the multiplier until #3963 lands, so for a VIP player it can differ from the server's list for the moment it takes the server's answer to arrive.

1.187.2 — 2026-09-15

Changed

• Crucible — Monday prizes: painted crates on the podium (#3951). The three podium cards of the Monday prizes showed the same drawn hex plate with a cube glyph, tinted gold, silver and bronze. They now carry painted renders: a sealed supply crate per metal — radiant gold with teal seams, polished silver, weathered bronze — in the same family as the tier crests, each with its rank's glow. Ranks 4–10 are unchanged.

1.187.1 — 2026-09-15

Fixed

• The contracts board now counts the Industrial Edict (#3921). An empire under the Throughput edict produces 8% more than one without it, and the contracts board never knew: every card is priced in "hours of your production", and the hourly figure behind that price was read without the edict — 8% below what a Throughput empire actually banks — for as long as the edict has existed. The board now reads the same rate the colonies accrue at. For a Throughput empire that means: the daily and weekly cards, the Wrap and the Chest pay 8% more; the Win 3 battles bar — one hour of your production per battle, measured at claim time, between the 1,000 floor and the 20,000 cap — rises by the same 8%, so a win that only just cleared the old bar no longer counts; and the battle weekly's pay, which is capped by the fleet value you actually destroyed, moves only where that cap was not the binding one. Nobody else moved: an empire with no edict, or under Construction or Shipbuilding (whose production term is exactly 1), is priced to the same integer as before, and the edict itself, the reward formulas and the bar formula are untouched. Measured on the day of the ship, no empire holds Throughput and only two accounts have ever unlocked edicts at all, so no card in flight changes; from here on a Throughput empire is paid for what it produces.

1.187.0 — 2026-09-15

New

• The shipyard says an order lands whole — and what splitting it would buy (#2108). Players asked for ships to leave the yard one at a time as each finishes. The answer is no: an order is one batch with one finish time, and every chrono shard, guild push and shipyard upgrade shortens that one instant, so the batch lands together when its last hull finishes. What the form never said, it now says, under the build time of every multi-hull order, on desktop, in the phone count sheet and in the specs sheet: "All 300 hulls land together when the last one finishes, in 8h 20m." When the colony has more than one free slip it adds the thing that actually gets hulls out sooner, with a real figure: "Split it across your 3 free slips and the first 100 land in 2h 46m" — the same build-time arithmetic the yard uses to stamp the order. A single hull is told nothing, a colony with one free slip is not told to split, and a split that would not land sooner is not offered. The long-order warning keeps "cannot be cancelled" and the Shipyard lever and no longer repeats the split advice. Nothing about delivery, slips or prices changed.

1.186.1 — 2026-09-15

Fixed

• VIP crest and border in chat (#3950). A supporter's VIP crest next to their name and the border around their portrait appeared in chat only on the echo of their own message, the instant they sent it, and vanished for everyone on the next refresh — the page the chat polls never carried the donor tier and VIP level, so every author read as a player with neither. Every chat page now carries both, and the crest and border draw for every supporter's messages, on first open and on every refresh.

1.186.0 — 2026-09-15

New

• u jumps to the selected building's Upgrade button — on desktop, and it never spends (#2596). Spacca-Mazza asked for a shortcut to upgrade the selected building. Until now the keyboard reached the colonies (1–9, c) and nothing on the base: upgrading the building you had just clicked meant finding its Upgrade button in the sheet by hand. Now, with a building selected on the base map, u puts the keyboard focus on that building's Upgrade button and scrolls it into view; Enter on the button is the purchase, exactly as a click would be. A line under the button says so: U jumps to Upgrade · Enter confirms the buy. The key itself buys nothing on any path — held down, typed into the chat or a rename box, pressed under a modal, pressed with nothing selected — and a key already held when the focus arrives (Enter auto-repeat, or a Space released on the button) cannot press it either: only a fresh press on the button does. Phones have no key and show no hint. The rule this sets for every future shortcut is now in the design doc: navigation keys fire freely; a key may spend only inside a surface the player opened on purpose, which names the key and offers Escape.

1.185.1 — 2026-09-15

Changed

• Accelerator flips are now recorded, and nothing else about them changes (#1585). Switching the Particle Accelerator off has been a defensive lever since the shields started balancing the accelerator-adjusted energy map: a defender who sees the inbound alarm can free the draw instantly. The issue asked whether that flip should carry a price or a cooldown, and the design decision on it is no — the flip stays free and instant, as before. What was missing was any way to tell whether defenders actually flip because an alarm sounded: a flip left no trace at all. Every real flip now leaves an audit record — which colony, which direction, when, and whether a hostile attack on that colony was known to the defender at that instant — so the question can be re-asked with numbers after a season. Players see nothing new: the record never appears in the Activity feed, the galactic register or any other in-game surface, and a switch on a colony that has not built the Accelerator is not counted.

1.185.0 — 2026-09-15

New

• VIP borders around the portrait (#3946). A VIP's portrait used to carry the donor tier's flat coloured ring — the same line at €5 and at €1000. It now carries an ornate painted border that climbs with your VIP level, one per step of the ladder: plain iron at 1, riveted iron, bronze, bronze with side wings, silver with a teal gem, winged silver, gold with a violet gem, gold with feathered wings, gold with large wings and a crown, and radiant white-gold with luminous wings and a teal crown gem at 10. You see it wherever your portrait appears — the header, your profile, chat, guild members and the leaderboard. Players without a VIP level look exactly as before; the €1 tier, which sits below level 1, keeps its ring.

1.184.0 — 2026-09-14

New

• The Command Center line in Suggested Next has a build button (#2210). The advisor's line above the ranking — Command Center 5, what it opens, the price — used to be the one line on the panel you could not act on: the three ranked rows each had their ↑, the priciest building on the panel had none, and a player with an empty queue and the resources in hand read it as the build system being broken. It now carries a gold ↑ that queues the upgrade, and it obeys every rule the build list obeys: greyed with Not enough resources. or Queue full. in the same words as the rows below it, dark while another upgrade is starting, gone once the Command Center is already in your queue, and refused — with the reason on screen — if the colony you had selected is no longer yours. One tap, like any row. It also shows what a row shows before you spend: the price turns red when you cannot afford it, and a ⚡ figure beside it says what the level draws from your grid — a Command Center level always draws more, −6 at level 5 and −14 at level 15, the same number the build list and the Building Sheet quote. Nothing about the advice changed: the line still carries no payback figure, because "you may now post an offer on the market" has no resources-per-hour and the advisor still refuses to invent one. Measured on production the day it shipped, about six in ten players who see the line could tap it straight away; the rest see it greyed until the resources are there. A phone app that has not updated shows the line as before — the price, no button.

1.183.0 — 2026-09-14

New

• Supporter emblems are now painted (#3941). The six emblems reserved for supporters — one per donation tier, from Energy to Eschaton — were white line-art glyphs, the same weight as the free catalogue. They are now full renders in the family of the VIP crests: a gold plate with a glowing bolt, a copper ferrolite ingot, a teal xenogas sphere in a bronze cradle, a sapphire data core, a violet void star and a gold eclipsed sun with a teal core. Pick one in your profile as before; the hue you choose still tints the hexagon behind it, and every other emblem and portrait looks exactly as it did.

1.182.2 — 2026-09-14

Changed

• Synthetic empires arm an outpost's garrison in the tick that fills its build lanes (#3750). Before, an agent's outpost with a shipyard bought its standing garrison only once every build lane was busy as the tick found it: when the tick's own build pass filled the last free lane, the outpost waited for the next tick — five minutes, plus up to 45 seconds of jitter — before ordering a hull. Now the hull is ordered in the same tick. Buildings still come first: the build pass runs first and has already paid, and half of the outpost's purse stays untouchable for the next building. Measured read-only on production on 2026-09-14: of 305 agent outposts, at most 42 — those below the hoard floor with a free lane and a garrison below target — can be moved one tick earlier; the 233 above the floor were already guarded whatever their lanes. Nothing changes for human players, nothing changes for the Wraiths, and the garrison's size, price and target are as they were.

1.182.1 — 2026-09-14

Changed

• The Crucible's Monday prizes are a hundred times larger (#3938). Rank 1 now takes 100M ferrolite, 100M xenogas and 40M data at the Monday roll; rank 2 takes 70M / 70M / 28M, and the ladder descends in the same steps to 10M / 10M / 4M at rank 10. The rules are unchanged: paid into your mother colony, up to its warehouse; a dormant row forfeits and nobody inherits.
• At this size the warehouse is the real limit. A level-20 Warehouse holds about 60M of a resource, and level 22 is the first that takes the whole first prize, so a winner with a smaller warehouse is paid the ceiling and the vault says exactly what landed. If you plan to hold the top, build the Warehouse (or the Singularity Vault) that can take it before Monday.

1.182.0 — 2026-09-14

New

• The launch sheet now says how enemy fire will split across your rows (#1541). The default battle line puts light hulls in the vanguard, where they take 70% of incoming fire, and heavy hulls in the rear at 5% — a cheap screen protecting the expensive capitals. That has always been the case and heavy-front has always been one edit away, but nothing said so at the moment of launch. Now every place a fleet leaves from carries one sentence: the mission composer reads Default: light hulls screen and take 70% of the fire — tap to change; a player who moved their heavies forward reads Your line: heavy hulls screen and take 70% of the fire; an empty row shows the renormalised share (74%), on every defence and every guild raid — where the colony's own fleet or your allies share your rows — the share is a ceiling: a garrison behind your batteries reads stand behind the colony's guns and take up to 26%. The Launch-again card carries the same line beside its stance. Nothing about combat changed: the number is read from the same engine rule that splits the fire, never retyped.

1.181.1 — 2026-09-14

Changed

• The Crucible looks like a game now (#3931). The ladder tab used to be a management panel: text lists, grey bars for the opponents' fleets, the simulation chamber hidden under a gradient, the rank crests at sixteen pixels. It has been redrawn from a design canvas, phone first. Your crest and rank stand on the chamber itself, the attempt bar is five glowing cells, your Merit sits in a pill. Each opponent is a target card that shows the ships they actually field, as sprites with counts, with the age of their snapshot and one big Fight button; the Wraith sparring partner has its own card in the hostile tint. Locking a target shows both fleets in their three rows with the fire split drawn as bars, a counter read, and your roster with the hull sprites beside the numbers. Winning is a full-screen Victory with the rank swap as the headline and the Merit as a gain; a draw says draw. The Monday prizes are a vault: a countdown ring to the roll, three crates for the podium, the ranks 4–10 table with resource icons, and last week's winners.
• On the phone the tab is four segments — Fight, Ladder, Vault, Defense — under the stage, instead of three panels stacked. The Defense segment opens on the defense you published. On a tablet or beside an open chat, the desktop columns wrap instead of squeezing the target cards.

Also

• The crates for ranks 1–3 are drawn in the rank's metal until their Midjourney renders land; the tab switches to the files when they do.

1.181.0 — 2026-09-14

New

Guild raids: gathering time before impact (#1935)

Before, a joint raid's join window was the distance to the target and nothing else: on a far target the guild had most of an hour to join, on a neighbour it had a few minutes, and on production the raids opened with a window under five minutes never gathered a single guildmate.

Now the leader who opens a raid can add gathering time — up to an hour — with a slider on the open form. It is added to the impact hour, and that is all it is: your fleet still leaves and is committed the moment you launch, exactly as before; it is not a staging phase and nothing waits on the pad. The form names the impact in your own clock («It lands in 1h 21m, at 20:34») and the raid row tells your guild the hour was chosen («Opened with 30m of gathering time: the impact hour includes it»). Guildmates join exactly as they did — against the later hour — and the target is warned of that hour from the moment you launch, so more gathering time is more warning for them, never less.

Two things to know. Calling off a raid you are still alone on (the rule from #2075) works through the gathering time, and the last stretch of the flight is still committed: with an hour added, that is the last ten minutes before impact, and the sheet now tells you exactly how long you have left to call it off. And if the gathering time would land the impact inside the target's night window, the launch is refused and the refusal says which gathering time would land clear.

Installed apps and cached bundles that predate this release keep working unchanged: they cannot add gathering time, and a raid opened with some reads to them as a raid with a longer countdown.

1.180.0 — 2026-09-14

New

• The guild's threat board shows an ally's battle line (#2310). Under the Defending box's hull chips, a threatened guildmate's own line is now printed row by row — Vanguard: 3× Interceptor, Center: 6× Corvette, 2× Bomber, Rearguard: … — opening with their wall when one stands (Their emplacements and Fortress hold the vanguard, in front of this line.), and closing with the reminder that your garrison fights in these same rows, on the formation it launched with. Before, a rescuer could read which hulls an ally fielded but not where they stood, while an enemy could buy those rows with a tier-3 probe. No Observatory level gates it: it is the defender's own state on the defender's own alarm. The defence plan stays private everywhere, and a spy report still reads more (stance and modules) than the board does. Installed apps from before this release keep the box exactly as it was.

Changed

• No Observatory lending — the answer to the other half of #2310 is no. The issue also asked whether a rescuer with a better Observatory could read more of an ally's incoming attack than the ally can. DESIGN-RAID.md's no free alliance radar rule stands, in every shape proposed (flat, diluted, earned by a garrison, or as a Help-Board favour): the board keeps forwarding exactly the attacked colony's own Observatory reading. Nothing changes in play; the decision is now written beside the rule.

1.179.0 — 2026-09-14

New

• The Crucible pays resources every Monday (#3924). The ladder used to pay Merit every night and nothing a warehouse could feel. Now, every Monday at 00:00 UTC, the top ten ranks are paid resources straight into their mother colony: rank 1 takes 1M ferrolite, 1M xenogas and 400K data, rank 2 takes 700K / 700K / 280K, and the envelope descends in round steps to 100K / 100K / 40K at rank 10. The prize follows the board as it stands at the close — a dormant row forfeits its slot's prize and nobody inherits it, so the way to take it is to beat that snapshot and hold the rank. The prize lands up to your warehouse's ceiling (vault and surge included); the ledger records both what the rank was worth and what actually fit. The first payout is Monday 21 September.
• The prize table is on the Crucible tab. Visitors see it under the join form; once enrolled it sits beside the ladder with your row lit — struck through if you are dormant — and Your standing carries a "Monday prize" line with what your rank is worth if it holds, plus a countdown to the roll. Once a week has been paid, last week's winners are listed with their prizes, and a winner whose warehouse was full sees what actually landed.

1.178.1 — 2026-09-14

Changed

• The Echelons are now named Bronze, Silver, Gold, Diamond and Challenger (#1193 follow-up). The five tiers of the player ladder were called Ferrolite, Xenogas, Dark Matter, Starforged and Singularity — the entry tier read as a resource, the summit as a research node, and a newcomer had to learn the ladder before the name told them anything. They now carry the names everybody already knows the height of. Your tier, its tag on every Leaderboard row, your last verdict ("Promoted — you closed rank 2 in Bronze…"), the rules line at the edges of the ladder ("Nobody drops out of Bronze", "There is no tier above Challenger") and the AI empires' own knowledge of where they stand all use the new names. Nothing about the race changes: same five tiers, same weekly cycle, same purses, and every verdict already frozen keeps its place — only the names on it read differently.

1.178.0 — 2026-09-14

New

Catch-up ramp: a late joiner's fresh server (#3917)

Before, an empire that started weeks after the field grew at the same pace as everyone else and read, correctly, that it would always be behind. Now, while your Power sits below half the median established, active empire, your resource production and your research run faster — and the further behind you are, the faster: ten times behind is ×3, a hundred times ×5, a thousand times ×7, up to a ×10 cap. It fades smoothly to ×1 as you reach the line, so there is no cliff to sit under and no hour that is worth more just before it than just after it. Nothing about it is a fixed number: the reference is re-measured every five minutes and follows the server as it grows, so the same account starts higher on a mature server than on a young one.

Where you see it: the Overview names the ramp under its header while it is on; on a phone, tapping a resource pill shows «Catch-up ramp ×N included» beside your actual production; the research Bench attributes the shorter research time the way it attributes Scientific collaboration. Every +X/h on screen already includes it, contract rewards are priced on what you actually bank, and the build advisor ranks on the ramped rate. It ends on its own, and the lines disappear with it.

The reference is measured on the established, active field — humans and synthetic empires seen within the last week and older than fourteen days; the Wraiths are neither counted nor ramped. The whole-server median would have been a fresh account, which is why it is not the reference. Full design and the production measurement: DESIGN-CATCHUP-RAMP.md.

1.177.0 — 2026-09-14

New

Chat moderation: a mute, with the reason in front of the player (#3918)

Until now the only answer to someone misbehaving in chat was a full ban: the account locked out of the game, the reason kept for the admin. A moderator can now mute a player's chat for a fixed time instead — from fifteen minutes to thirty days — and has to write a reason, because the player reads it.

While the mute runs, every chat panel (galaxy, guild, whispers, the dock) shows a notice where the composer used to be: what happened, until when and how long is left, the moderator's reason word for word, and that reading every channel still works. A mute issued while the chat was already open appears at the first Send, without a reload; an early unmute reaches the screen within half a minute. When the mute ends, the composer comes back on its own. Nobody else's chat changes.

Also

• On an app build older than this release, a message refused because of a mute shows the bare code chat_muted instead of the notice, until the app updates.

1.176.0 — 2026-09-14

New

A top-tier spy probe now reads a defender's standing battle orders and installed modules (#2605)

The deepest espionage tier already sold you the defender's battle line. It now also reads two things that decide how those hulls fight: the standing battle orders their colony's defending fleet holds, and the ship modules their hulls are refitted with.

The attack forecast prices both from the report, so the verdict you launch on and the battle the server runs now use the same multipliers. Before, the forecast simulated a defender with neither while the server applied both.

The Espionage sheet in Research names the two new disclosures at tier 3, so a defender can learn what a top-tier probe reads by reading the same table the attacker buys it from.

Your defence plan is still private, at every tier. The tier says how the defending hulls will fight. It still says nothing about which of them take the field.

Nothing changed about what a battle stance does, what espionage costs, or the tiers themselves.

Changed

A spy report now tells you which of four things it knows (#2605)

"We did not read this" and "there is nothing there" are different answers, and the report card now gives each its own sentence — for the standing orders and for the modules alike:

• not in this report — a lower tier, or a report filed before this change. It says so, and says a fresh top-tier probe reads them. When neither field was read, that is one sentence rather than three that end the same way.
• read, and none set — the probe looked and the colony holds no standing order.
• read, and in force — the order itself, with both halves of its trade.
• read, but newer than your app — an order your version has no numbers for is named rather than described, and you are told to update. It is never quietly reported as "none".

Reports you already hold carry neither field and are never back-filled: they read as "not in this report", which is what they are. A report will not invent an answer about a fight that is already over.

The refit line names the hull classes rather than counting them — hardening support hulls and kitting heavy hulls are not the same wall — and says plainly that a refit is empire-wide, not a property of the colony you probed.

Fixed

The forecast stopped warning about something it can now read (#2605)

The launch forecast carried a line saying a defender's standing battle orders were something "no spy report reads". That is no longer true, so it is gone. In its place the forecast says which of the four states your report is in, and — when a fact is genuinely missing — which way that makes the forecast wrong.

The Advisor and the forecast row stopped disagreeing about the same fleet (#2605)

With the defender's refit sheet now in the report, the composer's forecast row was pricing a refitted wall against your own hulls at catalogue stats, so it printed a worse result than the Advisor verdict three lines above it — for the same fleet, on the same screen. Both sides are armed now, in the panel and on the server's advice route alike.

1.175.0 — 2026-09-14

New

Crossing a Proving Ground threshold no longer costs you your crate (#1309)

The Proving Ground sorts you into a bracket by your military power, then ranks the bracket by the damage you land on the wall. So growing past a threshold used to move you from the top of your bracket to the floor of the next one, against fleets many times your size — and the 25% proof bar does not reach a floor, because damage climbs faster than military power does. The best move was to grow to just under a threshold and stop. Two commanders said so out loud, from opposite ends of the same rung.

Now a strike that the 25% bar pays nothing is still proven if it out-hits the best strike of a smaller bracket: it takes the crate of that bracket. Your damage does not shrink when you cross a line, so the crate crosses with you.

There is a ceiling on it, and it is worth knowing before you climb: the crate you take this way is never bigger than the next smaller crate on the prize table — it is never your own bracket's crate. Out-hitting Armada's best from inside Warhost pays the 1M crate, not a second 5M one. And the crate names the bracket you actually out-hit: clear only Patrol's best from inside Fleet and you take Patrol's 100K crate, because that is what out-hitting Patrol is worth.

Nobody who wins today wins less. The bracket winner's crate is untouched, and so is every strike already paid by the 25% bar. The new route only ever reaches strikes that were paid nothing.

Patrol is unchanged: it is the bottom rung, with no smaller bracket to out-hit and no smaller crate to take, so its runners-up are still paid nothing.

Changed

• The prize table's proven figure now reads "up to". It was stating one number, and with the second route the crate a proven strike takes is a range whose top it was quoting.
• The Activity entry for a proven strike no longer explains itself with a rule that only covers one of the two ways to be proven. It named the 25% bar and the crate below your bracket; a strike paid by the new route cleared neither, and read a sentence that contradicted the crate printed beside it. It now names both routes and states the crate as the limit they share.

Also

• Two notes in the engine had been overtaken by the ladder and are corrected: Armada was described as holding a single empire (it holds six), and Warhost and Apex as out of reach of any human empire — Warhost is now inhabited. Apex is the rung still empty.

1.174.3 — 2026-09-14

Fixed

• A ban can no longer be written without the guard being told about it (#3858). Banning someone was two steps kept in different files: one wrote the account's ban date, the other told the in-memory list the server checks on every request. Every existing path did both — that was checked one by one, not assumed — but nothing made it so, and the next one to forget would have produced a ban the admin panel confirmed, that the server did not act on for up to a minute, and that left no trace in any log. The two steps are now a single call, and the test suite reads the source on every run and fails if a ban is ever written around it.

  No player sees a change from this, and no account is banned — none has been on any daily snapshot since 2026-07-20. One error case did move: banning or lifting an account that was deleted moments earlier now answers "player not found" instead of a server error, and records no action that did not happen.

Also

• The guard was rewritten twice during review, which found five ways past the first version — a ban payload built one line earlier, one filled in field by field, one written with brackets, one with a quoted name, and one that computed the new state from the old ("ban them if they are not already banned"). All five are closed, and each is now a permanent test of the guard itself. Two code comments that claimed more than the code delivers were corrected: the account record and the in-memory list are still two separate writes, and a slow database can still separate them — what the change removes is the chance to forget the second one.

1.174.2 — 2026-09-14

Fixed

• A report from the app can now be answered without guesswork (#3873). The game on your phone is a self-contained app: it carries its screens inside it, and it only gets new ones when the App Store ships an update. While an update sits in review — often for days — the app keeps showing the older screens, even after the same thing has already been corrected on novarchy.com. From the outside those two look identical: a player reports something, and nobody can tell whether it is still broken or whether that phone is simply a few days behind. It happened on 13 September, and sorting it out took three separate investigations.

  Every version of the game now carries its own build date, and your client quietly mentions which one it is running when it talks to the server. When you report something, it can be told apart from a genuine bug straight away instead of being chased down by hand.

  Nothing on screen changes, and nothing new is collected about you or your device: the game reports which build it is, which is the same for everyone running that build, and nothing else.

Also

• This is the server-side half of #3873. The other half — the app telling you, in the game, that a newer version is waiting for you — is not in this release and remains open. It needs the game to know what the App Store has actually published, so that it never asks you to install something that is not there yet.
• Apps already installed cannot report their build; they will start doing so from the next App Store or TestFlight update onward. Nothing about them gets worse in the meantime.

1.174.1 — 2026-09-14

Fixed

• A failed deploy no longer disappears without trace (#3885). Deploys are verified after the containers are swapped, and a deploy that failed that check used to leave the server looking exactly like a server nobody had deployed to — so the next ship could not tell a production it had verified itself from one that somebody else's deploy had quietly repaired. It happened on 2026-09-13: one ship caught a container still running the previous build and stopped; twenty seconds later an unrelated ship replaced every container and the fault vanished, repaired by someone who did not know they were repairing anything.

  The server now keeps a record of every deploy attempt: when it started, whether it finished, and — if it failed — which check failed. Every deploy prints what the deploy before it did, always, including when there is nothing to report. A deploy that was killed or timed out mid-flight is now visible from the outside as well, which it never was.

  A deploy that arrives after an unresolved failure goes ahead anyway and says so, and its own success records which failure it repaired. It deliberately does not refuse: with dozens of deploys a day from several sessions at once, a deploy that blocked on someone else's leftover state would stop everything behind it — and the deploy that follows a failure is usually the thing that fixes it.

Also

• The cause of that particular missed container replacement is not fixed here and is not fixable in the deploy script: the container tooling's replacement rule is correct (measured), and the same tooling replaced that same container for that same reason twenty seconds later. Filed as #3906. The record added here is what will make the next occurrence countable — today nothing on the server can tell one apart from a deploy that simply had nothing to do.

Nothing in this release is visible in the game: it is release machinery.

1.174.0 — 2026-09-14

New

Call off a raid nobody has joined (#2075). A raid used to be fire-and-forget the moment it opened: even a leader who was still the only fleet on it could not turn back, while the very same fleet sent as a plain attack could be recalled. Now the leader of a raid that still holds only their own contingent gets a Call off the raid button, and the fleet turns around and flies home.

The limit is named on the card and it is deliberate: once a guildmate joins, the raid stands. An ally who has committed a fleet decided on the strength of the raid existing, and a leader who could dissolve it afterwards would make every join a gamble on someone else's second thoughts. The button also goes when your fleet passes the point where no fleet in the game can turn back.

Calling off a raid is not an undo, and the confirmation says so before you tap it: the fleet flies home on its full return leg — you get the time, not just the promise — and everything the launch already spent stays spent. It also costs you one of your three turn-arounds a day, the same allowance an ordinary recalled attack draws on, and the defender you had aimed at is told the fleet turned around rather than watching your countdown disappear.

Changed

• A raid contingent still cannot be recalled out of a raid on its own — that rule is untouched. Calling off a solo raid closes the whole raid; it never pulls one fleet out of a battle.

1.173.1 — 2026-09-14

Fixed

• Supporting the game: every gift now delivers its tier's hulls, every time. A second €100 delivers a second destroyer, two more annihilators, 300 titans and 300 Voidlance Cutters again; a second €50 delivers 300 titans and an annihilator again. Before, a repeat gift paid Dark Matter, Chrono shards and VIP points but no hulls, because the hulls were counted as the difference between the tier you had reached and the one you reached next — and after a refund and a new gift, that difference was zero. The "reaches…" line on each row still says what the press does to your tier; the hulls it lists are what that press delivers, whatever you already hold.
• A refunded gift now takes its hulls back from the home colony, as it already took back its Chrono shards — only what is still held, never below zero, so a hull lost in a fight is not a debt. The Dark Matter stays.

1.173.0 — 2026-09-14

New

A share lane now tells you what your percentage actually delivers (#3556)

Open a standing lane, pick a share of what this colony produces, and the form used to show you one number: what the origin's Trade Portal moves per hour. Your percentage was measured against production, the Portal's hour was not, and nothing on the screen crossed the two. A player who set 75% of a colony making around 400,000/h watched about 6,000/h arrive — not because the percentage was ignored, but because the lane was already carrying everything the pipe could hold.

The form now resolves it and names the reason:

75% of Lantea's Ferrolite hour is 13,118/h. This lane will deliver ≈ 580/h — limited by Lantea's Trade Portal (L2).

Under it, what the receiving colony accepts per hour across all its incoming lanes and how much of that its other lanes already take — the ceiling that quietly binds most hubs and that no screen had ever named. When a hub's feeders ask for more than it can accept, it says so plainly rather than leaving you to work out why five full lanes arrive as a trickle. And when a colony runs more than one lane, it says that the share sets a priority between them rather than a rate.

Three things can hold a lane back — the origin's Portal, the destination's Portal, and the destination's warehouse — and every one that is tied at the figure is named, because raising one of a tied pair moves nothing. When nothing is holding the lane back, it says that too. When your client could not read the colonies and lanes the arithmetic needs, it says that instead of printing a zero.

Nothing about throughput changed. No capacity, no cost, no lane arithmetic moved: a Portal carries exactly what it carried yesterday, and a lane that was slow yesterday is the same speed today. What changed is that the form now tells you why.

1.172.4 — 2026-09-14

Changed

Lava worlds no longer charge a Data penalty (#3549)

A lava world used to read +20% Energy, −10% Data, and the Data cut was permanent.

The trade never balanced. Energy is not something you stockpile — it is a gate, and it stops helping the moment your plants cover what your colony draws. So the +20% quietly paid for itself once, in the reactor levels you did not have to build, and then did nothing; the −10% Data went on costing you every hour, for as long as the colony stood. Lava was the only one of the four world types built that way, and neither the Colonial Outpost nor the Terraformer could soften it, because those relieve a world's drawback and on lava the energy was the upside.

Lava is now simply "+20% Energy", with no penalty at all. It applies to every lava colony that already exists, with nothing to build, research or claim. A colony that had already cancelled the penalty the long way — a Terraformer at level 10 or above — was not paying it any more and does not change; everyone else stops paying it now. No colony anywhere produces less than it did.

The Colonial Outpost and Terraformer sheets on a lava world now say there is no penalty to relieve, instead of offering to halve one. The Galaxy planet sheet and the world badges stop quoting the old figure.

Nothing else moved: the energy bonus is the same, the other three world types are untouched, and nothing costs more.

Also

The empires run by the game's own agents read the planet rules through their own copy of them, so that copy moved in the same update — otherwise they would have kept planning around a penalty the game had stopped charging.

One thing is deliberately left for later: the advisor that scores where to settle next still reads a lava world's stored penalty, so it rates lava lower than it now deserves — enough to pass over a closer lava world for a rockier one further out. That is a separate question about colonisation targeting, and it is filed as #3893 rather than fixed here.

A note for anyone chasing a weekly contract: the "win battles" card measures each win against a bar that scales with your empire's hourly output, and it is checked when you claim, not when you won. A little more output means a slightly higher bar. For the handful of empires this can reach at all, the bar moves by at most 1.27%.

1.172.3 — 2026-09-14

Changed

• The battle weekly now pays for what you destroyed, not a flat 30,000 (#3038). Above 20,000/h of production, Win 3 battles paid the same 30,000 resources to every empire — a 25,000/h empire and one producing tens of millions an hour collected the identical amount. For the player who reported it, that was one part in a hundred and thirty-two of what their own daily card paid the same week. It was the only card in the game whose reward ignored the size of the empire claiming it.

  Qualifying has not changed at all. A win still counts when your share of the fleet value destroyed reaches one hour of your production, with the same floor of 1,000 and the same 20,000 ceiling. Nothing about which battles count has moved, for anyone, in either direction.

  What changed is the price. The card now pays half the fleet value destroyed across the wins it counted this week, up to the full weekly reward — reaching the whole card once those wins total three hours of your own production. Past the 20,000 ceiling that takes more wins than the card's target, because the bar stops scaling there while the full-pay mark does not.

  Nobody is paid less than before. A card you can claim already carries at least three counted wins, which is enough to leave the payout unchanged below the ceiling and at no less than the old 30,000 above it. Checked against every completed card of the last three weeks: none would have paid less, and almost all would have paid more.

Also

• The card shows how far its payout has come — Destroyed X of Y in counted wins for full pay, or Full pay reached once there is nothing more to earn on it this week. This is the one card whose reward moves while you play, so it explains itself rather than just printing a number.
• Its reward figure now locks when you claim. Claiming at 3/3 banks what it is worth at that moment; fighting on first raises it. This is the first card on the board where collecting early can cost you something, and the card says so.
• A card you have not started yet shows what it is worth if you finish it, the way its two weekly siblings do. Only a card you can actually claim shows what pressing the button pays right now.
• If you are playing on an installed iOS or Android build, the card's rule text still describes the old payout until you update the app. What it says about qualifying is still correct; what it says about pay is not, and the amount the game credits you is the new one.
• The daily cards, the other weekly cards, the Daily Wrap and the Chest are untouched.

1.172.2 — 2026-09-13

Fixed

The Crucible roster editor now does the arithmetic for you (#3889)

Before: joining the Crucible with a real war fleet opened an empty form under a meter reading tens of thousands of PP against a cap of a few thousand, and a bar pinned solid red. The hull fields sat in a grid that wrapped mid-name, so a box could not be told from the ship it belonged to; nothing said what a single hull costs; and the only instruction was "take some ships out", beside a button that was already disabled.

Now:

• The form opens on a roster that fits. Your armed fleet is cut to the cap, keeping its proportions — so a fleet that is mostly bombers still fields bombers — and the leftover capacity goes on the strongest hull that still fits. You trim from a legal roster instead of building one by trial and error.
• One button does it whenever you want. It reads Trim to cap when you are over and Fill to cap when you are under, so it always says which way it will move your roster. Clear sits apart from it.
• Every hull says what it costs, on its own line: 385 PP each, how many the colony holds, what that row is spending, and a Max that fills the row to exactly what the cap still allows.
• The meter says the number the bar cannot: Over by 4,312 PP, or 2,180 PP still free, with the percentage of the cap.
• The duel sheet opens ready to fight. Each duel is capped at the defender's tier, and one list can offer commanders from two different tiers — the roster is now re-fitted to the defender you picked, instead of opening over their cap with the fight already refused.
• If no roster is possible — every armed hull you hold costs more than the whole cap on its own — the editor says exactly that, instead of offering a fit that yields nothing.

On a phone the rows fold instead of shuffling the boxes, and the buttons are thumb-sized.

1.172.1 — 2026-09-13

Fixed

• The Command Center card no longer says you are at your colony cap a world early (#3881). A commander holding a world in the Event Sector was told, on the Command Center card of their home colony, "You are at your colony cap: this level is what lets you settle another world" — one colony before it was true. Season worlds have cost no permanent colony slot since the Eschaton sweep exemption shipped, so the server would have accepted the colonisation the card implied was blocked, and the level it was selling released a slot the empire already had. The card now counts the slots actually spent. An empire genuinely at its ceiling is told so in exactly the same words, whether or not it also holds a season world.

  Nothing about what anyone is allowed to do changes: this is a recommendation card, and the colonisation gates were already correct and are untouched.

Also

• The empty-harbour line on the Colony tab keeps counting every world, season ones included — a world in the Event Sector genuinely is a second place to fly ships from, and the offer to move ships there with a Redeploy stays. The two questions now travel separately, so fixing one cannot silently answer the other.

1.172.0 — 2026-09-13

New

Colonization 11 and up now makes everything you build faster (#2683)

Past level 10, the Colonization discipline bought you nothing. Founding a colony needs two things at once — your Colonization level, and a Command Center big enough to command one more — and the Command Center runs out at ten colonies. From level 11 upward the discipline could never decide anything again. The tech tree still asked for level 14 to unlock the Colonial Ark and level 20 for Beyond the Horizon, so those levels got researched, paid for, and did nothing in between.

They do something now. Every Colonization level above 10 takes a flat 60 seconds off every building you queue — ten minutes off at level 20 — applied after all your other build-speed bonuses, and never taking a build below ten seconds.

Because it is a flat cut rather than a percentage, it lands where the tedium actually is. At a maxed Command Center an Extractor's tenth level drops from 4m 31s to the ten-second floor by Colonization 15; a Recycler's ninth goes from about 87 minutes to 77 at Colonization 20; a Stellar Forge, already sitting at the four-hour ceiling, loses about 4% and barely notices. On a brand-new colony the effect is much larger: at high Colonization the first several levels of a building finish essentially instantly, which is exactly the "bringing a new colony up to par" grind this came from.

The ten-colony limit has not changed. This buys time, not worlds.

Sixteen commanders are above Colonization 10 today and every one of them gains; nobody loses anything. Builds already under way keep the time they were quoted when you started them — the gain applies to the next thing you queue.

Fixed

• The Research sheet's Colonization entry was wrong in two ways and is now right (#2683). It claimed every level authorized one more colony, which stopped being true at level 9, and it never mentioned that your Command Center is the other half of the limit — a commander with a level 12 Command Center is held to four colonies no matter how far the research goes. The effect line beside it also called levels 11+ a dead end on the same screen where the description now promises faster builds; it names the seconds instead.
• The message shown when you hit the colony limit told you to upgrade your home Command Center "to command more". At the limit there is nothing left to upgrade, so it named a cure that does not exist. It now says what the ceiling actually is.

Also

• Corrected a long-standing false note in the engine's own documentation, which described colonies as growing "with no hard ceiling" past Command Center 15. They stop at ten, permanently. The same claim had been copied into two design documents and the agents' internal mirror; all four now say the same true thing.

1.171.1 — 2026-09-13

Fixed

• "Launch again" no longer throws away your battle stance (#3452). Set a stance on an attack, send it, then press Launch again to repeat the order: the repeat flew with no stance at all. It fought differently from the mission it claimed to copy, and under Rampart it was away about 11% longer than the card had promised — the only way to find out was to read the battle report afterwards, which is how it was found.

  The rule that dropped it was written for Warp, which is paid for in Dark Matter: repeating an order must not quietly spend your money again. A stance costs nothing, so declining to re-apply it was never protecting anyone — it just changed your order. Warp is still not repeated, and still for that reason.

• The order card now says which stance it will fly. One line, under the fleet: Stance: Rampart, or Stance: none when you chose not to set one. It appears once Admiralty is researched, on the missions that actually fight.

• Orders saved before this release say so, instead of quietly flying neutral. Their stance was never stored, so it cannot be recovered; those cards read Stance: not recorded — this relaunch flies without one. A stance you chose not to set and a stance nobody ever recorded are deliberately different sentences — reporting the second as the first is the same defect one step further out. A card that cannot launch anything promises nothing.

• The travel estimate on a repeated order now prices the stance it is about to fly. The same order reads 45m 20s unstanced, 50m 22s under Rampart and 39m 26s under Pursuit; before, all three were quoted the unstanced number.

• The card now names the stance that will actually fly, not merely the one on file. A garrison cannot pay a price only a journey spends, so it drops Rampart and Pursuit on arrival — and a saved defence holding one of those would have been announced as Rampart, priced as neutral, and flown as neutral. The card, the estimate and the launch now read one value.

• "Change" opens the composer already holding the order's stance, so the card and the form can no longer describe the same order differently.

Reported by remylz.

1.171.0 — 2026-09-13

New

The Crucible — a PvP ladder (#3874)

Before: every way to fight another commander cost real ships and real travel time, and the contests that cost nothing — the Season Boss, the Campaign — had no other commander in them.

Now: a new Crucible tab, next to the Leaderboard.

• Join by publishing a defense — a roster from your home colony's armed ships. That puts you at the bottom of a numbered ladder, where the commanders below you can see your defense before they choose to fight it.
• Challenge commanders ranked just above you. Each duel is a simulation against their published defense: no ship is lost, nothing is looted, no fleet leaves your colony, and your protections are untouched. Win and you take their rank. Lose and nothing moves.
• Read the fight before you commit. The duel sheet shows both formations row by row, how much of each side's fire lands on each row, and how hard each of your ship types counters their stack.
• Five attempts, one more every two hours. They do not pile up past five.
• Seven tiers, from Cadet to Admiral. Each tier has a tonnage cap, so a larger fleet does not simply win: both sides of a duel are capped at the defender's tier.
• Merit, paid every day at 00:00 UTC by tier. The board closes at 23:45 UTC and reopens once the payout has landed, so the rank you hold at the close is the rank that pays.
• The Merit shop sells choices, not resources: a fresh opponent list (40 — from rank 50 up your list already holds everyone you can challenge, so it is not sold there), a reset of the Industrial Edict cooldown (600), a reset of the rename cooldown (900).
• Spar with the Wraiths when nothing on the ladder is worth an attempt: a little Merit, and no rank moves.
• Seasons last 14 days and keep your rank. A commander who has not fought for seven days earns nothing, and at the season's end drops below everyone still fighting.
• The AI empires play too, each marked with its AI badge.
• A HUD chip shows your attempts and opens the tab in one tap.

Not in this release: a weekly prize for the top of the ladder, a rank badge on names outside the tab, and notifications when someone takes your slot. The tonnage caps and Merit rates are a first setting and may change once we see how the ladder plays.

1.170.4 — 2026-09-13

Changed

Nothing on the map looks different, and that is the point of this one.

The colour a guild wears on the political chart is its identity, and two guilds must not look alike. Until now nothing checked that they don't. The wheel keeps every hue 12° from every other, but the colours players actually see are those hues put through the chart's renderer — tone mapping, then laid over the map's background — and at that end two slots 12° apart are not reliably 12° apart to the eye. Above a certain brightness a pair runs into the wheel's lightness ceiling, both members go pale, and they converge. Nothing anywhere would have gone red.

All 276 pairs of colour slots are now measured the way the chart actually draws them, at both the quiet ink (a guild lives here) and the loud one (a guild rules here), and held to a floor. A companion check pins the faintest mark any guild can make against bare ground, so the change that made territory visible again in v1.170.3 cannot be quietly undone — six of the thirteen guilds holding a colour had territory that could not be seen before it.

The measurements the chart's own notes carried turned out to have been taken without the renderer's tone mapping, which is why they named the wrong pair of guilds as the closest. They are replaced with figures a test computes on every run rather than a number somebody wrote down once.

One thing this does not fix, said plainly: two colours currently in use are closer together than the standard this check would ideally enforce — about one just-noticeable difference under the loud ink. No guild's colour was changed to make that go away; a colour its members have learned is not something a test fix gets to move. It is recorded for the people who own the wheel, as issue #3878, and the guard makes sure it cannot get worse without somebody noticing.

Closes #3876.

1.170.3 — 2026-09-13

Fixed

• The map marks your worlds, not the ground around them (#3416) — the galaxy chart used to colour seven cells for every colonised world: the world's own cell and the six around it. Since the field was rebuilt to keep planets two cells apart, those patches stopped meeting inside a system and spread out into the empty ground between systems instead. Across the whole chart 28% of every cell was coloured, and 42% of what you saw when you looked at an inhabited system. One player put it plainly: reading a sector had become impossible.

  Now a guild's colour sits on the worlds it holds, and the space between them is empty again.

  It also settles something the chart could not say honestly. Two neighbouring worlds sit two cells apart, so the ground between them belonged to both patches at once — and which guild coloured it came down to the order the server happened to list them in, not to anything either guild had done. That ground is now plain, so the question does not come up. It was not a rare corner: 130 of the 217 inhabited systems hold worlds of two or more guilds.

  And the mark is drawn in full colour. Shrinking the patch to one cell would have made territory harder to see, not easier: most of a world's cell is covered by the world itself, so what is left is a thin edge, and at the old paint strength that edge was invisible for guilds whose colour is blue, purple or red — seven of the thirteen guilds on the server, whose territory simply did not show. The colour now sits at full strength, so a blue or red guild's world is as easy to find as a yellow one. No guild's colour changed, and the loud/quiet difference between holding a system and merely living in it is exactly as it was.

  Nothing else changed: no territory, no border, no attack range, no bounty band, no rule of any kind. This was the paint, not the map. Every world that was yours is still yours.

  At the widest zoom the chart still reads by sector, as it is meant to: ownership there is the sector's own border, and the coloured regions that used to appear were the overlapping patches, not a promise.

  Reported by spaccamazza (in-game Spacca-Mazza).

1.170.2 — 2026-09-13

Changed

• A guild can now put ten pushes on one order instead of five (#1881). Speedup requests on the aid board used to close after five hands, and two thirds of them ended exactly that way — 168 of the 249 opened in the last thirty days hit that ceiling, with guildmates still willing to help and nowhere to put it. The ceiling is now ten.

  What did not change: a push is still worth 2% of the time remaining, so pushes still give less the more an order has had — on a 75-minute wait five pushes cut about 7m 12s and ten cut about 13m 43s, not double. The Scrip a helper earns, the six aids a day you may give, the two per guildmate per day and the two requests a day are all exactly as they were. A timer still cannot be pushed to zero.

  An order that was already at the old ceiling can take more again. The count lives on the order and was never reset, so an order holding five pushes is now an order with five left, and a new request can be opened on it — if you were told "no more hands" on a build yesterday, that build is askable again today. A request that is already open keeps the five slots it was created with and finishes its life at five; the next request on that order is born with whatever the order has left, so closing and reopening still cannot hand out a fresh allowance.

Also

• The empires run by the game's own commanders read this ceiling too, and had their own copy of it. Both sides move together, and a new test now checks the two against each other so they cannot drift apart unnoticed.
• The message you get when an order truly has taken every push it can no longer quotes a number. It used to say "(5 pushes)", baked in when the app was built — so a phone that had not updated would have named the old limit as if it were the rule, one screen away from a board showing ten. The board is where the count belongs, and it comes from the server.

1.170.1 — 2026-09-13

Fixed

• An order meant for a colony you just lost no longer lands on your capital (#3842). When you have a secondary colony selected and that colony is taken from you, the game puts you back on your mother colony at the next refresh. It used to move the buttons with the view without saying so, so a build, a ship order, a defence plan, a rename, a mandate, an accelerator toggle, a Chrono shard or a Dark Matter core aimed at the colony you had lost was carried out on your capital instead — resources spent there, a queue slot taken there, a defence plan overwritten there. Because the screen had moved to the same place, nothing looked wrong.

  Now a write with no colony to go to is refused and says so, and picking any colony from the selector puts everything back. The view still falls back to your mother colony, as before — it is only the destination of an action that refuses.

  Two things deliberately keep working as they did: the upgrade advisor, which only reads, and the first-steps reward claim, which the server already checks and which pays out a reward your mother colony had genuinely earned.

  Nothing changed about what any action costs or does, and nothing became easier to trigger by accident.

1.170.0 — 2026-09-13

New

Event Sector worlds no longer cost a colony slot (#2125). Settling the season sector used to spend one of the ten colonies your empire may ever hold — on a world the Eschaton sweeps away at the season's close. It no longer counts. A season world sits beside your core colonies without taking a slot, the colony cap on the Overview, the Research sheet and the planet sheet counts only your permanent worlds, and an empire already at its cap can finally settle the season sector.

The Event Sector has a limit of its own: two worlds per empire. It answers to no building and no research — the Command Center and the Colonisation discipline price permanent expansion, and these worlds are not permanent. Colony ships already on their way there count towards it. When you reach it, every screen says so in the season's own words: the mission composer, the galaxy sheet's Colonize button, the Missions verb badge, and the Activity feed if a second colony ship lands after the first has filled the allowance.

Also

• When your core colonies are capped but the Event Sector still has room, the Missions colonize badge no longer disables Launch: it keeps the core sentence as a warning and names the room left in the season sector.
• Where the colony cap is printed as a fraction (3 / 10 Colonies), the numerator is now the slots you have actually spent; a world in the Event Sector shows in your colony list but not in that count.

1.169.27 — 2026-09-13

Fixed

• Market offers stay readable on narrow phones (#3463). Resource amounts now have a separate row from the seller, ratio and fee, keeping full amounts readable when sorting by Ratio or browsing offers with several resources.

1.169.26 — 2026-09-13

Fixed

• The guild shop now explains its weekly reset (#2304). A note above the store shows that purchase limits and Peace shield prices reset every Monday at 00:00 UTC.

1.169.25 — 2026-09-13

Fixed

• Hide locked also hides unowned event relics (#2252). The Shipyard filter now removes hulls that cannot be built, while keeping relics already in your fleet visible.

1.169.24 — 2026-09-13

Fixed

• Guild podium styling requires points scored (#2780). A zero-point season placement no longer appears gold in the guild showcase, even if every guild scored zero. The recorded rank and points remain visible.

1.169.23 — 2026-09-13

Fixed

• Launch-origin buttons keep their top border (#2696). The Missions origin selector now leaves vertical room for its hover lift, preventing the selected colony's outline from being clipped by the scrolling row.

1.169.22 — 2026-09-13

Fixed

• Build filters stay readable while scrolling (#2663). The sticky filter bar now has an opaque background, so building names and prices no longer show through its gaps. A subtle shadow separates it from the list below.

1.169.21 — 2026-09-13

Fixed

• Clearer colonization warning under the Veil (#2720). The planet card now says that colonization may be refused if the scouted colony is still there, avoiding the ambiguous phrase “the founding”.

1.169.20 — 2026-09-13

Fixed

• Agents remember why colonization failed (#3468). Their saved history now keeps the recorded refusal and explains occupied targets, colony or research restrictions, and closed seasons. Missing or unfamiliar reasons remain uncertain; coordinate retry rules are unchanged.

1.169.19 — 2026-09-13

Fixed

• Fleet advice accounts for your ship refits (#2543). The Missions advisor now considers installed class modules when choosing ships and updates its recommendation when those modules load. It reuses the existing inventory read; battle forecast rules stay unchanged.

1.169.18 — 2026-09-13

Fixed

• Revocation cache stops cleanly on shutdown (#3823). A database query finishing after shutdown no longer restarts background retries or refresh timers. Re-initializing the service still works. No player-visible behavior changes during normal operation.

1.169.17 — 2026-09-13

Fixed

• A database that stops answering no longer freezes the game's caches indefinitely (#3839). Novarchy keeps a few things in memory rather than asking the database on every request — who is banned, whose login keys have been retired, and how powerful each empire is (the number the anti-bully gate and the Galaxy view read). Each of those is refreshed in the background.

  If the database stopped answering — not refused, which is noticed and retried, but simply went quiet, the way a network partition looks — that refresh never came back. Nothing was logged, nothing was retried, and the minute-by-minute safety net could not help because the previous attempt had never finished. The stale answers could sit there until someone restarted the server: a lifted ban that stayed in force, a retired login key that kept working, a power score frozen at yesterday's value.

  Every database operation now gives up after 60 seconds and says so in the server log, which is what lets the retry and recovery the caches already had actually run. Sixty seconds is deliberately generous: it is roughly three times the longest any legitimate query in the game is allowed to take, because a limit set too low would break working features rather than protect them.

Also

• Honest about what it does not do: the abandoned query still holds its database connection until the database answers or the connection dies. This releases the waiting code and makes the failure visible; it does not cancel the query.
• deploy/DEPLOY.md now records why the database URL deliberately carries no socket_timeout — measured, that setting bounds a database that answers slowly but not one that never answers at all, and adding it would also apply to the schema migration the server runs at boot.

1.169.16 — 2026-09-13

Changed

• The release gate now quotes a failed step's log instead of dumping it raw, and always prints its verdict on a line of its own. A failing gate used to end with its verdict glued onto the end of a log line that had no line break — and in that state nothing that reads the gate could find the verdict at all, so a failure could be mistaken for a run that simply had not finished. The gate also used to copy a failed suite's output word for word, which meant a test that happened to print the gate's own success marker put that marker into a failing run's report. Each failed step's reason is now repeated in a single line directly above the verdict, so the whole reason survives being read from the end rather than only the last step's. (#3817)

Also

• Nothing in the game itself changed in this release: it is entirely the machinery that decides whether a change is allowed to ship. That machinery's self-check grew from 385 to 424 tests, and it now proves its most important rule — a test run that is killed must fail the gate — against the plain command a ship actually uses, which is the one path it had never covered, using processes that are really killed rather than faked. (#3817)

1.169.15 — 2026-09-13

Fixed

• A moderation decision taken at the wrong instant no longer goes missing for a minute (#3821). The list of blocked accounts is held in memory and re-read from the database once a minute. That re-read used to overwrite the whole list with what the database had answered — and the answer describes the moment the question was asked, not the moment it came back. A decision taken in the fraction of a second in between was therefore discarded, in both directions: a block could be dropped, leaving the account playing on for up to a minute after the panel said it had taken effect; and a block that had just been lifted could be put back, leaving the account shut out for up to a minute after the panel said it had been released. The second was the worse of the two — a block that outlives its own lifting is harder on the player it hits than one that lands late. The re-read now reconciles rather than overwrites, so a decision taken while it is running survives it either way, and the re-read still picks up changes made outside the panel exactly as before.

  Nobody was affected: no account is blocked now, none was blocked on any daily snapshot going back to 2026-07-20, and neither a block nor a lifting appears anywhere in the moderation history. The defect was real, but it had never had an occasion to fire.

1.169.14 — 2026-09-13

Fixed

A stopped clock no longer fires a round of overdue diplomacy when the empires restart (#3829)

The synthetic empires keep two clocks on disk: when each guild last held a council, and when the Observatory last published. Nothing limited how far behind those could fall, so after any gap — the service down for a day, a container that failed to come up, a long deploy — the first moment it came back read as every guild is due right now.

A council is the only thing that lets a synthetic guild declare war, demand reparations or surrender. So one restart after one outage could have fired all of those at once, across every synthetic guild, decided on a picture of the galaxy that stopped updating weeks earlier. A declared war cannot be taken back.

Now a clock that is more than three of its own intervals behind is read as stopped rather than late: it is set forward once and waits for its next normal turn. A guild that missed twelve councils holds one when it is due again, never twelve. A council that is merely late — a few hours, a night when nobody in the guild was awake — still happens exactly as before.

The Observatory no longer reports weeks-old colonies as freshly founded (#3829)

The galactic bulletin works by comparing the galaxy it has just surveyed against the last one it stored. If that stored survey had gone stale, every colony founded in the meantime came back as brand new — an empire announcing as news something that had stopped being true a fortnight earlier. The Observatory now knows how old its own survey is and will not report from one that is too old to compare against. The first bulletin after a long silence simply carries no new-colony line, and the ones after it are accurate again. One honest exception, worth stating: if a star system refuses to answer during that first survey, a colony founded there during the silence can still be reported late, on a later bulletin.

1.169.13 — 2026-09-13

Fixed

The Campaign tab finally uses a desktop screen (#2269)

The Campaign tab had no desktop layout at all. However wide your screen, it drew one 620-pixel column down the middle — the chapter map, and then everything else stacked underneath it. On a full-HD monitor that left roughly 1300 pixels of empty space either side, and pushed the outcome card, the Boss rematch desk and the parts footer below the fold, so reaching the rematch button meant scrolling past a map that had plenty of room beside it.

Now, on a screen with the room for it, the tab runs in two columns. The chapter map keeps the left column at the size it always had, and everything that used to stack underneath moves into a right column beside it: the outcome of your last engagement with its Watch button, the attempts ledger, the Boss rematch block with its ticket note and reward line, and the "Ship parts drop from Arc bosses" footer with its link into Inventory → Refit. On a full-HD screen the whole tab is now on one screen with no scrolling.

Phones and tablets are untouched. Below the width where two columns actually fit, the tab is exactly what it was — the same single column, the same order, measured pixel for pixel before and after at 768, 390, 375 and 320.

Three smaller points worth knowing. The tab decides from the width of its own box rather than the width of the window, so opening the chat dock — which takes 442 pixels off the page without changing the window — correctly puts it back to one column instead of squeezing two into the space left. If you have not engaged a level yet there is nothing to put in the second column, so the map stays centred exactly as it does today until you have a first result to show. And the Rematch button is now a secondary button rather than a second bright one: side by side with Engage they looked alike, and the one that spends your weekly ticket sat higher on the screen. Nothing about the ticket, the reward or when you can use it has changed.

Reported by spaccamazza, with a refinement from QQQQ on where the engage controls should sit that is not part of this change.

1.169.12 — 2026-09-13

Fixed

• The game no longer tells you that you have nothing when it has not looked yet (#3783). In the first seconds after opening the game — and any time the colony list is failing to load — the missions screen stated things it had not checked. The verb cards claimed facts about your whole empire from a list that had not arrived: "No Spy Probe in any hangar" to a commander holding three, "No ship with a hold", "No Recycler on any colony". That sentence also took the one line beside the Launch button away from the true reason underneath it. Now an unread list says so — "Waiting for your colonies to load." — and blocks nothing; a list that did answer, and really is bare, keeps its old sentence and keeps refusing.

• The warp row stops quoting a Dark Matter balance it has not read (#3783). Before the first reading arrived it printed "Not enough Dark Matter: 0.00/40 DM." to commanders holding thousands. It now says the balance is unavailable until it answers. The warp toggle still waits for a real number before arming — arming it on a balance nobody has read would fail the whole launch rather than cost a discount — and a balance that was read and is genuinely short still shows the shortfall with your own figure.

• The phone's destination step stops treating "not loaded" as "you only have one colony" (#3783). A four-colony empire was shown the sentence written for a one-colony empire, because an unread list and an empty one counted the same. It now says your other colonies have not loaded yet, instead of implying there are none.

Also

• Guild Pin and Invite were examined as part of the same sweep and deliberately left exactly as they are: there the unknown is whether you are in a guild at all, so showing the commands before the answer arrives would offer "Invite to guild" to every commander without one. The reasoning now sits in the code, so the next pass reads a decision rather than an oversight.

1.169.11 — 2026-09-13

Fixed

• A synthetic empire no longer remembers a raid's whole haul as its own (#3801). When several empires raid a colony together, the game records one loot figure for everybody: the coalition's total. A synthetic empire filed that number away as its personal take — and unlike a number on a screen, this is the memory it speaks from, so its chat could have credited itself with a haul three hundred times bigger than the one it actually flew home with. It now remembers its own share. Where its share was nothing, it says nothing about loot rather than rounding itself up; where the raid is too old for the game to still know the split, it says nothing rather than guessing. And when it does name its share, it now says plainly that a share is what it is, instead of a bare figure sitting next to the names of the empires it flew in with.

  Measured on the live game before the fix: 153 stored memories across 86 empires held a coalition total, and of the ones whose true split can still be recovered the worst was 371× too high. No empire had actually said one of those numbers in chat — every message ever posted was searched, and none of them quotes an inflated haul — so nothing any player was told needs correcting. The old memories cannot be rewritten, so they simply go quiet: an empire will not quote a haul it can no longer source.

1.169.10 — 2026-09-13

Fixed

• A ban that could not be loaded when the server started no longer passes unnoticed (#3784). The list of banned accounts is held in memory, because a login lasts 30 days and a ban has to take effect on the very next action rather than at the next sign-in. That list is filled once when the server starts. If that first attempt failed, the error was thrown away without a word and the list stayed empty, so every banned account was let through normally until the once-a- minute refresh happened to succeed. The server now reports the failure and retries every two seconds until the list loads, instead of waiting up to a minute.

  Nothing changes about what a ban does, who can issue one, or how one is lifted.

  No player was affected and nothing visible to players changes. No accounts are currently banned, and none have been for the last eight weeks, so this closes a gap before it could cost anything rather than repairing something that went wrong. Whether a start-up load has ever actually failed cannot be established either way — the server's logs do not outlive a restart, which is part of why the silence was worth removing in the first place.

1.169.9 — 2026-09-13

Fixed

• A warning that goes missing is not the same as "all clear" (#3782). Six screens used to treat a list that had not loaded — or that had failed to load — as if it had come back empty, and each of those lists decides a caution. The caution simply disappeared, and nothing on screen said why.
  • The system card could hide a rival guild's anchorage completely and still offer Anchor here, spending from the guild Fund on a system somebody else holds. It now says when it could not read who holds a system — and still lets you try, because the server has the last word.
  • The Peace shield row went quiet about fleets already in the air whenever that check had not come back, so a guild could spend Scrip and its weekly slot on a shield that cannot turn back what is already coming. It now tells you the check could not run.
  • The Market showed "0/5 open offers" for a board it had not read, and offered your whole warehouse as promisable — so Publish lit up for an offer the server then refused. It no longer prints a number nobody read.
  • The standing-lane form no longer stays silent about a duplicate it could not check.
  • The colony requirement ("you need ≥ N") is no longer quoted from a count of colonizations in flight that had not loaded. The refusal itself still stands — colonies you already hold are enough to be refused on, whatever is in the air — but the figure and the Command Center level are left unsaid rather than guessed low.
  • A pickup's "/ stock" line no longer ignores what your other fleets already have on order when the mission list has not arrived.

  Nothing here refuses anything the server would accept: every one of these now fails toward the warning, never toward a block.

1.169.8 — 2026-09-13

Fixed

• A pre-release check that ran out of time no longer reports itself as a failed test (#2528). The Python half of the checks we run before every release could be cut short — killed for running out of memory, or stopped by a time limit — instead of finishing. When that happened it printed exactly the same line as a genuine test failure, so whoever read it went hunting for a bug that did not exist. Once that cost two full check runs and a hold on an unrelated change. The line now says which of the two actually happened, and names the cause: killed by a signal, stopped at a time limit, interrupted, collected no tests, or a real failure with its count. The case where the check never even started — a dependency install that failed, or no network — now says so instead of claiming a test failed, which was the likeliest way to be misled of all.

  What has not changed is what passes and what fails. Everything that blocked a release before still blocks it, with the same verdict; only the explanation is different. That guarantee is not a promise in a comment — the checks now run the whole thing for real with the Python half killed mid-run, and confirm the verdict comes out red.

Also

• Three code comments describing test time limits that had been changed months earlier now describe the ones actually in force (#2528). No time limit was altered anywhere. One of those stale comments had already caused an issue to be filed a second time for a problem that had been fixed for weeks, and a second one would have invited a cleanup that silently halved a limit.

No player-visible change: this is release machinery, not the game.

1.169.7 — 2026-09-13

Fixed

• Make old arrival warnings read as history (#3400). Activity keeps the expected date and time after an attack is due, instead of saying it is arriving any moment. Future attacks still show a countdown, and Fast return receipts retain their scheduled arrival.

1.169.6 — 2026-09-13

Fixed

• Keep colony Veil calculations consistent (#3320). Server visibility checks and interface explanations now share one calculation. Current strength, balance and flight concealment are unchanged.

1.169.5 — 2026-09-13

Fixed

• Keep ship orders readable on phones (#3513). The BUILD button now wraps long labels so the full quantity and duration stay inside the button, including in the order sheet.

1.169.4 — 2026-09-13

Fixed

• Clarify agent reconnaissance limits (#2883). Updated internal documentation to describe the separate scouting and operation allowances accurately. Agent behavior and balance are unchanged.

1.169.3 — 2026-09-13

Fixed

• Correct contract rendering documentation (#3421). Fixed an inaccurate source comment about text separators. No player-visible behavior changed.

1.169.2 — 2026-09-13

Fixed

• Select the tenth colony with 0 (#2815). Desktop colony shortcuts now cover the tenth colony, and the selector tooltip explains the key.

1.169.1 — 2026-09-12

Fixed

Your raid summary now shows your share, not the coalition's (#3602). After a joint raid, the Activity headline printed the whole coalition's haul — everything every participant carried home — while the losses on the very same line were only yours. Nothing said the two figures were measured differently, so a joint raid could read hundreds of times richer than your own result: the reported case was 664× out, and the widest one still in our records is 40,135×.

No loot was ever missing and no payout has changed. The amounts always added up and every share always arrived; what was wrong was the sentence describing it.

The row now leads with your own share and names the coalition total beside it — your loot share 69,810 Fe · 416,237 Xe · 480,089 Dt (raid total 46,336,876 Fe · …). Raids fought before this change never recorded your share, so those rows say so and send you to the battle report, which has always held the split, instead of showing a number that was never yours.

Rows where there is only one figure to give stay exactly as short as they were: a solo battle, a raid you flew alone, and a raid that was repelled — where the haul is zero and so is everyone's share — all read as before.

The battle report now states a share of zero (#3602). A participant who hauled nothing had no loot line on their card at all, which looked identical to a report that simply does not carry the figure — so "you got none of it" and "this cannot tell you" were the same blank. The card now says Loot share: 0 Fe · 0 Xe · 0 Dt, which is what the older Activity rows point you to.

Reported by Elis, who measured her own share to the unit and was careful to bound the claim: a misleading summary, not missing loot.

1.169.0 — 2026-09-12

New

• Synthetic empires call relic expeditions again. For three weeks no agent-led guild had opened a single relic raid, and the board showed it: thirty-two of forty relics unclaimed, and the eight held ones untouched since 31 August. The cause was a rule inside the agent service, added on 22 August, that let a synthetic empire join a relic raid but never start one — and only once two allies had already committed fleets. Since almost every synthetic empire is in a guild made entirely of synthetic empires, the only commanders who could have started that raid were barred by the same rule, so the call was never made. The "waiting for 2 committed allied fleets" line is gone, together with the standalone call to arms it belonged to: an agent that wants a relic now opens the expedition, and its guildmates join the forming raid the way they join any other. A third of the relic board is contested again. (#3772)

Also

• Relic expeditions are still held by everything that holds an ordinary raid: the commander commits a share of its fleet and not its whole hangar, the battle engine forecasts the coalition before the call goes out and can refuse it, and a purchased Peace shield still keeps an empire at home. Agents go for unheld relics first, so no expedition is aimed at a site another guild is holding while free ones remain.

1.168.8 — 2026-09-12

Fixed

• Hear keyboard volume adjustments (#3372). Adjusting Effects volume with the keyboard now plays a short preview when you release the adjustment key, just as releasing the mouse already did.

1.168.7 — 2026-09-12

Fixed

• Home counts the ships being built (#3588). The Hangar summary now totals the hulls in your shipyard orders and labels them explicitly, instead of presenting the number of occupied slips as a ship count.

1.168.6 — 2026-09-12

Fixed

• Home names what is being built (#3589). The Colony summary now says “Building” before the active building's name, so it reads as construction in progress rather than a description of the colony.

1.168.5 — 2026-09-12

Fixed

• Discipline points named accurately (#3489). The leaderboard now labels its research score “Discipline pts” and explains that it counts completed Disciplines, excluding the Eschaton Tree. Scores and rankings are unchanged.

1.168.4 — 2026-09-12

Fixed

• The iOS app no longer prices supporter emblems it cannot sell. Before: in the native app the profile dialog showed the six supporter emblems, most of them dimmed and marked "locked", with a caption underneath reading "Unlocked by a gift in Support the game, below" and the six euro amounts — pointing at a store section the app deliberately does not have, because a gift with in-game rewards inside the app must go through Apple's own in-app purchase. Now: in the app the ladder shows only the emblems you already own (earned on the web), and the caption is gone; a player who owns none sees no supporter-emblem block at all. In the browser nothing changes: the whole ladder, locked cells and prices included, sits right above the store that unlocks them.

Also

• The rule lives in one function next to the store gate, so the two cannot drift apart again. This ship exists so the next App Store submission (1.168.x) does not carry the caption from build 483.

1.168.3 — 2026-09-12

Fixed

• Idle shipyards stand out (#3478). The Empire Overview now highlights an idle shipyard with the same gold pill as an idle construction queue, on desktop and phone. Colonies without a shipyard keep their plain status.

1.168.2 — 2026-09-12

Fixed

• The Home tab no longer crashes when the page changes size (#3512). Opening Home and then resizing the browser window — or simply zooming in and out with Ctrl+plus and Ctrl+minus — replaced the whole Home panel with the red «Signal lost» card. The resource bar and the bottom navigation stayed up, so the game looked alive with its first screen missing, and Retry brought it back only until the next size change. It now survives every crossing, in both directions, whether it comes from the window, from rotating a phone, or from browser zoom alone.

  Zoom was the important half: on a full-size desktop window, zooming past 200% makes the page report a screen short enough to count as a phone held sideways, and that was enough to trigger it. Anyone who plays zoomed in for readability could hit this without ever touching the window.

  One thing that changes with it: a desktop zoomed past 200% counts as a phone held sideways, so at that zoom Home now shows its phone layout — the hub and the next-move card, without the right-hand column. That was always the rule; players at that zoom had simply never got far enough to see it.

  Reported by Elis and by kingpaladinm, who between them pinned down both ways in.

Also

• The project's guard against this class of defect was covering only one of its shapes. It now fails on the other one too, across the whole web app, so the next occurrence is caught by a test rather than by a player.

1.168.1 — 2026-09-12

Fixed

• A guild help card no longer asks for a push nobody can give (#3770). When the ship order behind a request had left the queue, the card went on offering «takes N more» with a lit button. The first guildmate who tapped it got a refusal — and that same tap closed the requester's request for everyone, so she lost her one board slot to someone who was trying to help her, and was never told why. That card is now dark from the moment the board loads, and it says which of the two things happened: "That order is no longer in the queue" when the order has gone, and the older "has taken every hand it is allowed" only when the order really is full. The person who opened the request reads her own half on her own card, so she can withdraw it and ask on something else instead of waiting out the day. Neither line guesses at why the order left — the board knows it is gone, and says only that.

  The board only says this when it can prove it — when every order still in that shipyard queue carries a name and the one that was asked for is not among them. A queue it cannot read, or one holding an order old enough to have no name, goes on inviting exactly as before: a card that wrongly invites costs a tap, while a card that wrongly refuses costs a help nobody ever hears about.

• «Ask the guild» stops offering an order that has just been refused (#3770). Under the Chrono panel, asking for help on an order that had left the queue printed "pick another one" while the list above still showed the dead order, with the button armed on it. The queue now reloads, the way the shard buttons beside it already did.

1.168.0 — 2026-09-12

New

The battle report says what orders each fleet fought under (#3604)

A battle report is the only account you have of a fight you did not watch. It recorded the formation each commander chose before a launch — and dropped the stance, the other half of the same choice. You could see where your ships stood, and not what orders they stood under.

On a joint raid that mattered most: every contingent flies under its own stance, and nothing in the record could tell you whether a loss came from the order you picked or from something else. The multipliers printed beside each name (Attack ×3.01 · Hull ×2.55 · Shields ×2.40) are built from research and Tree alone, so a Rampart fleet's +10% hull and shields sat outside the number you were reading, with nothing on screen saying so.

Each commander's card now names the order that fleet fought under, and what it was worth in that battle:

• Rampart — hull ×1.10 · shields ×1.10, with the card stating plainly that the research multipliers beside it exclude that bonus.
• Pursuit — on a raid it names the hold it actually cost you. Pursuit buys speed with cargo, and that cargo is charged again when the loot is loaded, so a raider under Pursuit carried home 10% less than the same fleet without it. The report now says so, on the card that shows the loot.
• Fought under no stance — when that fleet carried no order into the battle. It says what is true of the fight rather than of the person: a defending colony never launched at all, and a garrison's Rampart is set aside while it sits still, so neither was a commander who "chose nothing".

Reports written before today say "Stance not recorded" and guess nothing. An old report will never be filled in from your current preference: a confident wrong answer is worse than an honest gap, and the two are now impossible to confuse. NPC defenders — the Beacon, the pirates — are never offered a stance, so their rows stay silent rather than carrying a blank that could never be filled.

Nothing about combat changed: no stance was rebalanced, and what each posture does is exactly what it did yesterday. This is the report finally saying it.

Reported by Elis.

Interceptions record both sides' orders too — the picket's and the prey's.

Also

• The orders are part of a public replay once its 24-hour embargo lifts, on the same terms as the formation shown beside them.
• DESIGN-SHIP-STARS.md §8.4 listed the stance among the things reports do not show. That sentence was true when it was written and is corrected here rather than left to go quietly false.

1.167.1 — 2026-09-12

Fixed

• A hand from your guild now reaches the order you asked about (#3761). When you ask the guild to help with a timer, the request remembers which job it is about — and it kept that memory as a position in the queue. If the job ahead finished while the request was still on the board, everything shifted up, and a second identical batch could slide into the slot the request still pointed at. The guildmate who helped then shortened that batch instead: one of its five pushes and one of the guild's daily budget spent on a job nobody had asked about, with no way to undo it. The board's "takes 4 more" was reading the same wrong batch.

  Shipyard orders have carried a name of their own since the Chrono shard learned to use one, and the aid board now uses it too. A push follows the name, so it reaches the order you meant even after the queue has moved under it. If that order has genuinely finished or been cancelled, the push is refused and nothing is spent — the request closes instead of quietly helping its twin.

Also

• Requests that were already on the board when this shipped, and requests aimed at a building or a research timer, keep resolving exactly as they did before. They were aimed under the old rule, and re-aiming them would have moved a push that was already promised.

1.167.0 — 2026-09-12

New

• Anomalies are yours to investigate — nobody can take them from you any more (#3765). Before: the galaxy drew 144 anomalies every Monday and the first fleet to land on one consumed it for everybody. With 500 commanders active, the whole board was gone by Monday night, "Investigate an anomaly" was completed on 4 deals in 100, and two dozen fleets a week flew to a signal that somebody else had already taken. Now every commander may investigate every signal once: another explorer's visit changes nothing on your map, the lost race is gone, and a second launch on a signal you already did is refused with its reason. The map's "nothing left" line now means nothing left for you.

• Your Observatory reading is yours. The outcome of a signal is now rolled per commander on the server, so what the Deep Observatory shows you is your reading and nobody else's — the planet sheet says so. Two commanders can see two different things on the same signal, and neither can work the other's out from outside the game. Dark Matter from anomalies is capped at 10 in any seven days; past the cap a Dark Matter signal pays the tier's resources into your hold instead, and the Activity row says which happened. The Dark Matter guide card states the cap.

• Denser board now, daily surveys from Monday 21 September. The weekly board grows from 12 to 18 signals per sector at once, without moving any signal already out there. From Monday 2026-09-21 00:00 UTC the Monday draw stops: six new signals per sector land every midnight UTC and each stays for three days, so the map changes daily and a signal you were shown at departure still pays when your fleet arrives after it expired. The calendar's "New anomaly survey" entry follows the new rhythm.

Also

• The planet sheet's blind-signal line now says "your outcome"; the Activity row for a repeated visit reads "you had already investigated this signal".
• Agents take no anomalies, as before.

1.166.19 — 2026-09-12

Fixed

• A synthetic empire no longer reads a call to arms it never made (#3758). When one of the game's AI empires decides to muster its guild for a contested relic, that decision is written onto the empire's own fact sheet — the record it composes every chat line from. The sheet said "0 allied fleets promised, not deployed" whether or not the call had ever actually been posted in guild chat. Read back, that is a guild that was asked and declined; in reality nobody had been asked at all. An AI empire could speak as though its guildmates had turned it down. It now says plainly when a muster has not been announced, and keeps the promise count only when somebody really did promise.

Also

• Investigating the report behind this fix turned up two answers worth recording, and neither needed code. Relic musters stopped appearing in guild chat on 2026-08-22 for a defect that was already repaired on 2026-09-07, and since 2026-09-01 the AI empires have been silent in every channel on purpose. Nothing about that silence changed here.
• What this release does not do: it does not bring relic musters back to guild chat. That is a deliberate open question — an AI empire is currently unable to open a relic raid at all, so the call it would post asks for something it cannot lead. That has been reported for a decision rather than changed here.

1.166.18 — 2026-09-12

Fixed

• The record of what the game's rate limiter refused no longer loses its first entry to a deploy (#3752). The server keeps a per-day count of every request it had to refuse, so that if a limit ever fires we can see it afterwards rather than guessing. Those counts were written to the database once a minute, so a refusal that happened in the last minute before the server restarted was simply lost — and the server is replaced around twenty-two times a day. The first refusal on each limit each day is now recorded immediately, so a limit that fired at all leaves a trace even if the server restarts a second later.

  Nothing a player can see changes: no limit, no window and no refusal behaviour was touched. This matters because a decision was resting on that record reading zero — whether to build a faster route for the travel-time estimates the missions and guild-raid panels ask for — and a measurement that can silently lose its only data point is a poor thing to decide on.

Also

• The note explaining that deferral used to say it rested solely on that zero. It now names what the zero cannot carry: one day of recording, no knowledge of anything before the record existed, a count that is a lower bound rather than an exact ledger, and a limit that needs two travel-estimate panels open at once to be reachable at all. The zero is weak evidence, not proof of safety, and it now says so where the decision is written down.
• The operator runbook was corrected in the same spirit: the immediate write is best-effort, not a guarantee, and the note on rolling back past the release that created this table now says the table fills on the first refusal rather than the first one to survive a minute.

1.166.17 — 2026-09-12

Fixed

• A Chrono shard sheet can no longer be re-aimed onto a different batch behind your back (#3669). Queue two batches of the same hull in one shipyard — two lots of 8 Freighters — open Speed up on the first and spend a shard. When that batch finished, the sheet stayed open reading the same label, with the same four live Use buttons, now pointing at the other batch. Nothing visible changed, because there was nothing that could: twin batches wear the same name, the same icon and the same everything.

  The sheet was remembering a position in the queue, and a position is not an order: when one order leaves the yard, the ones behind it slide up into the gap, so the slot you chose quietly came to mean somebody else.

  Orders now carry an identity of their own, and the sheet follows the order rather than the slot. The batch you opened stays the batch you opened, wherever the queue shuffles it to. If it genuinely finishes and leaves, the sheet says so and closes the four Use buttons until you choose again — and when the reason it left is the shard you just spent, it still tells you the shard landed.

  Nothing changed about what a shard costs, what it cuts, or the order your queue runs in.

Also

• Orders already in your shipyard when this went live keep the previous behaviour until they finish, and the warning that has been catching this since #3579 stays in place to cover them.

1.166.16 — 2026-09-12

Fixed

• A shielded empire no longer calls a relic coalition it cannot join, and one already holding a call now says why it is not flying (#3703). When an empire buys the guild's Peace shield, every door that would launch an attack closes for eight hours — striking alone, opening a guild raid, and joining one. The door that organises a relic expedition did not close. So a synthetic empire could post a call to arms reading "waiting for 2 committed allied fleets before I move" and then not move, for hours, without ever saying why. A commander whose words and deeds come apart in front of a guild being asked to commit fleets is the one thing these empires are never allowed to be.

  A call to arms now only ever comes from an empire that can answer it itself. And for the other order of events — the call goes out first, the shield is bought afterwards, which turns out to be the more common of the two — the empire now knows, whenever it speaks, that its own shield is why its fleet is not flying, and says so instead of promising a move it cannot make.

  The muster itself is deliberately unchanged: it still stands, still collects answers from guildmates, and still expires on its own clock. A call already made is explained, never silently dropped — anyone who answered one still gets a word.

Also

• Measured against production before shipping: two of the fifty-six musters open at the time had been called by an empire holding a live purchased shield. No guildmate has ever committed a fleet to any muster, so no player was left waiting on a call that could not be answered.

1.166.15 — 2026-09-12

Fixed

The header no longer wears a guild you have already left (#3707)

Change guild — leave one, join one, get moved or kicked — and the commander name at the top of the screen kept the old guild tag on it. The chat directly underneath already showed the new one, and so did the guild page, so the screen was not just behind: it contradicted you, and the natural reading was that the change had not gone through.

It cleared on its own the next time the app was started cold, but not quickly. Measured on a day and a half of real traffic, half of phone sessions carried the old tag for 19 minutes or more, and one in ten carried it for over five hours. The desktop was not spared either — it showed the stale tag beside your name while a second, correct one sat a few centimetres away beside the colony name.

The tag now comes from the same live guild data the rest of the game already reads, so it follows your guild: it disappears when you leave, appears when you join, and changes when you are moved. Nothing about joining, leaving or being moved has changed — only what the header tells you about it.

1.166.14 — 2026-09-12

Changed

• The guild raids panel's request budget now rests on measurements instead of an estimate (#3678). Nothing on screen moves: every travel time on the Joint Raids panel is printed exactly as before, and a note that is still loading or has failed still prints nothing while holding its space, so the join button never shifts under your finger.

  What changed is the reasoning behind the limit that protects the panel — the one that, if it ever fired, would make every travel time on the panel disappear at once with no message. It assumed a guild could have four raids open at once. Measured against the real game — 3,557 raids, 12 guilds, 45 days — the most a guild has ever had open at once is three, and that happened twice; 3,540 raids were the only one open in their guild.

  That turned out to matter less than it looks, and the note now says so rather than claiming the panel is comfortably safe: the limit can still be reached by one player in the ordinary single-raid case if they adjust their fleet about once a second. What actually supports leaving it alone is the refusal counter added in v1.160.5, which has recorded no refusal at all.

  So the planned batch request is deferred rather than built, with the condition that would revive it written down next to the limit: a single recorded refusal, or a guild seen with four raids open at once.

Also

• The panel's ETA requests are now pinned by tests that count them. Before this, a change that made the panel ask twice per raid row instead of once would not have failed anything.

1.166.13 — 2026-09-12

Fixed

An agent's outpost could be mistaken for an abandoned shell moments after the agent had paid to rescue it. When one of the synthetic empires bought a building for one of its own outposts, the price left the colony's purse immediately, but the order it had just paid for was not recorded against that colony until the empire next looked at the galaxy. For the rest of that minute the outpost read as a world with no income, no way out and an empty purse — so the empire's capital shipped it an emergency care package it did not need, and pushed a freighter up its build order to carry more of them. Two halves of the same empire disagreeing about one of its own worlds, over a building that was already on its way.

The order is now written against the colony the moment it is accepted, beside the payment that bought it, so an outpost that has just funded its own way out is no longer treated as stranded.

No resources were ever lost to this, and no purchase was ever made twice: the game's ledger records the charge and the order together or not at all, and a repeat order for a building already under construction is refused. What went wrong was confined to how the synthetic empires read their own colonies between one look at the galaxy and the next, and no human empire's resources were involved.

Also

Outpost garrisons are unchanged by this. They are only ordered when a world's build slots were already full when the empire looked — so that the money spent on hulls is money spare rather than money the build order had just claimed — and that reading now comes from a snapshot taken before the building work, rather than from a record the same minute's work had altered.

1.166.12 — 2026-09-12

Fixed

• The War Room no longer tells you a shielded enemy was "recently hit" (#3702). A protected target on the war board carried the line "Post-raid shield: someone hit them recently." Neither half was true: that shield has not existed for a commander since 6 August, and the protection the board was actually reading says nothing whatsoever about whether anyone attacked them. A player who had simply bought peace — or gone on leave — was announced to the whole enemy guild as freshly beaten, which is an invitation to strike a target who is not softened at all. The row now says what it can stand behind: the shield covers their whole empire and not just the colony you are looking at, it refuses new attacks and raids while the countdown runs, and any strike they make or raid they join lifts it sooner — with only the Pirate Beacon free.

• Recalling an attack now names everything the launch cost you (#3702). The recall sheet mentioned only "a newbie shield". Creating an attack gives up three protections at once — the novice shelter, your empire shield and the shield on every colony you own — and turning the fleet around buys none of them back. The sheet now says so, and says that striking the Pirate Beacon costs you none of it, which is the case the old line quietly relied on.

• A scouted target's shield drops when they join a raid, and the spy report now says so (#3702). The card said a target's shield "drops early if they attack anyone themselves", which left out the act that opens the same window: joining a guildmate's raid. The novice shelter line had the same gap. Both now name both acts, and both name the Pirate Beacon exemption.

Also

• Nothing about what any shield costs, does, or protects against has changed. These are corrections to what the game told you about protections that already worked this way.
• These three screens are part of the app itself rather than text the server sends, so an installed phone app shows the corrected wording once it updates.

1.166.11 — 2026-09-12

Fixed

• A colony that produces nothing is no longer mistaken for a healthy one (#3736). Before: the stalled-colony rescue counted an Extractor as an income even when the colony had no generator — so a colony producing exactly nothing, and unable to accumulate its way out of producing nothing, was read as fine and passed over. Now: a colony with a producer and no generator counts as stalled when it cannot afford Orbital Panels or a Reactor. The empires run by the game apply the same rule, and no longer send a supply run to an outpost whose paid order is already on its way.

Changed

• One rule instead of two (#3736). The rescue and the empires' own copy of it had drifted apart and disagreed on three situations at once. They are one rule again, and a cross-language test now runs both implementations over the same table of colonies and fails if they ever drift again.

Also

• Measured on production before shipping: no colony changes classification in either direction, so nothing is owed to anyone and nobody stops being owed. The rescue is operator maintenance that is run by hand; this release does not run it.
• The rule is now written down in the design docs, next to the starting-kit rule it depends on.

1.166.10 — 2026-09-12

Fixed

• The Promote to Officer button no longer waits for the server to say no (#3624). A guild that already held its three officers still showed a live "Promote to Officer" on every member, so the only way to find the cap was to press the button and read the refusal. The control now goes quiet before the tap and says what is in the way — "All 3 officer seats are taken. Demote an officer to free one." — and a leader's roster head states the seats beside the member count it has always shown. It stays visible rather than vanishing: a control that disappears suggests the member is the problem, when what is full is the bench.
• The same button stopped being offered where it could never work. Only a guild leader appoints officers, but the roster offered the promotion to officers too, and the server refused it every single time. Officers now read the reason instead of pressing into a wall.
• A guild can hold more officers than it has seats, and the panel now says so honestly. Handing command to a plain member leaves the outgoing leader as an officer, which can put a full guild one over the cap. There the old wording would have read "(4/3)" and told the leader to demote one officer, which frees nothing; it now counts the demotions it actually asks for.

1.166.9 — 2026-09-12

Fixed

• Synthetic empires were reading their own outposts short of the ships parked in them (#3713). Before every message and every decision, an agent empire is handed a summary of its real position, including how many combat hulls sit in each of its outposts. That count came from a list of "ships that do not fight" which had fallen behind the fleet catalogue: it treated the Shroudrunner as unarmed, even though the raider has an attack value and defends a colony by default, and it had never been told about the Continuum Hauler, so it counted that freighter as a warship.

  Ten agent empires were being told their outposts held 88 fewer combat hulls than they actually did; the worst-affected outpost was reported as holding 532 hulls when it held 573. An outpost garrisoned only by Shroudrunners was described to its own owner as empty. Agents speak and act on that summary, so an empire could talk down a garrison it really had, or pass up a defence it was able to make.

  The count now comes from the same rule the game itself uses to decide which ships defend a colony, so what an agent believes about its own fleet matches what the engine would actually put in the air. A cross-language check now fails the build if the two ever disagree again.

Also

• The module refit desk's three error codes had no player-facing text, so a player who asked the refit adviser twice in quick succession was shown the raw code refit_advice_throttled inside a game panel. They now read as sentences.

1.166.8 — 2026-09-12

Fixed

• The end-leave control now says that a peace shield you bought goes with the leave when you end it early, and that there is no refund (#3709). It already warned that ending a leave drops the empire shield, but a player read that as the shield the leave itself granted — a free thing. Declaring a leave quietly takes over a purchased shield and stretches it to the leave's end, because every leave on the picker (7, 14, 21 days) outlasts every shield on the shelf (8, 24, 48 hours), so hours paid for with Scrip end up inside the leave's clock and end with it. Both the HUD row and the phone Command sheet now say so before the button is pressed.

Nothing about leaves, shields, prices or durations changed. A leave that runs to its own end still takes nothing — only an early exit does, and that was already true.

1.166.7 — 2026-09-12

Fixed

• The Supply Convoy card can be closed (#3717). For the whole week of the supply line, the convoy card ("Supply Convoy 3 of 5 — docks in 23h 42m") sat pinned above every tab, the phone's Base included, with no way to put it away. It now has a close control. Closing hides the countdown until something actually changes: the moment the convoy docks, the delivery card comes back once, and each new convoy of the arc starts visible. The choice is remembered per player and across reloads. The last convoy's docked fork, the one decision that ends the arc, cannot be closed: it stays until you decide.
• The convoy has a permanent home on the Fleets tab. A convoy is a fleet inbound to your colony, so it now also appears there, first under «Active missions» on desktop and right under the segment pills on the phone, on every segment. There it has no close control, so the one place it must always be found is the one place it cannot be hidden. The banner no longer shows on that tab, so the card is never on screen twice.
• «No missions underway» is no longer printed directly under the convoy card for a newcomer who has launched nothing yet.

1.166.6 — 2026-09-12

Fixed

• Smarter ship configurations (#3731). AI empires now consider complementary modules, combat Cores and worthwhile replacements, using combat forecasts and the strength of their real fleets to choose upgrades.
• Complete refits (#3731). Up to three modules of a ship class can change together, with the normal component costs, refunds and cooldowns. An unaffordable or outdated plan leaves the whole configuration unchanged.
• Careful equipment spending (#3731). Agents defer refits during incoming attacks and avoid frequent reversals or repeated purchases after an uncertain response.

1.166.5 — 2026-09-12

Fixed

• The campaign adviser's refusals read as sentences, not codes (#3729). Since v1.166.3 the API could answer an adviser read with two new refusals — asked again within two seconds, or asked for a kind of advice the desk does not know — and neither had a player sentence, so the raw code would have reached the screen. Both now read in plain words. No screen calls the adviser yet; the sentences are in place for the first one that does.

1.166.4 — 2026-09-12

Changed

• The season podium pays ten times the Dark Matter (#3714). Until now the season close credited 40 Dark Matter to first place and 20 to second. From this season on, first place receives 400 and second 200. Third place is unchanged: three crates and no Dark Matter. The prize table in the Season panel shows the new amounts before you play, and the close credits exactly what the table says. Seasons already closed are not back-paid.

1.166.3 — 2026-09-12

Fixed

• Campaign progress for AI empires (#3725). Agents now assess actual battles and choose useful formations, allowing capable rosters to advance through several levels in a sitting. They remember failed orders, respect activity hours, and stop to prepare when their current force needs improvement.
• Campaign preparation (#3725). Agents can pursue useful combat research, equip earned modules and buy affordable ship-star upgrades while preserving colonization, urgent defense and existing savings commitments.
• Campaign forecasts (#3725). Forecasts include donor combat bonuses, and the campaign recognizes changes to unlocked ships and other combat upgrades when comparing previous attempts. Existing native clients remain compatible.

1.166.2 — 2026-09-12

Fixed

• The Python test suite is green again on main (#3721). Nothing changes for players. Four synthetic-empire tests had fallen behind two rules that shipped in v1.166.0: a producer without power is stalled, and a relic board that could not be read is not quoted at all. The fixtures now carry a Reactor, the income expectations follow the engine's curve for a powered producer, and the relic test pins the dropped stamp.

1.166.1 — 2026-09-12

Fixed

• A finished research lane no longer sits on «finishing up…» (#3720). When a research reached the end of its clock, the lane kept its full bar and «finishing up…», and the queued order behind it did not start on screen until you left the Research tab and came back. The research itself was done and the next order had started on the server; the tab just did not ask until its next 30-second poll, which a phone in the background suspends. The tab now reloads the queue the moment a lane's clock runs out, so the finished lane disappears and the next order takes its place with its own countdown.

1.166.0 — 2026-09-11

New

• Agents contest territory and bonus relics (#3716). Synthetic empires now scout worthwhile objectives, prepare missing probes, use battle forecasts before committing their fleets, and send limited watches to hold captured ground. Urgent defense of home and guildmates keeps priority.
• Occasional anomaly expeditions (#3716). Agents can investigate discoveries visible to their own sensors. A staggered weekly allowance keeps the roster from exhausting the shared supply.

Changed

• Campaign and Season Boss preparation (#3716). Agents save earned parts for useful ship modules, account for completed combat upgrades when deciding whether to advance, and revisit the Boss after equipment improvements.
• Guild finale participation (#3716). Leaders can contribute to the Eschaton Project even without a language model, using the current Fund balance while preserving money for other guild needs.

Fixed

• Reliable fleet commitments (#3716). New objective and expedition decisions respect existing reservations, current defense needs and missions already in flight, including when a launch response is lost or the service restarts.

1.165.2 — 2026-09-11

Fixed

• The Shroudrunner's card named the wrong one of its three counters (#3608). The hull's one-line description read "Demolishes spy probes" — and a spy probe carries nothing and has 50 hull, the cheapest and least consequential of the three things a Shroudrunner is good at. On the Shipyard card, where that line is the only sentence a hull gets, it was the whole of what the game said about the ship, so anyone reading it was told to build a light hull for chasing scouts rather than for what it is actually worth owning: the ×3 against freighters. The card now reads "Demolishes freighters", agreeing with the "Raids freighters and colony ships" the ship reference already showed underneath it. No counter, multiplier or combat rule changed — the ship fights exactly as it did — and no other hull's line moved. Reported by TartarusMkII, who asked in universe chat what Shroudrunners are for.

Also

• The description for a hull whose every counter target is unarmed is now derived the same way the economic line is, so the two can no longer disagree about which target is worth naming. The Shroudrunner is the only hull in the catalogue this applies to today.

1.165.1 — 2026-09-11

Fixed

• The peace shield now tells you that a raid ends it — joining one included (#3616). The guild store sells the shield in three lengths, and the row you press Buy on said only "Attack anything but a Pirate Beacon and the shield ends early." It never said the word raid. But a guild raid ends the shield too, and on the same terms whether you opened it or merely joined a guildmate's — so a player who bought eight hours of peace and then answered a guildmate's call spent those hours on a button that says Join, on terms the row never stated.

  The row now reads: "Attack or raid anything but a Pirate Beacon and the shield ends early — joining a guild raid counts." The same correction reaches the shield banner on your dashboard, the protection tooltip on desktop and the protection sheet on your phone.

  Nothing about the shield changed — not what it does, not what it costs, not one protection rule. Only what it tells you. And it stays exact in the other direction too: espionage, a garrison sent to a guildmate under attack, transports, transfers, pickups, colonisations, harvests, surveys, patrols and interceptions all still leave your shield standing.

  Found by QQQQ, who asked in the bugs thread what actually breaks the shield.

Also

• The synthetic empires' own briefing carried the same gap and has been corrected with it: an agent holding a peace shield now knows a raid it joins spends the hours it paid for, so it no longer offers its guild a fleet the game will refuse to let it send.

1.165.0 — 2026-09-11

New

• Pick your guild's colour on the map (#3700). Before, a guild's colour came from a wheel of eight, handed out at founding: the ninth guild onwards was painted the same paper-grey as empty space, and four of the twelve guilds in play sat there. The wheel now holds 24 colours, twelve hues in a vivid and a bright key, and a guild's leader or officers pick one from the Map colour row in the guild admin panel, under the emblem. Colours another guild holds are faded; press one and it tells you whose it is. Two guilds can never wear the same colour, and the chart, the chat tag and the leaderboard tag change the moment you save. The four grey guilds get a colour automatically with this release; nobody who already had one changes.

Also

• Phones on an older app build show a guild that moved to one of the new colours in paper-grey until the TestFlight build from this release lands.

1.164.0 — 2026-09-11

New

• The dock raid — the Shroudrunner finally has a verb (#3696). Before, a cloaked fleet escaped the map and the pickets and then arrived at a colony where the only thing it could do was fight with 40-attack hulls; in thirty days no human flew a cloaked attack. Now an attack by cloaked hulls alone (Shroudrunners, Corsairs) can be declared a Dock raid in the mission composer. At impact the flight's signature is compared with the target's Sensor Grid, empire-wide. If the grid cannot read it, there is no battle: the raiders run down up to a quarter (rounded up) of each grounded hull type their counters reach and take the loot their hold can carry, then leave. If the grid reads it, the raid is an ordinary attack, and a light fleet loses it. The report card, the recent-battles row, the alert and the Activity line say which happened, and name the Sensors level that reads a flight of that size. An unseen raid charges no siege tick, notches no Outpost breach and earns no Chrono fragment.
• Sensors have a daily reason to exist. The composer shows the level a chosen fleet hides under; a fleet too large for the empire's Espionage is told so on the same row instead of losing the option silently.

Changed

• The ship reference's rule line for the Shroudrunner and the Spectral Corsair now names the dock raid.
• Synthetic empires can declare dock raids too: they size the largest stack that still runs dark, use a scouted Sensor Grid when a spy report saw one, and remember for a week the empires whose grids read them. In chat they tell a heist as a heist, never as a battle won or lost.

Also

• No hull statistic moved: nobody's Power or points change with this release.
• Design record: DESIGN-DOCK-RAID.md.

1.163.0 — 2026-09-11

New

The Guild Duel: every day of the League week has a theme, and combat is the weekend (#3698). The weekly League fixture used to score battles and nothing else, so a guild with no appetite for fleets had nothing to play for six days out of seven. Now each day of the fixture is a contest of its own, on the pattern of Last War: Survival's Alliance Duel:

| Day | Theme | What scores | |---|---|---| | Mon | Construction | building orders you queue (1 point per 100 resources committed) | | Tue | Research | research you start (same rule) | | Wed | Shipyard | ship orders you queue (same rule) | | Thu | Expeditions | anomalies investigated (200 each) and campaign levels cleared (their war points) | | Fri | Mobilization | all four of the above | | Sat + Sun | Combat | battles against the paired guild, and nothing else |

Battles score the fixture on Saturday and Sunday only. A raid on a Tuesday is still a raid, but it moves no League score. The week goes to days won, then to total points: a guild that shows up every day beats a guild that shows up once with a huge fleet, because two combat days do not outweigh five.

In the fixture box you now see the days-won tally, seven day chips with each day's score and a mark on the days already decided, today's brief (the theme and exactly what scores, with the hours left), and your rank among your guild's scorers today. The match-room door tells you on weekdays that battles score at the weekend. A closed fixture in the history shows the days beside the points, and the verdict in your feed and guild log leads with the days.

The fixture that was already open when this shipped is judged on points as before: its day-by-day ledger started mid-week, and the panel says so rather than drawing empty days.

Also

• Not in this release: daily personal prizes (the rank is shown, the purse is still the weekly League purse) and bot empires favouring the day's theme.

1.162.0 — 2026-09-11

New

• Seasonal participation for agents (#3699): eligible agents now challenge the Season Boss after meaningful improvements and use ready weekly campaign rematches.

Changed

• Useful expansion (#3699): agents work through the actual colony-capacity blocker, save for one founder and send a starting kit that includes power. Existing colonies needing rescue keep priority.
• Research with a budget (#3699): investment decisions compare real economic gains, and selected research can accumulate its cost instead of losing the money to unrelated purchases.
• Guild cooperation (#3699): agents accumulate donation allowance, reconsider open aid requests and can defend guildmates from safe secondary colonies using the server's travel estimates.

Fixed

• Resource accounting and rewards (#3699): construction, research and ship orders no longer reuse an already-spent opening balance. Earned contract rewards are collectible on every active day, and refused guild maintenance is reconsidered without waiting a full successful-purchase cooldown.

1.161.2 — 2026-09-11

Fixed

• The Wraiths no longer forget a strike when two of their runs overlap (#3505). The faction's pressure ledger, the record of who it has struck and when, was written whole by each run; the long scheduled run, ending last, wrote back a copy that never saw the strike an operator's run had recorded minutes earlier, and on 2026-09-08 a strike vanished this way. A save now takes a lock, re-reads the file and applies only that run's changes: strikes add, the latest strike time wins, and a run adopts the merged file so a later save cannot subtract what another run added. A file that becomes unreadable mid-run keeps what the run had loaded instead of resetting to "nobody has ever been struck".

Also

• The lock protects the record, not the decision: two operator runs that overlap can still both strike the same player inside the six-hour cool-off, and the ledger then honestly says two. Both writers run on the operator's machine, so this takes effect when that checkout pulls main.

1.161.1 — 2026-09-11

Changed

• The €1 gift pays three Chrono shards of each size, not five (#3693). At five, the €1 rung paid twice the per-euro rate of the €100 rung, so splitting a gift into €1 charges collected more shards for less money. Three of each at €1 levels it with the €5 and €10 rungs; the five higher rungs are unchanged (15, 30, 60, 120 and 250 of each size). This corrects the figure published in 1.161.0.

Fixed

• A refunded gift now takes back exactly the shards that gift paid, read from the gift's own record, instead of whatever the ladder says at refund time. A gift paid at five and refunded after this change returns five; a purchase from before shards existed returns none.

1.161.0 — 2026-09-11

New

• Every gift now comes with Chrono shards (#3690). Each supporter package pays shards of every size, 5m, 30m, 1h and 4h alike: 5 of each at €1, 15 at €5, 30 at €10, 60 at €25, 120 at €50 and 250 at €100. They land with the Dark Matter on every gift, not only when a tier is first reached, and they sit outside the weekly shard pool: they neither count against it nor wait for it. The supporter strip, the post-checkout receipt and the Activity diary all say so; the diary shows one line per gift with the count.

Changed

• A refunded gift takes its shards back, down to what is still held. The Dark Matter and the hulls stay, as before.
• The Inventory shard panel and the Chandlery's pool caption now name the gift as the second exemption from the weekly cap, next to the bastion's prize, so a supporter who just received a thousand shards is not told their week should have stopped at four hours.

1.160.6 — 2026-09-11

Fixed

• The research Bench now quotes the price the game will actually charge (#3644). The queue strip on the Research tab priced each waiting order from the copy of the research tree built into the page you had open, while the card beside it already read the live figure. If a research price changed while your tab stayed open, the two disagreed about the same order — and the strip could show a "needs Dark Matter" wall for a toll that was no longer being charged. The strip now reads the same live figures as the card, and the "needs Laboratory lvl N" wall does too. Reloading the page was always the workaround; now it is not needed.

  Two things worth being straight about. Nobody was ever charged a wrong amount: the price is decided by the server when an order starts, and the strip only ever described it. And we checked before shipping — no empire is anywhere near the levels where the two figures currently differ, so nothing on screen changes for anyone today. This is for the next time a research price moves.

Also

• The Discipline cards still price from the page's own copy, so this is one panel made current, not the whole tab. That is deliberate, and named here rather than left to be discovered.

1.160.5 — 2026-09-11

Changed

• The game's rate limiter now keeps a record of what it refuses (#3672). Nothing changes for players: no limit, no window and no refusal behaviour moved, and there is no new screen. This is machinery for the people running the server.

  Until now, when the API refused a request for coming in too fast, it wrote a single line to the server's log — and only one line per player per minute, however many requests were refused. That log is erased every time a new version is deployed, which happens several times an hour. So "this limit has never fired" could only ever mean "not since the last deploy", and nobody could say whether a real player had ever been turned away.

  Refusals are now counted and written to the database, one row per day per limit, so the record outlives the deploy. What it deliberately does not keep is who was refused — that stays in the server log and nowhere else.

  Worth recording alongside it: the check that prompted this found the limiter had fired, once, on a real human player, hours before the record existed to catch it.

Also

• One of the emergency levers documented for server operators was reaching no server at all — it was named in the runbook but passed to no container, so setting it would have done nothing and said nothing. It is wired up now, with a test that fails if any lever loses its wiring again.
• The runbook's instructions for reading the new record, and for rolling this release back, were both written and then verified against the live server rather than assumed.

1.160.4 — 2026-09-11

Fixed

• Raiding agents can see their guild's Hypergate network again (#3681). When a guild holds a Hypergate at both ends of a route, the trip is replaced by a single jump. Agents worked their raid travel times out from the map instead, and the map does not know about the network — so on their own guild's routes they were reading a journey five times longer than the one they actually flew.

  Two things followed from that, and both are fixed. A guildmate joining a raid refused to bring its freighters along on raids the convoy would have reached comfortably, so the raid landed with a smaller hold and took less loot. And an agent opening a raid attached a slow convoy to a fast strike wing on a hop that was already short, which pushed back the hour it announced and handed the defender more warning than the strike was meant to give.

  Both decisions now ask the server for the same travel time the launch itself will use. Measured against what the server really flew, across 26,449 real missions from the past week: the agents' own figure was a median 5.09x too long on routes their guild gates, against 1.11-1.15x where no gate applies. Replaying the decisions, 44.2% of gated raid joins and 63.8% of gated raid openings were being decided on the wrong number.

  One visible consequence, stated because players will see it: on those gated routes the raid now lands sooner, so the window guildmates have to join it is shorter. The window has always been the distance — that is the design — and the distance is now read correctly; the wider window before was not a choice, it was the over-estimate. Measured across the last week, six raids would have flown differently, and every guildmate who joined them would still have arrived in time.

Also

• Two further travel figures in the agents' target-ranking code have the same blind spot and are deliberately left as they are for now (#3681). They only nudge which target an agent prefers — never a yes/no, never anything the agent says out loud — and correcting them would mean one server request per candidate target per agent per tick. That waits on #3672, and the reason is now written down where the next change to this code will find it.

1.160.3 — 2026-09-11

Fixed

• The war room and the league match room were over-estimating every flight time, and warning you off strikes you could have made (#3615). The estimated time to each enemy colony was worked out from raw distance alone — it did not count your Relay, your own Hypergates, or your alliance's Hypergate network. So on any route your network actually shortens, the room quoted a trip far longer than the one your fleet would fly. Measured against what fleets really flew over the last two weeks, the old figure was a median 2.42× too long, and over 1.5× too long on 73% of attacks.

  That mattered beyond the number, because the panel draws a conclusion from it: "even the fastest fleet lands after the war closes: points scored there would not count." An inflated figure fires that warning on colonies your fleet would have reached in time — the game talking you out of points that were there to take. On the live league boards, 43% of target rows now show a shorter time, and at the worst moment of a fixture's closing hours that warning was appearing falsely on 12% of rows.

  Both rooms now price targets with the same planner the launch itself uses, so what you are quoted is what you would fly.

  Reported by QQQQ (arcticpaladin).

Changed

• The "lands after the war closes" warning now says what it is measuring (#3615). It reads "from your capital, even the fastest fleet lands after the war closes", because the room prices every target from your mother colony — a fleet you already have parked closer can still make it. The line says so where you can read it, not only in a tooltip.

1.160.2 — 2026-09-11

Fixed

• Guildmates get the relief the machines were declining to send (#3666). A synthetic empire deciding whether to garrison a guildmate under attack worked out the travel time itself, and that arithmetic cannot see the alliance Hypergate network — which does not shorten a route by a percentage, it replaces the distance with a single jump. So on any route a guild gates at both ends the empire believed the flight was several times longer than the one its fleet would actually fly, decided the rescue would land after the battle, and often stayed home. Over the last fortnight 1,801 of 1,912 such defences flew exactly that kind of route; replayed against the flight the game really ran, the estimate was off by a median factor of 3.8, and 29% of the rescues that could be measured were called a miss when they would in fact have landed in time. They now ask the same planner the launch itself uses.

• And they no longer quote a relief time they cannot stand behind (#3666). When the server cannot answer, the empire still sends the fleet — it has to decide — but it states no clock rather than a wrong one. It still tells you plainly when it thinks the ships will not make the strike, or will miss the first blow and stand for the next; it just stops attaching minutes it has no way to know.

Also

• The offline simulator still uses the old local estimate, so conclusions drawn from it keep the old pessimism on gated routes. Four other internal travel estimates in the agent service are unchanged: none of them is ever spoken to a player, and each errs in the safe direction.

1.160.1 — 2026-09-11

Fixed

• Joint Raids now quotes the travel time the launch will actually fly (#3577, #3595). All three doors of the guild raids panel — open a raid, join a raid, send an ally defence — used to work their travel time out in the browser. That arithmetic could not see the alliance Hypergate network, and the network does not shorten a route by a percentage: it replaces the trip with a single jump. So on any route a guild holds gates at both ends, the panel printed several times the real figure.

  Two players measured it. Elis priced one fleet on one route through both forms without launching: Joint Raids said 16m, the mission composer said 3m19s, and the raid she had opened had about 3m20s left. gobbibomb read "25m to get there — it may land 2m after impact" on a rescue that was one jump away. Over the fourteen days before this shipped, 327 of 429 defences sent to a guildmate — 76% — flew a route the guild gates at both ends; on the open-a-raid door it was 2 raids of 22, and on the join row 10 joins. (Gate levels are today's, so the split is an approximation of what was true on each of those days.)

  All three doors now ask the server for the same figure the launch uses. Where the panel used to overstate a rescue, it now tells you the truth about it — and on the door that opens a raid, the number the whole guild times its joins against is the number the fleet will actually fly.

• While the answer is on its way, the panel says nothing rather than guessing. For a moment after you change the fleet, the stance or the launch pad, there is no travel time on screen; the line holds its place so nothing jumps, and the figure appears when it arrives. If the request fails there is no figure at all. That is deliberate: a placeholder several times wrong sits right under a sentence telling you it sets the impact hour for your whole guild, and there is no way to tell it from the real one.

• A raid whose impact has already passed no longer warns you about your fleet's speed. The row already says the raid has landed and the Join button is already grey; the warning above them was answering a question nobody had asked, with the wrong reason.

Also

• Supporters keep their VIP fleet-speed perk in these estimates. It is now read from your account on the server rather than claimed by the browser.
• Installed iOS and Android apps, and browsers still holding the old page, keep showing the old figure until they update. Nothing about the way the app talks to the server changed, so old builds keep working exactly as they did — they simply keep the old arithmetic until they refresh.

1.160.0 — 2026-09-11

New

• Every VIP perk is live (#3675). The VIP ladder shipped in 1.159.0 with four perks marked "coming". They now apply: shipyard builds finish sooner from VIP 4, fleets fly faster from VIP 6 on every door (missions, raids, transfers, supply lanes), warehouses hold more from VIP 9, a daily Dark Matter stipend lands on your Dark Matter screen once a day from VIP 1, one extra open market offer at VIP 5 beyond the Trade Portal's cap, and Dark Matter production rises at VIP 8, 9 and 10. The supporter strip no longer says "coming" anywhere.
• The stipend says where it came from. The Dark Matter tile and its phone sheet carry a line naming the amount and when the next one lands, so a balance that jumps overnight has an explanation on the same screen.

Fixed

• Previews agree with the launch for supporters. Every estimate the client draws — build-time quotes, mission ETAs, the "home undefended for" window, raid reach and the ally-defence verdict, intercept rows, the storage ceiling behind the attention alarm and the "full when the surge ends" countdown, the Dark Matter rate on the HUD and the Accelerator sheet, the Trade Portal sheet's slot rows, the market composer's overflow warning — composes the same VIP factor the server applies. Before this a VIP player could be told a join was too late, or a warehouse full, when the server would have accepted the fleet or kept accruing.
• The market composer no longer names a Requisition Surge you are not running. At VIP 9 and 10 the surge sentence fired permanently; it now asks whether a surge is actually active.
• The mission advisor sizes the fleet with the donor bonus a supporter's ships actually fight with.

Also

• The daily stipend is claimed and paid in one database transaction, so a restart between the two can no longer burn a day's stipend; two overviews racing on the same second see the same next-claim time.
• The hull sheet's Dark Matter preview prices the pending window at the VIP rate, like the Dark Matter screen does.
• The deploy runbook gains the rollback block for the stipend column and for the store columns 1.159.0 added.

1.159.0 — 2026-09-10

New

• Support the game (#3664). A new section in the commander card, on the web only, lets you make a one-off gift through Stripe at six amounts: €1, €5, €10, €25, €50 and €100. Before, there was no way to support the server at all. Gifts add up: the tier is the total you have given.
• Supporter tiers and their thank-yous. Every tier gives a coloured frame around your avatar, shown wherever your portrait appears, and a reserved emblem you can wear from the profile picker. From €10 up, a small permanent bonus to attack and HP (+1%, +3%, +5%, +10%). From €25 up, legendary hulls delivered to your home colony: 100 titans at €25; 300 titans and an annihilator at €50; a destroyer, a second annihilator and 300 Voidlance Cutters at €100.
• Dark Matter on every gift: 50 per euro, whatever the tier, so a second gift keeps paying. Each row of the support section shows what your next press delivers, priced against what you have already given, and marks tiers you already hold.
• VIP. Every euro is 100 VIP points for life. Ten levels, from €5 to €1000, each with a numbered crest next to your name on the leaderboard, in chat and in the guild roster. Two perks are live today: production, +2% per level up to +25%, and research speed, up to +20%. The other perks on the ladder (shipyard, fleet speed, storage, a market slot, a daily Dark Matter stipend and Dark Matter production from VIP 8) are printed as "coming" and are not yet in effect.

Also

• Hulls a gift owes are recorded on the purchase and delivered on your next visit if the first attempt was lost; a refund lowers your tier but never claws back hulls or Dark Matter, and never lets a later gift collect the same hulls twice.
• The attack forecast and the Advisor now fly a supporter's bonus, so the preview matches the battle.
• The AI empires have one more standing rule: they never claim or promise to have spent real money.

• The store stays closed until it is switched on server-side; nothing changes for you until then.
• The privacy page now names Stripe and what it sees.

1.158.12 — 2026-09-10

Fixed

• A raid whose strike has already landed no longer offers you a place in it (#3600). When a raid's impact hour passed while you had the War Room open, the row kept the Join with N ships button live: the countdown read impact now, the raid still showed as forming, and pressing the button sent a join that was then refused. Nothing was ever lost to it — a late join has always been turned away — but you could pick a fleet, commit to the click, and only find out afterwards that the race was over. And because nothing closes the row on its own, it could keep offering that click for as long as the panel sat there.

  The button now goes quiet the moment impact arrives, and the row tells you where it stands: "Impact passed — waiting for the report." It is not a dead row — the strike has landed, and the next refresh brings the report, which still carries your share if you were aboard. The sentence sits under the button rather than above it, so nothing shifts under your thumb at the second the button closes.

  The ally-defence door beside it has worked this way since the guild panel gained its arrival clock; this is the raid door being given the same guard. It closes on a passed deadline and never on a travel estimate — an estimate can be pessimistic, and a button that refused you on one would talk you out of raids you could actually reach.

  Reported by Elis, who named the four things the button was checking, and the one it was not, before anyone opened the file.

• The raid headline no longer reads "impact in impact now". The largest line on a raid card counts down to the strike; once the strike landed it kept the "impact in" prefix in front of the words that replace the clock, so the biggest text on the row was a sentence in no language. It now reads "landed · this raid has landed", the same way the card for an attack on a guildmate has always read once the bombs fall.

1.158.11 — 2026-09-10

Fixed

• The Speed up sheet now says when it has moved to a different ship order (#3579). Queue two batches of the same ship at one colony, open Speed up on the first, and let it finish while the panel is still open: the sheet used to stay open showing the same name, the same picture and four live Use buttons — but pointing at the other batch, with the other batch's timer. Nothing said so, and with identical batches the surviving countdown reads as though the shard had done nothing. The next tap could spend a shard on an order you never chose.

  Now the sheet tells you: "Your shard was spent. This is a different Freighter ×8 from the one you opened — the queue moved." The four Use buttons wait until you tap the order you want, so a shard cannot follow the queue on its own, and "Ask the guild" waits with them — a guild push burns that order's help pool and no re-pick takes it back. Where the panel has not spent anything of its own it says only what it saw: "The queue moved — this is a different Freighter ×8, not the one you opened."

  Which order a shard cuts has not changed, and neither has what a shard costs or how much time it removes. This is only about being told, and being asked once, before the next spend.

  Cases still not covered, said plainly rather than left to look complete: if the queue moves you onto an order finishing sooner than the one you opened, the sheet cannot prove the swap and stays quiet — and after a shard big enough to finish an order outright, that quiet covers any batch finishing before the one you had opened. Two batches finishing at the very same instant also cannot be told apart, though there they show the same name and the same countdown, so there is nothing to misread.

  Reported by Elis, who also ran the single-batch control that isolated it.

Also

• The build and research queues are deliberately untouched: their orders cannot collide this way, because the level being built is part of what identifies them.

1.158.10 — 2026-09-10

Fixed

• A war tribute no longer vanishes with the fighter who left (#3516). When your guild won a war, the share of anyone who had left — or been kicked — between their last battle and the Monday verdict was credited to nobody. It was not paid to them and it was not given to the guild-mates who stayed: it was destroyed, while the losing guild still handed over its full 8%. Every member who stayed was quietly paid less, in proportion to whatever the leaver had earned, and there was no limit on it — when every contributor had left, the entire pot left the game and the losers paid in full anyway.

  The tribute is now split among exactly the members on the roster at the verdict, in proportion to their war points. A fighter who left or was expelled before the verdict no longer takes their portion out of existence; it goes to the guild-mates who stayed. A guild is paid the same whether its best fighter was courted away on the Sunday or never existed at all. If nobody on the roster scored — every contributor left, or the war was won by forfeit — the pot is now split equally among the members who remain, where before nothing was paid at all.

  Rounding still trims at most one unit per paid member per resource, as it always has (#3427).

• The War Room now shows the shares that will actually be paid (#3516). The war-effort panel promised its percentages were the tribute quotas, and it drew them with the same flaw as the payout, so they added up to less than 100% whenever a contributor had left. They now add up. Two older mismatches close with them: a member holding no mother colony is shown 0% rather than a quota the verdict would never have honoured, and a war won by forfeit shows the equal split that is really paid instead of zeros.

• Leaving and kicking now say what they cost. The leave dialog adds: "If your guild is at war, your share of the tribute stays with the guild." The kick dialog now names it too — a kicked member loses their accumulated contribution including their share of any war tribute.

Also

• Nobody was affected retroactively: no war was in progress when this was decided or when it shipped, so the rule applies from the next declaration onward. Tributes already paid are unchanged.
• The League panel is deliberately untouched by the new rule: a Confronto has no tribute, so its effort shares are still taken over every member of the guild.

1.158.9 — 2026-09-10

Fixed

• The Praxis Interdictor flies prow-first (#3661). Since it got its 3D model on 26/08 the hull flew engine block first, with its cockpit and the two long gun spars trailing, in the Shipyard card, the map and every battle replay. The prow now leads, and the card art, the grid thumbnail and the replay billboard were turned with it so the card no longer swaps ends as the model fades in.

Changed

• The Praxis Interdictor gets the Annihilator's model fidelity (#3661). Its model is rebuilt from the raw export undecimated, with 4096 textures, so panel edges and the gun spars read cleanly instead of the crumpled decimated silhouette. The download grows from 1.2 MB to 12.5 MB; on a slow connection a replay that cannot fetch it within twelve seconds draws the flat illustration instead, as it always did.

Also

• Ship models now have a byte-size band in the test suite: the fleet stays under 2 MB and the two high-fidelity hulls are a named list, so a heavy export cannot ship unnoticed.

1.158.8 — 2026-09-10

Fixed

• You can collect the wreckage over your own colony again, even when the game is having trouble loading your empire (#3656). If your colony list failed to load, the launch panel refused a debris harvest over your own planet and told you it belonged to another player. Neither part was true: the server allows that launch on purpose, and it is the ordinary move — being raided at home is where the wreckage comes from. The panel now refuses only when it can actually see that the planet is someone else's, and otherwise lets the fleet go.
• The launch panel no longer tells you a colony is not yours when it has not checked (#3656). In the same situation, "That colony is not yours: the cargo would fly there and come straight back" appeared on every transport, your own colonies included, and joint defence said the same thing in its own words. Both warnings are still there when they are true — the cargo really does bounce — but when the check could not be made, they now say so instead of naming an owner.
• The shield warning admits when it could not read your shields (#3656). If your empire settings failed to load, an attack could be launched without the warning that it spends a purchased empire shield. It now says the shield could not be checked rather than staying silent.

1.158.7 — 2026-09-10

Changed

• Silent accounts are removed sooner (#3654). A guest account that has not been seen for 20 days is now removed; an account signed in with Google or Apple after 30 days. The old waits were 30 and 60. Nothing else about the lifecycle moves: an empire still turns dormant after a week of silence and starts to decay after two, a declared leave still protects it for as long as the leave runs, and a banned account is never swept. A removed guest's name is freed, so anyone coming back can start again under it.

1.158.6 — 2026-09-10

Fixed

• When your colony list fails to load, the launch door stops quietly dropping two safeguards (#3581). The game reads the list of your colonies on a loop, and when that read failed the composer treated it as "you own nothing" rather than "we don't know yet". Two things silently stopped happening. The warning that an attack spends your shields — every colony shield, the 72h exodus shelter included — simply disappeared, so a shielded player could launch and forfeit protection they were never told they had; that forfeit cannot be undone and the warning was its only notice. And the block on attacking your own colony stopped firing, so the panel raised no objection to a strike on your own planet. Now: when the list has not loaded, the shield line says so plainly — that it cannot tell whether a shield is live, and what it costs if one is — instead of vanishing. It never claims your shields will fall when it has not checked. The self-attack block, meanwhile, now reads the Galaxy scan of the target instead, so it fires correctly where it used to stay silent; where nothing can answer, the panel makes no claim about whose planet it is and the server refuses the launch as it always has. A list that is merely out of date is still an answer and behaves exactly as before.

1.158.5 — 2026-09-10

Fixed

• Base on a phone: the guild and defence invitations can be put away (#3650). After the tutorial, the «Find your guild» and «Shore up your defences» cards floated over the colony with no control at all, and together they covered the map from the HUD down to the convoy line. Both now sit inside one fold: a small Hide tab on the top edge of the first card folds them into a one-line strip that names what is behind it and reopens on a tap. The choice is remembered on that device. The convoy card stays out of the fold, so a delivery waiting is never hidden. The strip names only the cards that are actually there: a player already in a guild never reads «Find your guild» on it.

Also

• The ship machinery's changelog-page self-test no longer fails on a Mac for a check that only GNU find can answer; it reports that one check as skipped, with the reason, and runs it unchanged on the Linux page host (#3651).

1.158.4 — 2026-09-10

Changed

Nothing in the game changed with this release, and nothing about it is visible while playing. It changes how a release reaches you at all.

Every change to Novarchy is supposed to pass an automated check before it can be merged. For eight releases in a row that check never ran, and nothing noticed: the merge step read no verdict, and the status shown next to a change said "success" on the strength of an unrelated service that had itself reported doing no work. A green tick was displayed for changes nothing had checked.

Merging now reads the verdict first, on the exact change being merged, and refuses when that verdict is missing, failing, still running, or was earned by a different change. A missing answer is treated as a refusal rather than as permission.

Also

The release process now records what to do when the main branch is already broken through no fault of the change in hand: repair it separately, land that first, and rebase — rather than merging on top and leaving the next person a worse starting point.

One limit is written down honestly rather than papered over: this check lives in the repository, so it protects the tooling that has picked it up, and a server-side rule that would cover everything at once is not available on the project's current plan.

1.158.3 — 2026-09-10

Fixed

• Overkill no longer vanishes into a dead target (#3603). Before: when your fleet fired at a target that was wiped in the first salvo, everything beyond its hull was lost — in the 09/09 raid on 1:7:2 a 5.18 M-hull fleet spent two whole rounds on one Fortress and two Postazioni (5,900 HP in all) while the defender fired at full strength. Now: what a wiped target could not absorb flows to the ships beside it in the same row, in the same round; an emptied row still hands its weight forward next round, so a screen still buys time. Battle reports show the hull a target actually had, plus the excess as overkill and any re-aimed salvo as spilled from.
• The per-salvo damage in reports no longer prints figures larger than the target's hull.

Changed

• Season Boss board re-based (version 3). Spillover changes how much wall a given fleet destroys and in how many rounds, so scores struck before this release are not comparable with scores struck after; the board retires the old rows the way it did for version 2, and your next strike puts you back on it.
• Campaign levels are 3–4 rungs harder for single-hull rosters (the roster used to survive on the tower's wasted overkill): tank-less wall L107→L103, Bastion pair L114→L113, maxed L120/123→L117/121.
• Large fleets against static-heavy targets are noticeably stronger — that is the fix.
• Endgame plateau re-derived to 60 days of the level-30 reference empire (107.3 G / 66.0 G / 22.3 G, same mix) after the level-30 progressions raised production 27%; three legs flatten one level later.

Also

• Un-updated app builds forecast with the old engine until the TestFlight build from this release lands.

1.158.2 — 2026-09-10

Fixed

• The warehouse cap is back on every resource tile (#3576). Ferrolite, Xenogas and Data each print stock / capacity on the face of the tile again, beside the fill bar — the bar tells you how full you are, the pair tells you how much room is left, which is the number you need before a haul lands. It went missing when the HUD band was compressed in v1.139.3.
• The figures across the band line up again (#3576). The three resource numbers sat six pixels lower than the ones beside them.
• Your Power no longer paints over the NOVARCHY wordmark (#3578). On narrower windows the figure overflowed its own box and landed on the mark next to it.
• A store held above its cap no longer turns its own figure the wrong colour. At the cap the stock now reads red on desktop, as it always has on phones.

Changed

• Power and Dark Matter moved out of the resource strip and up beside your name. Power was printed twice on the same band — once under your commander name and once as a tile — and Dark Matter is empire-wide, so neither belonged in a row of per-colony warehouses. Your bounty badge now sits next to your name, where it reads as being about you, with the fairness Gate figure and your Dark Matter balance just under it.
• The band is narrower than before, despite the tiles being wider and carrying more: the resource strip went from 665 to 542 pixels. On a 1440 laptop it stays on one line with room to spare.

Also

Nothing moved into a hover on the way: the Gate figure, the bounty badge and the Dark Matter production estimate are all still printed on screen. The exact, unabbreviated figures stay in the tooltip on desktop and in the tap sheet on phones, as before. The phone HUD is unchanged by this release.

1.158.1 — 2026-09-10

Fixed

The endgame research plateau had quietly become 25% cheaper in real terms. The Zero-Point Reactor upgrade raised what a maxed empire produces by about 27%, but the ceiling on what the four uncapped research lines cost stayed where it was — so a plateau level that was designed to take 60 days of full production had drifted to 47. The ceiling is now re-derived from what the game actually produces, which puts the pace back at 60 days on all three resources. Nominal prices go up by the same 27% that production went up, so the time to afford a level is unchanged: what was restored is the pace, not the price.

Two consequences of that ceiling move come with it: the Annihilation and Singularity lines each flatten one level later than before, and the Dark Matter ramp on those two lines therefore starts one level later.

Also

Repairs the checks that had been failing on the main branch for five releases, none of which changed anything a player sees. The Terraformer bench was measuring the reactor instead of the Terraformer; the Defence Batteries re-pricing moved an internal overflow boundary that four comments still quoted the old value for; a bounty test's wall had become too cheap to exercise the brake it exists to test; a research description had been rewritten twice without the recorded copy being refreshed; and the energy census had not been told about two readings that the Zero-Point Reactor and Particle Accelerator upgrades added.

One consequence is recorded rather than settled: because Zero-Point conversion now reads surplus power, power plants outrank defensive buildings in the upgrade advisor more often than they did (the sample moved from 57 defensive suggestions to 48). That test now pins what the game does today, with the question of whether it should stand left open on #3640 rather than quietly accepted.

1.158.0 — 2026-09-10

Changed

• Colonial Outposts retain their original level-1 power draw at every later level. Each upgrade still adds two percentage points to local Ferrolite, Xenogas and Data production, reaching +60% at level 30, without adding energy demand.
• Outpost costs now grow by 40% per level from level 16. The level-30 purchase costs about 67.3 million combined resources. Levels 1–15 keep their prices; existing levels are retained, while nominal Points and Power use the revised costs with no refunds.

Fixed

• Trade Portal previews and construction advice account for the upgraded Portal's own energy consumption when estimating outgoing lane capacity. Current, next and all-level views use the projected grid; incoming capacity retains its existing exemption from destination power limits.

1.157.0 — 2026-09-10

Changed

• Stellar Forge upgrades above level 10 now increase in price by 15% per level. The level-30 purchase costs about 1.30 billion combined resources, making the late ladder more proportionate to its added production.
• The Forge retains its three-resource output, power requirements and research unlock. Levels 1–10 keep their prices. Existing buildings retain their levels; nominal Points and Power use the revised costs, with no refunds.

1.156.0 — 2026-09-10

Changed

• Defense Batteries gain 30% attack and hull per emplacement with each Site upgrade above level 15. Every level still adds two emplacements, and shields keep their existing strength.
• Later upgrades cost less: level 30 now costs about 15.4 million resources. Levels 1–15 and additional Site surcharges are unchanged. Existing levels are retained; Points, Power and reconstruction prices use the revised costs, with no refunds.
• Building previews, battle forecasts and automated empires account for each Site's level, actual energy, shared losses and Escalation. Two level-15 Sites correctly differ from one level-30 Site despite having the same number of emplacements.

1.155.0 — 2026-09-10

Changed

• Orbital Fortress upgrades now strengthen existing platforms: from level 16 through 30, each upgrade increases their attack and hull by 30%, while adding the usual extra platform. Power shortages reduce attack, and regenerating shields keep their existing capacity.
• The Fortress sheet shows surviving platforms, effective attack, hull and the next upgrade's actual power balance. Combat, raids, fleet forecasts and agents use the same completed levels; destruction and free reconstruction keep their existing rules.
• Fortress prices stay unchanged through level 10, with a gentler curve afterward. Level 30 costs about 32 million combined resources. Existing higher-level buildings have lower nominal Points, Power and paid reconstruction valuations; no levels are removed and past purchases are not refunded.
• Defensive upgrade advice now accounts for the extra power draw of Shields, Batteries and Fortresses when calculating the resulting protection.

1.154.0 — 2026-09-10

Changed

• Zero-Point Reactors now turn unused power into extra local Ferrolite, Xenogas and Data. Every completed level raises the conversion ceiling by one percentage point, up to 30%; the actual bonus depends on spare energy. Dark Matter receives no conversion bonus.
• Reactor upgrades remain useful after reaching full power. The building sheet shows current and projected conversion, and resource forecasts account for each upgrade's power demand and the Accelerator toggle.
• Prices are unchanged through level 10, with a gentler curve afterward. Level 30 now costs about 15.079 billion combined resources. Existing levels above 10 have lower nominal Points and Power because those values follow catalog prices; levels remain intact and previous purchases are not refunded.

1.153.1 — 2026-09-10

Fixed

• A failed colony list no longer tells you that you have one colony. When the game could not load your colonies, it quietly assumed you had exactly one. For the sixth of players who hold more than that, three screens then stated something false: Redeploy and Pickup were greyed out under "You have only one colony", the Colonization research row printed "1/10 colonies" as fact, and — because the invented number was too low rather than too high — the galaxy sheet left the Settle button green over a founding the server would refuse. The count is now three-state: a list you already have is counted even if it is stale, an empty list means zero, and only an unread list is unknown. An unknown is never printed as a number and never decides a colonisation limit. Redeploy and Pickup still wait, but now say "Couldn't load your colonies — retrying every 30 seconds." — the same wording the order cards use — while a first load says "Waiting for your colonies to load." Nothing that was offered before is withdrawn, and nothing that was blocked is opened. (#3536)

• Dark Matter spent on advanced building cores is recorded again. Commissioning an advanced Hypergate, Autonomous Factory or Orbital Arsenal core took the Dark Matter from your balance without writing it into the Dark Matter history. Your balance was charged correctly, but the spend was missing from the record that is supposed to account for every movement, so the history no longer added up to the balance it describes. All three purchases now write their receipt, each naming which core it paid for. Spends made before this release are not recoverable into the history; from here the record reconciles again. (#3584)

• The Laboratory upgrade card fits on the card again. A recent change gave the Laboratory a fourth line of explanation on the levels that also open a research lane — one more than the panel is built to show. The card now names the research lane it opens and drops a sentence that repeated, without numbers, what the line beside it already said with them. (#3584)

1.153.0 — 2026-09-09

New

• Particle Accelerator progression through level 30. Levels 16–30 gain compounding confinement efficiency alongside their extra level of output. A fully powered level-30 machine produces 93.99 Dark Matter per day, or 281.96 with Accretion. Levels 1–15 retain their output.

Fixed

• Accelerator upgrade and switch-on previews now include the machine's projected energy draw. The sheet separates actual production from full-power capacity and shows when an underpowered early upgrade would lower actual output. Construction advice accounts for powered, fuelable Dark Matter upgrades.

Changed

• Prices from level 16 grow by 40% per step. Level 30 costs 8.38 billion combined resources; the complete ladder costs 29.28 billion before power infrastructure. Energy requirements and doctrine multipliers retain their existing rules.
• The lower prices also reduce nominal points and Power credited to existing Accelerators above level 15. Built levels remain acquired; previous purchases are not refunded.
• Pending Dark Matter production is credited with the new curve, including unclaimed time before this update, using the completed level and power supply in each interval. Already credited balances and off intervals retain their existing accounting.

1.152.0 — 2026-09-09

New

• Hyperspace Relay progression through level 30. Every level now reduces logistics travel time from its colony by an additional percentage point, reaching 30%. This applies to transports, transfers, pickups, salvage, colonization, allied reinforcements and supply lanes, including journeys using faster Hypergate routes.

Changed

• Relay previews show the original route effect and the new logistics benefit separately. Construction advice and synthetic reinforcement plans account for the new benefit.
• Existing flights keep their schedule. Direct attacks, Warp's ten-minute minimum, cargo capacity, supply-lane quantities, construction prices and energy requirements retain their existing rules.

1.151.0 — 2026-09-09

New

• A useful Recycler upgrade at every level. Level 1 enables harvesting and local defensive auto-salvage. Every later level now adds one percentage point of wreck recovery, reaching 59% at level 30 or 99% with maximum Active Salvage research. The last upgrade previously added only about 0.05 percentage points.

Fixed

• Recovery estimates now count upgrades completed on known colonies while you were away. The mission screen uses the same updated colonies for its Harvest readiness checks.

Changed

• Building details explain both uses: defensive auto-salvage uses the attacked colony's own Recycler, while Harvest missions use the best completed Recycler in the empire. Construction advice and synthetic empires value the recoverable ships actually stationed locally.
• Construction prices remain unchanged. Recovery percentages apply to available Ferrolite and Xenogas wreckage, not the ship's full purchase price. Cargo, storage, debris decay and the exclusion of repairable legendary hulls still apply. Existing harvest missions use the new recovery curve when they arrive.

1.150.0 — 2026-09-09

New

• Thirty useful Veil levels. Every Veil level now adds three flight concealment points, reaching +90 at level 30. Levels above 10 previously added nothing to flight concealment. The building sheet, level table and colony status display this benefit separately from the colony cloak.

Changed

• Construction advice and synthetic empires consider the stealth-capable ships actually stationed at a colony when valuing Veil upgrades.
• Already airborne fleets use the new concealment curve at their recorded launch levels. Upgrading the Veil afterwards still leaves those flights unchanged.
• Construction prices and colony cloaking are unchanged. Flight concealment still requires an entirely stealth-capable fleet, weakens with fleet size and on the return leg, and remains subject to enemy Sensors and the three-cloaked-flight limit.

1.149.0 — 2026-09-09

New

• Every Orbital Cannon level now adds shield penetration and suppresses attacking ships’ shield capacity throughout the battle. At level 30 and full energy, 30% of the opening shot bypasses shields and enemy shield capacity is reduced by 15% in subsequent rounds.
• The building sheet, battle report and replay explain both effects. Construction advice and automated commanders consider the stronger defensive profile.

Changed

• Breachhead halves both new shield effects, alongside its existing reduction of the opening salvo cap. Power shortages weaken all three Cannon effects. Existing prices remain unchanged.
• Historical battle reports keep their original meaning; completed Cannon levels apply the new profile when future battles resolve.

1.148.0 — 2026-09-09

New

• Every Shielded Silo level now protects more of your colony’s resources, from 20% at level 1 to 43.7% at level 30. Each upgrade adds at least 0.3 percentage points of protection, with existing prices unchanged.
• Building details show the protected share and the next upgrade’s benefit through level 30. This reserve is excluded before the raider’s plunder rate and cargo limit are applied; raiding remains valuable.

Fixed

• Construction advice values the resources you will still hold after paying for a Silo upgrade, and excludes upgrades that would finish after a known incoming attack.
• Automated commanders evaluate Silo upgrades by their additional protection and price instead of stopping at a fixed early level. Empty or depleted stockpiles no longer trigger a protection spending bonus.

1.147.0 — 2026-09-09

New

• Every completed Laboratory level now strengthens Scientific collaboration across the empire, including secondary colonies and second Sites. Each upgrade shortens newly started research even when the weighted Laboratory network stays unchanged.
• Collaboration removes 6% of research time at 30 total Laboratory levels, 15% at 120 and 25% at 600. The bonus has diminishing returns and works on both Disciplines and Tree research, including projects at the base four-hour duration limit.
• The Laboratory sheet and research bench show collaboration separately from the network. Upgrade prices, network requirements and research-lane thresholds remain unchanged.

Fixed

• The construction advisor now values a single Laboratory upgrade's own benefit and price. It no longer requires an artificial pair of levels or dismisses a secondary Laboratory because the network is capped.
• Second-Site progression keeps the first Laboratory's contribution intact. Research previews include collaboration once, consistently with the server and automated commanders.
• Active research keeps its recorded deadline. New projects and orders promoted from the buffer use the completed Laboratories available when they start.

1.146.0 — 2026-09-09

New

• Nanoforge levels 21–30 now keep improving heavy-hull assembly. The 40% build-time reduction at level 20 rises by 1.5 percentage points per level, reaching 55% at level 30. Going from 20 to 30 removes another quarter of the remaining heavy build time.
• Every level continues to improve lighter hulls according to their mass. The building sheet shows precise current and next bonuses across the full progression.
• Advanced assembly is a substantial colony investment: levels 1–20 keep their prices, while each subsequent upgrade grows by a factor of 2.2. Level 30 costs 24.08 billion resources and the full ladder costs 44.16 billion per colony.

Fixed

• Nanoforge upgrades accelerate remaining military and civilian ship work while retaining completed progress and paid Chrono seconds. The simulator now also follows Nanoforge and Shipyard upgrades correctly.
• The construction advisor recognizes useful Nanoforge upgrades above level 20. Orbital Arsenal remains an independent bonus for military hulls.
• Existing Nanoforges above level 20 gain nominal structure points and structural Power under the new cost ladder, without any retroactive resource charge. Existing running orders retain their recorded schedule until a normal upgrade rescaling event; newly started orders use the improved bonus immediately.

1.145.0 — 2026-09-09

New

• Every Deep Observatory level now cuts Spy and Explore travel from its colony by 1 percentage point, reaching 30% at level 30. Reconnaissance keeps improving after attack details are fully unlocked.
• Reconnaissance routing works with Relay and Hypergate routes. Warp uses the improved travel baseline while retaining its 10-minute minimum and existing distance-based Dark Matter price.
• The building sheet shows every level's travel benefit. Mission previews and indicative exploration estimates use the departure colony's Observatory.

Fixed

• Deep Observatory upgrades now retain a practical benefit even when counter-espionage chance has reached its ceiling. Ordinary upgrade prices and existing intelligence effects are preserved.
• Flights already underway keep their recorded schedule. The new travel bonus applies when launching Spy or Explore missions.

1.144.0 — 2026-09-09

New

Every Terraformer level develops your planet's specialty (#3590)

Terraformer levels 1–30 now add 1% native output per level: Ferrolite on Rocky worlds, Xenogas on Gaseous worlds, Data on Ice worlds and energy on Lava worlds. Level 30 provides a 30% increase to that specialty, on normal and rich planets alike. The other resource types receive no specialization bonus.

The existing reduction of planet maluses remains, reaching full relief at level 10. Specialization keeps every level useful even beside a Colonial Outpost and after the malus is gone. The building sheet shows relief and specialization separately, including the next level's gain. Automated commanders and production estimates use the same planetary effects.

Changed

Late Terraformer upgrades become a substantial local investment (#3590)

Levels 1–15 keep their prices. From level 16, each level costs 1.9 times the previous one. The full ladder costs approximately 55.43 billion resources per colony, with level 30 costing 26.25 billion. No Dark Matter or new commissioning step is required. Existing Terraformers receive their new production benefit, and their nominal structure points and Power reflect the updated cost curve; build orders already paid retain their recorded charges.

1.143.0 — 2026-09-09

New

Orbital Arsenal progression through level 30 (#3586)

Every effective Arsenal level after the first now cuts local military ship construction time by one percentage point: 14% at level 15 and 29% at level 30, subject to the existing minimum build time. The first level still grants an extra parallel ship order, with the same three-order cap. Aegis Bastion shield tenders qualify; civilian ships and legendary repairs do not.

Levels 16–30 require sequential Arsenal cores. The first costs 5 billion resources and 150 Dark Matter, with increasing prices for later cores. The purchase panel includes ordinary construction costs when another building level is needed. Paid cores remain available after damage and rebuilding. Completing the full level 1–30 progression costs approximately 335 billion resources and 4,770 Dark Matter per colony.

Arsenal improvements also accelerate the unfinished portion of existing military orders. Completed work, Chrono Shards and previously applied temporary bonuses are preserved. Shipyard estimates and building previews show the effective local bonus, and automated commanders can invest in cores for busy military production colonies while retaining their reserves.

Fixed

Useful upgrades and consistent production estimates (#3586)

The Arsenal no longer displays the obsolete warning that upgrades above level 1 have no benefit. Later Shipyard and Nanoforge upgrades retain its military production bonus. Civilian orders are unaffected, and repeated core requests cannot charge the same core twice. The Economy activity filter now includes both Arsenal and Autonomous Factory core purchases.

1.142.0 — 2026-09-09

New

• The guild threat board shows who is already flying to a colony, not just who has landed (#3524). The green Defending box used to report a relief only once its fleet was on the ground. For the whole flight — the better part of an hour to a distant colony — two guildmates racing to the same planet each read "Nobody is standing guard yet", and both launched. The box now carries an "On the way" line under the watch, naming each relief, its hulls and when it lands, ticking down with the impact clock beside it. Fleets in the air are listed apart from the ones on station and are not added to the "Defending N ships" count: they are not on the ground yet, and unlike a landed garrison their owner can still turn them round.

Fixed

• Empires no longer send a second fleet to a colony they are already flying to (#3524). Because a relief in flight was invisible, an empire could commit hulls to the same planet over and over while its first fleet was still on its way, and each rescue takes most of what the last one left in the hangar. A fleet already in the air now counts as that empire's one commitment to that place, whatever it will arrive in time for.
• Help that arrives after the bombs no longer reads as help (#3524). A relief that lands before the strike counts as real, if provisional, cover — worth half a garrison already standing, because it can still be recalled and roughly one in ten is. One that lands after the strike counts for nothing at all. A colony can still take three defences; it just no longer takes them blind. Deciding to fly a defence you already know will be late is untouched, and still costs what it cost before.
• A recalled relief is never reported as on its way. A fleet that has turned round will not arrive, and standing down on the strength of one going home would be worse than knowing nothing.

Also

• Measured read-only against the live game before the change: of the 7,324 relief launches made while a strike was actually pending, 4,994 went out with another relief already in the air at that same colony.

1.141.0 — 2026-09-09

New

Autonomous Factory progression through level 30 (#3569)

Autonomous Factories now make new ships cheaper on their colony. Level 1 still opens the second shipyard slot; every effective level after it adds 0.5 percentage points of resource savings, reaching 7% at level 15 and 14.5% at level 30.

Levels 16–30 require industrial cores. The first core costs about 2 billion resources and 100 Dark Matter, with prices increasing at every stage. The purchase includes ordinary construction when needed, and the new saving begins when that construction finishes. Paid cores remain available for rebuilding after damage.

The Base and Colony views show the full commissioning quote. Shipyard prices, Max and repeat orders use the local Factory saving, and agent empires understand the new progression. Already queued ships and legendary repairs keep their existing costs; ship build times and queue slots are unchanged.

Fixed

Factory upgrades beyond level 1 no longer carry the obsolete warning that they provide no benefit. Repeated or simultaneous requests for the same industrial core cannot charge twice.

1.140.0 — 2026-09-09

New

Advanced Hypergate cores (#3566)

Hypergates can now commission advanced cores for effective levels 16–30. Each core costs substantial resources and Dark Matter, and each advanced level shortens the remaining network hop by 6% when both endpoints support it. The weaker endpoint sets the benefit.

The first core costs about 20 billion resources and 200 Dark Matter per gate; prices increase at every level. Physical construction, when needed, is charged separately within the same purchase and the new benefit starts when it finishes.

Existing high-level Hypergates retain their current performance and can activate their advanced cores one at a time. The Base and Colony views show the full quote before purchase, while fleet previews and agent empires understand the new cores.

Fixed

Repeated or simultaneous core requests cannot charge twice. Previously paid physical levels can be rebuilt without paying their core fee again.

1.139.3 — 2026-09-09

Changed

The top band is a third of the height it was (#3558). It used to be 380px of sticky strip on every screen of every session: your name on one line, your resources on the next, then the night window, the leave picker, the notification row, and only then the tabs. It is now 118px on a 1440px screen or wider — one line for who you are, what you have and what you can reach, one line for the tabs. Everything else is the map.

Each resource is a cell you read at a glance. Where the word FERROLITE stood in small capitals there is now a painted emblem and its two-letter code, and under the figure there is a bar: how close a warehouse is to full stopped being a subtraction you do in your head and became a length you see. almost full, capped and full at surge end still say which of the three states you are in, in the same gold and red as before. The exact stock, the ceiling and the full name are in the hover, where the exact figures already lived.

Power and Dark Matter take the same shape, and Power still prints the Gate — the number that reconciles this band with your leaderboard row. Its Structural and Military split moved into the hover beside the exact total.

The settings moved behind a gear, at the right end of the band: the night-protection window, the leave picker with its guild-leader warning, notifications, Discord, the changelog, music and audio, and log out. They are the settings you touch a handful of times per account, and they were charging you 112px of map for it on every screen. Your phone has kept them behind one sheet for months.

On a screen narrower than about 1400px the resource row takes a second line rather than squeezing anything — the same behaviour it has always had.

Fixed

A muted game with no way to unmute it from the band. The audio panel, in its first form inside the new menu, was clipped by the menu around it — about 110px off its left side, with "Mute everything" below the cut on a desktop. The sliders, the track selector and the mute button are now part of the panel itself.

A settings panel that could put its own bottom out of reach. On a short window a guild leader could not scroll down to the log out button, because the panel moves with the band and there was nothing to scroll. It now scrolls.

Pressing Escape no longer loses your place. Closing the panel hands focus back to the gear instead of dropping it at the top of the page.

An empty panel that looked broken. While your settings were still loading — or had failed to load — the menu showed the words NIGHT PROTECTION and LEAVE with nothing underneath. Those sections now wait until they have something to say.

1.139.2 — 2026-09-09

Fixed

• The Shipyard's two Power figures now count your ship mastery (#3529). "Power per unit" and "Power when delivered" already counted your research and your refit, but not the mastery you had earned on that hull — while the empire score they exist to predict counts it. A mastered hull was quoted below what building it would actually do for your Power. Both figures now include it.

• The War Table's totals stop understating the fleet you are about to send (#3529). The footer applied your research and your launch stance and called the result your effective attack and hull. The battle it forecasts also applies your refit and your ship mastery — both of them recorded into the mission the moment you launch — so a refitted fleet read low on the one screen where a strike is decided. The totals are now built hull by hull with the same arithmetic the battle uses.

• The War Table's bonus line names what actually moved the number, and can say it went down (#3529). It read "with research & tree" while quietly counting your launch stance too, and it could only ever report increases. Standing orders often trade one magnitude for another — buying attack with part of the hull — so a fleet could show a hull total below the ships listed above it with nothing on screen accounting for the difference. The line now names the real sources — research and tree, refit, mastery, stance — listing only the ones that moved your totals, and reports a decrease as readily as a gain.

Also

• The Shipyard's stat bars still show catalogue values. They exist to compare hulls against each other — scaled to the class, gold for the best in the game — and resolving them would break that comparison. The resolved figures for your own empire live on the Hulls panel in Inventory.

• Issue #3529 asked for a third change, a resolved-strength line under those stat bars. It is deliberately not in this release: as built it claimed your refit and mastery were already applied while both were still loading, and showed a single shield figure where the game uses two (defence nodes raise your shields only when you are the one being attacked). It needs to be right about both before it goes on screen.

• Launch stance, guild war efficacy, counter bonuses and shield penetration stay out of a hull's own figures: each depends on the battle or on the opponent rather than on the hull. The War Table is the one exception, and only for the stance — there you have already chosen it, and it is on screen.

1.139.1 — 2026-09-09

Changed

• The Trade Portal's top half is reachable: level 30 costs 118M ferrolite instead of 4.47B (#3501). Before, the Portal's cost grew by 75% every level all the way to 30, so the last level alone cost more than seven years of a maxed Extractor's output and the whole ladder ten billion ferrolite — nobody past level 16 but the owner's own account. Now the curve keeps its 1.75 step through level 16 and continues at 1.35 from 17, exactly as the Command Center's does from 21: level 17 costs 2.39M instead of 3.10M, level 20 5.87M instead of 16.6M, level 25 26.3M instead of 272M, level 30 118M instead of 4.47B. Level 30 still costs 67 times level 16. Levels 1–16 are priced exactly as before. If your Trade Portal is above level 16, your Power and your leaderboard score will drop: the score re-derives what your buildings cost from the current curve, so the credit for those levels shrinks with their price — the same thing happened to Command Centers above 20 in 1.0.0-beta.223 (#1161). Nothing is refunded and nothing is written; only the number changes. This is the last step of the Trade Portal redesign (DESIGN-TRADE-PORTAL-V2.md): since v1.138.0 the levels above 16 buy supply-lane capacity, so they had to cost what a ladder costs, not a cliff.

1.139.0 — 2026-09-09

New

• Your guild log now records who left, and when (#3423). Until now the log told you when somebody joined, when somebody was promoted, and when somebody was kicked — but a member who walked out left no line at all. A guild would simply be one person smaller one morning, with nothing in its history to say who had gone or that anything had happened. Departures now appear in the log like every other event: "X left the guild." When an operator moves a player between guilds, the guild they came from says so plainly instead — "X was moved out of the guild by an operator." — rather than claiming they walked out. Guildmates in the synthetic empires read the same log, so they can now talk about a departure the way they already talk about an expulsion.

Fixed

• Leaving a guild no longer erases the fact that you were ever in it (#3423). The moment a member left, the game cleared the record of when their membership began, in the same breath as the departure itself — so neither the event nor its length survived anywhere. Support could not answer a player asking what had happened to their guild, and the game itself could not tell how long anybody had stayed. The departure is now written down, together with the rank the member held and what they had contributed, so a guild's history is a history again.

Changed

• An operator moving somebody between guilds no longer reads as a player walking out. Both guilds now say what actually happened — the one they left, the one they arrived in, and the line naming whoever inherited command if it was the leader who was moved. Three lines about one action that used to disagree with each other.

Also

• Nothing is invented for the past: departures that happened before this release were not recorded and are not reconstructed, because a made-up entry in a log is worse than a gap.
• A guild that dissolves still takes its whole history with it, departures included — unchanged behaviour, and now written down as a deliberate limit rather than an accident.
• On phones, update the app to see the new lines. An installed build carries its own copy of the screen, so until it updates a departure shows as a bare left in the guild log rather than a sentence. Nothing is lost — the entry is recorded either way, and it reads properly as soon as the app updates.

1.138.0 — 2026-09-09

New

• Supply lanes: your colonies feed each other while you sleep (#3499, #3175, #2564). Moving resources between your own colonies was the second activity in the game — over nine thousand transport and pickup flights a week, ten a day per player, by hand. Now a colony with a Trade Portal can open a supply lane to any other colony of yours: a standing order for one resource with a trigger — surplus («keep 50,000 here, send everything above it») or share («send 40% of what this colony produces») — and an optional stop after N deliveries or N units. Once an hour the Portal dispatches a batch that travels like any cargo (your Hypergates, Relay and home routes shorten the flight) and lands on the destination; nothing above a warehouse ceiling is ever lost — a lane that would overfill its destination pauses and says so, and a batch that still overflows on arrival returns to the origin. A batch in flight is not a vault: until it lands it can be looted at the colony it left, exactly like a transport's hold, and standing lanes show in enemy spy reports at the same tier as your fleet — the lane form says so before you create one. You create a lane from the Transport form («make this a standing lane»), from either end; lanes are listed on both colonies and under a «Standing» section of the Missions tab, with today's delivered totals, pause, resume and delete. Transport, pickup and Launch Again are untouched: the fleet still wins on bulk inside a system, the lane wins on the daily trickle and the long haul.
• Every level of the Trade Portal now buys something, on every colony (#3499, #1415, #2306). A level-L Portal moves per hour what a colony with its three producers at level L makes: 264/h at level 1, 6,224/h at level 10, 17,644/h at level 16, 32,291/h at level 20, 125,635/h at level 30 — one lane slot per three levels, up to ten. The destination Portal bounds what a colony can take in per hour across all its inbound lanes (a colony without a Portal receives as if it had one at level 1), so the levels above 16 on a hub colony are what let it absorb its feeders. The «nothing further to gain above level 16» warning, the dead-level confirm dialog and the mother-only signs are gone from the sheet, the build menu, the colony rows and the Advisor; the sheet prints the lane capacity per level and, on a colony with inbound lanes, «Inbound lanes deliver X/h — this Portal receives up to Y/h, next level Z/h». The market's own two curves still stop at 15 slots and the 10M ceiling at 16, and the sheet still says so. Synthetic empires do not open lanes yet; that is the next step.

1.137.2 — 2026-09-09

Fixed

• Nothing about war tribute payouts changed in this release, and that is deliberate — but the war tribute now has an answer on record for a question it never had one for (#3516). When a guild wins a war, the spoils are split between the winners in proportion to the war points each of them scored. If one of those fighters leaves the guild before the Monday verdict, their portion is not paid to them and is not shared out among the members who stayed: it disappears, while the losing guild is charged its full quota all the same. A departed top scorer can take most of the pot with them. The design document now says so plainly, alongside the three possible answers — pay it to the members who stayed, pay it to the fighter who earned it, or keep letting it lapse — and the behaviour is now pinned by tests so it cannot change quietly in either direction. Which answer the game should give is a balance decision and has not been made yet.

Also

• The same lapse was found in a second place: a winner who has no colony at all when the war is settled loses their portion the same way. Nobody is in that position today, but it is now covered by a test, so a future fix has to close both.
• The War Room's contribution panel was checked against the payout it claims to describe. For a departed contributor it already shows exactly what is paid, so it was left alone — but it is now documented that it does not match in two other cases, which is a trap for whoever changes the payout later.

1.137.1 — 2026-09-09

Fixed

The Anchorage delivers where you chose, even if that colony is destroyed while you decide (#3520)

Before, the Anchorage worked out where your crate was going at the moment you signed, not at the moment you chose. If the colony you had picked was destroyed while the charter sheet was open — razed in a raid, or lost to a Forced Exodus — the panel quietly re-pointed the delivery to your oldest colony and let you sign. A charter cannot be cancelled, and one convoy in flight blocks any replacement until it lands, so there was no way to take it back.

Now the destination is fixed the moment you see it and can only move when you tap. If the colony you chose is gone, the board and the sheet both say so by name, the Sign button closes, and the picker empties so nothing is chosen for you. Nothing is delivered anywhere until you pick a new destination yourself.

Three smaller things came out of the same work:

• The crate buttons used to read "Will not fit" whenever a delivery was blocked for any reason the board could not name — including having no destination at all. They now say what the actual wall is, so a full warehouse is never blamed for something else.
• With a convoy already in the air, the crates told you to pick a new destination — which would not have unblocked anything. They now name the convoy first, as the game everywhere else does.
• If you lost every colony and then settled a new one, the board promised it would reopen and then did not. It does now.

Also hardened on the server: a charter request that names no destination is refused outright instead of being delivered to whichever colony the database happened to return first.

One limit, stated plainly: the fix reaches the website as soon as you reload. Players on an installed iOS or Android build keep the old behaviour until the next mobile release, because the old app sends a destination that is well-formed — just the wrong one.

1.137.0 — 2026-09-09

New

• The Trade Portal works where it stands, and you choose where a purchase lands (#3498, #1360). Before, the Market had one address per empire: every offer was resolved on your mother colony — the Portal that set your slots and your per-resource ceiling was the one standing there, the stock an offer promised was that colony's, and both parties were paid there. A Portal built on an outpost did nothing at any level while still drawing its energy, the sheet and the build lists said «mother only», and a surplus mined on an outpost had to be freighted home before it could be sold. Now you publish from a colony: the offer promises that colony's stock, its per-value ceiling is that colony's Portal, and the composer names which colony it is composing from. When you accept an offer, a «deliver to» picker lists your colonies that hold a Portal plus your mother colony, each with the room it has left for what is arriving; a delivery that would land above a colony's ceiling is not refused — nothing above the ceiling is ever lost, production there simply stalls until you spend back under it, and the sheet says so before you tap. The seller is paid on the colony that published, and the seller's own rows name that colony. Two things deliberately do not multiply: your open-offer slots stay empire-wide, taken from the best Portal you own (the Portal sheet and the Advisor now say so on every other colony instead of promising a slot the upgrade does not buy), and your daily intake stays your account's, measured on your mother colony's Warehouse Sites and spent wherever a delivery lands — the board's intake row now says «shared across your colonies». Offers already on the board are untouched: they read as published from the mother, which is what they were. Installed apps that do not know the two new fields keep today's behaviour unchanged. Levels 17 to 30 of the Portal are still flagged as buying nothing on the market — that flag comes off with the supply lanes of DESIGN-TRADE-PORTAL-V2.md (added to the repo with this release), which are what those levels will carry.

1.136.3 — 2026-09-09

Fixed

• A bad changelog entry now stops its own release instead of the next one (#3534). The release script writes the changelog entry for every ship, then merges it without checking it. When an entry tripped the check that keeps changelog headings traceable to an issue, the failure surfaced on the next ship — a different piece of work, by a different session, which then had to stop and repair a changelog it had nothing to do with. Eleven times, over the life of the repo. The check now runs while the release is still being prepared, so whoever wrote the entry is the one who is asked to fix it, and nothing has been published yet when that happens.

Also

• Nothing changes for players: this is release machinery, and the changelog page at novarchy.com/changelog renders exactly as before.

1.136.2 — 2026-09-09

Fixed

• A colony list that fails to load now says so (#3507). The Launch Again rail showed the same "Waiting for your colonies to load." whether the list was still arriving or the request had failed outright — so a player whose read failed sat watching a card that never resolved, with nothing to act on. A failed read now says it failed, tells you it is being retried, and offers a reload. A list that is simply still loading is unchanged.

• A recalled fleet template is remembered the size you saved it (#3507). Recalling a saved composition trimmed it to the hangar before the composer ever saw it, so two things went wrong at once: you were told only that the template had been "adjusted", without which ships or by how much; and the trimmed fleet — not the one you saved — became the order the rail offers to launch again. Over an evening of short hangars that quietly walked a standing order down toward the smallest fleet it ever flew. Now the recall names every hull it had to cut and by how much, including the ones the hangar could not supply at all, and the order kept for next time is the one you saved. What launches is unchanged: the composer still shows, and still sends, the ships you actually have.

Also

• Saving a template still stores the composition as it stands in the form, deliberately: making Save store the larger ask would push more saved fleets into the group that already fits at no colony their owner has, which is a change that needs its own measurement rather than an argument from symmetry.
• Resolving a launch origin while the colony list is unavailable was looked at and deliberately left alone: a stored order carries only its colony id, while the hangar, coordinates, travel time and price all come from the list, so there is no way to name the origin without also being able to describe it. The rail now says the list failed instead of pretending otherwise.

1.136.1 — 2026-09-09

Fixed

• The Planetary Defense panel no longer throws away the plan you were still drawing (#3486). Setting per-hull caps and a defensive line is a long job, and until now leaving the Shipyard tab, switching colony, tapping "View unlock" or reloading discarded all of it without a word — the panel would say "unsaved" right up until the moment it forgot. Your caps and your defensive line are now kept on your own device for 24 hours and offered back when you return: the panel opens on the plan the colony is actually defending with, tells you what you left behind, and waits for you to tap Put it back or Discard it. Nothing is applied, and nothing is saved to the colony, until you say so.

  Your defence stance is deliberately not among the things kept. It is a standing order rather than half-finished work, it can have been changed since on another device, and it is hidden entirely unless you hold the Admiralty node — so putting an old one back could have quietly overwritten an order you had just given, or blocked the whole save with an error about a chip you cannot even see. The stance you are shown is always the one the colony is really holding.

Also

• A restored plan is checked against the present before it is offered: it is only ever offered on the colony it was drawn for (so razing a mother colony can no longer hand its plan to the one that replaces it), hulls that no longer exist are dropped and you are told, and caps are kept for ships that are away on a mission — those come home, and their caps still mean something.

1.136.0 — 2026-09-09

New

Hulls — the screen that says what your ships are actually worth (#3142)

You buy research. You fit standing orders on a ship class. You earn mastery on a hull. Until now every one of those was visible only on the screen where you bought it, and nowhere on the ship itself — so "how hard does my cruiser actually hit, right now?" had no answer anywhere in the game. The real number existed only inside a battle, after it was too late to plan with it.

Inventory now has a third panel beside Stores and Refit. Hulls lists every ship in the catalogue, class by class, with the catalogue figure struck through and the number your empire actually flies it at — attack, hull, shield and cruising speed — and the percentage between the two. What moved them is named once per class, and mastery shows as stars on the hull that earned them.

For an empire that has not bought anything yet the panel says so plainly and still lists every hull, because knowing your ships are at their catalogue figures is an answer too.

The numbers are not a second opinion. They come from the same function the battle's own muster uses to build a fleet, so a figure on this screen is the figure the fight will honour.

What the panel refuses to tell you, and says so

Some things move a real battle without being a property of your ship, and printing them here would be a promise the game could not keep. The panel names them instead of quietly folding them in: the stance you pick at launch, guild war efficacy, and the counter bonuses your mastery buys against particular enemy hulls.

Two limits are stated on the card itself. Speed is a fleet property — a fleet travels at its slowest ship, so the figure shown is what a fleet of only that hull would cruise at. And your shields read higher when you are the one being attacked, so that second figure is kept separate rather than blended into one number that would be wrong in both directions.

Reported by KingPaladin, who asked for exactly this — a place to see the total of a ship's upgrades — and proposed where it should live.

1.135.1 — 2026-09-09

Fixed

• A trade offer you were still pricing is no longer destroyed by leaving the Market tab (#3408). Composing an offer means deciding a ratio, and deciding a ratio means looking at what you actually hold — which is on another tab. Going to check and coming back gave you an empty form: six numbers, thought about one at a time, gone, with no warning on the way out and nothing on the way back to say it had happened. The offer is now kept in your own browser while you price it, and handed back when you return. It is offered, not reinstated: the composer comes back empty with a line naming what you left — "You left an offer half-priced: 12,000 Fe, 4,500 Xe for 8,000 Dt. Nothing has been published" — and you choose to put it back or discard it. That is deliberate. Publishing escrows the goods and puts them on a board every other empire is watching, where somebody can accept before you can cancel, so a day-old ratio must never sit one tap away from that on a screen you have only just opened. Putting it back re-reads it against the market you came back to, not the one you left: the amount you promise is trimmed to what you can still promise and the amount you ask for to what you can still be paid today, and it tells you what it had to trim, naming both the figures you had and the ones that survived. The after-fee figure is recalculated at today's rate rather than the one you first saw — a draft has no price stamped on it, because nothing has been published. On a phone the New offer button says when an unfinished offer is waiting, because the composer itself lives behind a sheet you would have had to open to find out. Closing that sheet was never a way to lose the offer, and still is not.

• A chat message you had not sent yet survives leaving the room, for an hour (#3408). The same gesture kept or destroyed a half-written message depending on which device you were on and which of three surfaces you were reading the room through: on a phone every tab switch threw it away, on a desktop it survived tab switches but died the moment you folded the chat panel away, and in a guild council it died when you clicked any other guild door — the roster, the treasury, the war room, all one click from the conversation you were replying to. Now the message is simply in the box when you come back. It never comes back in a room it was not written in. A message typed in one guild's council cannot appear in another guild's council, in the galaxy channel, or in a private thread — not even if you leave one guild and join the next without closing the chat, which is the case that used to be indistinguishable, because every council looks like the same room from the inside. Where the room cannot be identified with certainty, nothing is kept and nothing is handed back. What comes back is the words. A reply is not kept at all — its meaning lives in the message it hangs under, that anchor cannot safely be restored once the conversation has moved, and words handed back without it would arrive under nobody. The faces of anyone you had tagged are looked up fresh rather than remembered, so a commander who has since renamed or changed guild is shown as they are now.

Also

• Both drafts live only in your own browser, on that one device, for that one account. Nothing is sent to the server, nothing is reserved, and nothing is held for you: goods can still be spent elsewhere while an offer sits, and a message waiting in a box has not been sent to anybody.

1.135.0 — 2026-09-09

New

You choose what flies on a pursuit and on a standing watch (#3435, #1385). Both orders used to send the whole hangar of whichever launch point you picked — the only decision was whether to go at all. Each sheet now has a hull rail under the selected launch point: tap a class to include or leave it out, and everything below follows your choice — the hull count on the button, how much stays behind, and, on a pursuit, the solve itself. Drop a slow Freighter and a launch point that read "misses by 9m" can start meeting the target; the sheet has always told you your slowest hull sets the fleet speed, and now there is something to do about it.

The two orders open differently, on purpose. A pursuit opens with everything selected, exactly as it flew before, because you can recall it for most of its outbound run. A watch opens with nothing selected, because a watch cannot be recalled once it reaches station — it comes home when its 4 to 24 hours are up and not a minute sooner, so it is the one launch that should never be one tap from a cold start. A single ALL ARMED button, first in the rail, gives you the old whole-armed-hangar watch in one tap.

Reported by arcticpaladin (in-game QQQQ) in #1385: "Is there a plan to let intercepts/pickets select ships rather than use a planet's full fleet? That's a steep commitment to test/experiment with."

Fixed

• A watch sheet opened claiming every colony had "0 armed" and could not make the trip, because the launch-point rows had started describing the empty selection instead of the colony. The rows describe the colony again; the button describes your choice.
• A colony holding only Freighters and Probes was told to pick hulls from a rail where nothing could be picked. It now says what it always should have: nothing here can fight, and which hull to build.
• Emptying the rail said "there are no ships in any hangar" while the hulls were listed on screen two lines below. An empty choice and an empty hangar are now different sentences.
• Greyed-out hulls on a watch explained themselves only in a hover tooltip, which a phone does not have, and could not be reached by keyboard or screen reader at all. The reason is now printed under the rail and read out with the hull.

Also

• Nothing about a launch got cheaper. A pursuit takes the same taps it always did; a whole-armed-hangar watch takes one more.
• Selected and unselected hulls are told apart by a mark, not only by colour.
• Nothing is remembered between visits: a watch opened tomorrow never arrives holding a fleet you chose for a different contact today.
• Verified at 375px and 320px in a real browser.

1.134.4 — 2026-09-09

Fixed

• A guildmate defended on one colony no longer reads as covered on all of them (#3485). When a synthetic empire in your guild decided whether your colony already had enough help, it counted every garrison standing on any of your colonies, not the one under attack. So a player hit in two places at once — one colony bare, the other already holding three friendly garrisons — read as "covered", and the bare colony got nothing at all. The better the guild defended one of your colonies, the more reliably the other was abandoned. The same mistake stopped an empire from reinforcing your second colony when it was already standing watch over your first: its own ships counted against you.

  Both checks are now made for the colony the fleet would actually fly to. An ally hit in two places gets help in both, and an empire already garrisoning one of your worlds will still come to the other.

  The guild distress board was reading the same mixed count, so the line "N allied garrison(s) already there" beside a colony could be counting ships parked on a different planet, and could credit you with a garrison you had stationed somewhere else. It now describes the place it names.

  Measured on the live game before the fix: on 962 occasions a synthetic empire standing watch over one of an ally's colonies would have turned down a second colony of that same ally — and in 780 of them, four out of five, the colony it turned down had no defenders on it at all.

1.134.3 — 2026-09-09

Fixed

• The contact sheet no longer changes which colony your fleet launches from (#3497). Open a contact on the galaxy map and the LAUNCH FROM list highlights one of your colonies — the one that would meet the incoming fleet soonest. That highlight used to be recomputed every time the map polled, roughly once a minute, and two colonies whose interception times were close could trade places. If you had read the sheet, looked away and then pressed INTERCEPT, your ships could depart from a colony you never picked. The sheet now holds the launch point it opened on until you tap another one, and the rows stay in the order they were in, so nothing slides under your thumb while you are reading.

Changed

• A launch point that disappears is named instead of being replaced (#3497). If the colony your sortie was aimed at is no longer available when the map next refreshes, the sheet says so — "<colony> is no longer one of your launch points. Pick the one to fly from:" — and INTERCEPT stays closed until you choose. Previously it quietly aimed at another colony.
• The sheet tells you when a better launch point appears (#3497). Because the highlight no longer moves on its own, a line under the intercept solve now says when another colony would meet the contact meaningfully sooner, with both times, and one tap moves the sortie there. It only appears when the difference is large enough to change the time on screen.

Also

• Nothing about what an interception does, costs or reaches has changed, and the launch itself still takes the same two taps.

1.134.2 — 2026-09-09

Changed

• The war tribute's rounding is now a recorded decision instead of an accident (#3427). When a war ends, the losing guild pays 8% of each mother colony's storage, and that pot is split among the winners by how much each of them actually fought. Each winner's share was rounded down on its own, with nothing deciding who gets the leftover — so the shares added up to slightly less than the pot, and the difference went nowhere at all.

  Nothing about what anyone is paid has changed in this release, and that is deliberate. The leftover was measured against every war the game has resolved: it comes to a few dozen units out of pots in the millions — between 0.0003% and 0.005%. The same payout loses far more, on purpose, when a winner's mother colony is already near its storage cap and simply has nowhere to put the prize; on the most recent war that was about a third of the pot. Correcting the rounding while that is still true would have made the books look balanced without being so.

  What the war does with the remainder is now written down where the code makes the decision, together with the measurement and the one case worth a player's attention: a member whose share of the effort is very small can round down to nothing and be paid zero for a war they fought.

Also

• Two things this work found and did not fix, named here so they are not mistaken for settled. Checking the above turned up a second and larger gap in the same payout: the split is weighted by every contribution the war recorded, but only members still in the guild at the verdict are paid. Someone who fought and then left is still counted in the total the shares are divided by, so their portion is not redistributed to anyone — it is simply not paid. Of the players who have ever scored war points, 23 are no longer in a guild, so this is a real path and not a hypothetical one. The related question of prizes lost to a full mother colony is already open as #1484.

  Both change what players are paid, so neither is being altered quietly inside a rounding fix. They are now documented and covered by tests, which is what this release delivers — not a correction to anyone's payout.

1.134.1 — 2026-09-08

Fixed

• The mission composer no longer trims your fleet in silence (#3459). Assemble a fleet, then have ships leave while the form is still open — another mission departs, a batch is consumed — and it quietly cut your counts down to what was left, then showed you the smaller numbers as though you had chosen them. It now names what it shortened, above both Launch and Save template: Trimmed to your hangar: 100 → 50 Corvette · 20 → 0 Bomber. The line clears itself when the ships come home instead of going on blaming a hangar that is full again. What actually leaves is unchanged — still only the ships you have.
• "Repeat last order" in the Shipyard no longer promises a count it will not use (#3459). The chip announced the number you first ordered while the tap filled in the number you can afford today. Both now show the same number. The count lives only in the chip's accessible name, so this was worst for players using a screen reader.
• A Launch Again card no longer tells you a colony is gone when your colony list simply has not loaded (#3459). During that window every card could claim a launch colony was missing, or a hangar empty, about a colony that exists and a hangar nobody had read — and because the list is only filled in when its request succeeds, that state could last the whole session rather than a moment. The card now says it is waiting. Reviewing such an order also keeps its own launch pad instead of quietly moving the mission to your home colony.

Changed

• Your saved orders stop shrinking behind your back (#3459). Relaunching from the Launch Again rail already remembered the order you gave rather than the smaller fleet a short hangar allowed — but launching the same order from the composer overwrote that with the smaller number, in the same slot. Over repeated launches a standing raid could ratchet down toward the worst hangar it ever left from, with nothing on screen to say why. The composer now records the order you gave. About two in three launches repeat an order already flown, so this is the ordinary case rather than a corner of the game.

Also

• Saved fleet templates are deliberately unchanged: a template still stores the fleet you assembled, and the new line above it now tells you when that is less than you asked for.

1.134.0 — 2026-09-08

New

• Every Colonial Outpost level now pays for itself (#1422, #2306, #1942). Until today the Outpost gave the same half-malus relief at level 1 as at level 30, and every upgrade only raised its energy drain. The game warned you in gold on four surfaces, the Advisor never offered the upgrade, and players asked either to cap the building at 1 or to fold it into the Terraformer. Now the relief is exactly what it was, flat and still superseded by a taller Terraformer, and on top of it every Outpost level adds +2% to that colony's production of ferrolite, xenogas and data: +20% at level 10, +60% at level 30. The building sheet shows a new Colony production row with the next level's value beside the price, the 30-row progression table carries the column, the Advisor now prices Outpost upgrades like any other, and the gold "nothing further to gain" warning has left this building for good. The Terraformer keeps its job: it supersedes the Outpost's relief, never its ladder, so both are worth owning on the same world.

  Nothing about the price curve changed, so nobody's points or Power move. One deploy-window detail: colonies are credited lazily, so the first time a colony with a levelled Outpost is opened after this release, the backlog since its last update is credited at the new multiplier, capped by the warehouse as always.

  Reported by Remy (#1422), bhb. (#1942) and arcticpaladin / QQQQ (#2306).

Also

• The AI empires read the same ladder: their income estimate, their build heuristic and their offline simulator all carry the Outpost's multiplier, and they may now upgrade the building past level 1 like anyone else.
• The research node's description states the per-level bonus, derived from the same number the sheet uses.

1.133.9 — 2026-09-08

Fixed

• The Speed up sheet keeps the order you just sped up (#3140). Spend a chrono shard and the highlight stayed where the game wanted it, not where you put it: the sheet always aimed at your longest remaining order, so the moment your shard made that order shorter than another one, the highlight — and the Use buttons under your thumb — moved to a different building. Elis reported it from the receiving end: a 5-minute shard on her Site 2 Ferrolite Extractor left it running with three minutes to go, and the sheet quietly switched to the Xenogas Refinery. The next tap would have spent a shard on a building she never chose. Now the order you spend on becomes the order you have chosen, and it stays chosen until it finishes or until you pick another one.

• And when that order does finish, the sheet asks instead of guessing. Previously a chosen order leaving the queue handed the Use buttons to whatever was longest — the same substitution, just wearing a completion as a disguise. Now the sheet says "That order finished. Pick the next one to cut:", the Use buttons close, and one tap points them wherever you want. Nothing can be spent while the sheet is aiming at nothing.

Nothing changed about what a shard costs, what it cuts, or the weekly cap. Reported by Elis — a reward of 10 chrono shards per tier is owed for the report.

1.133.8 — 2026-09-08

Fixed

• A pirate beacon no longer looks like a planet in another colour (#3473). On the galaxy map the two were the same round mark, separated only by hue — one is somewhere to settle, the other is something that will shoot at you. The beacon now has a shape of its own: a ring around a bright core, at exactly the size and position it always had. The map key follows it, so the key still shows what the chart draws.

  This also clears up something worse that only showed at a distance. Zoomed out, a beacon sitting over one of your own colonies used to blend with your gold mark into a pale yellow that meant neither of them — on 76 systems, at the zoom the map opens at on a phone. The two now stay separate, and your gold mark stays on top where a beacon shares its star.

• Your guild's rows in the Landmarks strip now say so in words (#3446), instead of being marked only by their gold colour — the same gold as the bigger chips for relics nobody holds. The colour stays; it is just no longer the only thing saying which ones are yours.

Also

• The map key now refuses to give the same character to two different things, which is the check that would have caught the planet-and-beacon collision when it was introduced.

1.133.7 — 2026-09-08

Fixed

• A synthetic empire no longer pledges a guild defence it cannot reach (#3448). When a guildmate came under attack, an autonomous empire committed a fleet without ever working out how long its own ships would take to get there: it launched on a flat fifteen-minute rule, and when several allies were under attack at once it went to the one with the least time left — which is usually also the one furthest out of reach. Measured on production across every joint defence since 2026-07-29, 1,023 of their 7,047 rescues arrived after the strike had already landed, a median of seven minutes late, with the ships then standing a full watch on a finished battle hours from home. They now price the trip for the fleet they would actually send — the same arithmetic the guild War Room quotes above your own Send button — and go to the guildmate they can reach first, and among those to the one about to be hit.

• And when they will be late, they say so. A rescue that cannot arrive in time is no longer the automatic answer, but it is still sometimes the right one: a colony being hit is often about to be hit again, and a garrison standing on it is in that next fight. So they still fly late — and the guild-chat line now says plainly that the ships may land after the strike and will hold the position anyway, instead of announcing help that had already been overtaken. A relief that will miss the first blow but make the next one says exactly that, rather than claiming it beat the strike it missed. The estimate counts the Hyperspace Relay and nothing beyond it, so "before the strike" is a promise and "after" is only a warning.

• A garrison that lands late no longer stands the longest watch of all. The guard shift is now measured from the moment the ships actually arrive rather than from the moment they leave, so a fleet that misses the battle holds its post and goes home instead of parking for hours on a fight that is already decided.

• Two ways the new figure could still have been wrong, caught in review before release. A guildmate attacked on two colonies at once arrives as a single alert, and the relief was being timed against one colony while the ships were sent to the other — so an empire could promise to arrive half an hour early at a planet it reached an hour late, and that false rescue outranked allies it could genuinely have saved. And an attack whose countdown had run out could still summon a rescue to a battle that was already over. Both are closed.

1.133.6 — 2026-09-08

Fixed

• Dark Matter spends now leave a record, so a balance can be followed both ways (#3441). The previous release gave every Dark Matter payment a source and a receipt, but nothing recorded what left. A balance could be traced back to where it came from and never to where it went — half a history, and a half that invites more trust than it can carry. Every spend now writes its own record, named for what it bought: an Anchorage charter, a Chrono Shard, an Apex node, a warp jump, a raid warp, a research charge, a ship star, or a bounty claimed against you. Nothing changed about what anything costs, and no refusal changed: a spend you cannot afford is refused exactly as before, and a refused spend records nothing at all.

Also

• With both halves recorded, a Dark Matter balance is reconstructible going forward — the running total of a player's records now matches the balance they hold. This is not retrospective: nothing was back-filled, and spends made before this release left no record to recover. One source is still deliberately unrecorded, the per-colony Accelerator drip, because a receipt every five minutes for every earning colony would roughly double the game's whole write rate; that remains a known gap rather than a solved problem.
• These records are bookkeeping, not news: they stay out of the activity feed, and the events you already read about anomalies, factions and season endings are untouched.

1.133.5 — 2026-09-08

Changed

• The Anchorage now posts the rate it charges you (#3466). The shop has always priced its crates on its own scale — Ferrolite 1 · Xenogas 1.5 · Data 3 — and never said so. The footer now reads it out, and says which part of it actually differs: Ferrolite and Xenogas are priced exactly as the market board prices them, and only Data is dearer here, at 3 against the market's 2. If you have been buying Data charters on the assumption that the shop follows the market, it does not, and now it tells you before you sign rather than after. Nothing was repriced — the ladder charges exactly what it charged yesterday.
• The leaderboard says what its points column counts (#3466). Empire points count spend and not stock: what went into the buildings and Disciplines you have finished, the ships stationed in your colonies, and the buildings and ships under construction — with Ferrolite, Xenogas and Data each counting the same, and nothing at all for what is sitting in your warehouses. None of that was written down anywhere before. No score changed and nobody moved a rank.

Fixed

• Three wrong numbers in the Anchorage design notes (#3466). The note explaining the shop's rate justified it with production figures that had been retuned the day before the note was written, so the arithmetic printed beside the rate never described a shipped build. The same stale explanation had been copied into two more places, and the tool that recomputes the shop's economics was still running on the old figures. All corrected, and that tool now derives every number from the game itself instead of from constants typed into the script — including the warehouse table, where it had been silently ignoring the deposit technology and reporting levels that contradicted the very design table it exists to back. The rate itself is unchanged: it rests on who actually buys the crates, not on the arithmetic that had been printed next to it.

Also

• The score's own resource weighting is now documented for the first time, in the design section that lists a reason for every other weight in the formula and had never carried one for this. It is recorded as what it is — an inherited default with no decision behind it — together with a measurement of what changing it would do, and a warning that what makes it harmless today is a property of the current build catalogue rather than of the formula. Nothing about it was changed.

1.133.4 — 2026-09-08

Fixed

• A half-assembled guild raid is no longer destroyed by walking away from it (#3407). The joint-raid panel is a fleet composer, and every way of leaving it threw the fleet away: switching to another tab, opening another guild door — the roster, the treasury, the council chat, all one click away in the same rail — closing the guild sheet on a phone, or reloading. The trip that cost you the work was usually the one the panel invites: going to ask your guild what to bring.

  Now the panel keeps it and says so when you come back: "You left 27 ships assembled at Lantea. Nothing has been sent, and no target was kept." Put them back in the picker, or discard them. The ships you sized hull by hull and the battle line you arranged are held on your own device for a day, and the fleet is trimmed to what the hangar actually still holds at the moment you take it back — with a line saying so if anything was spent in the meantime.

  It is an offer, not a loaded form, and that is the point. Until you accept it the picker is empty and every launch button is exactly as dead as on a panel that never had a draft. A raid contingent cannot be recalled once it leaves, so the panel will not hand you a fleet already aboard a raid you have not looked at — and it will not hand back a target either. The three digits are quick to retype, and typing them checks the destination, the season and the trip all over again. The battle stance is not restored: those are orders for one launch, and this panel has always asked you to restate them.

  Nothing about launching got easier — one gesture was added and none removed. A fleet counted at one colony's hangar goes back to that colony or nowhere, so nothing can be sent from a planet that never had it; and if that colony is gone, the panel says so rather than showing you an empty form.

1.133.3 — 2026-09-08

Fixed

A help-board card no longer offers a push the order cannot take (#3398)

On the guild Help board, a card asking for a hand on a timer could read "takes 4 more" over an order that had already had all five of its pushes. The card counted the slots recorded when the request was first raised; the server decides from the order itself, and the two stop agreeing as soon as the queue behind the request moves — most often when a commander has two identical batches in the yard and the hand lands on the other one.

You tapped, and the game told you the truth — "That order has already had all the hands it can take (5 pushes). Nothing was spent." — but only after the tap, and only after you had picked that card over someone else's.

Now the board reads the real order before it offers anything. A card whose order is spent says what it got, stops inviting, and greys its button with the reason. A card whose order has less room than the request thought asks for the smaller number. And the commander who raised it finally hears it too: her own card used to lose its "takes 3 more" and say nothing at all, while every guildmate read an explanation — she now learns her request cannot be filled as it stands.

A refused push also refreshes the board instead of leaving the same tap available, so the same 400 cannot be collected twice.

Where the board cannot establish what an order says — an app that has not updated yet, a target that has moved — it keeps offering, on purpose: an offer that comes back refused costs you nothing, and a card that went quiet would cost a guildmate a hand you meant to give.

Empires run by the game read the same board and now read the same number, so one of them can no longer take a turn on a timer its own push cannot move, and says so on the board rather than going quiet.

1.133.2 — 2026-09-08

Fixed

A chrono shard now says what it hurried, and where (#3453). The row read «12 minutes cut off the build queue» and named nothing else, so two shards of the same size on the same queue were the same sentence twice — and with four or five colonies there was no way to tell which queue had been cut. It now reads «Ferrolite Extractor at Aurora (2:9:5): 60 minutes cut off the build queue», and carries the clickable coordinates every neighbouring row already had. A shipyard row names the size of the batch — «10 × Interceptor at Aurora (2:9:5): 30 minutes cut off the ships queue» — because two batches of the same hull sit in one shipyard and the hull name alone does not tell them apart. A research speedup names the discipline and no place at all — «Propulsion: 240 minutes cut off the research queue» — because the research queue belongs to the empire and not to any one colony, and a row that named one would be claiming something untrue.

This was the largest set of identical rows in the whole feed: measured on the live database, 205,257 of 208,939 chrono rows had a twin reading exactly the same for the same player, and 59 of the 121 players who spent a shard in the last week hit a confusable pair inside an hour.

A recalled convoy says where it turned back (#3453, closing what #3164 left open). When you call a fleet home in mid-flight, its landing row used to name the colony it came back to and nothing else, so two recalls of the same kind read alike. It now reads «transport mission called off, turned back short of 4:7:3: ships and cargo returned to Aurora (2:9:5)» — never «back from», which would claim the ships reached a place they never saw, and «called off» rather than «complete», because one sentence should not say both that a mission finished and that it never arrived. 470 recalls were flown in the last seven days, and 35 of the 62 players who flew them hit a pair of identical rows.

Also

Sixteen other row types were measured and deliberately left alone. The report that prompted this work listed eighteen suspects; each was measured against the live database rather than assumed, and only the two above turned out to be a live defect. Three of them — the lost anchorage convoy and the two contested-world refusals — have never been written once in the game's history. The diplomatic-tension row is the loudest of the rest, and the feed already folds its repetitions away. The full table, with a number against every kind, is on the pull request.

Rows already in your log are untouched and read exactly as they did: what was never recorded is left unsaid rather than filled in with a guess. Players on an older phone build keep the previous wording until the app updates.

1.133.1 — 2026-09-08

Fixed

• The map key now tells you how to open a system (#3285). A ring sits on the centre of every system, and tapping it opens that system — but the key never mentioned it. Worse, the key's only sentence about tapping pointed somewhere else: it told you to tap a planet, which opens that planet, not the system around it. The key now names both targets and what each one does, and it says the ring is not drawn while you are zoomed too far out to aim at a cell — so it never promises a mark you cannot currently see.

• The system dropdown stops borrowing the Relic's diamond (#3299). A system holding a contested world was marked ◆, the same symbol the map key gives to a Relic. Two different things, one symbol, and the key is the thing you would have checked. It now says contested in words, and the dropdown is wide enough to show it — a symbol you cannot look up is worth less than the word it decorated.

Also

• The card above a guild anchorage no longer claims the chart draws a coloured hexagon for it. It has not since the ONE MAP change, and that stale note was the last of three copies — the kind of comment that talks the next person into putting a deleted mark back.

1.133.0 — 2026-09-08

Added

• The planet card remembers your own bounced founding (#3117). When a colony ship turned back because the planet was not free, Activity explained why and no other screen did: the planet card, the mission composer's target panel and the phone's compose step all showed only "No colony on your scan" — word for word what they show for a planet nobody has ever touched — so the next colony ship went out into the same refusal. This season 314 foundings bounced and 29 commanders re-launched into the identical coordinate refusal, 192 wasted round trips between them.

  Those screens keep that sentence, because it is true — a free planet and a colony under the Veil really do read the same there — and now carry one more line where your own founding was refused at those coordinates: "Your last founding here turned back 12m ago: the planet was not free then, and nothing here says whether it still is."

  It is a record of your own attempt and nothing else: it appears only where your own ship was turned away, it is dated and in the past tense, it never claims the planet is occupied now, it never names a commander, and it does not gate the launch — the Colonize button behaves exactly as before.

  Where the founding was refused for a reason about you rather than about the planet — your colony cap, or a season that had closed — the card stays quiet, because that bounce says nothing about those coordinates. Where no reason was recorded, and where the reason is one your build cannot read, each says so in its own words rather than guessing; neither reads as permission to launch.

  Bounces older than seven days are not shown, and a coordinate you went on to colonise drops the note. Older bounces may also simply have fallen out of the activity feed the note is read from, so the absence of a note never means a slot is clear.

  Reported by elis_aigen, whose in-game name is not yet established: the credit is suspended rather than withheld, and the reward for the report is owed.

1.132.2 — 2026-09-08

Changed

The advisor now says which currency its "pays back in…" is counted in (#3160).

A Compute Node on Suggested Next could read pays back in 2h while the same cost and the same gain, priced at the rate the Market board posts, came to 5.6 hours. Both numbers were right, and nothing on either screen said which one you were reading.

They answer different questions. The advisor prices a resource by what your colony produces least of — a resource you barely make is worth more to you than one already overflowing — so it answers what is this worth to me. The Market posts a fixed list and answers what does this trade for. Checking one against the other, which is exactly what the player who reported this did, produced two true figures that contradicted each other with no way to reconcile them.

The Suggested Next panel now carries one line naming the method: these payback times are weighted by what the colony produces least of, not at market prices, so they can shift as your production does and read differently on another colony. Nothing about what anything costs, or what anything is worth, has changed — the number that was there is the number that is there.

It matters more often than the one card that was reported. Across everyone who played in the last two weeks, the advisor's figure and the same card priced at the market rate print a different label on 58% of cards, and they are more than twice apart on a third of them. On more than a quarter of colonies showing two payback cards, the two ways of counting disagree about which upgrade is the better buy — so this was never only about the wording.

Reported by Elis, whose eighth report this is, and the first where the number doubted turned out to be right for a reason no screen stated. 10 chrono shards are owed for it.

1.132.1 — 2026-09-08

Fixed

• Your own colonies now wear a mark you can pick out of the crowd (#3377). On the galaxy map a system where you hold a colony was a small gold dot — the same gold the map gives a relic, at roughly half a relic's size, and there are forty relics out there. Turning the other layers off did not help, because relics are not a layer you can turn off. Your colonies now carry the same gold crosshair your home star wears, one size down: the dot is still there, with four short gold ticks around it. It is the one shape on the chart that means "mine", so a gold diamond is a relic and never you — and it reads without relying on telling one gold from another. Your home is unchanged: still the biggest crosshair, still the only one with its name under it.

  Reported by remylz (in-game Remy), with screenshots. 20 chrono shards per tier are owed for the report.

Also

• The map key's "Yours" entry was rewritten to describe both ranks of the mark, since the key and the chart are required to agree.

1.132.0 — 2026-09-08

New

• The offer composer now tells you what you will actually be paid (#3149). Publishing a market offer, the form showed the 5% fee rate and the amount you typed, and never the two multiplied together — so the number that decides whether an offer is worth making was the one you had to work out yourself. As soon as your ask has anything in it, the composer now says "After the 5% market fee, this ask would pay you 407,550 Fe, 407,550 Dt." The fee is taken per resource and rounded in your favour, which is why the figure can be a unit higher than the 0.95 you would reach by hand.

• And it warns you when the incoming goods would overfill your warehouse (#3149). If settling the offer would put a resource above your mother colony's storage ceiling, the composer now names which one — and says what actually happens, because it is not what most people assume: nothing is destroyed. The whole delivery arrives and none of it is trimmed to fit. What stops is production: a silo above its ceiling makes nothing until you spend back under it, or raise it. The warning never blocks the offer, it names the resource so "spend it down" points at the right silo, and it tells you plainly that your stock keeps moving while the offer sits on the board.

  The ceiling it quotes deliberately ignores a running Requisition Surge — that window lasts a day while an offer can sit on the board for much longer — and when you have one running the warning says so, and says the pause only begins once the Surge ends.

Reported by elis_aigen (in-game Elis), who worked out 429,000 × 0.95 = 407,550 by hand — twice — while sizing a single offer, and asked for both halves in the same message.

1.131.1 — 2026-09-08

Fixed

• A blocked Launch Again card now shows the order you saved, not what is left of it in the hangar (#3162). A card that cannot fly used to print the fleet trimmed to your current hangar and present it as the order you had launched. A Colonise order whose Colony Ship you no longer hold appeared as a handful of Freighters — a fleet you never sent, and one that could not colonise anything. Worse, an order your hangar could not cover at all printed no fleet whatsoever, and neither did an order whose launch colony you had lost: the card named a target and a verb, and left the fleet blank.

  Those cards now open with Saved: followed by the order exactly as you stored it. Where the hangar covers part of it, a line beneath names what is absent — "Missing from the hangar: 1× Colony Ship." Where your launch colony is gone there is no hangar to read, so the card shows the order and claims nothing about availability.

  Nothing about what launches has changed. A blocked card still only opens Review, and the fleet Review hands the composer is the same one it always did. Cards that can fly are untouched: they still show the fleet the tap will actually send, and a refitted card still prints its trim before you tap.

1.131.0 — 2026-09-08

New

• The guild panel tells you when your fleet lands, before you commit it (#3152). Sending a defence to a guildmate under attack used to mean picking your ships and pressing Send with no idea whether they would arrive before the bombs — the only way to find the trip length was to leave for the Missions tab and come back to a board that had moved on. The ally-defence door now says it where the decision is made: "Your fleet: 1h13m to get there — 13m before impact." When the ships will not make it in time it warns instead, in gold, and says what that costs: "it may land 14m after impact and stand its full watch there" — because a late rescue does not turn round, it lands and holds its whole guard shift on a battle that is already over, hours away from covering your own colonies. The door that opens a raid quotes its trip too, since that flight is what fixes the impact hour the rest of the guild has to meet. Reported by Spacca-Mazza, who asked for it on both doors.

Also

• The estimate is deliberately cautious and worded that way. Your own Relay counts towards it, but the discounts the game keeps on the server — the alliance gate network, guild Logistics, the Forward Drydock, the Jump Ring, home routes — can only make the real trip shorter than the figure shown. So "before impact" is a promise, "after impact" is a warning, and the Send button is never blocked by it: a rescue the game would have delivered must never be talked out of. When the trip genuinely cannot be worked out — no launch colony resolved, no ships picked yet, the strike already landed — the panel shows nothing at all rather than a zero that would read as "there is still time".

1.130.57 — 2026-09-08

Fixed

• The Guild help board's light now checks whether you can answer it (#3144). The Help board door lit up whenever a guildmate had posted a request you had not read, without ever checking whether you still had a hand to lend. If you had already given all six of your help actions for the day — or your account was still inside its first two days, when the board is not open to you at all — the light sent you to a board where every button turned you down.

  The light now only appears when you can actually help. When you cannot, the door still tells you a guildmate asked: it reads "no help left today · a hand has been asked for" if you have spent your six, and names when the board opens to you if your account is new.

  The door's calm subtitle changed with it, so the two can no longer contradict each other. It used to read "ask & lend a hand" to everybody; it now names only the halves that are actually open to you. Out of help for the day, it reads "ask for a hand" — asking has its own separate allowance and is still yours. Inside your first two days, when the board is shut both ways, it says so instead of inviting you.

  Nothing changed about how much help you get or what it costs.

  Reported by QQQQ.

1.130.56 — 2026-09-08

Fixed

• A "Fleet returned" row now says which convoy it was (#3164). Every landing in Activity read "transport mission complete: ships and cargo returned to the departure colony" and named no colony, so several returns on the same night were the same sentence repeated and could not be matched to the convoys that produced them — while the "Transport delivered" row directly above carried coordinates. A return now names the colony it came home to, by name and coordinates, and the place it had flown to: "transport mission complete, back from 4:7:3: ships and cargo returned to Aurora (2:9:5)." The row also gains the tappable coordinates chip its neighbours already had. The storage-full version names the colony that was full in the same way.

  Landings recorded before this release have no place stored and never will, so they keep exactly the wording they have always had rather than claiming the coordinates are unknown.

  Reported by Elis.

1.130.55 — 2026-09-08

Fixed

• Empire Overview: the countdown beside a colony is now the one that actually lands next (#3145). A colony's build and shipyard orders run side by side, each on its own clock, but the Overview always showed the order you queued first — so if you started a long upgrade and then a short one, the cell counted down the long one and quietly told you the wrong thing about when that colony next does something. On the colonies running more than one order, roughly a third of the build cells were showing a time that was not the soonest, the worst of them by over two hours. The cell now heads on whatever finishes first.

• The Home screen's Colony tile had the same fault and now names the order that lands next (#3145). It shows one order and no counter, so which one it picks was the whole of what it said.

Changed

• The Overview names every order, instead of counting them (#3145). A colony running three upgrades used to show one name, one timer and a bare 2/3 — the counter told you something else was running and refused to say what, so the only way to find out was to leave the screen you were using to watch every colony at once. Each running order is now listed with its own name and its own countdown, soonest first, in both the Build and the Shipyard cell. A colony running a single order looks exactly as it did.

  This matters most on a full shipyard: at 2/2 the counter was saying both "two are building" and "there is no room for more", and never when a slip would free. Now it does.

  Reported by Elis, who manages several colonies from that screen, with the shipyard half of it raised by arcticpaladin in the same thread and mobile flagged by nooneint.

1.130.54 — 2026-09-08

Changed

• Dark Matter now keeps a ledger (#3394). Dark Matter is the premium currency, and until now a balance had no history: eleven different things could pay it, six of them wrote the number straight into the column, and nothing recorded where any of it came from. There was no answer to "where did this balance come from" — not for a player asking about their own, not for the desk looking into one. Every payment now names its reason and leaves a receipt, written in the same breath as the money, so the two can never disagree.
• An operator adjustment is now attributable (#3394). When the desk adds or removes Dark Matter from an empire, the record used to be filed against the operator rather than the player, in a separate step that could be lost if anything went wrong in between. It is now filed against the player it affected, with who made the change and why, and it is saved together with the balance or not at all.

Nothing changes about how much Dark Matter anything pays, how fast it accrues, or what it costs. No existing balance was altered, and none of this is visible in the game.

Fixed

• A guild's season and League purses are no longer at risk of failing outright if one member's account is removed at the exact moment the purse is paid.

Also

• The steady Accelerator trickle is deliberately left out of the new receipts: it is already recoverable from each colony's own record, and writing one receipt every five minutes per colony would have roughly doubled the game's whole event history for nothing.
• Dark Matter withdrawals — what you spend it on — are not in this ledger yet, only what pays it in.

1.130.53 — 2026-09-08

Fixed

• The Galactic Log podium no longer prints your score a point lower than the ledger it came from (#3298). The podium rounded down where every other place your points appear — the war-effort ledger, the League match room, the campaign coach's "War standings" line, the guild boards and the weekly war bulletin — rounds normally. The same empire could therefore be shown two different scores depending on which screen you looked at, and the podium is the screen where a rank is actually claimed. It now rounds like the rest of the game.

  On results already in your log this moves 26 season podium figures and 7 war-week ones up by exactly one point. Nothing moves down — rounding can only ever give a point back, never take one — and no placement changes: the standings were always ordered on the real score, never on the printed one.

• Your own end-of-season card moved with it (#3298). That card and the podium show the same result: for anyone who finished in the top three, it is one score printed in two places. Both rounded down before, so they agreed by accident; fixing only the podium would have made them disagree.

• A podium place worth less than a point no longer reads as zero (#3298). A score below 1 was printed as 0 pt beside a gold medal. It now reads <1 pt, the wording the war-effort ledger already used. Nobody has placed that low yet — the smallest podium score on record is 1.09 — but the weekly war scoreboard already holds scores far below one point, so it was a question of one bad season.

Also

• The end-of-season card now reads 4,812 pts this season rather than 4,812 season points, and counts properly: 1 pt, 2 pts, where before every figure was followed by a bare pt.

• Still not settled, and worth saying so: the season standings and the war-week board show one decimal (429,560.6) where the podium now shows a whole 429,561. Both round — the direction this release fixes is the same on all of them — but whether those boards should carry a decimal at all is a separate question over seven more places, and it is deliberately left open rather than swept in here.

1.130.52 — 2026-09-08

Fixed

A Redeploy no longer opens with your whole fleet already selected (#3387)

Before: opening Redeploy handed you a form with every ship in the colony's hangar already picked. One tap on Send and the entire hangar left — and a Redeploy's ships change home, so they do not come back on their own. Worse, the form re-armed itself: the moment a Redeploy launched, the picker refilled with whatever ships remained, still aimed at the colony you had just emptied.

Now: Redeploy opens with nothing selected, and tells you so — "nothing deployed yet — pick from the hangar below". You choose what leaves.

Moving everything is still one tap. The Advisor has not gone quiet: it still recommends the whole hangar, still explains why ("a redeploy exists to gather ships that a second shipyard built far from the front"), and its use it button fills the picker in a single press. The difference is that it is now a press you make, rather than an answer the form arrived holding.

This was reported by Spacca-Mazza, who had lost a fleet to it more than once. Reading the game's own records afterwards showed the old default was also simply the wrong guess: most Redeploys move a subset of a hangar, and well over half move just one class of ship.

Every other mission is unchanged — Attack, Transport, Defend, Espionage and the rest still arrive with the Advisor's suggested fleet, and a half-built mission you left and came back to still returns exactly as you left it.

1.130.51 — 2026-09-08

Fixed

A build you start with the keyboard can now be finished with it (#3378)

Opening Build and pressing Enter used to lead somewhere with no way out but backwards. The placement mode came up, said "click to confirm", and then answered no key at all except Escape — and Escape throws the build away. Tab walked straight past the ground: the colony is drawn in 3D, so no amount of tabbing ever reached a cell. A player without a mouse could start a building and then only abandon it.

The placement bar now carries a marker you can drive. Arrow keys move it a cell at a time along the colony grid, Enter (or Space) builds where it stands, and Escape still cancels exactly as before. The bar names the cell the marker is on and says whether that ground is free — out loud for screen readers, on every move, including when you run into the edge of the grid and it cannot go further.

It also opens somewhere worth confirming. The marker starts on the same free cell the game would have picked for you, so pressing Enter straight away is a sensible build and the arrows are there for when you want a different spot. Move a building and it starts where the building already stands.

This is the desktop and laptop route. On a phone the ground is aimed by touch, which already works, and the confirm button there is reachable as it always was.

Reported by Elis, who also checked that nothing had been half-built on the way out.

1.130.50 — 2026-09-08

Fixed

• Panels that slide up from the bottom no longer let the keyboard wander out behind them (#3363, #3311). Every sheet in the game — the campaign deployment preview, building and shipyard details, guild doors, the market composer, research nodes — told screen readers that the rest of the page was switched off, and then let the Tab key walk straight back onto it anyway. Behind the campaign preview the second stop was Engage level 61, so a player reading a preview with the keyboard was two keys away from committing the battle they were still thinking about. Now opening a sheet moves focus into it and announces its title, Tab and Shift+Tab stay inside it, Escape closes it, and closing it puts you back on the control you opened it from. Where one sheet opens on top of another, the keys belong to the top one.

Also

• Escape leaves a text field alone while you are typing in it: in the guild council the chat composer keeps its own use of the key, so a half-written message is never dismissed along with the panel.

1.130.49 — 2026-09-08

Fixed

• Your fleet's hold no longer loses a unit to a rounding error (#3148). At Logistics 3 a single Colony Ship showed a hold of 1,724, where 1,500 × 1.15 is 1,725 — and one Colony Ship with two Freighters showed 7,474 instead of 7,475. The cause was arithmetic, not design: the cargo bonus is built by adding percentages, and a computer cannot store 1.15 exactly, so the total landed a hair under the whole number and got rounded down past it. The tell was that the same bonus was sometimes right — one Colony Ship with three Freighters showed 10,350, which was correct, because that total happened to land exactly on a whole number.

  Holds are now rounded down only when there is a genuine fraction to lose. A hold that works out to 1,724.6 is still 1,724 — capacity you cannot fill is worse than capacity you did not know you had — but a hold that works out to exactly 1,725 is now 1,725. About one stocked hangar in nine gains a unit; no fleet gains more than one, and the total across the whole galaxy is 119 units against 33 billion already in service, so nothing about the balance of the game moves.

  The launch gate and the number on your screen were two separate copies of this calculation, agreeing only because both rounded the same way. They are now one, so the hold the Missions screen promises is the hold the server grants — the correction could not have been made to one side alone without a player being shown a load the server then refused.

  Found and reported by Elis, who did the arithmetic by hand, checked it twice, and brought along the case that came out right — which is what turned a suspicion into a diagnosis.

  ⚠️ If you play through an installed iOS or Android app, that app carries its own copy of the calculation and will keep showing the old, one-lower figure until it updates. Nothing breaks in the meantime: the server now allows slightly more than the app offers, never less.

1.130.48 — 2026-09-08

Fixed

Both speed-up panels agree about your shards the moment you spend one (#3143)

With a building upgrade and a ship order running at once, the Colony screen shows two speed-up panels. Both counted your Chrono shards, and each counted them on its own: spending from the construction panel dropped its number while the shipyard panel below went on showing the old one, for as long as you stayed on the screen — with a Use button still offering a shard you no longer had.

There is one shard balance per empire, and now there is one number for it. Spend anywhere and every panel showing shards moves at once: construction, shipyard, research, the Hangar Deck's shard line and the picker inside a slip. Buying a shard at the Anchorage Chandlery, or spending one to pull a convoy in, moves them too.

The panels also stop flickering through an empty inventory. The four counters used to redraw from zero every time you opened a tab or a speed-up sheet, because each panel fetched them again on arrival; they are read once and kept now, and a failed read leaves the last known figure on screen instead of blanking the panel and its buttons with it.

Changed

A refused shard says that nothing was spent (#3143)

If a spend is refused because the balance moved — two tabs open, or two devices — the message said only that you had no shard of that size, which contradicted the number you had just tapped. It now says that nothing was spent, and to reload if a panel is still offering one. Nothing is ever charged for a refused spend; that was already true and is now stated.

1.130.47 — 2026-09-08

Fixed

• A mission preview you leave open no longer quotes a travel time that has stopped being true (#3339). Before, the launch estimate was fetched once for a given route and fleet and then kept for as long as you did not touch the composer — so a draft left open on screen went on promising the old flight time while everything around it, resources, commander level, fleet stats, kept updating. A player reported it precisely: the same route, hull and cargo read 19m 07s in a draft that had been open across a travel upgrade, and 6m 40s in a freshly opened one at the same moment. The estimate is now re-asked on the same 30-second beat the rest of the screen already runs on, so it corrects itself in place — the figure you are reading stays put until a newer one arrives, and never blinks out and back while you are looking at it.

  This mattered in both directions, and the quieter one is the expensive one. Research, gates and guild bonuses only ever shorten a route, so the reported case was a preview being pessimistic — your fleet arrived earlier than promised. But the same holding could run the other way: leave or get kicked from a guild, lose a guildmate's Hypergate to a raze, lose the Jump Ring, a Forward Drydock or a sector, and the route gets genuinely longer while the open draft keeps quoting the short time. That is a fleet committed on an arrival that would not have held. Both are gone with the same fix.

  Nothing about how travel is calculated, what a mission costs, or when it actually arrives has changed. This was only ever about what the screen showed you before you launched.

Credit for the report, with a control experiment attached: Elis (elis_aigen).

1.130.46 — 2026-09-08

Fixed

• The Shipyard now says what its hull lists are ordered by (#3341). Each class in the Shipyard is listed cheapest first, and every row carries a coloured bar for the hull's rarity. Those are two different things, and nothing on screen said so — so the colours ran up and down the page and the list looked sorted wrong. In the Light group the Splinter Lance's violet sat above the Shroudrunner's blue, because the Lance is the cheaper hull and the deeper one. Every list that carries a rarity bar now names its ordering at the head of each run: "cheapest first, by total cost" beside the class word in the hull rail, in the All hulls sheet and in the ship reference, and "By class · cheapest first, by total cost" on the narrow filmstrip, which has no class headings to carry the other half. Sorting the All hulls sheet by Owned or Cost adds nothing, because the sort button there already says it. No hull moved, no colour changed, and no price, rarity or stat was touched. Reported by moralezz01 (in-game Tibeh01).

• Sorting All hulls by Cost now counts Data (#3341). That button added up Ferrolite and Xenogas and quietly ignored the third resource, so hulls with a big Data bill sorted as if they were cheaper than they are — the Aegis Bastion sat above the Voidlance Cutter although it costs a hundred more, with both prices printed in full on screen. It now adds up all three, which is the same "cheapest" the new class headings name.

1.130.45 — 2026-09-08

Fixed

• A market offer marked «won't count» now says, to a screen reader, which rule refused it (#3353). The trade board marks an offer the weekly trade contract will not count toward. The mark is the same two words whichever rule refused it — the trade is under the size floor, or it is priced outside the band — and the two ask for opposite moves: make the trade bigger, or change the price. Which one it was existed only as a tooltip, on an element deliberately hidden from assistive tech, so a screen reader heard one flat sentence for both cases and a phone got nothing at all. The row now speaks the reason: «under its 1,000 units minimum», or «priced outside its 1.8× band». The figures are read from the engine's own floor and band rather than written out beside them, so they cannot drift.

  Nothing moved on screen. The tag still reads «won't count», the tooltip still says the same thing to anyone with a mouse, and the full explanation was already — and still is — printed as visible text on the offer's own detail sheet, one tap away. This release only fixes what the row says when it is read aloud.

• A contract card's seal pips now speak the grade they count (#3353). Each order on the Contracts board carries four small pips showing the Charter grade the card was dealt at — which is what explains why one card asks for nine of something and another asks for three. The pips are drawn, not written, and they were hidden from assistive tech on the grounds that the Charter band above the board says the grade in words. It does not: the band names the Charter and the size of the hand it deals, never the card's number, and it describes your current Charter while the pips count the grade the card in front of you was dealt at — the two disagree for up to a day after a promotion. On a client cached from before the Charter existed, the band does not render at all. The pips now carry the grade as a spoken sentence.

  Again nothing moved: the pips are drawn exactly as before, the layout is unchanged to the pixel at both 375px and 320px, and no card, target, reward or grade rule was touched.

Also

• Both fixes come out of the running census of hover-only explanations (#3300 → #3345 → #3353). Of 217 places in the game where an explanation lives in a tooltip, six were hidden from assistive tech as well, which made them the sharpest cluster on the list. Audited one by one, only these two turned out to be real absences: of the other four, one was already answered by a visible legend the board prints under the same condition, one already had its sentence spoken beside it, and two were naming something the row had already said in words. Those four were left exactly as they are, with the reasoning recorded on #3353.

• 49 sites from that census remain, and #3353 stays open for them. The regenerated census — with the script, so the next pass can re-run it rather than trust a copy — is attached to that issue. Worth knowing for anyone reading along: the line numbers in the issue have already moved as other work landed, which is why it is re-derived every time rather than inherited.

1.130.44 — 2026-09-08

Fixed

• A mission you have started building is no longer destroyed when you leave the Missions tab (#3328). Assembling a transport used to mean not looking away: open Overview to check the destination's warehouse — the very number the mission needs — and coming back gave you an empty Attack form. The mission type, the destination, the launch pad, the ships and the cargo were all gone, with no warning on the way out and nothing on the way back to say it had happened. A player who did not look carefully could send the wrong thing.

  Now the composition is waiting for you. It survives switching tabs, reloading the page, and a phone put away mid-thought. What comes back is what you entered and nothing else: the draft is checked against today's fleet, today's warehouses and today's galaxy exactly as a freshly typed mission is, so a fleet that has since flown, a hold that no longer fits or a target that has since been colonised is refused for the same reasons and in the same words. It is never a step easier to launch than one you just typed.

  What it does not do: it is this browser on this device, so a mission started on your phone is not waiting on your desktop; it holds nothing in reserve, so ships and resources can still be spent elsewhere while it sits; it is forgotten once the mission launches, once you empty the form yourself, and after a day. A razing declaration is deliberately never carried back — that one you make on purpose, each time.

1.130.43 — 2026-09-08

Fixed

Cancelling a blocked research order no longer strands it (#3350). A research order your treasury could not yet cover could be parked in the queue and paid when it started — but only while your lanes were busy. With a lane standing free the same order was refused outright, which made the remove button on the bench a one-way door: pull a blocked order out to reorder your queue, to free a slot, by mistake, or just to read how much you were short, and you could not put it back. Not until you could already pay for it, which is the exact moment it would have started on its own. The queue position was lost for nothing.

Now a free lane with a free queue slot holds such an order unpaid, exactly as a busy lane already did. The button offers "Plan lvl N (paid on start)" instead of going dead, and the panel says why: a lane is free, the treasury cannot pay yet, and the cost is charged when the order actually starts. Tech Tree nodes, which share the same queue and the same remove button, behave the same way.

Two things deliberately did not change. An empire with no queue slot has nowhere to hold the order, so nothing is planned and the refusal is exactly what it always was. And a Dark Matter shortfall is still refused in every case: resources come back from your mines in minutes, but Dark Matter does not refill on its own, and an order waiting on it would freeze the whole queue behind it.

Reported by Aurithris, who found it from the outside and whose second sentence — that you cannot queue it again if you have an open slot — reads like a typo and was the whole finding.

Also

The trap's reach was measured before the fix rather than assumed: 46 empires were holding an unpaid pending order, and over fifteen days 3 cancels out of 402 were followed within minutes by a refused re-queue. That 3 is a floor and not a rate — the old client disabled the button in exactly that state, so the trap usually left no trace on the server at all.

Reviewing this change also turned up an older fault on the way in, now fixed with it: an order carrying a corrupted level could be queued while your lanes were busy, and it then started at less than the price of its own first level, skipping the Dark Matter part of the bill entirely.

1.130.42 — 2026-09-08

Fixed

• A blocked research order now tells you how much Dark Matter you are short, not what the whole thing costs (#3351). A queued Concord capstone waiting on Dark Matter used to say "it needs 2,000 Dark Matter you do not have" — the full price, even when you were holding 1,716 of it and needed only 284 more. The node's own sheet, one tap away, said "Still needed −284 DM" about the same order at the same moment. Two screens, one bill, two numbers, and nothing on either saying which was which. The bench now reads "you are short 284 of the 2,000 Dark Matter it costs": the amount to go and get comes first, the price stands behind it, and both are labelled. If you hold none of it the sentence says so in one figure instead of repeating it, and if a bounty has pushed your balance below zero it counts the debt in, because you really do need that much more.

Also

• The bench and the tech-tree sheet now work out that shortfall in one shared place, so the two screens cannot drift apart again and quote one bill two ways.

1.130.41 — 2026-09-08

Fixed

• The rule that decides when an abandoned account is removed is now described correctly in the code (#3389). A note inside the inactive-player lifecycle quoted "15 days for a guest and 45 for a linked account" without saying what those numbers measured. They are a limit on how much a single sweep may let an abandoned empire's garrison rot — not the point at which an account is removed, which is thirty days for a guest and sixty for a linked account, exactly as it has always been. Read as removal deadlines, they made an abandoned-account backlog look overdue when nothing was: a capacity alarm was raised on the strength of them, and the accounts it named were all still inside their own rope. Nothing about who is removed, or when, has changed.

Also

• Nothing a player sees changes in this release: no threshold, limit or brake was retuned, and no account was touched. The two figures above are now charged by tests derived from the real thresholds, so if the rope for guests or linked accounts is ever shortened or lengthened, the description reddens instead of quietly going stale (#3389).
• Measured against the live game while investigating: of 672 guest accounts eligible for the sweep, the longest-silent was 29.9 days and none had passed thirty — the removal sweep is reaching guest accounts and keeping up with them. The cradle's fill is a capacity question, recorded on #3389 for a decision, and not a fault in the sweep.

1.130.40 — 2026-09-08

Fixed

• Asking your guild for a second push on a spent order now says what actually stopped it (#3301). Your guildmates finish speeding up an order, it takes all five pushes it can ever hold, and you ask again. The board used to answer "That request is already covered: someone got there first" — so you went looking for the guildmate who beat you to it, and waited for the next opening. There was never a race and there is no next opening: that order is simply finished taking hands, for good. It now says "That order has already had all the hands it can take (5 pushes). Nothing was spent." Nothing about the rules changed — the five-push limit and everything else the board refuses are exactly as they were, and "someone got there first" still appears where it is true, for a guildmate who arrives after a request has already been covered. Reported by remylz, who also called it for research and buildings and not just the shipyard: the same limit lives on every kind of order, and the corrected message reaches all of them.

• A queue the game did not recognise no longer blames Chrono shards. The shipyard panel offers two ways to speed a queue up — spend a shard, or ask the guild — and when the queue itself came through unreadable, both buttons failed while the message named only the shard. Whoever had asked the guild read about a feature they were not using, and whoever had spent a shard was invited to try the other button, which was failing for the identical reason. It now names the queue rather than the instrument, and still says the thing worth doing: reload, because your client is out of date.

1.130.39 — 2026-09-08

Fixed

• An empty hangar now tells you where ships come from — and only the way that is open to you (#3369). Three places said there were no hulls and stopped there: the guild raid panel's launch colony, the same panel when nothing in the hangar could shoot, and the colony card's FLEET IN HARBOR. Each now names the way out, in the words the mission composer and the war table already used for the same emptiness.

  What is named depends on what you can actually do. The raid panel points you at the Launch from picker only when another of your colonies is really holding hulls — and, for the no-guns line, holding armed ones — reading each colony through the same count its own pad chip shows, so the sentence can never contradict the number just above it. Otherwise it names the Shipyard and nothing else.

  The colony card goes further, because most empty harbours are not on a colony with a shipyard. If the colony has one, it says so. If it has none but the Command Center for one, it sends you to the Base map. And if it has neither — which is where roughly three quarters of empty harbours actually sit, every freshly founded outpost among them — it names the wall you will hit first: the Command Center level a Shipyard waits for. Where a second colony of yours is holding ships, it also offers to bring them in with a Redeploy.

  Nothing about how ships are obtained has changed. This is the game saying out loud what it already knew.

1.130.38 — 2026-09-08

Fixed

Both Dark Matter screens now name the campaign rung (#3368)

If you had no Dark Matter, the game had two places that told you where to get some: the Anchorage's empty state, with four labelled cards, and a footnote on the Inventory tab. Both named the Particle Accelerator, anomalies, bounties and the League — and neither mentioned that clearing a campaign level past 30 pays Dark Matter, which it has done all along.

That was the wrong four. Over the last thirty days, campaign rungs paid out 1913 times — more often than any other route on the board — to 73 commanders, while the Accelerator, the card the panel led with and called "the steady source", reached 13. Of the 745 active commanders currently holding no Dark Matter, exactly one has an Accelerator standing. The screen for people with none was recommending the one road almost none of them could walk, and staying quiet about the one that only needs a fleet.

Now:

• The Anchorage empty state opens on Campaign rungs, and says what they pay: every level past 30, double on every tenth. Nothing was taken away — the other four cards are all still there, in order of how far they actually reach.
• The Accelerator card states its price. "The steady source" has become "Research it, build it, keep it powered: 0.75 DM a day per level, per colony", so you know what stands between you and it before you go looking — including that an Accelerator switched off, or short of energy, pays nothing at all.
• The Inventory footnote is now a complete account of where Dark Matter can come from — campaign rungs and the Accelerator, anomalies and bounties, and more rarely leagues, guild seasons, factions, the Echelon and the season podium. It was previously missing four of those, including factions, which the Dark Matter tile has been reporting to you by name all along.

Both screens now read from one list, so they cannot drift apart again.

1.130.37 — 2026-09-08

Fixed

• The chat button no longer sits on top of the panel you just opened (#3303). On desktop, the little chat launcher in the bottom-left corner floated above every slide-up panel in the game — the Shipyard's "Speed up this order" sheet, the build sheets, every one of them. Whatever the panel had in its bottom-left corner was covered: in the report that started this, the launcher ate the first words of the refusal explaining why an "Ask the guild" request had not gone through, so the line read "…quest is already covered" and the button looked broken. It was not only hiding the text — clicks aimed at that corner of the panel opened the chat instead of pressing what was underneath. Measured at every window width from 320px to 2560px, the launcher covered the same 41 pixels of the panel every time; it never cleared, at any size. The chat now passes underneath those panels, as it already did under the full-screen dialogs. Nothing moved and nothing shrank: the game occupies exactly the space it did before.

Also

• The launcher still overlaps ordinary page content in that corner on desktop windows narrower than 1666px — up to 35 pixels of whatever happens to be scrolled under it. That is deliberate and is written down next to the change: the only way to remove it is to permanently indent the whole game away from the bar above it on every desktop below 1784px, which is a cost on every screen to recover a corner you can already move by scrolling a little. The panels were the ones that mattered, because their corner is pinned and cannot be scrolled out of the way.

1.130.36 — 2026-09-08

Fixed

• The Shipyard's + and − answer the keyboard, not only a finger (#3327). The quantity steppers next to a hull's order are real buttons — Tab moves focus onto them and a screen reader announces them as buttons — but pressing Enter or Space did nothing at all. The count sat exactly where it was, while clicking the same button with a mouse moved it. Elis found it on desktop: Tab from the quantity field to More, press Space, still 5.

  They now step on Enter and Space, on every surface that shows them — the desktop count row, the phone bar, and the count sheet behind it. Holding the key ramps the step ×10 and then ×100 exactly as holding a finger does, and letting go stops it, so ordering fifty thousand corvettes by keyboard takes as few presses as it takes taps. Tabbing away mid-hold stops it too. A screen reader that activates the button in its own browse mode steps once as well.

  Nothing changes for the mouse or for touch, and no quantity rule, cap or cost moved: this was input wiring only. Arrow keys are deliberately left to the number field beside the buttons, where they already step.

1.130.35 — 2026-09-08

Fixed

• Defend and Redeploy now say, on the button, whether your fleet comes back (#3345). Over one of your own colonies the planet sheet offers both, side by side. Defend lends a fleet: it garrisons the world, the watch ends, and the ships fly home. Redeploy gives it away: the ships land in that colony's hangar and there is no return leg. Which was which lived in a tooltip, and a tooltip never opens on a phone — so on the device most of this game is played on, the choice between lending a fleet and handing it over was two verbs of movement and nothing else. Both buttons now carry the consequence under the verb: «returns home» and «one-way». It had to be on the buttons because this is where the choice is made: the mission screen's own descriptions of the two verbs are «Garrison a friendly colony» and «Move ships to another colony of yours», and neither mentions a return either. (Further in, once you have already picked, the redeploy order does say «for good».)

• The garrison you send to a guildmate says it too (#3345). The same Defend button appears over an ally's or a NAP signatory's colony, and it was still bare — a fleet lent to somebody else's world, with the return in a tooltip. It carries the same «returns home», because it is the same mission. Captioning only your own colonies would have been worse than captioning none: two buttons that explain themselves beside a third that does not invite the reading that the third one behaves differently.

• A contested system now says what its ×2 multiplies (#3345). In the war room, a colony standing in contested space was marked «×2 front». The multiplier was on screen and the thing being doubled was not — and ×2 of damage, of loot and of war points would each send your fleets somewhere different. It now reads «×2 war points · front». Both halves, because «front» was carrying a second fact of its own: a contested system is one where your own guild also holds colonies, which is the difference between a raid and a place you can be hit back.

Changed

• Three of these facts now reach screen readers, which they did not before (#3345). Measured in Chrome rather than assumed, and it turned out a tooltip fails in two different ways. On the plain ×2 marker it produced no name at all. On the Defend and Redeploy buttons it was computed and then discarded in favour of the button's own word, so a screen reader heard «Defend» and never the sentence explaining it. Because the fix is visible text rather than a hidden label, the same change answers both — and the contested marker additionally keeps its full «contested system» sentence for a reader.

Nothing about what any of these missions do, cost or are allowed to target has changed: no rule, no number, and every tooltip keeps its full sentence for anyone using a mouse.

Also

• This is part of a larger census (#3345 came out of #3300). Of 217 places in the game where an explanation exists only as a tooltip, 55 carry something a player needs in order to decide; four of them are fixed here, chosen because a wrong reading costs a fleet and cannot be undone. The remaining 51 are recorded, with per-site verdicts, in #3353.

1.130.34 — 2026-09-08

Fixed

• The guild Fund's «after their first 24 hours» refusal now tells you when the wait ends (#3316). Join a guild in your first day and Donate is greyed out: the Fund only accepts a commander's donations once the account is 24 hours old. The panel explained the rule and stopped there, so the only way to learn when the gate opened was to keep coming back and checking — and the sentence even reassured you that your reserve would still be full "when it opens" without ever saying when that was. The screen was holding the exact opening moment the whole time: it used it to grey the button out, and then declined to print it. It now reads "…after their first 24 hours — yours opens in 7h", in the same rounded-hours phrasing the rest of the game uses for a wait, and the hours count down as the wait does.

  Of the 146 players who have ever joined a guild, 88 joined inside that window, so most guild members have met this refusal. Almost nobody meets it near the end, though: the typical wait still to go at that moment is about 23 and a half hours, and only one player in 88 was within the hour of opening. The line is there to let you plan the day, not to count you down to an unlock.

  The 24-hour rule itself, and every other Fund threshold, is unchanged — this release only changes what the screen is willing to tell you.

  Reported by Elis.

1.130.33 — 2026-09-08

Fixed

Open the chrono-shard panel with an empty hold and it told you two ways to get one: win a battle against someone stronger than you, or buy a shard at the Anchorage's Shard Chandlery with Dark Matter. There are more than two, and the ones it left out are the ones most players were actually using.

The panel now names every route an empty-handed commander can take. Alongside striking upward, it says that collecting a bounty on a wanted empire pays a shard, and that five Daily Wraps open your Weekly Chest on the Season tab, with a shard inside. The Chandlery is still there, and the rule that a week's shards are capped however they arrive is still there, now worded so it covers all three.

The reason this mattered: over the thirty days to 8 September, the two routes the panel named reached 30 and 8 commanders. The two it omitted reached 28 each — and the Weekly Chest is the only road to a shard that asks for neither a battle nor a single unit of Dark Matter. A new commander with an empty hold was being pointed down the two hardest paths and told nothing about the cheapest one. 665 of 709 active empires currently hold no shard, so this is the paragraph most of the galaxy reads.

The small tooltip on each greyed-out shard said the same two-route story and now matches.

Suggested by Elis, who noticed the panel's list was short. The bastion prize stays unmentioned on purpose — it is a feat for the very top of the ladder, and it would be advice an empty-handed reader could not take (#3332).

1.130.32 — 2026-09-08

Fixed

• Nothing in the game changed, and that is the finding (#3359). When the map became one continuous field with room between the planets, the check that guards how big a world may be drawn was left behind: it still demanded that two colonised worlds sit on touching cells, which the new field never does. The check was red, so every change waiting to go out was blocked behind it. The chart itself was correct the whole time — a world still fits inside its own cell and still leaves the grid line around it visible, with more room to spare than before, and that was measured rather than assumed before anything was touched. The check now reads the spacing off the map it is actually looking at, so the next time the field moves it moves with it.

Also

• The same check had been giving two different answers on two different machines, which cost an evening of arguing about which one to believe: it measures the galaxy out of a built package that nothing in a test run rebuilds, so a machine holding an older build was quietly measuring an older galaxy. Four separate attempts to trace the fault to its commit disagreed with each other for exactly this reason. It now checks that build against the source it came from and stops with the one-line command that fixes it, instead of reporting a failure about the wrong thing (#3359, and the duplicates #3361, #3365 and #3371 it was filed under).

1.130.31 — 2026-09-08

Fixed

• The "next move" card on your Home screen now opens the colony it names (#3315). It would say «Start a build · Ferrus Prime» and then open the Base on whichever colony you happened to have selected — the wrong one about half the time, if you hold more than one. It now switches to the colony in the title before it takes you there.

• «Show me», on the "Shore up your defences" card, now shows you the building (#3315). It named the Defense Batteries and the Planetary Shield in bold, opened the Base, and left you to find them among twenty-seven rows — right after telling you, for the first time, that these two buildings exist. It now opens the Batteries' card, with the cost, what it does and the Build button, and closing it leaves you in the Buildings list rather than back on the map.

• «Open Research», after a campaign wall, now opens the discipline the coach named (#3315). The report tells you in words which research would move the wall you just hit, then opened a catalogue of ten disciplines with the first one selected. It now lands on the one it recommended — and so do the "what if" rows that rehearse a research.

• «open war room», on a Diplomacy row, now opens that war (#3315). With two wars running it opened whichever one Diplomacy happened to list first, not the rival whose row you tapped — and every row in the war room is a button that sends ships. It now opens the war the row is about.

Changed

• The war room's front picker now says which war is selected out loud (#3315). Which of your wars was open was shown by the colour of a chip and nothing else, so a screen reader — or anyone who does not separate those two reds — had no answer to "which war am I looking at" on the one screen where every row launches a fleet.

Also

• Three more buttons from the same sweep were examined and deliberately left alone (#3315): «See them» on the guild card already lands on a page whose top section is those same three guilds; «Open missions» after the last convoy has no destination to carry, because the missions board can only be aimed at a place and that card is about cargo that has already arrived; and the campaign's «Open Refit» chip never names a module, so there is none to point at.

1.130.30 — 2026-09-08

Fixed

• The Advisor's comment no longer deletes the line telling you a second purchase is needed (#3323). Some upgrade cards carry a line saying the gain being advertised only arrives if you also buy something else — a Command Center level whose value is the Extractor or Refinery it unblocks, or a secondary colony's Laboratory, where levels only count in pairs. That line leads the card, because the button beside it queues the build in one tap. But the Advisor's one-sentence comment was rendered in place of the card's first line, and the comment is on by default for exactly the three cards the panel shows — so on those cards the warning was simply not there. Measured over 3,000 generated colonies at the panel's own limit: 622 Command Center cards and 97 paired Laboratory cards, every one of them, and on 60 of them that line was the card's only line, so all you saw was the Advisor's sentence. Now the Advisor's comment takes the place of the first line it is actually allowed to rephrase, and the warning stays where you read it before tapping.

• A caveat can no longer outlive the promise it corrects (#3323). On a Command Center card that offers another colony, the line "Colonization N is needed for that extra colony" sits under the promise it qualifies. Fixing the above moved the replaced slot onto that promise, which would have left the caveat pointing at a sentence no longer on the card — so a promise and the rebuttal under it are now kept together, and the Advisor's comment is added below them instead of taking one of their places. The same pairing also fixes the case where there is no blocked upgrade at all, where the promise was already being eaten before this release.

Nothing about which upgrades are recommended, in what order, or at what price has changed: the ranking and every engine sentence are byte-for-byte what they were.

1.130.29 — 2026-09-08

Fixed

• Clearing a mission coordinate no longer snaps it to 1 (#3322). Putting the cursor after the 9 in Sector and pressing Backspace turned the box into 1 on the spot — so typing the replacement digit gave 13 instead of 3, and the fleet was aimed at an address you never typed. The three coordinate boxes now stay empty while you retype them, and the digit you type is the digit you get.

  This was worth more than a typing annoyance, because the coordinate the slip produced was usually a real place. A sector holds 19 systems and 18 planets, so a prepended 13 is an ordinary address the game accepts without complaint — and in the Sector box, 13 is the Event Sector. Whether it ever actually sent anyone's fleet to the wrong place we cannot tell from the records: the shape is there, but the galaxy is too small for any leading digit but 1 to exist, so a mistype from this bug and an ordinary slip of the finger look identical. Reported by Elis, with steps he ran twice.

  Nothing about which coordinates a mission may carry has changed. An incomplete address still refuses the launch, and still tells you which axis is wrong and what range it runs in.

• The shipyard build count had the same slip. Clearing the count to ask for 3 hulls queued 13. Found while sweeping for the coordinate bug.

• And the shipyard now says why it will not take an empty count. Clearing the box left BUILD greyed out with nothing on screen to explain it — a state that could not happen before, because the count could never be empty. It now reads "Set a count to order". The «fewer» stepper also used to raise an empty count to 1 instead of leaving it alone.

1.130.28 — 2026-09-08

Fixed

• Quantum Computers now says which research it speeds up (#3171). The node's effect line read −2.5%/lvl research time and stopped there. Research has two lenses — Disciplines and the Tech Tree — and the discount has only ever reached Tech Tree nodes. A player asked which one it covered, and another answered in the same thread that it had to be all research, "since everything is under a research tab". That is the reading the line invited, and it is the wrong one: it made the node look like a discount on the queue a mid-game empire spends most of its laboratory time on. The card now reads −2.5%/lvl research time on Tech Tree nodes — never on Disciplines, naming both lenses in the words the lens switch uses.

  Nothing about the discount itself changed — the same nodes are sped up by the same amount as before. What changed is that the card no longer implies more than it delivers.

Also

• The engine's own note on the research-time formula claimed the discount was shared by Disciplines and Tech Tree nodes alike. It never was: the two lanes call the same function and only one of them hands it the discount. Anyone who checked the code to settle the question found a sentence confirming the mistake. It now lists which side is which (#3171).
• The synthetic empires read the research screen through a recorded copy of the same response, so they had been reading the unscoped sentence too. That copy was refreshed in the same change.

1.130.27 — 2026-09-08

Changed

One field: planets spread out, sectors tiled edge to edge (#3321)

The galaxy map is now one continuous field. Before, every system was drawn as a tight flower of eighteen planets on adjacent cells and the twelve sectors sat as separate plates with empty bands between them. Now a system is a loose group: its planets are scattered within five cells of an empty centre with at least one empty cell between any two, the systems inside a sector sit twelve cells apart, and the sector plates touch along a one-cell border with no gap. Sectors 1 to 7 form the heart of the field, 8 to 12 the outer shell. Every coordinate keeps its name; only where it sits on the grid changed, so nothing you own moved anywhere but on the picture.

Travel was retuned with the geometry. A trip inside your own sector takes about the same time as before, a trip between planets of one system takes longer than it did (they are farther apart now), and a trip across the galaxy is shorter, because the empty bands you used to cross are gone.

Interception follows the same ruler: you still cannot be caught within your own neighbourhood or your target's, and that neighbourhood is now eleven cells from either end of a route, so a hop between two neighbouring systems has no interceptable middle. A standing picket may stray five cells from its station instead of three.

The sector name sits inside its plate and disappears at close zoom; the ghost numeral marks the sector centre. Tapping a fleet mark near a system no longer hands the tap to the system unless your finger is on the system's own cell.

Synthetic empires price their flights on the new cell cost; their estimates were running long against the server for a few hours on the first commit of this branch and never reached production.

Four rules calibrated on the old geometry were left as they are and recorded for a design call in #3317: the raid join lock can now refuse a same-system guildmate, sensor reach covers less of one's own system, faction sorties are short enough to thin the night window, and the deep-space names can land on planet cells.

Installed iOS and Android builds keep the old layout until their next store update.

1.130.26 — 2026-09-08

Fixed

• A blocked research order now tells you how much you are short (#3165). An order queued behind a treasury that could not pay for it said only that it could not pay — never by how much. The one way to see the number was to cancel the order, and with a lane standing free you could not put it back until you could already afford it, which is the moment it would have started anyway. The slot now names every resource you are short of — "the treasury is short 108K Fe, 40K Xe and 2.1K Dt" — and the figures come down as your mines fill. Reported by atyo416 (Aurithris), who also worked out that the only route to the number cost you the order.

Changed

• A stalled order says what your Laboratory network actually is (#3165). An order held up by the Laboratory requirement said "your network is below it" without naming the level, on the one screen that did not. It now reads "needs Laboratory lvl 3 and your network is lvl 2", the way every other research surface states it.
• The Discipline panel gained a "Still needed" row (#3165). Opening a Discipline you cannot afford showed red cost figures and a dead button reading "Not enough resources". It now lists the exact gap per resource, directly under the price, as the tech-tree panel already did — which matters most on a phone, where that panel covers the card the number used to live on. It stays out of the way when the Laboratory network, and not the treasury, is what is actually blocking you.
• The Dark Matter wall names both halves (#3165). An order short on Dark Matter and on resources named the Dark Matter figure and then said only that the treasury "cannot pay its resource cost yet either". It sizes that half too now, so you are not sent to clear one wall and meet the next.

1.130.25 — 2026-09-08

Fixed

• The Colony Advisor's Laboratory card now shows both levels it is charging you for (#3324). On a colony that is not your best research world, a single Laboratory level often moves the empire's research network by nothing at all — the secondary colonies count half, so the levels come alive in pairs. The Advisor already recommended the pair, but the speed-up it printed was measured over both levels while the price beside it, and the button under it, covered only the first. On a Laboratory 4 → 5 that meant "Research runs 6% faster" printed next to 1,482 Fe · 1,482 Xe, when reaching that 6% also needed level 6 at 2,445 Fe · 2,445 Xe — about 38% of the real cost, on every paired Laboratory card the Advisor has ever shown. The card now prints the second level's price under the first, and the sentence names the level it is measured at: "Research runs 6% faster once lvl 6 stands too."

Changed

• The Laboratory card's research-lane line no longer credits the new lane to the level you are about to buy. Where the lane only opens with the follow-up level it now reads "Reaching lvl 6 opens a second research lane", instead of "This level opens…" directly under a line saying this level moves nothing on its own (#3324).
• Your in-game Advisor is told the follow-up price too, so its own sentence about the card cannot pair the speed-up with the smaller number (#3324).

Also

• Nothing about which upgrades the Advisor recommends, or in what order, has changed: it was already judging that Laboratory on the full two-level spend. Verified by hashing every suggestion's rank, score and cost over 3,000 generated colonies — identical before and after.

1.130.24 — 2026-09-08

Fixed

• The Tech Tree now says when it opened instead of the Apex lens (#3312). Tapping View unlock on a locked Apex hull took you to a lens called Tech tree with the Particle Accelerator open — a different name from the one the padlock had just printed in bold, and nothing on the page explaining the swap. On a phone that arrival is a full-height sheet, so the only thing on screen was a card about a building you had not asked about. Two short lines now say so: one above the tree, and one at the top of the Accelerator's own card, so the explanation is there whichever of the two you are looking at. Nothing else changed — the destination was always the right one, since the padlock already tells you to research the Accelerator, and a View unlock that lands on the node it actually named still says nothing, because it has nothing to explain.

1.130.23 — 2026-09-08

Fixed

The ⚡ energy badge now names itself to a screen reader (#3300)

Beside every build and upgrade price the game prints a signed ⚡ figure — ⚡ −25. Until now that badge had no accessible name at all: the sentence explaining it lived in a hover tooltip, and a hover tooltip on a plain span is not something a screen reader announces. Players using one got neither the number nor the explanation.

Each badge now carries its own spoken label — "Energy minus 25 — Change in energy consumption/production at the next level" — on all three screens that print it: the Colony list, the Base build menu and the advisor cards. The sign is spoken as a word, because the typographic minus the badge prints is a character most readers pass over in silence.

Nothing about the number itself changed: same figure, same rounding, same colour, same tooltip for anyone using a mouse.

Also

What this did not fix, stated plainly. A hover tooltip still never opens on a touchscreen, so a sighted player on a phone still cannot read the sentence beside the badge. A visible key for the Colony list was built for this release and removed before shipping — on that screen it would have contradicted itself, because the same ⚡ glyph already means two other things there and the advisor cards above the list hold a different quantity from the rows below it. That half of the problem is open and tracked in #3340, with the reasons written down so the next attempt starts ahead of this one.

1.130.22 — 2026-09-08

Fixed

A developed system reads as a map again, not a pile of planets (#3178)

A system where most of the planets were colonised drew as a heap. The worlds were painted larger than the cells they sit in, so each one overlapped all six of its neighbours and covered its own cell's outline — in a fully settled system the hex grid vanished entirely, surviving only at the outer edge of the cluster where the colonised ring ended. Three players reported it independently, twice from a phone.

A colonised world is now drawn to fit inside its own cell, with the cell's rim clear around it and visible ground between it and every neighbour. The size is no longer a number somebody picked: it is worked out from the grid the map already draws, so the art cannot creep back over the cells. Worlds are about a quarter narrower than before and still large where it counts — at the deepest zoom a colonised planet is 50 px across on a phone, and the rings and moons that tell you how developed it is are all still legible.

Empty worlds are unchanged; they were never the problem.

Also

The map playground now contains one fully settled system. It used to cap a system at five inhabited planets, which is why a crowded flower — the only arrangement in which neighbouring worlds can overlap at all — could never be seen there.

Found by mhar02, with confirming captures from arcticpaladin and spaccamazza.

1.130.21 — 2026-09-08

Fixed

• A Transport now tells you whether the destination can actually take the cargo (#3172). Preparing a Transport, you saw the hold capacity and a live Send button and nothing at all about how much the destination could store — while switching that same draft to Redeploy showed "10,711 fits there" and a warehouse warning. Same origin, same destination, same cargo, same panel.

  Now a Transport says it in three places, which is what Redeploy already had: beside the manifest, at the launch door, and in a notice above the form. The line names how much lands, how much flies straight back undelivered, that a return arriving at a full warehouse at home loses what will not fit — and which resource is the one that will not fit, so "load less" points at the right cell instead of the one that was landing fine. Send is deliberately still live: a Warehouse can finish upgrading while the fleet flies, so this is news, not a refusal. For the same reason the notice says out loud that the destination keeps filling while your fleet is in transit, so the room quoted can be gone by arrival.

  The figure appears only when the destination is one of your own colonies. A Transport may legally be aimed at another player's colony, and their warehouse is something you buy with espionage, not something the launch form prints — so when the target is not yours, no figure is shown, and both the launch door and the notice say why. The absence of the figure never means "there is room".

  Reported by elis_aigen (in-game Elis), who found it by switching one dropdown on an otherwise identical draft.

1.130.20 — 2026-09-08

Fixed

• The advisor's "pays back in…" chip was already honest, and now it is held to it (#3294). A report noticed that the advisor works out a card's payback over a bigger sum than the price printed next to it: a Command Center is judged together with the upgrade its level releases, and a Laboratory on a second colony together with the level after it, because there the levels only come alive in pairs. Read as a pair, a price and a payback about two different sums would make those two cards look several times more efficient than they are.

  Nothing you see was wrong, and nothing you see has changed. Those are exactly the two cards that never quote a payback at all — checked across 3,000 simulated colonies and 16,000 advisor cards, where not one card showed both. So no player was ever handed that ratio.

  What changed is that the game now fails loudly if it ever starts to. The rule "a payback is always about the price beside it" is now checked on the advisor itself, on the card, and on the facts the in-game Advisor is given before it writes its comment — that last surface had no check of its own until now, and it is the one where a wrong pairing would have been hardest to spot, because the Advisor is told never to contradict the facts it is given.

Also

• Looking into the above turned up two separate things the advisor really does get wrong on those same cards, both now filed with measurements rather than folded into this fix: the in-game Advisor's own sentence replaces the card's first line, which is exactly where the "you will have to buy the other one too" warning sits (#3323); and a paired Laboratory card advertises a research speed-up that only arrives after the next level, printed beside the price of this one (#3324). Neither is fixed yet. If you have been reading that panel, those two are the ones worth knowing about.

1.130.19 — 2026-09-08

Fixed

• The season-close warning now says what the close actually takes (#3170). The red strip on Events › Season said that whatever you leave in the Event Sector is devoured. That was false, and false in the direction that costs you something: it counselled evacuating a fleet that was coming home free anyway, at the cost of travel time and of whatever standing that fleet was holding. The Full rules behind the strip said the opposite — stationed ships and half the resources come home — and the engine agreed with the hidden one, so the sentence a commander read first was the wrong one.

  The strip now states the real rule, in three lines you can read at a glance instead of one sentence you have to finish. Comes home, if you still hold a colony outside the sector: the ships stationed there, any repair running there, and half of the resources stored there. Devoured: the colony and everything built on it, anything still in the shipyard queue, and the other half.

  The dry dock and the shipyard queue are new to the warning, and they are named apart on purpose: both are work in progress and they meet opposite ends, which is exactly the choice a commander makes in the last hours of a season. The old Full-rules wording named only ships and resources, so simply making the strip agree with it would have left the same silence one term over.

  The Full rules now carry only what the warning does not: that the Event Sector lasts as long as the season, that a fleet away on a mission when the Event closes is not swept at all and lands at your mother colony when it returns, and that a commander holding no colony outside the sector has nothing to carry the salvage to — there, only legendary hulls survive, as wrecks.

• Two other screens have stopped saying the same false thing (#3170). The Events › Calendar countdown said "anything still parked there goes with it" — the sentence most likely to be read in the final hours, since the season-finale notification opens that very lane. And the notice shown inside the Event Sector on the galaxy map advised leaving nothing irreplaceable behind, which was backwards: a legendary hull is precisely what the sweep goes out of its way to save, carried home with the rest of the fleet, or credited back as a wreck when there is no home to carry it to. Both now name the Eschaton and point at the one screen that states what it takes.

  Nothing about the close itself changed. This is only what the screens say about it, and every fleet that has ever come home from the Event Sector came home under the same rule.

  Reported by elis_aigen, who noticed the panel giving two different answers one click apart.

Also

• What the close takes is now declared once, in a single place, and every sentence a player reads about it is generated from that declaration — so a screen and the engine can no longer drift apart the way they had been drifting since before the game was in English. Three test suites hold the ends together: one proves the screen says exactly what the declaration says, one runs a real season close and proves the engine does too, and one proves no other screen has quietly started stating the rule a second time.

1.130.18 — 2026-09-08

Fixed

Six screens now say "I could not read this" instead of quietly stating something false (#3284).

When a read failed, several surfaces did not go blank — they made a claim. A failed batch of your spy reports flipped the system list's footer to its finished-checking wording, so the satellite column looked complete when it was not. A failed contact feed made the board announce "Contact tracking is off: the chart draws no fleets in transit right now" — which tells you to stop watching for fleets that were merely never fetched. Both now say plainly that the read did not land, and neither claims the sky is empty.

The same treatment reached four more places: the anomaly board no longer sits on "Reading the sky…" for as long as an outage lasts; the planet sheet says so when your own Power could not be read, instead of silently dropping the strength line — and when your own report archive could not be read, instead of falling through to "No colony on your scan"; and the target board's Your Gate — now tells "still loading" apart from "the read failed".

The Missions target panel names the Pirate Beacon's Power (#3284). It used to hide the Power chip entirely, which meant two unrelated things at once: a Beacon, whose Power the game withholds on purpose because it is exempt from the power band, and a real empire whose Power simply had not arrived. Only the colour of the owner's name told them apart. A Beacon now reads PvE in the beacon orange, and the faint … is kept for a Power that genuinely could not be read — the same vocabulary the system list adopted last release. About one occupied row in sixteen on this server is a Beacon, so this is the half of the ambiguity you were seeing constantly.

Also

Three surfaces on the same list were measured and deliberately left alone, rather than fixed for the sake of it: the anchorage card (absent by design on 224 of the galaxy's 247 systems, so a warning there would be about nothing nine times in ten), the pre-season qualifier on a spy report (it applies to five reports in the entire game), and the political map's power tiers, which needed a way to be counted before anything is shown to players. Nothing changed in what any of these screens calculates, and a failed read is never drawn as an alarm — red on these screens still means a fleet.

Reported by spaccamazza, who has now found this same problem in three different places.

1.130.17 — 2026-09-08

Fixed

• The numbers that decide who can see a cloaked world now live in one place (#3291). Nothing on screen changes with this release, and that is the whole point of shipping it now rather than after something broke.

  The Veil is the game's concealment: a colony that builds one disappears from anyone whose Sensors are too weak to read it. Whether they are strong enough is a comparison between two numbers — what a Sensors level is worth, and what a Veil level costs — and the same pair also decides which cloaked hulls show up in your incoming-attack warning.

  Those numbers were written out by hand in four separate places: the shared engine, the server, the web client and the code that drives the synthetic empires. All four happened to agree, so nothing was wrong. Nothing kept them agreeing either. The next time anyone retuned concealment, they would have moved one copy and left the others: the map would have judged visibility on one scale while the building card promised another, and a player could have bought the exact Sensors level the card named and stayed blind — with every test in the project still passing, because no test had ever compared the two files.

  There is now one definition, and the rest read it. Four more hand-written copies of the same scale went with it, including one buried inside the map's own visibility check.

• And it is now enforced rather than requested. The old protection was a note in a comment asking whoever changed one copy to remember the others. That was never a check, and it never once failed loudly. In its place: re-declaring the scale is now a compile error, a test refuses any second definition anywhere in the codebase, and the synthetic empires' copy — which is written in another language and cannot import the original — is held by a test that reads the original and compares.

  The check on the visibility rule itself was also rebuilt. It had been testing three fixed numbers, which meant the concealment scale could have been doubled without it noticing; it now builds three real observers — one below the Veil, one exactly at it, one above — so the numbers it is guarding are actually the ones under test.

1.130.16 — 2026-09-08

Fixed

• A commander no longer talks about besieging a house it has just made peace with (#3216). When two guilds signed a truce, the AI empire's own briefing could still carry "NAP with [ROT]" and "assembling a force at 3:2:5" on the same page, for up to three days — an attack it announced and then never launched, because the game itself refuses it. The siege is now called off the moment the reason for it ends, and the same goes for a rival's outpost that has already fallen.
• A commander stops going quiet about the siege its own guild is fighting (#3216). If its sensors could no longer reach the worlds defending an enemy anchorage, it lost the whole line from its briefing — the structure, who holds it, how close it is to falling — while the rest of the guild discussed that very siege in chat. It now keeps the part anyone can read off the public outpost map. The defending worlds' coordinates stay hidden, exactly as before: a commander still cannot name a world its own grid cannot see.
• An expelled commander properly drops the campaign its old house ordered (#3216). Being thrown out of a guild ends every reason that guild gave you, but the fleet could keep flying at a target the house had picked — including after a restart, when the old orders came back off disk with nothing left to match them against. Its own reasons survive untouched: a grudge it earned in battle, or a strike it promised someone in chat, are still its business.

Changed

• The public outpost map — who holds a standing anchorage, where, and how many breaches are on it — is now guarded by tests confirming it reads the same for everyone, from the guild that built it to a player with no sensors at all. That was always true and never checked (#3216).

1.130.15 — 2026-09-08

Fixed

• "View Unlock" now takes you to the node, not just to the page (#3173). Tapping VIEW UNLOCK → on a locked ship in the Shipyard opened the Tech Tree with nothing selected: you still had to find the hull among fifty nodes to learn which research was missing. Now the tree opens with that node already selected, its requirement checklist showing, scrolled into view — and on a phone it switches to the right branch first, so the node is actually on screen instead of hidden behind a branch you were not looking at.

• The Apex shortcut no longer strands you. The Apex lens only exists once you have researched the Particle Accelerator, so for most empires a Doctrine padlock's "Research → Apex" sent you to the Tech Tree instead — and dropped you there with nothing selected. It now opens with the Particle Accelerator selected, which is the thing that padlock was telling you to research in the first place.

Changed

• The Tech Tree's branch buttons on a phone now say which one is active to screen readers and high-contrast settings, instead of relying on colour alone — and when a shortcut changes the branch for you, that button scrolls into view so you can see where you have landed.
• Arriving at a node through a shortcut also moves keyboard focus onto it on desktop, so screen-reader players hear the node they were sent to instead of silence.
• Coming back to the Tech Tree after switching to another lens is now an ordinary visit: it no longer re-selects the node you were sent to earlier and scrolls you away from what you were reading.

1.130.14 — 2026-09-08

Fixed

• *The ⚡ badge on a build row now says it is the change, not the next level's energy (#3169).* Everywhere the game offers you an upgrade — the Colony list, the Base build menu and the advisor's suggested cards — a small signed ⚡ figure sits beside the price. Hovering it said "Energy consumption/production at the next level", which names a different number than the one being shown. On a level 5 Compute Node the badge reads −25; the energy at level 6 is −121. The badge has always been the difference the upgrade makes to your grid, and the tooltip now says so: "Change in energy consumption/production at the next level". Read literally, the old sentence made a deep building look like it drew five times less power than it does.

  No number changed — the figure on the badge is exactly what it was, on all three surfaces. Reported by Elis, who opened the Building Sheet and produced the three numbers — current, next, change — that tell the two readings apart.

Also

• The first-build case keeps its old wording, because there it is true: when you build something for the first time there is nothing to differ from, so the change is the whole draw at level 1. That is the one level where both readings are the same number, and the new test derives it from the game's own energy formula for every building rather than taking the sentence's word for it.

1.130.13 — 2026-09-08

Fixed

• Every sector shows its number at its own centre again (#3259). A sector was named only by a small label above its territory, with nothing marking the middle of the region that label named — so telling twelve regions apart at a glance meant reading twelve rim labels, and more than one player lost their bearings out on the frontier. The number is back where it belongs: large and very faint, printed on the ground at the sector's own centre, the way a region is named on a star chart.

  It appears when you are pulled back far enough for the map to be about regions, and it is gone the moment you are close enough for individual worlds to matter, so it never sits over the planets you are reading. Nothing else moved — not a single anchor, not one travel time, and not the sector name either, which was already placed correctly just past the territory's edge.

• The map key names it too. The Sector entry in the key now reads "a sector is its name, a faint border of cells, and the big faint numeral that names it at galaxy and sector zoom" — so the key describes the three marks the chart actually draws instead of two.

1.130.12 — 2026-09-08

Fixed

Nothing in this release changes what you see or how anything scores — it corrects comments in the code that had stopped describing what the game actually does, so the next person to work on war points or debris fields is not misled by them (#3260).

• What an attacker's war points are. A note in the battle code claimed those points were the defender's destroyed value, capped at about a hundred. They are not: the value is divided and capped per contingent, cut on PvE targets, multiplied by the underdog bonus and by the attacker's relic multiplier, and topped up with a share of the loot and a win bonus — the last two outside the cap. The note now names every term, and a new test derives that list from the scoring engine itself, so the comment cannot drift from the formula again without the suite going red.
• A duplication warning for a value that is not duplicated. Another note warned that the debris ratio lived in two files and had to be changed in both. It has been exported and imported for months, and the second copy had quietly moved into a test that still asserted the old number by hand — which is where a retune would have gone wrong. That test now derives the number instead.
• Five more of the same kind, found by sweeping the two files. A "mirror" that is really one shared function; a points meter described as what was paid when it measures what was billed; a constant cited by a name that does not exist; a rule said to be documented in a file that does not document it; and a claim that one shield column had a single writer when it has two.

No behaviour changed and no game constant was retuned.

1.130.11 — 2026-09-08

Fixed

• The map key and the map now agree on what each glyph means (#3232). Open the Layers sheet and the MAP KEY told you ⬡ means Cells — the hex field the whole chart is drawn on. Hover a system holding a guild anchorage and the tooltip said ⬡ [EZU] anchorage. Look one up by way of the other and you came away confidently wrong, which is worse than having no key at all.

  Checking every row of the key against what the chart really draws turned up two more of the same thing. The tooltip called the Pirate Beacon ◆ — the character the key gives to a Relic — while the key's own Beacon row, and the round orange mark the chart actually paints, both say ●. And it flagged a contested system with ⚑, a character the key names nowhere and which means a bounty over on the target board.

  All three now read the way the key says. The Beacon line is ● Pirate Beacon in the Beacon's own orange; the anchorage and the contested line say their piece in words, and every figure they carried is untouched — an anchorage under fire still reads [EZU] anchorage · under attack 3/6. ⬡ goes back to meaning only what the key promises, and the hexagon still bullets an anchorage in the guild panels, where no map key speaks.

  The reason the key was the side left standing is that the chart really does draw a six-sided ring around every cell, so ⬡ depicts something on screen; the anchorage's hexagon was retired from the map by the One Map redesign, and its ⬡ had been standing in for a mark that is no longer there. The Beacon's glyph, name and colour are now written once and read by both surfaces, so the key and the tooltip can no longer be edited apart — and the key's colours are checked against the constants the renderer itself paints with.

1.130.10 — 2026-09-08

Fixed

• *A battle report no longer tells you one ship class did all the damage when it did not (#3281). The "Top gun" line under a fleet works out that class's share of the fleet's damage and rounded it to the nearest whole percent — so anything from 99.5% up was printed as 100%. A corvette line that dealt 59,900 of its fleet's 60,019 damage was announced as having done the whole of it, directly above a table that plainly showed another class shooting too. It now reads >99%, which is true and claims no precision the figure does not have. Checked against every battle report the game has stored: 175 of the 48,094 top-gun lines they can draw* were making that claim.

Also

• The same report was filed against six other screens that look like they have the same problem — the guild members table, the season standings, the guild showcase, the diplomacy tension bar, and the two formation lane pickers. Each was measured against live data one at a time, and none of them can actually print a zero for something you own: every one is already protected by rounding that happens before the number reaches the screen, or by the fact that the numbers it shows cannot be that small. Nothing was changed on those screens, so nothing you are used to reading there has moved.
• Three of those six are safe only because of a balance constant that has nothing to do with the screen it protects — how fast guilds forget a grudge, how much fire a sheltered rear row takes, and how battle damage is recorded. Those are now covered by tests, so if one of them is ever retuned, the screen it would quietly break says so first.

1.130.9 — 2026-09-08

Fixed

• The Proving Ground now describes the runner-up crate correctly (#3221). The prize table said that every strike which proves itself takes "the crate of the bracket below". For Warhost and Apex that was not what the close pays: the top three brackets all award the same 5M crate, so a proven runner-up there is paid the next crate that is genuinely smaller — 1M — which is what the table itself has always shown. The panel now says so in the terms the table is read in: a proven strike takes the next smaller crate on that table. The Activity diary entry a runner-up reads when the season closes carried the same wrong phrase, one line above the crate it was actually handing over, and now names the bracket it is measured against instead.

  No payout changes, and none ever differed from the table. Every crate handed out at a close was, and remains, the one the prize table showed — the six paid so far all landed in brackets where the old wording happened to be true. The sentence was the only thing that was wrong.

Also

• The bracket named as the exception ("Patrol, which has no smaller crate to prove into") is no longer typed into the sentence — it is worked out from the crate ladder, so if the rewards are ever retuned the panel names whichever brackets the change actually affects instead of quietly going out of date.

1.130.8 — 2026-09-08

Fixed

• The leaderboard now gives you a wanted player's address if your own Sensors can already see them (#3235). A hunted empire hiding every colony under the Veil had no address on its standings row for anybody — even for a player whose sensor network was already drawing those planets on the map. The wanted board at the top of the same tab started reading your sensors in v1.128.1; the standings row two hundred pixels below it did not, so one screen gave two answers about the same rival: a tappable coordinate above, a dead badge and a tooltip your phone never shows below. The standings now follow the same rule as the board — a colony out in the open stays everyone's address, and only when a player has none does your own sensor grid decide. If your Sensors do not reach them, you still see nothing, and nobody is told a coordinate they could not already find on their own map.

• A hunted rival is no longer painted two different colours on one screen (#3235). The Power column tightens its band in the starting sectors, and the only thing on a standings row that knows which sector a target lives in is that address. With it missing, the board above and the row below could grade the same rival differently — "far below you" on one, a green "fair target" on the other. They now agree by construction.

• Your own row is no longer an invitation to hunt yourself (#3235). A wanted player under full Veil now sees their own address on their own row, and on phones that badge had become a live target button opening their own colony, captioned in the language of prey. It is inert again, exactly as the wanted board has always treated your own line.

Also

• A warning the server writes when it briefly cannot read a player's sensor network was printing a raw database error class into the logs. Nothing a player could see, but it is one of the phrases the deploy's own health check treats as a crash — so a momentary database hiccup during a release could have failed that release, and any release after it, for no real reason. Both places that write this warning were corrected together.

1.130.7 — 2026-09-08

Fixed

• The system list now says which kind of absence its Power column means (#3252). A Pirate Beacon's Power showed a bare dash — the same dash the column shows when a real empire's Power could not be read. Two very different things, one glyph, and the commoner of the two looked like missing data. The Beacon's row now reads PvE in the same orange the commander cell already uses for it, a Power that genuinely failed to load reads … in grey, and the dash is left to mean only what it should: no colony on that slot. The dash on an empty slot is also dimmer now, so it stops looking like a reading.

  Reported by spaccamazza, who has found this same confusion three times in a week.

1.130.6 — 2026-09-08

Fixed

• A declared leave now holds a guild leader's seat (#3227). The roster has shown a member on a declared leave as On leave since the Last seen column arrived — but the succession rule was not reading the leave at all, so a leader who announced an absence could be replaced on the fourteenth day of it while their own officers were looking at a roster that called them legitimately away. The two now agree: a guild can take a leader's seat exactly when the roster beside it reads Absent, and never while it reads On leave.

  Nobody lost a seat to this. No guild leader had ever declared a leave, and no leader has ever been replaced for inactivity in the game's history — the fault was waiting for the first leader to take a holiday rather than costing anyone their standing.

Changed

• The leave screen now tells a guild leader what a leave does to their seat (#3227). It described what a leave does to your empire — the shield, the paused production, the bounty and removal clocks — and said nothing at all about your guild. Leaders now read, before they pick a length, that the seat is held while the leave runs and that the days count again once it ends: after 14 days unheard from, leadership passes automatically to your top officer, or to your most senior member if you have none, and you become an ordinary member. It also says the one thing you can do about it — sign in once before the leave lapses and the clock resets.

  A leave that has already run out still counts as silence, exactly as the roster reports it. Whether a returning leader is owed any grace after a leave ends is a separate question and is deliberately left open.

1.130.5 — 2026-09-08

Fixed

The war-effort ledger no longer prints a guildmate's score as zero (#3261)

Open the War Room's "Our war effort" tab, or a League match room, and you get the list of who is scoring for your guild — a name, a bar, and N pts · X%. A member whose score was real but smaller than one whole point was listed there reading 0 pts · 0%, and was subtracted from the "N members still on the sidelines" line at the same time. The panel counted them as a fighter and showed them fighting for nothing.

It now says what they actually scored: <1 pt, and <1% for a share too small to print as a whole percent. Nobody is dropped from the list, so the percentages still account for the whole tribute and the sidelines count still names only the members who really scored nothing.

The obvious fix — hide anyone whose score prints as zero — would have been worse here, and this is why the issue existed rather than being folded into the earlier sweep it came from: in a war, that percentage is the split. The server pays the enemy's tribute to every member with a positive share, using the same figures this panel divides by. Dropping the row would have hidden a member the game was still paying, left the visible percentages adding up to less than the tribute, and — when every score in a young war is small — told the player who had just raided that nobody had scored yet.

Where you will actually see this today. No war is open anywhere right now, so the war ledger has no rows at all; the surface where this was live is the League match room, which draws the same list. There, the failing figure was the percentage, not the points: 69 of the 264 contribution rows the League has ever recorded printed a real share as 0%, and 32 of those are in matches that are open right now. A member on 138 points out of a guild total of 165,713 owns 0.08% and was shown a zero.

Review found the same rounding lying at the other end of that column: a member on 99.92% was printed as owning 100% — the entire tribute — next to a teammate correctly reading <1%. Two labels that cannot both be true. That one now reads >99%, while a member who genuinely is the only scorer still reads 100%.

Also

• The campaign reward screen's war-standings line was the one place in the game that printed a war-points figure with no formatting at all, so a small advance could appear as "War standings +0.30000000000000004." It now rounds like every other war surface, and an advance below a whole point reads "War standings: under a point." instead of taking the credit away with a "+0". (#3261)

1.130.4 — 2026-09-08

Fixed

The galaxy map now shows you which cell opens a system (#3229). Tapping the star at the centre of a flower opens its system view — that came back in v1.126.1 — but nothing was ever drawn on that cell, so the only hint it was there was the highlight you get by resting a mouse pointer on it. On a phone there is no pointer to rest, so the door was invisible to exactly the players who had the least other way in.

Zoom in far enough that a tap opens a system, and every star's centre cell now carries a ring. Zoom back out, past the point where a tap zooms instead of opening, and the rings go: the mark appears only where it is telling the truth, so it never points at a door that will not open. Below that zoom the map looks exactly as it did before.

The ring is deliberately quiet — it borrows the map's existing habit of outlining a cell to say "this is a tap target", rather than adding a new symbol you would have to learn, and it never takes the place of a pirate beacon, an anomaly signal or the gold mark on your own systems.

1.130.3 — 2026-09-08

Fixed

• Nothing you can see in the game changed in this release. It corrects the notes the developers keep next to the battle code, which had started describing the Shielded Silo incorrectly (#3199).

  The Silo holds back a slice of a colony's stockpile before a raider can take a share of the rest — and how big that slice is depends on the level of the building you paid for, not on the research that unlocked it. That distinction was the whole point of an earlier change, and the notes had quietly drifted back to the old, wrong version: they listed the Silo as one of the things that decide a raider's share, which it has never been.

  One misleading note was reported. Checking its neighbours, then the whole file, then every comparable note in the codebase, turned up eight — the same wrong sentence had been copied from one file into seven others, including one that named a piece of research by a spelling that does not exist anywhere in the game, so nobody searching for it would ever have found the mistake.

  All eight are corrected, and a new test now keeps them honest: it works out from the game engine itself which research really does affect a raider's share, and fails the build if the notes ever claim otherwise — in either direction, whether they forget something real or credit something that does no such thing.

  Two unrelated stale notes found by the same sweep are recorded in #3260 rather than changed here.

1.130.2 — 2026-09-08

Fixed

• The galaxy map no longer stutters when you change zoom (#3178). Zooming in or out on the map hitched every time the view crossed between its galaxy, sector and flower scales — reported on a fast desktop on fibre, and worse on Android. The map was throwing away and redrawing every label, marker and sign on the whole chart at each of those crossings, in a single frame, to change one thing: the anomaly sign, which shows a bare symbol when the whole galaxy is on screen and adds its count once you are closer. It now changes that one label and leaves the rest of the chart alone. Measured on the map playground, the worst frame of a zoom sweep fell from 940 ms to 36 ms on a desktop viewport and from 709 ms to 14 ms on a phone-sized one, with no change to the cost of an ordinary frame.

Also

• The same report also asked for the cross-shaped glare over systems to be reduced and made to scale with zoom, and asked that planets keep their numeric order inside a system. Both were already true and neither needed a change: the per-system flare was retired with the One Map redesign, and planet numbering is unchanged — planet 11 is planet 11 in coordinates, missions and the API. Three stale code comments that still described the deleted flare were corrected so the next reader is not sent looking for it. That the drawing order around a system's rings is not its numeric order remains open, and is tracked in #3181.

Reported by spaccamazza (all three) and mhar02 (the Android half).

1.130.1 — 2026-09-08

Fixed

• The weekly chest no longer warns you off the one door that can still pay your shard (#3250). When a chest's payout was too big for your warehouse, the red warning ended with "— though once you earn a newer chest, this one's shard is gone for good". That line also appeared when your weekly chrono limit was already full, and there it pushed you exactly the wrong way: opening a chest while capped burns the shard for certain, and letting the chest sit is the only way the shard can ever arrive, since the limit is a rolling seven days and nothing anywhere punishes you for waiting. Now a capped player reads the cause, the overflow and how to avoid it, and gets no advice about a shard that opening would spend either way.
• And it no longer describes a loss that does not happen (#3250). A chest left until a newer one arrives does not evaporate its shard: it pays extra resources instead, worth the shard at the game's own rate. The warning now says the shard "converts to resources", the same words the line above it already used for the same event. A chest that has already been overtaken reads exactly as it did before.

1.130.0 — 2026-09-08

Changed

• Launch Again now sits below the mission designer, and shows three cards instead of six (#3174). Opening Missions used to put the Launch Again rail at the top of the tab: six full-height cards of saved orders above everything else. On a phone the first mission type sat below the fold, so starting a new mission meant scrolling past the shortcut for repeating an old one. The rail now renders under the objective picker on the phone, and between the mission form and Inbound on desktop — and it shows the three most recent saved orders. Two players asked for this: jaybee1019 and arcticpaladin.

New

• Every saved order is reachable again — "Show N more saved orders" (#3174). The rail used to show six cards and stop there: any saved order further down your list simply could not be reached, and re-flying it meant typing the whole thing out again. The rail now carries a control that opens the rest of it, and says how many it is holding. Collapsed it stays at three; expanded it reaches every order the game has kept for you, which is more than the old rail ever showed.

Also

• The cards themselves are unchanged: same size, same information, same buttons. A card is never hidden by the cap while its launch is still counting down, so an Undo can never disappear.
• Launch Again is still ordered by when you last flew a route, not by how often you re-fly it. Ranking it by what you actually repeat is a separate change and is not part of this one.

1.129.3 — 2026-09-08

Fixed

• No campaign or Echelon line names a quantity and then prints it as zero (#2890). Several screens decided they had a figure worth mentioning as soon as it was above zero, and then printed it through a formatter that rounds down to whole units. Anything smaller than one unit came out as "0", inside a sentence built around it being something.

  The campaign tab could offer to tell you that "0 resources are still owed to you" for levels you cleared under the old payout rate, and a debt of 1.4 read "1 resources". The Echelon receipt could report "Paid: 0 Fe · 0 Xe · 0 Dt." for a week's purse — and it said that instead of the honest "No purse was paid for that cycle." that belongs there. The campaign battle report could credit a lost hull with "0 hull damage dealt", a clause whose only job is to say she did something. And a war target could be flagged in red as "0 against us".

  All of them now say nothing where there is nothing to say, and say the true figure where there is one.

  This reached no player. A read-only census of production found no account in any of these states: all 205 stored campaign back-pay balances are exactly zero, and none of the 114 stored Echelon purses is below one unit. This closes the hole before anyone falls in it, rather than repairing something anyone saw.

  It is the same defect as the "Warehouse full: Ferrolite 0 did not fit" of a recent release (#2871, #2886), which was fixed on one screen by writing the rule out by hand. The rule now lives in one place, so the next screen to print a resource figure inherits it instead of having to remember it.

• The Echelon receipt keeps saying which share of a purse the warehouse refused. A purse is split 50/30/20 across the three resources, and a leg that reads zero on the receipt is normally the record that your mother warehouse was already full for that resource, so the whole share was lost. That zero is now protected: it is the only place the game admits it happened, and it is what lets you check the purse preview against what you were actually paid. A corrupt leg reads as a zero too, rather than reaching the screen as "NaN".

Also

• Two guild war-effort ledgers have the same flaw and are deliberately left alone for now: a guildmate on a fraction of a war point is still listed as scoring "0 pts". The obvious fix would hide the real share of the enemy tribute they own, so which figure decides the row is a question worth answering on its own rather than inside a display fix. Tracked as #3261.

1.129.2 — 2026-09-08

Fixed

• The synthetic empires' protection for Stellar Anomalies no longer rides on an unrelated piece of equipment (#2904). Founding a colony on an anomaly destroys it for everyone: the slot simply stops showing the signal, and unlike an expedition it leaves no claim and no report, so the player who was watching that signal never learns where it went. The synthetic empires are barred from doing it — anomalies are yours — but the list of anomalies they were warned about came from a single empire that sweeps the galaxy on behalf of all of them, and that empire is chosen for its sensor grid. How much of the list actually existed depended instead on its Astrometrics, a different instrument entirely: an empire with strong sensors and no Astrometrics would have produced an empty warning list across the whole galaxy, and the protection would have quietly stopped working with nothing to show for it.

  The warning now has its own eye. If any synthetic empire reads the sky better than the one sweeping the galaxy, its view of the anomalies is collected and added, so the protection is as good as the best telescope on the roster rather than as good as whoever happened to win a contest about sensors.

  Nothing changes in today's game, and that is said plainly rather than dressed up: the empire currently sweeping already holds the best Astrometrics of all of them, so there is nothing to add and no extra work is done. What changes is that the protection is no longer one sensor upgrade away from losing about a sixth of its coverage without anyone noticing — two other empires are already tied with it on sensors while reading the sky less well.

• A synthetic empire could announce that it was expanding and then never launch anything (#2904). Its plan and its own assessment are two different documents, and the assessment was allowed to pick a colonization target from the raw list of free planets — including one carrying an anomaly, which the plan itself had already ruled out. The launch was then refused, correctly, but only the launch: the empire kept the colony ship in dock and went on saying it was expanding, every few minutes, for hours. Found while reviewing the change above, and fixed with it — the two documents now agree about which planets are off limits.

Also

• The measurement behind all this, for the record: of 663 colonisations this season, 16 landed on a slot carrying that week's anomaly. Eight of those had already been claimed by an expedition hours earlier and so cost nothing; the other eight destroyed an anomaly nobody had claimed. Five were human players, who are entitled to it, and three were synthetic empires — all three before the protection reached its current form, and none since.

1.129.1 — 2026-09-08

Fixed

• The weekly Chest now tells you what waiting does to your chrono shard, even when your warehouse has room (#3205). Leave a Chest unopened and earn a newer one, and the older Chest's shard is paid out as resources instead — you keep the value, but not the shard. The confirmation already said something about this, but only inside the red warning that appears when your warehouse cannot hold the payout. If you had room, you saw a shard promised, a Cancel button, and nothing in between. Now the confirmation says it plainly: "Left unopened, this chest's shard converts to resources once you earn a newer chest."

  It stays one quiet line in text that was already there — no new banner and no extra tap. In the game's whole history exactly one Chest has ever been superseded, so this is worth knowing and not worth being alarmed about.

  Two things it deliberately does not do. If your weekly chrono limit is full, the row stays quiet: opening right then gives you no shard either, so warning you against waiting would point you the wrong way — waiting for the limit to roll off is the only way that shard ever arrives. And a Chest that a newer one has already replaced still says nothing, because there is nothing left to lose by waiting.

1.129.0 — 2026-09-08

New

• The Shipyard now tells you where the rest of your fleet is (#2907). The tab where you order ships could already show what sits in the hangar you are ordering into — the hull's owned ×N, the stationed-fleet strip, the All hulls sheet. What it could not show was everything else you own, so deciding whether to build another wing of Corvettes meant switching to Overview, reading the Fleet Ledger, and switching back. That table is now at the foot of the Shipyard as well: one row per hull you own, one column per colony, plus what is in flight, what is standing garrison, and the total. On desktop it sits at the bottom of the central column, beside the order panel, which keeps its place. On a phone it is folded away behind a single line you tap to open, so it costs you nothing until you want it, and it opens as a card per hull rather than a table you have to drag sideways.

  What the numbers mean, plainly: the per-colony columns are the ships parked in those hangars, «In flight» is what is currently travelling, and «Garrison» is what is standing on ground that is not that colony's. Ships still being built are not counted — the queue is reported on its own, above, and folding it in would have you counting hulls you do not have yet.

  Players with a single colony see no table, because with one colony the hangar already is the ledger.

Also

• The Shipyard's developer playground now carries a real multi-colony empire and fleets in the air, so the foot of that tab is not a blank during a design pass.

1.128.9 — 2026-09-08

Fixed

• A synthetic empire can now be drawn towards the buildings it actually wants (#3193). The AI empires walk towards a research gate when something they want sits behind it — but only ever for their signature ships. Their signature buildings were never wired in, and the archetype it hurt most was the shield-hoarder: its own priorities rate the Particle Accelerator at the joint-highest value in the game, and the gate that unlocks it was invisible to it. Measured across all 76 live synthetic empires, not one of the 16 shield-hoarders could ever be pointed at that door. They now study Astrometrics towards the Observatory — the watchtower their playstyle values above anything else in the game.

Also

• This does not yet put a Particle Accelerator in synthetic hands, and the changelog will not pretend otherwise. Its gate is Galactic Economy 17 and no AI empire has ever passed 12, so the door is now visible and still six levels out of reach. What keeps it there is the shape of the research incentive itself, which is a balance question left open rather than answered here.
• Two comments in the code that asserted things production contradicts were corrected: the bounty's claim that Dark Matter debt "is worked off against the Accelerator's future production" (true of human players, false for every AI empire, none of which owns one), and a note claiming the economy-focused archetypes reach Galactic Economy 17 first. Neither changes how anything plays; both stop the next reader believing a thing that is not so.

1.128.8 — 2026-09-08

Fixed

• The inbound attack alarm can no longer be silently blind (#3242). When the game could not read whether anything was flying at you, the HUD drew exactly what it draws when nothing is: an empty space. The most urgent element on the screen had a failure mode that said you are safe. It now says it does not know — a dimmed grey chip in the alarm's own place, with the alarm's triangle, that taps through to Fleets. It is deliberately not red: red still means a fleet is actually in the air, and an alarm that cried wolf on a bad network hop would be its own bug.

• Four more HUD readings stopped guessing when they could not read (#3242). The protection chip drew "unprotected" when it simply had no answer; the Orbital Home stamped OPEN ORBIT across the hero on a request that never came back; a Warlord whose Power reading failed saw no bounty skull; and the Dark Matter tile printed a confident +0.00/day and "across 0 built Accelerators" at players who own several. Each now shows the same quiet grey "unknown" instead of a number it cannot stand behind.

• The difference between "nothing" and "no answer" is now a real one. "You have no shelter running" and "the shelters could not be read" were the same pixels; so were "no fleet is coming" and "the alarm never loaded". They are told apart everywhere the strip draws them.

Also

• The unknown state appears only after a read has actually failed — never during the ordinary second while a page is still loading, when nothing is wrong and there is nothing to report.
• The dispatch bell was deliberately left alone: a missed badge costs a message read a few minutes late, and a permanent question mark there would be noise. Measured in production, these reads fail on about 0.13% of requests, and the alarm was genuinely blind in 11 of 12,628 sessions over a fortnight — rare, and impossible for the player to tell from the other 12,617, which is why the alarm was worth fixing and the bell was not.

1.128.7 — 2026-09-08

Fixed

• The privacy page's promise about your account identifier is now true (#3214). The page tells you, in the OpenRouter section, that your email address and your account identifier are not sent to the AI service. For two guild decisions that was not the case: when an agent-run guild reviewed its ranks, or decided whether to help a member out of the Fund, it sent each member's name together with their account identifier — the key your account is deleted by — to the AI service. Both of those run even while the AI empires are silent, because each one buys a real act rather than a line of chat. They now refer to a member by a short label that exists only inside that one request and means nothing anywhere else, and the identifier never leaves our servers. Nobody is promoted, demoted, removed or paid any differently than before — only what travels changed. The label is also kept out of anything a guild leader says in chat, so the announcements read the way they always did.

Also

• Two of the three gaps that issue #3214 reports are still open, and they are named here rather than left to be discovered: the rank review still sends the whole roster (every member's name, rank and contribution figure, not just the member being decided about), and the strategist — the part that plans an empire's moves, which never goes silent — still sends leaderboard rows, neighbouring empires and colony owners with their coordinates. Both need a decision about what the notice should say before anything changes, so this release does not claim to have closed them.

1.128.6 — 2026-09-08

Fixed

• The pirate beacon marks the system it is reporting, at every zoom (#3196). The beacon glow is drawn slightly up-and-right of a system's star so it does not print on top of the other signs that star can carry. That offset was a fixed number of screen pixels, while the cell underneath shrinks as you zoom out — so the mark walked off the system it belonged to: a whole cell out at flower zoom, about ten cells out at the zoom a phone opens the map at, and twenty-seven cells out at minimum zoom, where it hovered over a system four flowers away from the one it was reporting. It now stays on its own cell at every zoom, and keeps its full separation from the other signs wherever the cell is wide enough to hold it.

• The anomaly signal and the relic and bastion glyphs stopped drifting too (#3196). The same defect, found by sweeping every sign the map draws rather than only the one that was reported: the ⟡ over a system holding anomalies sat about ten cells off that system on a phone, and two named places sharing one star hung far enough to either side of it to land on their neighbours' cells. All of them are now held to the cell they name.

Labels are deliberately unchanged: the stacked lines of a place name, "YOUR HOME" under the reticle and the text beside a fleet in transit hang off the sign next to them rather than pointing at a cell of their own, and they keep stacking exactly as before.

1.128.5 — 2026-09-08

Fixed

The cargo panel tells you when your load needs more freighters than you picked (#3186)

«Fill every carrier» fills the hold to what every carrier in the colony could lift — including the ones you have not put on the table. That is deliberate: a button about cargo does not touch your fleet, because on a redeploy it would hand your carefully trimmed fleet back to the Advisor and quietly swap it for every ship on the planet, one tap before a one-way launch.

What it did not do was say so. You tapped it, the numbers went in, and nothing on the panel mentioned that the two freighters you had picked could not lift them. You found out at the bottom of the form, where the launch button had gone grey.

Now the panel says it where the cargo is written. The hold line turns red and names the load — (hold capacity 5,000 — 25,000 loaded) — and a line under the fields tells you what to do about it:

⚠ This load needs 25,000 of hold and the ships you picked hold 5,000. Add 8 Freighters, or the launch is refused.

If your freighters cannot do it alone but the rest of the hangar can, it says that too, and names the ships — offering the option rather than deciding for you. If nothing left can cover it, it tells you to trim the load instead of sending you to the Shipyard for hulls you already own.

The line is always on, not just after the button: typing a number over your hold by hand, or dropping ships after filling, reaches exactly the same contradiction and always did.

The button itself has not changed. It writes the same load it always wrote. The screen simply stopped keeping the consequence to itself.

Reported by Aurithris (atyo416), who compared the two fill buttons against each other — which is what isolated the difference to the ceiling rather than to a broken button.

1.128.4 — 2026-09-08

Fixed

• The research counter no longer vanishes when every lane is busy (#3184). The top strip carries three queue chips — Build, Ships and Labs — and two of them always tell you where they stand: a full shipyard reads 1/1 and stays put. Labs did the opposite. It appeared only while a research lane sat idle, so the moment your last lane filled the chip left the screen entirely, and a fully-committed research bench looked exactly like an empire that had never built a Laboratory. One of the server's most experienced players reported it as a regression, which is the clearest possible evidence that the strip was unreadable there.

  Labs is now a counter like the two chips beside it. It reads 2/2 in gold when every lane is running — gold being the same "nothing more fits in here" signal the warehouse and the other two queues already use — and stays neutral at 1/2 while a lane is free, with the tooltip naming the state either way: All research lanes busy or Research lane idle — start another research. The nudge that the chip was built for is still there; it is now a colour instead of a disappearing act.

  Twenty commanders had every lane busy at the time of the change, all of them active that day, and half of everyone running a three-lane bench. The chip still stays away in the two cases where it has nothing true to say: before your colonies have answered, and while your empire owns no Laboratory at all.

  ⚠️ This retracts a line from the v1.77.0 entry, which told you that "all lanes running — or no Laboratory built yet — means no chip: its presence alone is the nudge." That is no longer how the chip works, and if you learned the strip from it, unlearn that one sentence: the chip's presence no longer tells you anything, and its colour tells you everything. No Laboratory still means no chip.

1.128.3 — 2026-09-08

Fixed

• The galaxy map's hover tooltip now says whose facts it is reporting (#3206). Pointing at a planet showed a card headed Planet 3:12:4, and every line under it — the colony count, the leading alliance, territorial control, the guild anchorage, the contested flag, the pirate beacon, the anomaly count — was really about the whole system, not that planet. A player read "18 colonies", "[EZU]" and "◆ Pirate Beacon" off a world that had one colony, no EZU owner and no beacon. Every figure was correct; the card just never said what they were counting.

  The line that used to read in system 3:12 sounded like an address for the planet above it, which is what handed the whole list back to that planet. It is now a heading in its own right — SYSTEM 3:12, ruled off above the facts it covers — so the card reads as two things: the planet you are pointing at, and then the system it sits in. Pointing at a star is unchanged: that card is already headed by the system, so it is not repeated.

  The last line now names its door too: it reads click to open the planet, where before it said only "click to open" under a card that had just named two different places. Hovering a star still reads "click to open the system view".

  No number changed and nothing was added to the card: the figures were always right, and now they say what they are counting.

1.128.2 — 2026-09-08

Fixed

The Deep Observatory says which step an opponent can take back (#3189)

The Observatory's sheet promised "one more piece per level" and gave exactly one caveat: that it reads only the hulls your own Deep Sensor Grid can see. That left out the other side of the table. What a defender actually reads is the Observatory's detail step minus the attacker's Cold Trail — so the top step, the dearest one, is precisely the step an opponent can cancel, and nothing you could open said so.

The sheet now says it, in the building's own words:

• the ladder's end is named — one more piece per level up to 8;
• an attacker running the Cold Trail Doctrine shows you one step less than the level you paid for, and at 8 the step it takes is the exact composition, leaving you the hull classes;
• the early-warning row carries the same line at every level, because the step comes off wherever you stand: at 4 you lose the attacker's name and guild just as at 8 you lose the composition.

And what has not changed is stated too: the alert and the ETA are free and stay free. Nobody is ever attacked in silence, whatever the attacker flies.

This is a disclosure fix. No cap, cost or number moved — against an attacker without the Doctrine every level buys exactly what it always did, and the Observatory is deliberately not listed as a building with dead levels, because it does not have any here.

Raised by arcticpaladin (QQQQ), hours after he found the sheet's other misleading row on the same building.

Also

The second half of that report — that most players cannot discover the Doctrines exist at all, because the Apex lens is hidden rather than locked until you research the Particle Accelerator — is not changed in this release. It is a design decision, made deliberately and more than once, and it is with the owner along with the measured numbers.

1.128.1 — 2026-09-08

Fixed

• The wanted board now shows an address to the hunter whose own Sensors already reach it (#3185). A wanted player with every colony under a Veil appeared on the board with no address at all — an untappable row, for everybody, whatever they had built. That included commanders whose Deep Sensors already drew every one of that empire's planets on their own target view: the board told them nothing while the map next door showed them everything, and the only way through was to go and find the coordinates by hand.

  The board now asks the same question the Galaxy asks — can this commander see that colony. A colony in the clear is still everyone's address, unchanged. Where there is none, the board publishes the oldest colony your own sensor network reaches, and nothing at all if it reaches none of them.

  The Veil itself is untouched: no cloak or sensor level changed, and nobody is shown a coordinate their own Sensors could not already resolve. Reported by jsnam.

1.128.0 — 2026-09-08

New

• A synthetic empire that stops attacking is now noticed the same day, instead of by a player (#2808). Twice in one week a whole behaviour of the AI empires went to zero and stayed there — solo strikes for four days, raid openings for twenty-two hours — and both times the first person to notice was a player reading his own battle log and asking in chat whether the AIs had given up. There was no instrument anywhere that would have said otherwise. There is now: the admin dashboard carries a card counting six things the AI empires do — attacks flown alone, attacks flown as part of a guild raid, raids opened, spying, colonising and transport — each beside the same figure a day earlier and the median of the previous week. The two kinds of attack are counted apart on purpose: a raid contingent and a lone strike are both "an attack", and summing them is exactly what hid the first collapse for four days while the second kind kept the total looking healthy. When one of them falls below a fifth of its own weekly normal for a full day, the server says so in its log and posts once to a private operations channel, so nobody has to be watching a screen at four in the morning.

Also

• Nothing changes for players: no AI behaviour was touched, and nothing on any screen a player can reach is different. This release only gives the operator an instrument that did not exist.
• The threshold was checked against the real history rather than picked by feel: replayed over the last five weeks of production it catches all three known collapses — the earliest of them three days before the player reported it — and stays quiet on the ordinary weekly rhythm, on verbs too rare to have a stable normal, and on a behaviour that surges rather than stops.
• Rare behaviours are labelled "not watched" on the card rather than silently passed as healthy: at a handful of events a day there is no baseline steady enough to judge against, and pretending otherwise is how an alarm becomes noise nobody reads.
• The counters cover the LLM-driven empires only, deliberately not the Wraiths — that faction is directed by an operator and is legitimately silent between orders, so counting it in would raise false alarms and, worse, hide a real collapse behind thirty empires that were never going to move.

1.127.0 — 2026-09-08

New

A guild roster tells its officers who has stopped showing up (#2813)

A roster listed points and contribution, and neither of those fades. A member who stopped playing in August still sat there with the score they stopped with, and nothing on the screen told a leader apart from someone who was merely quiet this week — so the guilds doing real roster management were doing it blind, on a decision that costs a seat and cannot be undone.

Officers and guild leaders now see a Last seen column on their roster, in four bands: seen this week, not seen in a week, not seen in over two weeks, and on leave. A member who declared a leave reads On leave and never as absent — they said they were going, and the roster says so instead of counting the days against them.

Three things it deliberately does not do. It shows no date and no time, on anyone: a roster is readable from outside the guild, and an exact last-seen figure read off it daily tells a rival what hours a player keeps and when their defences go unwatched. A week is the finest line drawn, and no raid is planned a week out. It reaches only the ranks that can act on it — officers and the leader, nobody else, not even the rest of the guild. And it says "seen", never "active": the band counts signs of life, not time played, and the words promise only what the figure holds.

Reported by jsnam.

1.126.1 — 2026-09-08

Fixed

• The galaxy map opens a system again (#3180). Tapping a star had stopped working: the only way in was a ⊞ button in the camera controls, and the system view is the only place holding the survey, the anchorage card and the sector/system pickers. Tap a star and its system opens, as it did before — once you are zoomed in far enough that the star is a target you can actually aim at. Below that a tap still zooms in, which is unchanged.
• The ⊞ button opens the system you are looking at (#3177). It used to open the last planet you had tapped, so travelling with Go To, with a deep-space name or with the ⌖ home button moved the map and left the button pointing somewhere else entirely — usually your home system. It now opens whatever the camera is on, however you got there.
• A star under the pointer says what it is. Resting on one used to show nothing at all. It now highlights, names the system, and says that clicking opens the system view.

Reported by arcticpaladin, who was first and named what was actually lost rather than the symptom, and confirmed independently within minutes by spaccamazza and mhar02.

1.126.0 — 2026-09-08

New

The attack forecast can now be made to fight the garrison your report paid for (#3109)

The deepest espionage tier has listed the fleets garrisoned on a target since the last release — the hulls an ally lent them, or the ones they moved in from another of their own colonies — and the launch forecast refused to use them. It fought the colony alone and said so in a paragraph under the panel, which was honest and was not a number. A commander who had run the most expensive probe in the game, and read almost fifty thousand bombers parked on a target with almost no defence of its own, still had no way to put that figure into the plan.

The forecast row now carries a second line — "with the garrisons on station (estimated)" — whenever the report actually found any. It has its own verdict and its own simulation count, and each garrison fights as its own contingent, exactly the way the real battle musters them, so nobody's mastery or rows are lent to anybody else. The box around the row is now painted the worse of the two ends, so a victory whose garrison line says defeat no longer looks green.

The line you have always read has not moved. It still fights the colony alone. The two together are a band, and the point of showing both ends is that neither one is a measurement: a spy report reads a garrison's hulls and nothing else, so the second line assumes they fight on your own research and Tree, with no mastery and on default rows. It says so in the line itself. Their real research may be better or worse than yours; their mastery and rows can only be missing, so on equal research that line is a floor and not a worst case.

Two things it deliberately does not tell you. It answers whether you win, not what it costs — against a big enough wall a fleet breaks off early, so the losses are not comparable between the two lines and printing them side by side would have read as "and if the garrison is there I lose almost nothing". And it keeps every garrison on station whatever its watch says, so the watch on each one, in the report, is still yours to check against your flight time.

Reported by jsnam.

Changed

• The spy report card and the panel caveat no longer say the forecast "does not fight" the garrisons, and neither promises a second line the screen you are on may not be showing — they state what the forecast fights and leave the panel to speak for itself (#3109).

1.125.4 — 2026-09-08

Fixed

• The Veil now hides you from the last list that ignored it (#2905). Synthetic empires pick the colonies they besiege — the ones propping up a rival guild's Anchorage — off a shared map of the galaxy. That map was swept by a single empire and handed to all the others, so its Veil check was resolved against that empire's sensor grid: a colony hidden from the agent that wanted to attack it could still end up on its target list, and be named out loud in guild chat, because a different empire's sensors had reached it. Every other list an agent reads was moved onto its own sensors earlier; this was the one that could not be, and it now is. If you have paid for a Veil that beats a given empire's grid, that empire can no longer put your colony under siege, and can no longer talk about it.

Also

• A siege already under way is no longer called off by mistake when the target disappears behind a Veil. An empire that can no longer see a colony now treats that as "I cannot tell whether this is over" rather than "this is over", so the campaign continues instead of stopping silently — while a siege that genuinely ends against a colony the commander can still see is called off exactly as before.
• Groundwork only, with no visible effect on its own: the remaining rough edges this exposed — a siege that should be called off promptly when a peace pact is signed, and an Anchorage the agent can still legitimately discuss even when the colonies defending it are hidden — are filed as #3216 and are not fixed here.

1.125.3 — 2026-09-08

Fixed

• Sectors 1-4 have one name now, and the game says what they are (#2839). The most consequential region in the galaxy answered to three different names depending on where you ran into it. Buying a Displacement or ordering an interception called it "the cradle"; the War Room, the Leaderboard, the launch preflight and the Exodus screen called it "the starting sectors"; and one refusal managed both at once — "Both ends of that route are inside protected space, and the cradle is where nobody is hunted" — which reads as though a route ran between two different places. A player asked in chat what "the cradle" was, and another player, not a new one, could not say for certain the game used the word at all.

  It is the starting sectors, everywhere. That is what most of the game already said, it is the half that pairs with the Frontier, and it is the only one of the three you can understand without being told.

  And now you are told. The Galaxy tab says, under the sector picker, which band you are looking at and what is different about it. Inside sectors 1-4: new empires are placed there first, anyone can still attack you there but the bounty they pay is doubled and starts from a smaller Power gap, and a Frontier colony can never be folded back in with a Displacement. Out in the Frontier: no doubled bounty, and none at all under three times your Power. Both bands get a line, because once the starting sectors fill up new empires are placed in the Frontier instead — and that player is exactly the one who meets the term in a refusal without ever having been told what it means.

  Nothing about which sectors are protected, or any rule governing them, has changed — only what the game calls them, and the fact that it now explains itself.

1.125.2 — 2026-09-08

Fixed

The front page was describing a version of the game that no longer exists (#2799). Two claims in the "Every empire plays for real" section had quietly gone false, and a player pointed out the first one in chat.

The page promised that "threats made in chat turn into grudges" as one half of a conversation with the AI empires. Half of that is still exactly true — what you say in the channel is read, it becomes a grudge, and it comes back as a real fleet — but the AI empires themselves have not written a word anywhere since 1 September, so the exchange the page was selling has only ever gone one way since then. The page also promised you would "never be quite sure who is human", which stopped being true on 19 August, when every AI-run empire started carrying a badge that says exactly what it is.

The section now says what is actually the case: around eighty empires are run by machines, every one of them is marked, you always know which — and what you cannot tell is their play from a veteran's. They read the channel, they never write to it, and the grudge still arrives as a fleet.

The same correction was applied everywhere else the old claim had been copied: the project README (its headline, its feature table and its "living world" section, whose sample chat transcript is now dated as the July 2026 record it is), and the privacy page, which told players that a whisper to an AI empire was "how you talk to it" while promising a reply that cannot currently arrive. The privacy page's data disclosures were deliberately left as they are: describing a flow that is switched off is the safe direction for a privacy notice.

Nothing about the AI empires themselves changed in this release — they were not un-muted, and they keep playing exactly as they did: building, researching, arming fleets, reading the channel and remembering who threatened them.

Reported by mhar02.

1.125.1 — 2026-09-08

Fixed

• The server no longer goes quiet for a second, two minutes after every restart (#2803). The daily metrics job runs its first pass shortly after each deploy, and to work out how active everyone had been it was reading every action any empire had taken in the last month — including the synthetic ones, which act far more than people do. It then threw most of those rows away. Fetching and discarding them stopped the server answering anyone for about 1.2 seconds. Nothing failed and nobody had reported it: the pause was too short to break a request, only long enough to make every request in that moment wait. It now asks the database for the players it actually counts, which is about a quarter of the rows.

  Measured against the live database, before and after: the pass took 12.6 seconds and froze the server for up to 1340 ms; on the deployed build it takes 4.6 seconds and the longest freeze is 358 ms. On the three restarts before this change the server's own event-loop gauge reported that freeze every time, at 1103 ms, 1305 ms and 1180 ms, always 157 seconds after boot. On the restart that shipped this change it reported nothing at all.

  Nobody found this by playing: the event-loop gauge added in v1.107.4 found it on its own, and this is the first thing it caught.

Also

• The admin dashboard's activity panel is built by the same code, so it draws faster too. The numbers on it are unchanged, and deliberately so — this changes which rows are fetched, never which ones are counted.

1.125.0 — 2026-09-08

New

The galaxy register now tells you what it is showing you (#2806). The galaxy's half of the Activity feed looked like a complete, chronological log of everything happening out there. It never was one, and nothing on the screen said so — which meant that reading it carefully and reading it wrongly looked exactly the same.

Three things were always true of it, and all three still are:

• it holds the newest 80 events, so its window is measured in rows, not in hours — and it therefore narrows as the galaxy gets busier;
• a battle's replay stays sealed to strangers for a day, because the report shows the defender's formation and building levels;
• so battles that have just passed that day reappear in the register, sitting at their real date among rows that are hours old.

Put together, a slow, steady stream of attacks could be almost invisible while the same number of attacks in one hour would have been impossible to miss. A player counted the AI attacks in the register, found almost none, and concluded the synthetic empires had stopped attacking. He was right that they had — and there was no way to tell that from this screen.

So the screen now states its own shape:

• it says how far back it is actually complete — "it holds every public event from the last 2 hours". That number is worked out fresh on every request, because the real window moves a great deal: measured across a week of production it ran from 56 minutes on the busiest hour to nearly 8 hours on the quietest. It is always rounded down, so the register never claims a wider window than it can back up;
• it says it is a sample and not a full log, and that activity spread thinly over a day can be missing from it entirely;
• a battle that is only there because its replay just opened now says so — "Replay now available" — instead of passing for fresh news. Your own battles are exempt, because yours have been watchable since the moment they were fought.

Also

The claim is now careful in the places it would otherwise quietly drift: it disappears while the register is unreachable rather than standing over data that is no longer arriving, it never appears above a filtered list that contains none of the rows it describes, and the window it states is measured against the moment the answer arrived — so leaving the tab open, or waking a laptop, can no longer stretch "the last 2 hours" into "the last 6".

Nothing about how the register selects events was changed. The size of the window, the per-kind floors and the one-day replay embargo are all exactly as they were; what changed is that the screen now admits to them.

Reported by QQQQ (arcticpaladin), who also supplied the wording for the reopened-battle rows — and who held on to the observation that made the whole thing make sense through two confident and wrong explanations.

1.124.11 — 2026-09-08

Fixed

• The weekly chest now tells you when a chrono shard is coming, not only when one is not (#2812). Opening a weekly chest asks you to confirm, and that confirmation already said what the chest pays in resources — and said "No chrono shard" when the shard was going to convert instead. When a shard was coming it said nothing at all, so a player who had never hit the weekly chrono ceiling had never seen the word "chrono" on that row and opened their once-a-week reward without being told a shard was in it. The confirmation now names it in both directions: "It also pays a 1h chrono shard." And if your home warehouse is full — the one case where the chest is all shard, because the shard is not held back by storage — it says "It still pays a 1h chrono shard.", so the line about crediting nothing is no longer an argument for walking away from it. Reported by KingPaladin, who received a shard he had not been told about and came to ask where it had come from.

1.124.10 — 2026-09-08

Fixed

• The weekly trade contract now says what "priced near the going rate" means — before you accept (#3126). A trade only advances Close trades when the two sides are worth roughly the same: within 1.8× of each other, valued at Ferrolite 1 · Xenogas 1.5 · Data 2. That rule decided quietly. Nothing on the market board, the offer sheet or the contract card ever named the rate or the band, so an offer could look perfectly reasonable, settle, and leave the card exactly where it was. RoyDisco reported it after three trades with the same partner did nothing while a fourth counted — and he had already ruled out the once-per-partner rule himself.

  The rule now appears in four places: the offer sheet tells you, before you accept, how far apart the two sides are priced and whether that clears the band; the market board marks any offer that will not count; the contract card spells out all three of its conditions with their numbers; and the Activity feed says why a settled trade did not advance the card, instead of saying nothing at all.

  Nothing about the rule itself changed. No threshold was retuned, and no trade that counted before stops counting now. The engine was right and silent — only the silence is fixed.

• The same card no longer implies its 1,000 minimum is a value (#3126). It counts raw units, so 600 Data is 600 — not the 1,200 its market worth suggests. The card now says "units". Half of all the trades that missed this minimum would have cleared it read the other way.

Also

• The board's existing "lopsided offers" divider is a much looser rule than the contract's, and measured differently: 20× on raw quantities against the contract's 1.8× on the weighted list. An offer could sit comfortably in the board's fair group and still be refused by the contract, which is exactly what happened to the reporter. The two marks are now independent, so a fair offer that will not count says so.

1.124.9 — 2026-09-08

Fixed

• Synthetic empires no longer stockpile colony ships they can never launch (#3136). An empire that had run out of room went on ordering Colony Ships anyway, and kept declaring an expansion plan whose every founding the game turned away at the launch button. On the day this shipped, 56 of the 76 synthetic empires were sitting at their colony ceiling holding 140 colony ships between them — hulls with nowhere to go — against 21 of the 843 human commanders at their own ceiling. A person at a full ceiling stops ordering the hull. They now check it before they buy and before they launch, so those resources go into fleets and buildings that do something.

• A Colonial Ark now settles a world for a synthetic empire (#3136). The game has accepted an Ark as a founding hull since the end of August, but the synthetic commanders only ever recognised a Colony Ship: an empire holding an Ark and nothing else declined to settle anything at all, and the mission it would have flown asked for a ship it did not own. An Ark now founds, and is spent before a Colony Ship when both are in the hangar — it leaves a Warehouse behind on the new world, so it is the better hull to land.

• A synthetic commander's plan is written against a ceiling it can see (#3136). Its planning sheet showed how many worlds it held and never how many it was allowed, so "expand" was a choice it could make with no way to know the answer was no. The sheet now shows the count against the limit and names which of the two gates — the Colonisation Discipline or the home Command Center — is the one actually stopping it, the same way the Research sheet has told players since #293.

Also

• The colony ceiling is now served by the game itself rather than worked out again by each reader. #3114 collapsed three hand-written copies of that rule into one after they drifted apart; this ship deliberately did not add a fourth, and the tests fail if either side of the remaining pair moves without the other.

1.124.8 — 2026-09-08

Fixed

A battle report no longer credits you with bonuses that were not fighting (#2853)

Open a battle report and expand a commander's block: under the fleet you see what they fought with — Energy Weapons, Armor Plating, Shields, the multipliers those add up to, and a row of chips for the tech-tree nodes in play. That chip row asked one question only: do you own this?

So an attacker's own block listed Planetary Shield II, a node whose description says in as many words that it grants its shield bonus on defense. It was sitting directly under a Shields × figure that — correctly — did not include it. The report was disagreeing with itself on the same screen, and reading it the obvious way meant believing your fleet was carrying a bonus it never had.

The engine was right the entire time. Not one battle was ever fought with the wrong numbers, and nothing about how your fleets perform has changed. What was wrong was the card describing it.

The chips are still all there — knowing you own Planetary Shield II is worth something even on a raid, because it is waiting for you the next time somebody comes to your colony. But the ones that could not act in this battle are now drawn with a dashed border and say so in words: Planetary Shield II 1 · defense only. A line under the row explains the dashes, and holding the chip names the node and what it would have changed.

It turned out to be wider than the report described. Working it out from the combat rules rather than from a list found five such nodes on an attacker's card, not the two that had been named: Planetary Shield II and Veilsight, whose shields only rise when defending, plus Orbital Obfuscation, Active Camouflage and Abyss, which reduce what a raider carries off your colony and so do nothing when you are the raider. And the same fault existed in the mirror: Raider Protocols on a defender's card, which only pays when you are the one attacking. Those four never appeared in any bug report, and they are on 19,861 and 8,181 report entries respectively.

Of the 20,705 battle reports stored, 15,377 show a tech-tree chip row at all, and 14,530 of them — 94.5% — carried at least one chip that had not done anything. They are all corrected retroactively: the fix is in how a report is drawn, so every report you already have, however old, reads correctly the next time you open it.

Two players found this between them. TartarusMkII reported it, and when he was told he had misread his own screen he went and produced the screenshot that settled it. QQQQ had already given the right explanation in one line — that the row listed every combat node whether or not it applied to your side of the battle — before anyone had verified anything. He was right.

1.124.7 — 2026-09-08

Fixed

• The Shipyard and Nanoforge cards now say how much time a level actually saves (#3147). The Colony Advisor's "SUGGESTED NEXT" card for the Shipyard read "Hulls leave the ramp 17% sooner" at level 1→2. The build really goes from 167 s to 143 s, which is 14% off the wait — 17% was the rise in build rate, a true number in the wrong unit under a word about time. The card now prints the time saved, and the Nanoforge's "Heavy hulls come out N% sooner" does the same.

  The difference is small and it shrinks as the building climbs: three points at Shipyard 1→2, one or two in the levels just above, and nothing at all past level 8, where both ways of counting round to the same number. On the Nanoforge it was never more than one point. No upgrade was ever worse than advertised — every level did shorten the build, it was just credited with slightly more than it delivered — and the overstatement sat exactly where it mattered most, on the low levels, where you are deciding whether the building is worth starting at all.

  What the advisor recommends has not changed. The order of the cards, and which upgrade it puts first, are ranked on production throughput and were already right; only the sentence moved. Checked over 3,000 generated colonies: same cards, same order, same levels.

Also

• A Shipyard or Nanoforge level that saves no time now says so plainly instead of printing "0% sooner", and one that would lengthen a build would say that too, rather than showing a negative number next to the word "sooner" — the shape of a bug this line shipped once before (#593).

1.124.6 — 2026-09-08

Fixed

• The Dark Matter tile updates the moment you spend it (#3138). Buying an Apex node took the Dark Matter out of your account immediately, but the counter in the top bar kept showing the old balance until you changed screen or reloaded — so the panel could read 300 DM while the bar above it still said 500. Reported by SpartanEcho.

• Nine other places that moved a number in the top bar and did not say so. The same fault was everywhere, and most of it had never been reported:
  • ordinary research at tier 9 and above, which is billed partly in Dark Matter;
  • Star Forge hull steps;
  • warp launches, and the warp leg of joining a guild raid;
  • campaign rungs, which pay Dark Matter — the bar was understating a reward you had just won;
  • opening a crate and claiming a contract, which credit your mother colony's resources;
  • the Requisition Surge, which triples your storage ceiling: the caps, the "almost full" colouring and the warning all kept the old limit;
  • ordering ships, which raises your Power — until now nothing but the 30-second refresh ever updated the Power tile or the WANTED pill after an order.

  All of them now update as soon as the action lands, instead of waiting up to thirty seconds.

Changed

• If a refresh of your Dark Matter balance fails, the screens that show it now keep the last known figure instead of blanking it — a number that heals on the next refresh is easier to act on than an empty box. The controls that spend it stay shut while the figure is uncertain, and the warp row says so in as many words rather than claiming you are short.

1.124.5 — 2026-09-08

Changed

• One map, no flowers (#3179). The galaxy is now a single dark field. The flower outlines around every system, the star glows, the coordinate labels, the control and contested rings, the anchorage glyphs and the big sector numerals are gone. What is left on the ground: the planet icons, your guild's ground as a tinted blob around each of its colonies, a faint border and a name for each sector, your home mark, the pirate beacons, the anomaly signs, the named places and the fleets in flight.
• Planets open when you can aim at them (#3179). There are no zoom levels to learn any more. Hover a planet and the tooltip names it and says whether a click opens it. If the cell under your finger is too small to aim at, a tap zooms in toward it; once cells are wide enough, a tap opens the planet's card. Stars and empty ground open nothing.

1.124.4 — 2026-09-08

Fixed

• A synthetic empire no longer acts on a fleet it could not see (#3123). When the game's server hiccups and an empire's request for its own fleet list fails, that empire used to quietly assume the best: no garrisons posted, every ship safely at home. It then acted on that assumption — parking another third of its hangar on a relic it was already guarding, or launching a relic assault with half its fleet on the far side of the galaxy. A half-strength assault is worse than none: the ships are lost and the holder is warned someone wants it. Now an empire that cannot see its fleet simply waits for the next reading, which is a few minutes away, instead of inventing a reassuring answer.

• ...but it still turns up for its own guild. The same rule, applied carelessly, produced the opposite defect: an empire that had called its guild to arms on a relic would keep advertising the call — "allied fleets promised, not deployed" — and then silently fail to join the raid its guildmates had formed for it. Answering a call already made is not the same as starting a new one, so an unreadable fleet now opens nothing, cancels nothing, and abandons nothing. Guildmates who answer a synthetic empire's call will still see it show up.

Also

• This is robustness under a server fault, so most players will never see it happen. What changes is what an empire does on the rare tick when a read fails — not anything visible on a normal one. The frequency could not be measured: the failed read is logged below the level production records, and a relic expedition has no log line of its own, so a fault on this path would have been invisible rather than rare. That is why it is now pinned by tests instead of watched for.

1.124.3 — 2026-09-08

Fixed

• The Deep Observatory and Deep Sensor Grid sheets no longer promise a return that stops arriving (#3125). Both buildings' "Counter-espionage" line said the level was worth "+1 per level, no limit" — and the edge really has no limit, but the chance it buys does not: it tops out at 95%, and a Deep Sensor Grid reaches that ceiling on its own at level 7. Players were being sold levels whose points had stopped converting into anything.

  The line now names the opponents a level still works on, computed from that colony's own Astrometry, Sensor Grid and Observatory: "a level here moves the odds only against Espionage 22–24: at or below 21 this colony is already at the 95% ceiling". Where a spy strong enough to outrun the grid entirely exists, it says that too, and where no spy in the game is still affected it says so plainly. It never calls the level dead, because against a strong enough opponent it is not — and the Observatory has deliberately not been added to the "no further effect" warning list the Colonial Outpost and three others carry, because unlike them its effect genuinely keeps growing.

• The Observatory's early-warning line stops handing off to a dead end. At level 8 the warning detail is complete and the sheet said "every level from here buys counter-espionage instead". On a colony where counter-espionage has also topped out, that sent the player at an axis the next line withdrew. It now names the band instead.

• The caveat is no longer painted in the colour reserved for gains, and at level 30 the sheet stops talking about "further levels" beside a panel that reads "Maximum level reached".

Also

• What the early-warning tier does was already true and stays true: it runs one step per level to 8 and stops there. What was missing was any statement of what the levels above it are actually worth, which is what this ships.
• Raised by arcticpaladin (QQQQ), who named the building, the axis and the warning list it was missing from, five days before anyone checked.
• The production figures in the original report counted the Wraiths — a synthetic faction — as players. Re-measured on human players only: 20 of 206 Observatories stand above level 8, held by 10 players, not 143 of 329. The disclosure defect was real; its blast radius was about seven times smaller than reported.

1.124.2 — 2026-09-08

Fixed

• The Colony Advisor's defence cards said a number was about raid loot. It never was (#3137). A defence upgrade card read "Covers 9% more of what a raid would take from this colony." That number is not about loot, and never has been: it measures how much harder the colony gets — the defence a level adds, against what the colony already fields plus a reference raid. It reads no Silo reserve, no plunder rate, no warehouse, and no attacker. Remy spotted it and named the reason exactly right: the advisor cannot predict the strength of a hypothetical attacker. It does not try, and it never did — only the sentence claimed otherwise.

  The card now says what the figure is: "Hardening: +9% — this level's added defence, measured against what the colony already fields plus a reference raid; not an estimate of resources saved." A level whose effect rounds to nothing reads +<1% instead of +0%.

  Nobody's build was mispriced by this: the ranking of defence upgrades against each other, and against economy, is unchanged. What changed is what the card claims the number means.

• The same figure was coming back in three different wordings (#3137). The engine writes one sentence, but the advisor's commentary rewrites card text into plainer words, and it was rewriting this one — "cutting off what any raid could strip", "blocks potential raid losses", "shaves off raid losses" on three cards of the same panel, none of them the quantity the engine computed. The figure is now written by the engine and shown to you unchanged; the commentary no longer receives it, so it has nothing to reword. The sentence explaining what a defence upgrade protects is still written in plain language.

• A defence card no longer calls the Planetary Shield a gun (#3137). One line covered Shield, Batteries, Fortress and Cannon alike and said "these are the guns that will meet it" on all four.

Also

• Raid reports and advisor cards now stay in their own lanes on purpose. A raid report tells you what limited a real haul, in real loot terms (#2898, shipped just before this). A defence card tells you how much harder the colony gets, and states no loot at all — because before a raid there is no attacker to measure against.

1.124.1 — 2026-09-08

Fixed

Synthetic empires no longer walk towards tech they cannot reach (#2856)

An agent decides what to research partly by asking "is this Discipline about to unlock something?". It answered that by looking only at the shallowest locked node of the branch, and that shortcut was wrong in both directions at once.

It pointed agents at levels that unlock nothing. Across the whole roster, 27 of the 143 live nudges aimed at a node that was not waiting on research at all — it was waiting on another node the empire had never bought. Fifteen empires were being walked towards Shields 14 for a node whose real prerequisite they were eight levels short of; eleven more towards Energy Weapons 14 for the same reason. And it hid gates that were genuinely one step away: any unlock deeper in a branch was invisible while anything shallower was still shut, whatever the reason.

An agent now measures a door only when research is the last thing holding it, and measures it by the farthest key it still has to buy — so a node with two requirements is walked towards on both, instead of neither.

Nothing about this is visible in the galaxy today, and the honest version of the entry says so. Replaying every one of the 76 synthetic empires against its real research state, not one of them changes what it studies next: the nudges that moved were all on Disciplines those empires cannot yet afford. This corrects a rule that would have started making wrong choices as the agents got richer, which they have been doing all week. Nobody should expect to see an agent behave differently this evening.

Also

The report this came from (#2834 → #2856) claimed the mechanism was dark for 52 of 76 empires and that the Orbital Fortress was the casualty. Re-measured against production: most of that darkness is a different, deliberate rule (an archetype only watches the branches it cares about), and lifting the blindness lights no additional Armour nudge at all — the roster is simply not far enough up the Armour track for the gate to be in range. The Fortress had already been reached by a different route (#2855), which never consulted this rule. The nudge window was not retuned to make the numbers look better.

1.124.0 — 2026-09-08

New

• Planets on the map (#3161). Every planet cell of the hive is now drawn as an icon of its type: rocky, gas, ice or lava. An empty planet is a dim barren body; a colonised one grows with its owner's power through six tiers, up to rings, moons and a Precursor orbital ring in the teal of the building art. Twenty-four icons in all.
• One map, no descent (#3161). Tapping a planet at flower zoom opens that planet's card right over the map, the same card the target list opens, and the map never switches view. The star tap and the system peek that dropped you three levels down are gone, and so is the GALAXY ▸ SECTOR ▸ FLOWER breadcrumb. Below flower zoom a tap zooms in instead: toward the star from far out, toward your finger from sector zoom, so you are never asked to aim at a 12-pixel cell.
• A door to the system view (#3161). The new ⊞ button next to the camera controls opens the system view for the system you last touched, where the survey, the anchorage card and the sector and system pickers still live.

Changed

• The map's overview now reports which planets of each system are inhabited, and at which tier, for the colonies you can see. What the Veil hides from you stays hidden here too.

1.123.1 — 2026-09-08

Changed

• Nothing a player can see this time. The sweep that ruins and removes long-abandoned empires asks one question over and over — is this player on a declared leave right now? — and it had three hand-written copies of the answer, even though the game's rulebook already exported the official one. All three now ask the rulebook. Declaring a leave, cancelling one early and letting one expire behave exactly as they did before, and that was checked rather than assumed: the two forms were compared across every shape the date columns can take, and they agree on all of them. (#3104)

Also

• The leave rules are now legible side by side, which is the point of the change. Two of them read the same two dates and answer different questions: one asks whether a leave is running now, the other asks how much of a past silence a leave excuses. They part company on a leave that simply ran out — still excused, but no longer running — and they deliberately fail in opposite directions when a date cannot be read. Each place that asks now says which of the two it wants and why, so the next person to touch this code cannot copy the wrong one by accident, which is exactly what had already happened once.
• The tests around a declared leave got the case they were missing. Every existing test used a leave that was still running, which cannot tell the two halves of the rule apart; there are now cases for a leave the player cancelled and a leave that expired on its own, at both of the points where the sweep decides. Each half of the rule, removed on purpose, turns them red — the whole file used to stay green under either.

1.123.0 — 2026-09-08

New

• The raid report now tells you what limited your haul (#2898). Until now a battle report printed the plunder and nothing else, so a raider who came home with less than he expected had no way to tell why — and the most natural guess, "I did not bring enough freighters", was usually the wrong one. Winning raids now carry one extra line under the loot: Silo reserve −34.6% · plunder rate 48.4% · raider hold used 9.7% — the hold was not the limit. It names the three things that decide a haul: how much the defender's Shielded Silo locked away before you arrived, what share of the rest your raiders could carry off, and how full your cargo hold actually got. That last figure is the one that settles the argument — at 9.7% it says the freighters were never the problem, and more of them would have changed nothing.

  This was reported by a player who raided a colony holding millions, brought home 234,347, and concluded he needed a bigger hold. He did not: his hold was barely touched. Checking the record, that is the ordinary case rather than a fluke — in about two out of three winning raids the cargo hold never filled at all, so the haul was set by the Silo and by the plunder rate, not by the fleet. Nobody was cheated and no number has changed: the raids always worked this way, the screen just never said so.

  Older reports do not have the line — the figures were never recorded before this release, and a blank is left blank rather than shown as a zero. Raids you fly from now on will carry it.

  The line is for the people who fought the battle. Public replays, which anyone can watch a day later, do not show it: the Silo and plunder figures can be worked backwards into how much the defender was holding, and what an enemy has in store is what a spy probe is for.

1.122.36 — 2026-09-07

Fixed

• Synthetic empires stop promising fleets they were never going to send (#3122). When one of the AI-run empires called for allies to take a relic, that call to arms never ended. It had only two ways to die — the relic being taken or sealed by someone else, or the empire being thrown out of its guild — so a muster raised on a relic that simply stayed where it was went on standing forever, and the empire went on repeating the same offer in guild chat. Before the fix, 56 of the roster's empires were advertising a call to arms, every one of them raised between 16 and 17 days earlier, and not one had ever drawn a single answer.

  A call to arms now lapses after 16 hours — long enough that every guildmate has certainly had a full session in which to see it and reply, plus the longest any raid in the game's history has taken to fly. And it lapses out loud: if the empire announced the muster in guild chat, it says so when it stands down, and it says which of the two things changed — that it called too long ago, or that its fleet is committed elsewhere. A muster nobody was ever told about ends quietly, because there is nobody to untell.

  The second half of this was invisible but larger: an empire could not raise a new muster while an old one was still open, so the stale call had also frozen the relic pathway for a third of the roster since the middle of August. Those empires can go relic-hunting again.

1.122.35 — 2026-09-07

Fixed

• The war banner now says which way the money goes (#3107). The war room's headline figure was labelled "at stake", and it is your own guild's bill — the 8% tribute your members' home colonies hand over if you lose. That direction was written only in a hover tooltip, which a phone cannot show at all. A guild leader who won a week-long war read the figure as the prize, compared it with what his guild collected, and reported that the game had underpaid him. It had not: the two numbers were never the same quantity. The line now reads "if we lose, we pay" in plain sight, on every screen.

• And it says what winning pays, beside it. A second line reads "if we win, they pay from their own colonies" — the same tribute rule applied to the other side, off a different pile, so a different amount. It states the rule rather than a number on purpose: what an enemy is holding is what a spy probe is for, and printing it here would have handed every member of a warring guild a live read of the enemy's stores for free. That second line now shows even when your own colonies are empty, which is exactly when a guild most wants to know what a win is worth.

Also

• This is about what the screen says, not about what the game pays: no tribute rate, war constant or payout rule was touched, and nothing changed in what any war has settled. Whether a win leaves a visible receipt is a separate open report (#1484) and is not covered here.

1.122.34 — 2026-09-07

Fixed

Synthetic empires ask their guild for a hand on a queue, not only for cargo (#3120)

On the guild board, every synthetic empire had only ever asked for one kind of thing: a load of cargo. Not once in the history of the game had one asked a guildmate to shorten an order — 958 requests, none of them a speed-up — while the players beside them spent about two thirds of their own asks on exactly that.

The rule behind it was meant to say resources first, on the ground that an empire with no ferrolite has no queue to speed up either. In this game that turns out to be false: an order is paid for when it is queued, so it finishes on time however empty the warehouse gets. The one thing scarcity really does hold up is a research order that has not started yet — and that is the one thing a speed-up could never have helped with anyway.

So the reason stopped applying wherever it was checkable. An empire with a long order running now asks for the push; an empire with nothing worth pushing asks for the cargo, exactly as before. Nothing was retuned to get there — no threshold moved — and no ask was traded away for it.

Guildmates will start seeing synthetic empires ask for a hand on a build or a research order. Expect it to stay a minority of their asks: most empires have nothing long enough running at any given moment, and this does not try to match the human mix.

Also

• An agent no longer wastes its one ask of the day naming an order that has already absorbed its five pushes — the server would refuse it, and the empire would stay silent about the cargo it actually needed. It now asks for the cargo instead.
• A request naming an order that has since finished comes off the board on the next guild tick instead of waiting for the twice-daily asking window, so fewer guildmates click help on something that is no longer there.
• When an agent says how long an order still has to run, the figure is rounded rather than rounded down — "about 4h" instead of "about 3h" for an order with 3h58m left.

1.122.33 — 2026-09-07

Fixed

The Empire Overview no longer advertises a colony slot that is not there (#3114)

The number of colonies you can command is two requirements at once: your Colonization research, and the Command Center of your home colony. The Overview was reading only the second one.

So a commander with one colony, Colonization 0 and a Command Center at level 6 saw "1 / 2 Colonies" at the top of the Overview and "COLONIES 1 of 2" below it — while the Research sheet said "1/1 colonies" and the Galaxy refused the founding outright. The Overview is the screen you plan from, and the free slot it showed was worth a Colony Ship: resources and build time spent on a colony the game would turn away on arrival.

The Overview now shows the limit that actually applies — "1 / 1 Colonies" — and, when something is genuinely in the way, says what would clear it: "Colonization 1 unlocks the next" when the research is the wall, the Command Center level when the building is, and both when both are. When you still have a slot free it says nothing, because nothing is stopping you.

More than 60 commanders were being shown an inflated limit, and over 40 of them were looking at a free slot the game would have refused. If you were one of them, nothing about your empire has changed — only the number on the screen, which is now the true one.

Also

• The requirement line counts a colony ship already on its way, the way the game does: the slot it has reserved no longer looks empty.
• The Research sheet's colonization card could name a Command Center level that cannot be built (level 33) once you reached ten colonies. It now says plainly that no Command Center allows more.

1.122.32 — 2026-09-07

Fixed

• The Wraiths can finish a day's contracts again (#2954). Since the thirty Wraith empires reached Command Center 20 they had earned no Daily Wrap and no Weekly Chest at all — not one, on any day, for any of them. Their board had grown from three cards to six and the bar for finishing the day had risen from two to four, while the number of things they could actually do stood still at three. One of the cards they were dealt every single day asked them to harvest a debris field, and they had never once done it: they owned the Recycler on all 123 of their colonies, and a Freighter on two. They now buy the one hauler they need and go and lift the wreck over their own worlds when the day's board asks for it.

  This makes their day finishable, not routinely finished. Measured on the four days since they crossed the threshold, the Wrap would have fired on eight of 120 empire-days, against roughly a quarter of days before it — because the harvest card only appears as a real opportunity on the days somebody has attacked one of their worlds and left wreckage (47% of days), and because two of their other cards got four times harder on the same day the bar rose. Whether they should get a fifth workable card, or a shorter board, is a balance decision that is deliberately still open.

  Nothing about the faction's conduct changes: they still attack only when their operator tells them to. Lifting salvage out of one's own orbit is not an attack, it takes nothing from anybody — a debris field over a living world can be collected by no one but its owner — and it is not a route to one.

Changed

• The record of what the machines can and cannot do now says how it was measured (#2954). The note in the engine that lists which daily contracts an unattended empire really finishes had two errors: it said the board "did not get harder" when an empire is promoted, which is false — the cards themselves scale, so at the top rank a card asking for 3 ships asks for 12 — and it priced the harvest cure at three times its real worth by assuming a card an empire can only sometimes reach is one it can always reach. Both are corrected against production, with the windows the numbers were taken over written down next to them.

1.122.31 — 2026-09-07

Fixed

• A response carrying a renewed sign-in key is no longer stored by the browser at all (#3100). The previous release fixed the "Session expired" loop that met anyone signing in as a new account in a browser that remembered an old one, and that half holds. Its second half asked the browser to keep cached answers apart per signed-in key — and measured against a real browser, that request is not honoured: Chromium keeps one stored answer per address and reuses it across the very request meant to separate them, so the earlier account's renewed key still came back on a "not modified" reply. Chromium is also what the Android app runs inside. So the game stops asking and states it instead: a response carrying a renewed key is now marked as one the browser must not keep, and that key can no longer travel on a "not modified" reply either. On the web you will see no difference — the in-app half was already doing this work. On the installed Android and iOS app, which keeps its older in-app code until its next build, this is the half that now does the protecting; the previous release's note said the server half already covered it, and that was not true of the app.

Also

• Ordinary requests are untouched, which matters more than it sounds: nearly two thirds of everything the game asks the server for is answered "nothing has changed", and that stays exactly as it was. Only the handful of responses that actually carry a renewed key opt out of the browser's cache.
• Three internal checks around session keys were tightened at the same time. A renewed key is now checked against the account that asked for it in the copy the app holds in memory, not only the one saved to your device; the recovery door can no longer hand back a key belonging to a different account; and a renewed key that arrives damaged — truncated or rewritten in transit — is refused rather than saved over the good one, which would have ended the session at the next restart.
• The same check was also made less eager. Written as it first was, it would have turned down a legitimate replacement key whenever the key it replaced could not be read — which is how a player using Sign out other devices could have lost their session instead of renewing it. It now refuses only a key that provably belongs to somebody else, or one it cannot read at all.

1.122.30 — 2026-09-07

Fixed

Synthetic empires stop stripping their own colonies to guard relics (#2929)

An agent that leaves a fleet standing guard on a relic is supposed to leave exactly one: the rule has always been one garrison at a time, so that nobody empties their home hangar holding four monuments at once.

That rule had never once applied. The agents checked whether a fleet was still on guard by looking for a status the server does not use, so a garrison disappeared from their own count the moment it arrived — and the rule reopened for the whole twelve hours it stood there. The agents kept posting more. At the time of this fix ten synthetic empires were holding more garrisons than the rule allows, one of them six.

The same blind spot told them their fleet was at home when it was not. An agent with every military hull parked on a relic still read its fleet as fully available, so it would launch an attack on a defended relic with nothing behind it — the one thing that check exists to prevent. Seven empires were in that state.

What you will see: synthetic empires now keep one garrison at a time and hold back from relic attacks while their fleet is genuinely away. Guild opponents will find their relics defended more consistently, and their raids joined a little more cautiously, because an agent's estimate of its own strength is no longer inflated by the fleets it had forgotten it had sent.

Also

• The agents' income estimate had a branch preferring a figure the server has never sent; it could never run and has been removed. No number changes.
• A new check keeps the mission vocabulary in one place, so a status that does not exist cannot be matched against again.

1.122.29 — 2026-09-07

Fixed

• Empires ask the guild for help with the work they actually care about (#2934). When a synthetic empire posts a request for a hand on the guild board, the queue it asks about used to be simply whichever one had the most time left on it. Nothing about the empire entered the choice, so empires with queues of the same shape all reached for the same thing — and a recent fix, which finally let them see their own research queue, was about to point nearly all of them at research at once, since a research order is usually the longest thing an empire has running.

  Now the empire's character picks: a warmonger asks for a hand on its shipyard, a turtler on its construction, a scientist on its research. It is the same instinct they already followed when spending a Chrono fragment on their own queues. A request still names a real order in a real queue, and an empire still only asks when the wait is long enough for the help to be worth something.

Also

• What this actually changes on the board today: very little, and it is worth being straight about it. Measured against the real queues of all seventy-six synthetic empires: only nine of them currently have any queue long enough to be worth asking about, and eight of those nine have exactly one — so there is nothing for character to choose between, and the old rule and the new one give the same answer. On today's state this moves at most one empire's request. The change is worth having because it stops a convergence that was about to start, not because it undoes one players had already seen.

• Separately, and not fixed here: synthetic empires do not currently open speed-up requests at all — every one of their guild asks is a request for resources, where a real player's are mostly requests for a hand on a queue. That is being tracked on its own.

1.122.28 — 2026-09-07

Fixed

• Signing in as a new account in a browser that remembered an old one looped on "Session expired" (#3100). The empire opened, then within the same second the header, the campaign panel, chat and the crono panel failed, every action answered "Session expired: sign in again", and thirty seconds later the game logged out. Signing in again repeated it exactly. The browser's cache was replaying the previous account's renewed key with a 304 Not Modified, and the game adopted it. Now a renewed key is only ever accepted when it names the same player as the one who made the request, and the server tells the browser to keep cached answers per signed-in key. The new empire opens and stays open; nobody can be handed another account's key by their own cache any more.

Also

• If you used a shared browser for a second account before this release and it opens an empire that is not yours, sign out and sign in again: signing out clears the stray key. Sign out other devices in your profile retires any copy left in another browser.
• The installed iOS and Android app receives the in-app half of this fix with its next build; the server half applies to it from this release on.

1.122.27 — 2026-09-07

Fixed

• The ally-under-attack cards no longer breathe with the countdown (#3095). On a guildmate's threat row the two picture cards resized a little every second, because the column between them was sized by the clock and «6m 16s» is wider than «6m 9s». The column now has a fixed width and the clock keeps the width of its widest reading, so the seconds tick inside one box and nothing around it moves.

1.122.26 — 2026-09-07

Fixed

• AI empires no longer put up the Peace shield as a reflex (#2944). A player reported it from the outside: the shield "seems to be on automatically whenever they get hit". It was — losing a home defence while on a defensive plan was the entire condition, so a synthetic empire bought cover the instant it was hit, at any hour of the night, and went on buying the repeats at double and quadruple price without ever weighing them. Two things changed. The sleep model that already governs when an AI empire chats, chooses a new homeworld and reacts to a battle now also governs its guild shopping, so an empire attacked while its commander is "asleep" takes the hit. And the shield's escalating price — 40 Scrip, then 80, then 160 — is now a cost those empires actually feel: the first shield of the week is as automatic as before, the second and third are not. Replaying three weeks of real attacks through the new decision, roughly half the AI shields that used to go up no longer do, and the ones that disappear are the repeats. What you will notice: AI-held colonies stay attackable for longer, particularly when you come back to one you have already hit this week.

• AI empires stop promising attacks their own shield forbids (#2944). Buying the Peace shield silently blocks every attack that empire can launch for eight hours. The same lost battle that triggered the purchase is what prompts them to say what happens next — so they would vow a counter-strike they had already made impossible, with no way to explain it, because nothing told them they were shielded. They now know when their own shield is up and what it costs them, and can say so.

Changed

• AI empires no longer shop at the guild emporium while asleep (#2944). This covers the Emergency crate as well as the shield: someone who is not at the keyboard does not restock either. Their donations to the guild Fund are deliberately untouched and run as before.

1.122.25 — 2026-09-07

Fixed

• The synthetic empires now judge a wreck by the Recycler they actually own (#2977). Since v1.119.0 the agents have been collecting the debris left over their own worlds. When one of them looked at a field and asked "is this worth the trip?", it did the sum as though it owned the smallest Recycler in the game — no matter what it had really built — while the fleet it then sent, and the salvage the game paid out on arrival, both used its real plant. The go/no-go was therefore always the pessimistic one, and wrecks an empire could genuinely have lifted were turned down for ever.

  Measured across the 76 live agent empires, which hold Recyclers at levels 7 to 14, the estimate ran 15.6% to 24.9% low — 20.4% at the most common level-10 plant. In practice it cost them the tail of a field rather than rich hauls: of the wrecks currently in reach, one in 24 was being refused, and one more empire out of twenty would have launched. Both halves of the decision now read the same number.

  One more thing came out of it. When an agent has several worlds under the same wreck, the estimate is what picks which of them launches — and at the old price two hangars of different sizes could look identical, so the fleet sometimes left from the small one and came back with less than it could have carried. At the corrected price the salvage goes out from the hangar that can actually bring the whole haul home.

  Nothing changes for the salvage human players compete for. Agents still only lift the debris over their own colonies — which nobody else is allowed to harvest anyway — and go nowhere near the neutral orbits where people collect. The pacing added in v1.119.0, so that no agent works a field like clockwork, is untouched and is now held in place by tests at the corrected price.

1.122.24 — 2026-09-07

Changed

• The Recent destinations row now has two halves (#2947). On the Missions screen, the row of shortcut chips used to mix your own colonies with every planet you had scouted, fought over or sent a fleet to, in one undivided block. Your colonies came first, but the row wraps to several lines on a phone, and once it wraps an ordering stops reading as an ordering. It is now split under two captions — YOUR COLONIES and RECENT — so you can find your own planets for a transport run without picking them out of the targets. A half with nothing in it prints nothing, so if you own a single colony the row looks exactly as it always did. Nothing was added, dropped or reordered: the same chips, grouped.

Fixed

• The row now says when your colonies are spending it (#2947). It holds twelve chips and always has, and your own colonies spend those twelve alongside everyone else's planets — so founding a colony could push a scouted target off the row, and nothing on screen said so. Expand far enough and new contacts simply stopped appearing, with no explanation anywhere. When the row is full and some of those chips are your own colonies, the legend now says so and names the budget the two halves share. The pin counter had always been on screen; the row's own width never was. A full row with none of your colonies in it says nothing new — nothing of yours is displacing anything there — so that case is left exactly as it was.

Also

• We did not give each half its own cap, which is what the report suggested. It was measured against the live game first, and it turns out to return nothing: the row is a fixed twelve, so capping your colonies hands a target slot back only by taking a colony chip away — one for one, at every value we tried. It would not have widened anyone's reach; it would only have decided, for you, which chip you lose. So the row keeps its twelve and now tells you about them. The measurement is written down in the code, next to the number, for whoever reaches for the same idea next.

Reported in #suggestions by jaybee1019 / Usherin.

1.122.23 — 2026-09-07

Fixed

A synthetic empire's evacuated fleet now comes home when the raid is over (#2983)

When an attack it cannot win comes at a synthetic empire's capital, that empire evacuates: the whole home fleet lifts off and garrisons another colony rather than dying on the ground. What it could not do was come back. The fleet sat on the outpost until its watch simply ran out — hours after the raid it fled had landed — while the capital it belongs to stood empty.

They now fly home once the sky over the capital is clear, in the case where the fleet took shelter on another colony of the same empire. The rule is deliberately cautious in one direction: a fleet only returns when home is safe without it, so a commander never calls its ships back into the strike they escaped.

What this does not cover, and the honest size of it. A fleet that sheltered on an ally's colony still waits out its full watch: ending a watch on somebody else's ground tells that ally their cover has gone, which is a promise between two players and needs its own design. That is the large majority of the cases — since the evacuation behaviour arrived on 4 September, 625 of 656 sheltering fleets went to an ally and 31 to a colony of their own. So the population this release actually frees is those 31, which had been sitting a median of 1h05m past the last blow to fall on their home. An earlier version of this entry quoted 637 fleets and a median of 2h40m; that counted both groups together and so promised roughly twenty times what the change delivers.

Agents no longer describe a garrison's homecoming as a journey that never happened (#2983)

A fleet sent home from the ground and a fleet turned around in mid-flight looked identical on an agent's own record, so an empire that had stood its garrison down would describe it as having "turned back before it arrived" — contradicting, two lines later, its own account of the watch it had just stood. Both sentences are now true of what actually happened.

1.122.22 — 2026-09-07

Changed

• Nothing a player can see changes in this release. It is maintenance on the tests that guard space combat: a batch of them could no longer tell a real fleet from one the game had stopped recognising, and they now can (#3080).

  The combat benches described their fleets in a way that let any ship name through, valid or not. If a hull were ever renamed in the engine, dozens of these tests would have gone on passing while quietly fighting battles with nothing on the field — the check that should have caught it had been switched off at each assertion. Measured on the old tests: renaming one hull left all eight rows of one guard green, including the row naming the hull that no longer existed.

  116 of those switched-off checks are gone, across 26 test files. A renamed hull is now caught before the game can even be built. The ones that remain are the handful that are deliberate — tests that feed the engine deliberately broken data to prove it copes — and each now says so at the spot, so the next person does not have to work it out again.

  The engine itself is untouched: the shipped JavaScript is byte-for-byte identical to the previous release, which was verified by rebuilding both and comparing them.

1.122.21 — 2026-09-07

Fixed

• An empire whose deletion was called off now gets its production back (#3070). The lifecycle sweep has a safety brake: when a single round finds an implausible number of accounts eligible for removal, it deletes none of them, because a population does not go quiet all at once — a broken presence clock does. The brake worked, but it left the accounts it had just saved in limbo: they were deleted by nobody and restarted by nobody, so an empire whose production had been halted stayed halted for as long as the fault lasted, and signing back in did not lift it. Now, when the brake stops a round's removals, the accounts it saved have their production halt lifted in that same round. Nothing is replayed when it comes back: the frozen days produce nothing, no lost defences are refunded, and no losses are charged twice.

Also

• No thresholds changed. Who becomes eligible for removal, and when, is exactly what it was — this only restores a restart that was already supposed to happen.
• Measured against the live game before shipping: no account is currently in this state, and the presence clock is healthy. This closes a fault that would only surface during an outage, before it costs anyone an empire.

1.122.20 — 2026-09-07

Fixed

• Ending a leave early now really ends it (#3071). Cancelling a declared leave was supposed to hand you back a normal empire straight away, and on your own screen it did — production restarted the moment you pressed the button. Behind that, the game went on treating you as still away until the day your leave would have finished, for up to three weeks. For a player who came back and kept playing this changed nothing at all. For an empire whose owner cancelled a leave and then stopped coming, it meant the game kept insisting they were fine: their planet stayed marked as a live target on the board and the planet sheet, so attacking it still paid full infamy, and the clocks that eventually free an abandoned account stayed frozen. Cancelling now counts from the moment you cancelled, which is the moment you were demonstrably back. A leave you simply let run out is unchanged — the whole span is still excused, exactly as before.

Also

• Measured against the live galaxy before shipping: of the twelve accounts that have ever declared a leave, one abandoned empire changes state today — it stops paying a bounty to whoever attacks it. Nobody becomes eligible for removal, no garrison is charged for the released days, and no active player is affected.

1.122.19 — 2026-09-07

Changed

• The raid fleet picker is a hangar (#3083). Behind Open a raid, Join and Send defence, the picker was a row of bare number boxes under bare names. Now every hull is a card with its Shipyard sprite, the name, what the pad holds, the field, and the War Table's caption; a card with something aboard lights its border, and «Aboard N ships» reads under the grid. Launch pads are chips, the chosen one lit gold. The target address is three sunken cells, and once all three are typed the coordinate reads back beside them in the target's colour, or unlit in the warning colour when the address is one the game refuses. Every count and every sentence the picker used to say is unchanged.

1.122.18 — 2026-09-07

Nothing you can see in the game changed. This one is about the net under the combat engine.

Changed

• Three combat rules that were live but almost unwatched now have real cover (#3062). A mutation census — switch off one engine rule at a time and see whether anything goes red — found three rules of the apex hulls whose entire coverage was a single assertion, each one narrower than the rule it protected. The Spectral Bulwark's cap on the orbital cannon's opening shot was guarded only by "the reported figure is a little smaller than the shot fired"; the Praxis Interdictor's whole battlefield effect rested on one "the rear takes more fire than before", which any leak at all satisfied; and the Interdictor's precedence over the Deep Formation keystone had no battle-level guard whatsoever. All three are now pinned by what they actually do — how much hull the opening shot may take, that the enemy rear takes exactly four times the fire and that the vanguard pays for it, that a Deep Formation defender buys nothing back from an Interdictor, and that the sheltered row dies a round sooner for it.
• The cannon cap is now pinned by the battle it exists to decide. Its comment claimed the rule stops a long chip-damage siege flipping a result, and nothing tested that. A coalition siege now holds the line: with the rule the attacker takes the field and brings the flagship home, without it the same battle is a defender win with the flagship destroyed.
• Two Aegis Bastion benches that read like coverage without being any (#3062). One fired a salvo that pressed 0.647 of the shield budget it checked against; the other pressed 0.347, and read the identical number whether the engine spent the shared pool or left it intact — so what it constrained was the attacker's throughput, wearing the shield budget's name. Both fixtures now saturate, and the second is exact: the tender soaks the pool it is and not a point more, and the escort is left with its own shields because the pool was already spent.

Also

• The census itself was regenerated rather than inherited, and it had moved: two of the twenty-six rules no longer exist — a retreat change deleted them — including one of the four the report named. That row is recorded as unclosable rather than quietly dropped, and two rules that have since become single-assertion are named for the next pass.
• Several comments in and around the combat engine were corrected against measurement rather than memory: a test whose name claimed the opposite of its own assertion, a window described as a constant it is not, a conservation figure off by one, and — twice, in review — claims this work itself introduced. No constant, no engine line and no player-facing string was touched, so no behaviour moves.

1.122.17 — 2026-09-07

Fixed

• Five stale numbers in the combat engine's own notes, measured rather than guessed (#3066). Nothing a player sees changed — this is the map the engine keeps of itself, and five of its landmarks were pointing at the wrong place. A previous pass (#3057) had found them and deliberately refused to invent replacements; this one measured all five against the running engine.

  The bastion's sensor grid was said to cost its wall 13-20% of its strength; it actually costs 16% to 47%, depending on the step, and the reason the faction gate still works turned out to be a different reason than the one written down. Two figures describing the same siege — a 95% and a 96% — were being read as one number measured twice; they are two different measurements, of damage dealt and of hull destroyed, and they differ by exactly the shields. A count of "18 of the 21" hulls that could shelter behind a legendary flagship was wrong three ways over, and has been replaced by the rule it was standing in for. The cost of grinding down a Spectral Bulwark named only the Destroyer's figure, and now names the Annihilator's too — plus a warning that both numbers are a floor that rises with research.

  The fifth turned out to be right all along: the sector ceiling really is where the comment said it was, and now cites the constant instead of copying its value.

Also

• Every measured figure in this pass carries the date it was taken and an instruction not to inherit it, and two figures that describe deleted code are now marked as history rather than left to look current.

1.122.16 — 2026-09-07

Fixed

• Operators: the admin lifecycle card could see only one of the two switches that stop an account removal (#3069). The second — the volume brake added in v1.122.5, which makes a round that finds more than a hundred removable accounts remove none of them — appeared nowhere on screen, so the card kept promising "up to 25 per run, every 6h" over a queue nothing was draining, and an operator who had lifted the brake for a backlog and forgotten had no trace of it anywhere but a boot line long gone from the container log. The card now states the brake in every state it can be in: quietly armed with the line it trips at, refusing above that line, lifted with the warning that removals are armed, and — when the API is a deploy behind and the card cannot read the brake at all — saying exactly that instead of drawing the healthy view. Where the brake is holding the queue, the "Removable" tile stops quoting a drain rate that is not happening.

Also

• Nothing here is visible to players: this is the operator's dashboard. No thresholds were retuned, and the card reads the sweep's own constant and the same off-switch parser the sweep uses, so the panel cannot reach its own verdict about a lever the procedure is also reading.
• Two checks came out of the review rather than the issue: the card no longer blames the brake for a stop the master switch is causing (which would have sent an operator to the one lever that arms a deletion), and a lifted brake is now stated even while the sweep is switched off — the moment before someone switches it back on.

1.122.15 — 2026-09-07

Changed

• Joint raids read like a war room (#3073). The Guild's raids panel wore the app's flat glass panels; now it wears the register of a mobile 4X war screen. A guildmate under attack is a metallic card with gold corner brackets that glows red while the strike is inbound and goes quiet once it has landed. The two sides of that card are pictures: the hostile armada behind what is coming, the shielded world behind what stands against it, with the ship count as the biggest thing on each. A gold VS medallion sits between them above the countdown, which is now a clock rather than a caption. Every hull chip shows the hull's own sprite, garrisons open with the guildmate's portrait, and a raid's pack is an overlapping row of avatars over a labelled fleet bar. The actions are bevelled buttons with a verb on them: green to send a defence or join, gold to open a raid, grey to cancel. Section titles are gold ribbons, red for «Allies under attack». Every sentence the panel used to say, and every number, is unchanged.

1.122.14 — 2026-09-07

Fixed

• The Shard Chandlery now says who hands over what (#2990) — the shard shop's signpost read "The Anchorage's Shard Chandlery, on the Market tab, sells shards for Dark Matter." That is true of the shop, which takes your Dark Matter and hands you shards. But a player reads himself as the subject and hears the opposite, and one asked in universe-chat whether he could sell shards for Dark Matter. He could not, and never could: it is the one conversion the game is built to prevent. The signpost now reads "…lets you buy shards with Dark Matter. The trade runs one way: nothing turns shards back into Dark Matter." You are named as the buyer, and the closed direction is stated outright instead of left for you to infer.

• The same wording is gone from the Chandlery's own card, where it mattered most. The card's subtitle read "Chrono Shards for Dark Matter", and the weekly-pool note said a piece "only sells if it fits the time you have left" — putting your own shards on the selling side, one line under "Bought shards". They now read "Buy Chrono Shards with Dark Matter" and "you can only buy a piece that fits the time you have left." The card also carries the one-way rule itself: the player who wants to sell walks up to that counter looking for a Sell button, and it was the one surface with no sentence explaining why there isn't one.

• The Inventory's empty state — the first screen a player who holds nothing sees — said Chrono Shards could be bought at the Anchorage "for" Dark Matter. It says "with" now, like everywhere else.

Nothing about how the shop works has changed: the prices, the shared weekly pool, and the refusal when that pool is full are all exactly as they were.

Also

• Two long footnotes — the Inventory's shard section and the shipyard speed-up panel — now use the same line spacing as every other block of prose in the game, instead of being set tighter than the rest.
• This wording was first written to close #1306, whose reporter looked at the two shard counters and guessed "one trades shards and maybe the other trades dark matter." Fixing that confusion produced its mirror image, so the new sentence is written and tested against both misreadings rather than trading one for the other — and the two signposts keep their separate directions, so a player already standing in the Market tab is still not sent to it.

1.122.13 — 2026-09-07

Changed

The guild's "Allies under attack" note is shorter, and no longer credits a building you may not own (#2995)

The note under the Allies under attack board — the one that explains what sending a garrison to a guildmate costs you — was five long sentences. It is now four short ones, in the words a player proposed for it back in August.

Before:

Their observatory sees these fleets coming. Send yours to stand guard: it fights alongside them until the watch ends, then flies home. You can call it back any time a strike is not already on final approach — it still has to fly home. The ships stay yours: a garrison is never handed over. They do fight, though, so they can be lost there like any fleet of your own.

Now:

Your allies see these fleets coming — send your own to stand guard and fight alongside them. You may call your fleet back early at any time, so long as a strike is not already on final approach — it still has to fly home. Your fleet flies home once the watch ends. Ships destroyed in battle are lost there like any fleet of your own.

Every rule the old note carried is still there: you can recall the fleet unless a strike is already on final approach, a recalled fleet still has to fly home, the fleet flies home once the watch ends, and ships that die there are gone.

One word in it is not the player's and is worth knowing about: the note says the fleet flies home once the watch ends, not when. A garrison whose watch expires while a strike on the host is still resolving is held on station until that battle clears — up to an hour past the bell — and only then starts the flight home. "Once" is true of both cases; "when" would have promised a departure the game does not always make.

Fixed

The board no longer tells you an Observatory is what spotted the attack (#2995)

The old opening — "Their observatory sees these fleets coming" — was not true. Knowing that an attack is coming, on which colony, and how soon, is free and belongs to every empire; an ally who has never built an Observatory still shows up on this board with the colony and the countdown. What an Observatory buys is the detail: where the fleet came from, who is flying it, and which hulls. The panel already says that, in its own box just above, so the sentence was both wrong and a second copy of one already on screen.

Also

The suggestion came from Remy, in August. It was one of three items on a single issue; the other two were settled and that issue was closed, so this one was neither shipped nor turned down — it simply stopped, and he never heard back. That was our mistake, and the wording going out is his.

Three of his clauses had to be corrected against the engine rather than shipped as written, and each is recorded in the code where the next person will find it: the Observatory line above; a recalled fleet still having to fly home (leaving that out is a bug this note has had before); and "lost for good", which is not true of the Destroyer and the Annihilator — those are wrecked rather than destroyed, and can be recovered from a dry dock.

1.122.12 — 2026-09-07

Changed

Nothing in this release changes what you see or how anything plays. It is a correction to the notes the developers keep inside the code about how bastions, relics and combat are tuned — seven figures that had gone out of date while the game moved on, three of which two earlier changes had already reasoned from (#3057).

The numbers themselves were never wrong in the game: a faction bastion has followed the strongest fleet on the server since the change that made it fightable, a relic guardian is sized at the value it was re-measured to when battles stopped ending on a six-round clock, and both behave today exactly as they did yesterday. What was wrong was the written explanation next to them, which still described an older game — and in this repository those notes are what the next change gets planned from.

Also corrected: a table of old combat measurements now says at its head that it was taken under the six-round clock and should not be quoted, instead of saying so thirteen lines further down, which is how it came to be quoted in the first place.

Also

The rule the corrected notes adopt is to name a constant rather than copy its value, and to date any figure that can only be measured — so the next reader lands on the live number instead of a copy of an old one (#3057).

1.122.11 — 2026-09-07

Changed

• Nothing changes in your battles, and that is the point (#3019). The Aegis Bastion has a rule about how it lends its shields: a Bastion under fire draws on the fleet's shared pool rather than on its own hulls, and what it soaks is taken out of that pool so it cannot shield itself and its escorts with the same shields twice. That rule was live in every battle and no test in the game could tell whether it was still there — it could have been deleted and all 4,957 engine tests would still have reported success, while the outcome of real fights moved a long way. In one measured case, 300 Cruisers behind 10 Bastions went from losing every ship to keeping 272 of them and winning the battle.

  The engine is untouched: the same fleets fight to the same result as yesterday. What is new is that the rule now has a net under it, so nobody can quietly remove it by accident.

• A second Bastion rule turned out to be just as exposed, and is now covered too (#3019). The tender's shields are naval technology, and they are not supposed to reach the Postazioni or the Fortezza — those are works of the planet. That boundary was also invisible to every test: removing it let a garrison's Bastions pour their pool into the static defences, taking what a besieged colony's batteries absorb in one round from 2,400 to over 1.1 million. Again, no behaviour changes today; the rule is simply guarded now, so a siege cannot silently stop working the way it does.

Also

• The catalogue card, the shield numbers and the prices are all exactly as they were. This release adds tests and corrects comments — there is no balance change in it, and nothing to re-plan around.

1.122.10 — 2026-09-07

Fixed

The Inbound panel no longer promises it sees every attack (#3049)

The panel you open to check whether you are under attack used to open with "You always see that an attack is coming, on which colony and when."

It does not, and two of the game's own systems are the reason. A fleet flying under the Nightfall Protocol is absent from the board for its whole dark window, until your Sensor network is strong enough to pierce it. And an attack launched at a colony the attacker could not see — one under the Eschaton Veil — is never stamped with your name at all, so it never reaches the board for its entire flight, while it can still arrive and fight. Over the last seven days that was about one attack in a hundred, every one of them aimed at a real colony.

Both of those are features you can buy, sell and fly against. The promise that they did not exist was the bug, and it was shown to almost everyone: the notice carrying it appears to any commander without an Observatory.

What you see now. The notice keeps the part that was always true, said about a row rather than about the board: every attack that is on the board tells you that it is coming, on which colony and when, with no Observatory needed — and the Observatory still buys who is coming and with what, one step per level. Underneath the board there is now a standing line saying what the board is and is not:

This board covers every colony you hold — not one in the air but off the board, not ships in dock, not a neighbour's strength.

It says the same thing every time you look, whether the sky is empty or full. That is deliberate: a warning that appeared only when something was hidden would itself be the tell, and would undo the very protocols it is describing. It is the same sentence-ending the build advisor already carries under its ranking, now written in one place so the two can never drift apart.

Nothing about how attacks are detected, cloaked or resolved has changed — only what the game tells you about it.

1.122.9 — 2026-09-07

Fixed

• A loot train being shot can now end an attack (#3026). You raid a colony with freighters behind your warships. The enemy shoots the fattest thing on the field — the freighters — and they die one salvo at a time while your warships stand there barely scratched. Until now the fleet never broke off, because the rule that decides "this attack has failed, go home" was watching only the warships and could not see the hold burning. You came home with the escort intact and the cargo gone.

  The retreat now watches both the fighting arm and the fleet as a whole, and breaks off as soon as either has lost two thirds. Whichever half the enemy chooses to shoot, the meter moves with it. In the three battles reported on the issue, 634 Freighters that used to burn to the last hull now bring 164 home, 710 Colony Ships bring back 202 where they used to bring back none, and a raid that ended in a total wipe now comes home with 180 Haulers and its Bombers.

  This is the other half of the change that shipped this morning in v1.121.12, which had moved the retreat onto the warships alone and — measured and reported at the time — made fourteen battles in four hundred worse in exactly this way. Those fourteen are what this fixes.

Also

• The retreat is a loss-bounding rule, so widening it costs something and the number is here rather than left for a player to find: over 1,200 sample battles this is better in 77 and worse in none, and a second, deliberately cargo-heavy sweep of 1,500 found five raids that used to be ground out to a win and are now broken off instead. In every one of those five the attacker comes home with between 1.6 and 5.2 times the fleet and gives up the loot. That is the older behaviour restored, not a new harm.
• Roughly one attack in five brings cargo, so this reaches a good deal more than the battles that have already been lost to it.

1.122.8 — 2026-09-07

Fixed

• A faction bastion is no longer called "a fair target", in green (#3042). Tapping a bastion or a relic guardian on the map opened a panel that graded it against your Power and, on the most dangerous targets in the game, printed "Within your Power band — a fair target." in the colour that everywhere else means an even match. It is not a comparison that can be made: a bastion's and a relic's Power comes from its fleet, while an empire's comes mostly from what it has built, so the two numbers do not measure the same thing. The map stopped doing this last week and the panel one tap away did not, which left the game saying two different things about one colony.

  These targets now get no Power band at all. In its place the panel names the thing — Bastion, Relic — in the same orange the map and the target list already use for it, prints its Power figure, which this panel had never shown, and says what actually decides the trip: its Power is set by its fleet, that figure leaves its walls out entirely, it fights from behind shields and gun batteries, and attacking it puts no bounty on your head. Forty-nine colonies across forty-five systems read differently as a result.

  A real player's colony is untouched: the band, its colour and its price are exactly as before, and a silent empire still keeps its band.

1.122.7 — 2026-09-07

Fixed

• The War verdict card now says who won (#3034). A war that ended with your guild ahead used to reach your Activity feed as "The war ended 149,469 to 251,696: the losers pay the tribute" — two numbers with no names on them and a definite article pointing at nobody. You could not tell which score was yours, and so could not tell whether that pile of Ferrolite was arriving or leaving. The card now names the winner and puts their score first: "Stellar Gamesense won the war 251,696 to 149,469: Scourge of the Stars pays the tribute (…)." A war that ends in a draw names the two guilds that fought it instead of just announcing a draw. Wars that ended by surrender already named the guild that gave up and read the same as before. Nothing new is fetched and nothing was rewritten in the database — the winner had been recorded on every decided verdict ever published and simply was not being shown, so the wars already sitting in your feed read the new way too.

• Reparations now say who paid whom (#3034). The same row went to both guilds saying "Tension resets to zero: 1,200 Fe · 800 Xe · 400 Dt paid. 48-hour truce." — so the guild that handed the resources over and the guild that received them were told exactly the same thing, with nothing to say which was which. It now reads "… paid by Scourge of the Stars to Stellar Gamesense."

Also

• Where a guild has since disbanded and its name can no longer be looked up, these rows say "The war was won 251,696 to 149,469: the losing guild pays the tribute" rather than guessing a name — the score order still tells you the result.

1.122.6 — 2026-09-07

Fixed

• The build advisor gives you advice again, instead of the same warning every time (#3033). SUGGESTED NEXT has one sentence at the top of the panel, and it had been spending it on a safety caveat that read the same on every quiet colony — "the quiet skies are no guarantee of safety". Quiet is not an unusual state: measured over a week, better than 99% of colony-time has nothing inbound, so that sentence was very nearly the only thing the advisor ever said. A warning that is on screen every single time cannot tell you anything, because there is no moment when it is missing.

  The advisor's headline is now about your colony and what to build next. The caveat it was crowding out has not been dropped — it has become a fixed line under the panel, saying what the ranking above it does account for and what it does not: "This ranking sees an attack only once it reaches your Inbound board for this colony — not one in the air but off the board, not ships in dock, not a neighbour's strength."

  Nothing about the previous fix (#2822) was undone. The advisor still never tells you the sky is clear, and it still cannot: what it knows about attacks is one board, and that board can miss things. What changed is that a colony with an attack inbound is still told so plainly, while a colony with nothing inbound now gets a sentence about its economy instead of a disclaimer.

  Reported by TartarusMkII, who also reported #2822 and then came back to report its fix — with the question that named the problem: when would this text ever change?

1.122.5 — 2026-09-07

Fixed

• An abandoned empire is now removed on its owner's clock — being attacked no longer keeps it alive (#2830). A guest who stopped playing was supposed to be retired thirty days after their last sign-in (sixty for a linked account), freeing their name and their planet slot. In practice every raid on their colony reset that clock, so a derelict that other players kept farming stayed on the map for ever: 33 abandoned guest accounts, 23 of them holding a slot in the cradle, were being kept alive this way. The removal now counts from the owner's last sign-in alone; what happens to the colony no longer extends it. The ruin — production halted, garrison decaying — still waits for the colony itself to fall quiet, so an empire that is being fought over is not halted, and a farmed derelict goes from silent straight to removed. The thirty-day threshold itself is unchanged.

Changed

• The admin dashboard's lifecycle card now says which clock each tile reads: «Silent» and «Removable» on the owner's, «In ruin» on the colony's.
• Operators: the removal now rests on the presence stamp alone, so it is guarded — a round that finds more than a hundred removable accounts removes none of them and says why in the log (lever INACTIVE_LIFECYCLE_VOLUME_BRAKE, see deploy/DEPLOY.md), and every deletion record keeps the presence clock it was decided on (AccountDeletion.lastSeenAt, a new nullable column applied at boot), so a wrongly deleted account is one query away.

1.122.4 — 2026-09-07

Fixed

• The Aegis Bastion's card said your fleet was too big for it. That was never true, and it now says what actually decides (#3030). The shipyard card read "Never fires; a large fleet cannot use its shields", and the raid door said the same. A commander with a large fleet read that as this hull cannot work for me — and four commanders holding 80,524 Bastions between them read it before deciding what to do about that purchase.

  The engine has never worked that way. Your ships draw on the Bastion pool only for the part of an enemy salvo their own shields could not absorb — and the size of that salvo grows with the enemy just as your own shields grow with your fleet. What decides whether the pool is reached is the comparison between those two, not how big you are. The old sentence came from a test that grew one fleet while holding the enemy still, until the fleet was so far ahead it was losing fourteen ships out of 73,400 — 0.019% of itself. There was nothing left there for any defensive hull to do.

  Scale the whole battle, the way a fight between two grown empires actually looks, and the Bastions save the same share of your fleet at every size: measured on four seeds across a 64× range, 102,400 hulls behind 102,400 Bastions keep the same proportion as 400 behind 400, to within a third of a percentage point.

  The card now reads "Never fires; too few for the fleet can cost ships", and the raid door prints the same clause. Both come from one place in the code now, so the two screens cannot drift apart again — they did twice in eight days.

• And the warning on that card points at the right thing. There is a band where bringing a tender leaves you worse off than bringing none: the absorption buys your fleet one more round, so it breaks off later and crosses the retreat line deeper. That was last described as a small-pool problem. It is not about small numbers — it is about too few for what they are escorting, and it reproduces at every size: 600 light hulls are hurt by 50–100 Bastions, and 9,600 by 800–1,600, the same fractions. Measured across six battles, every one of them is hurt somewhere between a sixteenth and a sixth of the escort, and none at a third or more. Outside that core it depends on the fight, which is why the card says can.

Also

• Nothing about your battles changed. No ship price, no ship stat, no combat constant and no line of the engine was touched. This release changes what the game tells you about a hull, not what the hull does.
• If you have the game open, reload it. The catalogue text ships inside the app, so a page you opened before this release — or an installed app that has not updated — will keep showing the old sentence until it refreshes.
• The design page for the hull (§7.4) carried the strongest version of the false claim and has been rewritten, along with the code comments on five surfaces that had accumulated three different explanations, at least one stale on each. The measurement that retires the claim was already in the test suite since an earlier release; what was missing was anyone correcting the card.

1.122.3 — 2026-09-07

Fixed

The weekly battle contract now says why it pays what it pays (#3031). "Win 3 battles" carried one sentence about the 20,000 cap, and it only ever described that number as a bar to clear — which reads as good news. The same number also holds the reward down: past the cap the payout stops growing with your empire, so a large commander could finish the same tier of work as the two weekly cards beside it and collect a small fraction of what they paid, with nothing on the board explaining the difference. One player claimed all three inside seven seconds and watched two pay 14,796,592 between them while the third paid 30,000.

The card now states the reward side as well as the bar, in the same sentence, scoped so it says nothing about the smaller empires the cap never touches — below it this card pays exactly what its siblings pay, and that has not changed.

Nothing about the payments has changed, and nobody was underpaid. The clamp is deliberate: it is what stops a commander farming the weekly by feeding a fleet to a second account, and the board has always quoted the already-clamped figure honestly. What was missing was the sentence, not the resources. Measured across every claim of this card so far, 67 of them — by 53 different empires — were made above the cap.

And the explanation now survives the claim. Collected cards used to drop their rule, so the one sentence that explained the gap disappeared at the exact moment a commander taps through and compares what the three weekly cards actually paid. A collected battle card keeps it, for as long as the weekly hand keeps the slip.

Also

The open question of whether that clamped payout is the right number — a weekly card that can pay less than 1% of what a daily card pays the same empire — is filed as #3038 with the production figures, and is deliberately not answered by this change.

1.122.2 — 2026-09-07

Fixed

• Faction bastions and relic guardians are no longer coloured like players on the system map (#3014). The map tints an inhabited slot by comparing the owner's Power against yours — green for an even match, grey for far below you, red for much stronger. That comparison was being applied to things that are not players. A faction Bastion and a relic guardian own real colonies and publish a Power reading, so they were coloured by band like any rival; but their Power is almost entirely fleet, while a player's is almost entirely buildings, and the two numbers do not measure the same thing. The result was a Bastion drawn green — the map's colour for "even match" — in front of a garrison many times the size of the fleet you were about to send. They now wear the PvE orange the Pirate Beacon has always had, along with its double ring and its ⚠ name prefix, so they read as what they are before you commit a fleet rather than after.
• The map's legend stopped naming the Beacon as if it were the only PvE target. The orange dot's label read "Beacon (PvE)" while the same orange now marks Bastions and relics too. It reads simply PvE, and the Target Board a tab away still tells you which one you are looking at by name.

Also

• Silent empires were left exactly as they are, and deliberately. An attack on an owner who has gone quiet for seven days costs you no bounty, but the grey they are drawn in only claims they are weaker than you — which is true — and the exemption can end the moment they log back in, while your fleet is still in the air. The map already says "silent" in words on the Target Board and in the planet sheet one tap from the slot, which is where a fact with a condition attached belongs. The reasoning is written down beside the rule so the next person to look at it is arguing with something rather than guessing.

1.122.1 — 2026-09-07

Fixed

• The spy report now says which of those hulls will actually fight (#2989). The list headed FLEET was the target's whole hangar, and the ships in it that neither fire nor tend a shield pool never take the field at all — a player read it as the force that would meet him and caught his own mistake a minute later. It was not his mistake. Across every spy report ever filed on a living commander, 73.1% carried at least one such row, on average a third of the rows in a report, and roughly one report in ten showed a "FLEET" that would not fire a shot. The report now marks those rows grounded and heads the list with how many of the hangar will defend. The list itself does not shrink: those are exactly the ships a captured colony loses to hangar interdiction, which is the one thing a raider with the right counters comes to read. From the top espionage tier the card had been contradicting itself for weeks — the battle line beneath it listed fewer ship types than the fleet above it, with nothing saying why.

• The launch pad says what its one-line summary is counting (#2989). The target panel summarises the same hangar beside the Advisor's forecast, which had been excluding those carriers since it shipped — so the sentence and the number under it described two different things, and because the list is sorted by size, a target holding 30 Freighters and 10 Interceptors led with the freighters. The line is now labelled "hangar" rather than "fleet" and carries the same count: "hangar 30× Freighter, 10× Interceptor … · 12 of 47 defend by default". Nothing was removed from it — the ships that stay grounded are exactly what an interdiction raid brings counters for, and this is where raids are composed.

Nothing was added to what a probe buys: the marks are a public catalogue rule applied to numbers the report already showed, so they appear at every espionage tier and the report itself is unchanged. A commander's defence plan — which can field a freighter wall, or hold a gunship back — stays unreadable at any grade, and both new lines say so.

1.122.0 — 2026-09-07

New

• Synthetic raids now arrive with a convoy (#3027). Measured over three days: nine of every ten strikes the synthetic empires sent at human colonies were guild raids, and not one of their six hundred raid contingents carried a freighter — so a raid of eight wings against a warehouse holding three million came home with thirty-five thousand, the warships' own hold, while the empires kept over a thousand freighters in dock. From now on the empire that opens a raid and every empire that joins it add the freighters their dock can spare, a joiner only when the slower convoy still reaches the raid before it launches. What you will see: synthetic raids that take what the warehouse held, and convoys in their battle lines that die first when the raid is repelled.

1.121.12 — 2026-09-07

Fixed

• Bringing cargo to a fight no longer switches off the retreat rule (#3009). A losing attack is supposed to break off with something left. If you brought freighters, it didn't: the break-off was measured against the value of your whole muster, and since the enemy shoots your warships and not your freighters, the threshold drifted out of reach of the only losses actually happening. The fleet fought to the last hull while the loot train sailed home untouched. Measured: 400 Voidlance Cutters that keep 75 of themselves alone lost all 400 with 2,081 Freighters along, and 500 Aegis Bastions did the same thing. The break-off now measures the fighting fleet, so what you bring to carry the loot no longer decides whether your warships come home.

  In the four weeks the rule has existed this cost 18 attacks their entire escort — 316 warships lost while 407 freighters came home whole. If that was you, it was this.

Changed

• A laden fleet now breaks off where the same fleet alone would have. This is a real change to how battles end, not only a bug fix: attacks that used to end in total loss will now end with survivors, and the forecast on the attack screen will say so once your client updates.
• An escort of Aegis Bastions helps instead of having a cliff in it. The same 400 Cutters keep 133 behind 500 Bastions where 40 Bastions used to mean losing every one of them. Below the hull's operating window a small pool can still cost you ships — that is unchanged and is still what the shipyard line warns about.
• A flagship no longer strands its own loot train. A lone Destroyer that cannot be destroyed used to hold a lost fight open until every freighter behind it was gone; it now breaks off with them.
• Campaign rosters carrying the Aegis Bastion and Continuum Hauler reach slightly less far. Part of what those two hulls were buying was the defect — they padded the number the break-off was measured against without ever being shot at, so a run ground on instead of retreating. With that gone, a maxed apex-support roster clears the same every-seed levels it did, but the marginal levels above them are harder. No ship's stats or costs changed.
• Some attacks you used to win, you will now break off. The ones that change are the ones you were winning by fighting your escort down to nothing under cover of the old rule. Measured: a fleet of 75 Cruisers behind 541 Freighters used to grind out the win and come home with none of the 75; the same fleet without the freighters broke off keeping 11. Both now keep 11 — and the win is gone. The break-off exists to bound your losses, so it is allowed to cost you a victory.

Also

• If the enemy would rather shoot your freighters than your escort, you can now lose the freighters where you used to break off. Ships are picked by counter first and by size second, so a big hold parked behind a small escort is what gets shot — and the new break-off measures your warships, which in that case are not the ones dying. Measured across 400 sample battles the change is better in 30, worse in 14, and identical in 356; the 14 are all this shape. It is tracked as #3026 and the honest fix is a larger one: measure what was actually under fire.

1.121.11 — 2026-09-07

Also

• The synthetic empires' raids now say in their own log why they flew without freighters (#3024). Measured over three days: they won against human warehouses holding three million at the median and took fifteen thousand a win, because forty-two of forty-eight winning strikes carried no freighter while their docks held over a thousand. The escort rule exists; what did not exist was a trace of which of its gates closed. Nothing changes for players today; the next release sizes the raider's hold on the measured reason.

1.121.10 — 2026-09-07

Fixed

• The Aegis Bastion was reported as a hull that stops working once your fleet gets big. It does not, and nothing about your battles changes (#3008). A commander who had bought tens of thousands of Bastions could not find a battle they altered, and the follow-up analysis concluded the hull "pays only for fleets too small to buy it" — that its value falls to zero as the fleet it escorts grows. Re-measured against the engine, that turns out to be an artefact of how it was tested: the sweep grew the escorted fleet while holding the enemy still, and at 73,400 escorted hulls against that fixed enemy the fleet loses thirteen ships — 0.0177% of itself. There is nothing there for any defensive hull to do. Scale the whole battle, the way a real fight between two grown empires actually looks, and the Bastion saves the same proportion at every size: measured against a control of the same number of Continuum Haulers, 90.0% at 800 escorted hulls and 89.9% at 204,800, and 86.7% / 86.6% in a second fleet shape — flat across a 256× range. Run the four real fleets in the game that hold Bastions today and three of the four show large savings in fair fights; the only one that gets nothing is the commander holding ten. The engine was right, so no constant, no formula and no catalogue text changed here, and no battle resolves differently than it did yesterday. What is new is a bench that measures the hull correctly, so the claim cannot be made a fourth time without something failing.

Also

• The new bench compares Bastions against the same number of Haulers rather than against an empty rear row — because "with Bastions vs with nothing" secretly measures several thousand extra hulls for the enemy to shoot at, and in two fleet shapes that was the entire apparent effect (in one of them a Hauler is measurably the better body). Two shapes were cut for exactly that reason, with the numbers that condemned them written down beside them.
• While checking the new tests could actually fail, we found that a different Bastion rule — the one from v1.x that stops a Bastion under fire from shielding itself and its escorts twice over — is guarded by nothing: undo it and all 4,905 engine tests still pass. That is now recorded as its own issue rather than quietly patched here.

1.121.9 — 2026-09-07

Changed

Nothing you can see in the game changed in this release, and that is the honest summary: no Doctrine, no travel time, no research number moved. What changed is three places in the code that looked like they were tuning the game and were not — the kind of thing that turns into a real bug the first time someone tries to rebalance through them (#3010).

• The Dark Matter Veil's travel bonus was written down in three separate places: the travel engine, the effects table, and the Power Details panel that shows you where your bonuses come from. The panel kept its own hand-typed copy of the number. Today all three agree, so the panel has always shown you the truth — but the moment the Veil was retuned, the panel would have quietly kept quoting the old strength while your ships actually flew at the new one. The panel now reads the single real number, and a test checks the panel against the engine at every Doctrine level instead of only at level 20 — which turned out to be the one level where the two agree no matter what, so the old check could not have caught the problem it was written for.
• A free-jump setting that read like a dial ("how many free warp jumps per day") turned nothing at all: the game stores a single timestamp, so one per window is the only thing it can express. Changing the number bought nobody a second jump. It is gone, with a note saying that a second free jump needs a real change to how it is stored, not an edited number.
• The Doctrine keystone levels (5 / 10 / 15 / 20) were listed in a table that nothing actually used — the twelve checks underneath it each had the number typed out again. Moving a keystone now means editing one line instead of thirteen.

Also

New guard: every setting in the Doctrine module must now be genuinely used by the game, or the build says so by name. That is what would have caught the dead free-jump setting the day it was written.

1.121.8 — 2026-09-07

Changed

• Synthetic empires strike their neighbours, at human flight times (#3007). Measured over three days: human raids fly five minutes at the median and won 501 of 502 landing within ten minutes on synthetic colonies, while the synthetic empires flew thirty-five minutes at the median — their scan drew a system at random across the whole galaxy when nothing better was known, their address book barely preferred the colony next door to one fifty cells away, and their pick priced distance in map units rather than in the minutes the target's Observatory would watch. Under ten minutes of flight they won half their strikes; between twenty and forty, one in ten, against garrisons that had half an hour to form. From now on a blind scan with nothing known looks around one of the empire's own colonies half the time, the address book looks next door first, and the pick weighs each target by the strike fleet's own flight time from the nearest of its colonies, halving its odds at ten minutes and quartering them at thirty; a colony other synthetic empires just struck weighs less to the rest, so a neighbourhood's raiders spread out instead of piling on one warehouse. War, siege, plan and farm targets keep their own doors; a far target alone in reach is still taken. What you will see: synthetic raids that arrive from nearby, on a dock you have not had time to fill.

1.121.7 — 2026-09-07

Fixed

The Aegis Bastion was sold on a sentence that was not true, and we are saying so plainly rather than filing it as a tidy-up (#3000).

The shipyard described the hull as "Shields the fleet around it, and never fires." A commander bought tens of thousands of them on that sentence, put them behind his line, and ran the battle forecast with and without. The two results were identical, down to the last ship. He was right, and the report was right: for a fleet that size the Bastions were doing nothing at all, and no number of them would have changed that.

Here is what the hull actually does. Its shields are a loan, not a wall. Your ships draw on the Bastion pool only for the part of an enemy salvo their own shields could not absorb — and your own shields come back to full every round and grow with every hull you add, while the enemy's salvo does not grow just because your fleet did. So the bigger your fleet, the more certainly its own shields already cover everything shields are allowed to cover, and the Bastions behind it are never reached. Past a certain fleet size that is not "less useful" — it is exactly zero, and buying more cannot move it.

Nothing about your battles has changed. No stat, no formula, no combat rule was touched: fleets fight exactly as they did yesterday, and nobody's losses move by a single hull. What changed is that the game now tells you the truth about the hull before you spend on it. The card reads:

Never fires; a large fleet cannot use its shields

The same promise was live in three other places — the raid fleet picker, the defence planner and the campaign roster note — and all of them now agree.

Two things we found while investigating and are not fixing quietly:

• A tender can cost you ships. Far below the size where its shields are reachable, in a fight you were going to lose, the Bastions keep your fleet on the field past the point it would otherwise have broken off and come home — so you lose the ships you would have kept. Measured: 200 light hulls escorted by 50 Bastions lose 186 ships without them and all 200 with. That is the retreat rule counting hulls that cannot fight, it affects freighters and loot trains too, and it is now filed as its own bug.
• Whether the hull is worth its price at all is a real balance question, not a wiring one, and it is filed with the numbers rather than quietly retuned on the way past. Fixing the sentence does not make the purchase good; it makes the purchase informed.

Thanks to jsnam for the audit and for the A/B screenshots that started it. If you are holding Bastions, nothing has been taken away from you — but you can now read the card and decide.

1.121.6 — 2026-09-07

Fixed

• A note in the engine kept asking for a fix that had already shipped (#3001). When a colony ship is turned away on arrival, the game stores why — the slot is taken, you had no room for another colony, the season closed under the fleet — and since v1.105.2 the diary has printed a different sentence for each, instead of one "planet no longer available" that was wrong about all three. The comment at the place that writes those failures never got the news: it still said the reason was read by nobody and kept on the wire against the day something would render it. It was then quoted, as the code's own admission, in a bug report about a defect that had already been fixed and released — which is how a stale sentence costs more than the line it occupies. It now says what actually reads the reason, and warns the next person that a fourth cause added there is text a player will read, which nothing in the build will remind them to write.

Also

• The test that stops a bounced colonisation from blaming a rival now pins the sentence from the coordinates rather than from the middle, closing the gap where an attribution could have been slipped in front of the clause without turning a single test red. Nothing a player reads changed in this release.

1.121.5 — 2026-09-07

Fixed

• The screens no longer disagree with the engine about what a strike costs (#2986). Ten places told you the price of hitting someone far below your Power, and each had worked the rule out by hand on a different day — so each was missing something. The ladder promised a bounty for hitting a wanted head, the one strike that is always free, on exactly the rows its own Bounty sort puts at the top. The raids panel listed the exemptions and stopped one short. The target board's header named two of the four exceptions its own rows were already printing underneath it. The galaxy legend, the War Room target hint and the War Room Power tooltip stated the price flatly on lists that routinely contain a silent owner or a head already wanted.

  Now the rule lives in one place and the screens ask it. Where a screen has the information it answers precisely; where it structurally cannot — the leaderboard receives no dormancy flag at all — it names the free cases and sends you to the planet sheet for the exact figure instead of promising something it cannot know.

• *The free cases now say when they are measured.* A wanted head and a silent owner are both judged at the moment your fleet lands, not when you read the screen: a bounty that decays mid-flight, or an owner who signs back in while you are in the air, is billed in full on arrival. Every surface that names those cases now says so — which matters most at the raid door, where joiners ride the opener's arrival and the gap between reading and impact is longest.

• Synthetic empires no longer recite a rule the game does not have (#2986). Asked in chat what it costs to hit someone much weaker, an agent would say it always puts a bounty on your head. That is false for a wanted head and false for a silent empire, and it is the one surface where a machine explains the game out loud to a person. It now names the free cases — including that returning a blow is free only when you strike alone, never for your slice of a raid — says it cannot tell a silent empire from where it stands, and will not claim a specific player has gone quiet.

Also

• The galaxy map's legend for the grey dots now reads simply "weaker". It used to promise a bounty, which is false for the silent empires the same grey covers, and there is no room in that strip to qualify it honestly — the planet sheet one tap away has read the whole rule since it was written. Whether those slots should be coloured differently is a separate question, left open rather than decided inside a copy fix.
• Four defects in twenty-four hours came out of this one duplicated list (#2900, #2958, #2960, #2979). The list is now written down once, with tests that redden when it grows or when a screen states a price it cannot keep — so the next exemption reaches every screen by existing instead of by being remembered.

1.121.4 — 2026-09-07

Fixed

The Eschaton Veil now tells you, before you buy it, everything it does not hide (#2997)

The two screens you read while deciding whether to research and build an Eschaton Veil described it more narrowly than it works.

The tech-tree card said the Veil was pierced by a spy mission and that attacking from a veiled colony "puts your address in the enemy battle report". Read plainly, that promises two things the game does not: that only attacks give you away, and that without a battle nothing leaks. The building sheet was narrower still — it named a spy mission as the only way through and never mentioned fleets at all.

Neither is how concealment actually breaks. An attack's origin reaches the defender from the moment it launches, so a feint you launch and recall has already published the coordinates with no battle ever fought. A garrison sent to help a guildmate carries its launch pad into the report the attacker reads. And a fleet in open space can be read from where it set out — which is true of every mission you send, not only the ones that end in a fight.

All three Veil screens — the tech-tree card, the building sheet and your own colony's Veil status — now say the same thing, in the same words:

A successful active spy mission still pierces the cloak, and anything you launch from a colony under the Veil can still be traced back to it: an attack's origin reaches the defender from the moment it launches, and the battle report names where every fleet came from, garrisons included.

Nothing about the Veil itself changed — the same colonies are hidden from the same empires, and the threshold each screen quotes is untouched. What changed is that the screens you buy it on no longer promise more than it delivers. If you own a Veil, the sentence above is the one to plan around: it was always true, and now it is written where you can read it before you spend.

Also

The wording is now stored once and drawn by all three screens, so they cannot drift apart again and quietly disagree about what your Veil covers.

1.121.3 — 2026-09-07

Changed

• Synthetic empires buy the research level they can already pay for, before the next building (#2998). Two and a half hours after the previous release taught them to save for a nominated level, their treasuries were filling up and the level was still not being bought: fifteen of the empires with an empty research queue could pay their next Galactic Economy level and spent it, every tick, on batteries and cannons instead — the construction pass runs first, and the saving rule only held the money while the empire was still short of it. From now on the price is set aside whether or not the empire is short of it, and the research pass buys the level in the same tick — one level, not a cheap extra on top; a capital hit at home in the last two days keeps the money with the battery unless the level is the defence itself; and a level the server refuses is left alone for an hour rather than held for. In chat the empire can now say what it is holding money for. What you will see: the synthetic empires' Galactic Economy climbing, and the Shielded Silo appearing over the warehouses raiders farm, within the day.

1.121.2 — 2026-09-07

Fixed

A synthetic empire told "no" about a fleet recall now stops asking (#2957)

When one of the game's synthetic empires turns an outbound attack around — because you signed a pact with it while its fleet was in the air, because your garrison grew past what it sent, or because its own home colony came under attack — it asks the server to recall the fleet. If the server said no, it used to ask again five minutes later. And again, and again, for as long as the fleet was flying: a mean of six requests for one decision, and as many as twenty-one on a long flight. A person told "no" presses the button once and stops.

Now a refusal ends the attempt. The reasons a recall can be refused are no longer treated alike: the ones that cannot change while the fleet is still in the air — the recall window having closed, the fleet having already landed, a raid, a garrison that needs a different order — stop after a single request. The ones that genuinely can lift, chiefly the three-recalls-a-day limit whose window rolls rather than resetting at midnight, are tried again a few times, spaced by a pause drawn from that commander's own habits rather than from a clock the game publishes.

That last part is the same rule the night-shield retry (#2588) and the Council directive reroll (#2633) were given: an empire must not come back at the first instant it is allowed to, because striking on the first legal second looks like foreknowledge even when the opponent is blind. Measured over 40,000 simulated flights, the share of successful recalls landing within one tick of the moment the limit released falls from 93% to 37% — level with a maximally spread retry. The honest cost, stated because the changelog is a claim: spreading the attempts out means some now arrive after the fleet has landed, so of the recalls the server would eventually have allowed, 87% used to get through and 66% now do.

An empire could describe a fleet as still attacking you after it had already turned back (#2957)

If the recall order reached the server but its reply was lost on the way back, the empire concluded it had failed — while the server had already told you, its target, that the attack was withdrawn. Its own account of itself said the strike was still committed against you, and it would say so in chat. The two sides now agree.

Also

A network timeout on a recall used to escape the code that handles it, instead of being treated as a refusal worth one more try. And if a recall order reached the server while the reply was lost on the very last try, nothing used to tell the empire its own order had gone through; it now reads that off the fleet's own row.

One limit, stated because the numbers above are a claim: an empire remembers a refused recall only until the game's servers next restart, and the game ships several times a day. A flight that spans a restart starts asking again, so the real average sits somewhat above the six-to-two figure measured here. Each stretch is still capped.

1.121.1 — 2026-09-07

Changed

• The guest upgrade now leaves a trace, so it can finally be counted (#2985). Nothing changes for anyone playing: this release adds no button, no screen and no message, and the upgrade itself works exactly as it did before. What changes is behind it. Until now, when a guest account linked Google or Apple to keep its empire, the game simply overwrote the address and kept no record that the upgrade had ever happened — a guest who converted became indistinguishable from a commander who had signed in with Google on their very first day. Two facts are now recorded at the moment it happens: when the account stopped being a guest, and which of the two doors it came through.

  It matters because of what it makes answerable. Nine out of ten new commanders arrive as guests, and a guest empire lives only on the device that made it — clear the browser, change phone, and it is gone. The game already warns guests about this and offers them the upgrade, but nobody could tell whether that warning was persuading anyone, which made it impossible to judge how much the sign-in experience is worth improving. From this release on, that question has an answer.

• What it cannot do is recover the past (#2985). Every upgrade made before this release left no trace at all, and no amount of digging brings it back. The 707 guest accounts and 228 signed-in accounts standing on the day this was written stay uninterpretable for ever: there is no way to know how many of those 228 were rescued guests, and there never will be. The counting starts at this release and covers only what happens from here.

Also

• The operator notes gain the query these two facts exist for, together with the traps in reading it — including one that was got wrong on the first pass and corrected in review. Upgrading an account does not put it beyond the reach of the sweeps that remove abandoned empires: it roughly doubles how long an idle one is kept, and counts for nothing at all against the sweep that clears accounts which never really started. So a commander who upgrades and then never comes back is still removed, and the record of their upgrade goes with them.

1.121.0 — 2026-09-06

New

Your own colony now tells you what its Eschaton Veil is hiding it from (#2876)

You could buy a Veil and never find out whether it was doing anything. The mechanic worked; no screen in the game said so. As the player who reported it put it: it's hard to test in game.

Open one of your own colonies on the galaxy map and the planet sheet now says it outright — the Veil's level, and the line it draws:

Eschaton Veil L9 Hidden from any empire whose best Sensors are below level 9 — at level 9 or above they see it on the galaxy scan like any other colony.

"Best Sensors" is the strongest Deep Sensor Grid anywhere in that empire, not the one nearest you — a single well-equipped outpost extends their whole detection grid.

The sheet also names what a Veil does not stop, because a cloak you trust too far is worse than none:

An active spy mission still pierces it, and a fleet you send from here can carry these coordinates out — an enemy Observatory reads an attack's origin at launch, and the battle report names where every fleet in it came from, garrisons included.

Both halves of that are worth reading twice. The launch is what publishes the address, not the battle — so sending a fleet out from a hidden colony and recalling it does not take the coordinates back. And it is not only attacking: a garrison sent from a hidden colony to help a guildmate carries that colony's address into the report the attacker reads.

The level shown is the one you have finished paying for, not the one last written to the colony — so a Veil that completed while you were away is counted from the moment it completed.

Nothing changed on anybody else's colony. Looking at another player's world tells you exactly what it told you yesterday, whether or not they are cloaked, and whether or not your Sensors are what let you see them. That is deliberate: a marker that appeared only over a cloaked colony would, by its absence everywhere else, hand out a free map of who owns a Veil — and two players at different Sensors levels could compare notes and pin a third player's cloak between them without ever spending a probe. The Veil's price is a probe and a flight, or Sensors of your own, and it stays that way.

1.120.2 — 2026-09-06

Fixed

• Garrisoning a guildmate no longer closes their Emergency crate (#2975). You send a fleet to help a guildmate hold off a raid, the defence holds — and your guildmate could no longer buy the Emergency crate. The crate opens on an attack in the last 48 hours that was not dismissed as a token, and "too little against the defence that met it" counted every ship standing there, yours included. So your help raised the bar the attacker had to clear, and an assault that was real against your guildmate's own colony was written off as a token, shutting the window they would otherwise have had.

  The test now asks twice, and both answers have to agree before a battle is dismissed: was the attack slight against the whole force that met it, and was it slight against what the defended colony itself had standing — its own ships, a fleet it parked on its own colony included, and its emplacements. An attack too big for what you had there yourself is no longer called a token, however many allies stood with you.

  Nothing became easier to fake. The second reading counts only the defender's own ships and emplacements and cannot be moved by anybody else's fleet, so a throwaway hull is still a throwaway hull however many allies are in orbit. Checked against every battle of the fortnight: no battle that would have been dismissed without allies present starts counting because of them.

  Two limits worth stating rather than leaving to be discovered. The crate's store row now spells the second measure out, including its exception: if you had nothing of your own on the planet, only what met the attack is weighed — so a defender whose fleet was away and who has no emplacements can still be refused, and the row says so instead of promising otherwise. And the change reaches battles fought from now on: a battle already dismissed keeps that verdict for the rest of its 48 hours.

  The guild Anchorage's repair relief deliberately keeps the old, stricter measure. It asks a different question — whether there was really something to overcome before a repelled assault earns a notch back — and there the wall that met the attack is exactly what won the battle.

1.120.1 — 2026-09-06

Fixed

• Attacking a silent empire no longer says you are close to a threshold you are past (#2979). Lining up a strike on a colony whose owner has stopped signing in, the warning beside the Launch button read "You are close to the bounty threshold on this target: a bit more Power and the strike starts costing you" — even for a player at twice the band, who was well past it. The planet sheet, one tap behind, said the opposite and said it correctly. The prediction was the worse half: growing your Power never starts costing you a bounty on a silent empire, at any size, so anyone pricing their expansion off that line was reading the opposite of the rule. Where that sentence used to appear, the launch screen now says what the planet sheet says — "Silent empire — its owner has not signed in for days. Attacking costs you no bounty at all — checked when you land, not now." — and the two screens now draw it from one place, so they cannot drift apart again. 163 strikes by 28 attackers had settled against a silent empire since the waiver shipped; every one was launched off the wrong sentence.

• Both screens now say that a silent empire waking up mid-flight still bills you (#2979). The planet sheet had promised "no bounty at all" flatly since the waiver shipped, and the game decides it when your fleet lands, not when you launch: an empire that stirs during the flight charges you the full bounty. Neither screen said so. Both do now.

Also

• The launch screen had two separate branches deciding what to say about the bounty, each with its own hand-written copy of the rules that waive it — and the second copy was missing one. They now both ask the same list, so a rule added in future reaches both. No other wording changed, and the new line is shorter than the one it replaces, so nothing else on the screen moved. The wider version of this problem — the same list re-derived on several other screens, each shorter than the engine's — is written up as #2986.

1.120.0 — 2026-09-06

New

• Synthetic empires now remember the colony that paid, and come back for it (#2981). Measured on 2026-09-06: human raiders return to the same colony in a quarter to a third of their strikes, and the return pays more than the first visit; the agents returned in six percent of theirs and took a tenth of the loot a human takes, because they remembered every defeat and every refusal and not one success. From now on a won strike that filled the holds earns one return: hours later, on the empire's own rhythm, it scans that system again and prefers that colony over the others in its reach — almost always for a hawk, half the time for a cautious one — after the targets its plan already named and inside every check it always made: its power band, its reach, the enemy's fleet, its own defeats and refusals, its guild. It keeps coming back only while the warehouse keeps paying; a loss or a poor haul ends the visits. What you will see: an agent that found your warehouse comes back for it within the day, like a human raider does; an empty warehouse or a Shielded Silo over it is what stops it.

1.119.2 — 2026-09-06

Fixed

• The test suite can now tell a live agent read from a dead one (#2967). Seven defects in two days shared one shape: agent code that looked alive over a test fixture that had invented the server, so a read of a key the game has never sent stayed green forever. The agents' research reads are now checked against a real recorded server response instead of a hand-written one, and that recording is re-derived from the live code on every test run — so if the server's answer ever changes, the suite says so instead of the agents quietly reading nothing.
• A decision the agents never actually take is now a failing test. A decision function could be written, tested and wired to nothing, and no test noticed — which is how the debris-harvest decision sat unused since the day it was written. Every decision must now have a real caller, and a caller that exists only in the offline simulator does not count.
• A changelog entry that closes two issues no longer reddens the suite. The page's own guard wanted exactly one issue number per entry, so the bounty-copy entry (#2960, #2958) read as an entry citing none.

Also

• Nothing in this release changes what a player sees in the game. It is machinery: the guards that decide whether a change to the synthetic empires can ship.

1.119.1 — 2026-09-06

Fixed

• The diplomatic tension card now tells you which guilds it is about (#2962). The row you meet most often in Activity used to read "The crisis between the alliances rises to 100/100 after a battle" — and never said between whom. If you wanted to know, the only way was to dig into Guild → Diplomacy. It now names both guilds: "The crisis between Kepler Circle and Stellar Gamesense rises to 100/100 after a battle — the top of the scale." This applies to the rows already in your feed, not only to new ones — the card had the information all along and simply was not showing it.

Changed

• One word for a guild, everywhere (#2962). A handful of screens called your guild an "alliance", which nothing else in the game does — and which was confusing for a real reason: an alliance here is a specific thing, the pact two guilds sign in the Diplomacy panel. Using the name of the treaty to mean the group of players made both harder to read. Fourteen pieces of text now say guild where they mean the group: the tension and war-declared cards in Activity, the Diplomacy panel and War Room, the war target list, the Empire power dialog, the tag beside your name, the account-linking note, and the home screen hotspot. Where the word genuinely means the pact — "Propose alliance", "sign an alliance" — it is unchanged.

Also

• Long guild names no longer push the Activity card off the edge of a narrow phone screen.
• The surrender confirmation now says your guild pays the tribute, which is who actually pays it.

1.119.0 — 2026-09-06

New

• The synthetic empires clear the wreckage in their own orbit (#2953). Until now no agent, Wraith or NPC had ever flown a single debris-collection mission — 0 of them against 1,389 flown by 35 people in the preceding fortnight — while 74 of the 76 agent empires stood on a Recycler they had built and never once used. Their contract board kept dealing them "Harvest a debris field" (410 times in that fortnight) and they never finished one. Now, when a battle leaves wreckage over a world an agent owns, it sends freighters up to collect it, the way you would.

  This takes nothing from anyone. Debris over a living colony can only ever be lifted by that colony's owner, so the 21.4 million tons that piled up over agent worlds in a fortnight were decaying to nothing where no other player could reach them. The rich fields over the neutral Beacon orbits — the ones people actually compete for, 18.85 billion tons of them — are deliberately left alone: agents will not race you for salvage you could have had.

  You will see it in the mission log and on the war board: expect a visible number of new debris runs a day across the roster, spread out rather than arriving in bursts. An agent waits six hours before returning to the same field and opens at most three new ones a day, so it collects like someone tidying up after a raid, not like a machine on a timer.

Fixed

• A shipyard that re-ordered your freighters while they were out. The count of hulls "away on a run" knew about deliveries and pickups but not about debris runs, so an empire with freighters in the air looked as though it had none and kept ordering more. Harmless while nobody flew debris runs; not harmless now (#2953).

Also

• Two safety switches that were not actually connected are now connected: the ones that let the operators turn the new debris behaviour, and agent chat commitments, off without rebuilding the game. The check that is supposed to catch a disconnected switch was only looking at half the places switches are declared, and now looks at all of them (#2953).

1.118.3 — 2026-09-06

Fixed

The game was telling you the wrong thing about who collects a bounty (#2960, #2958)

Every screen that priced a strike far below your Power ended on the same promise: "Priced on impact, and anyone who stops you collects it."

Nobody who stops you collects anything. A bounty is claimed by attacking a wanted commander. There is no payout on the defending side of a battle anywhere in the game: repelling a wanted attacker pays nothing, joining a guildmate's defence pays nothing, and an interception pays nothing at all. A player who had just helped repel a wanted attacker went through his own event log looking for the dark matter we had promised him, and it was not there, because it was never earned.

We are sorry. That sentence had been on screen for a long time, and #2900 had only just put it back on the launch line after finding it had been hidden in a fold since August — so the warning players finally started reading again was the one that was wrong about who gets paid.

The same line also stopped one word short of the rule that decides the risk. The bill is computed from the Power gap the moment your fleet lands and never looks at how the battle went, so a repelled attacker is wanted for exactly the amount a victorious one is.

Both screens now read: "Priced on impact — win or lose — and only someone who attacks you can collect a share of it."

Six surfaces said this, in three different wordings, and each had its own copy of the sentence: the launch composer, the planet sheet, the Attack button tooltip, the WANTED tile in your HUD, the same sheet on a phone, and the diary row you get the moment a bounty lands on you. They now share one sentence, so they cannot drift apart again. The raid join door gained the "win or lose" half. The synthetic commanders were reading the same false rule out of their own briefing and could repeat it in chat; that is corrected too.

Also

• "a share", not "all of it". No single battle ever clears a head — one claim takes at most half the pot and usually far less. The WANTED tile always said this and the first draft of the fix dropped it; it is back, on every surface that now carries the rule.
• The launch warning in a starting sector is longer than the game's own limit for that row, and was before this change. It is not fixed here — it is now measured, and recorded as owed.

1.118.2 — 2026-09-06

Changed

• Synthetic empires now save for the research their defence needs, instead of only wishing for it (#2949). Measured on 2026-09-06, six and a half hours after the empires began nominating the Shielded Silo's technology: not one of them had started a level of Galactic Economy toward it, and two thirds of them were researching nothing at all. Their capitals had the Ferrolite and not the Xenogas or the Data, because every five minutes the shipyard and the build queue spent both before the research pass ran, though the empire's own income would have covered the level within hours. From now on an empire whose defence nominates a technology within a day of its income sets that price aside, its builders and its shipyard spend above it, its research buys nothing else in the meantime, and its freighters bring the missing Data and Xenogas home from the outposts. What you will see: agent empires that reach the Shielded Silo and the Fortress's technologies within days instead of never.

1.118.1 — 2026-09-06

Fixed

A Veil that has finished now hides your colony immediately (#2892)

Buildings finish quietly. A Veil level that completed while you were away was applied the moment anything looked at your colony, but the saved copy of that colony kept the old level until some other action happened to rewrite it — a build, an attack, a transfer. Fifteen screens read the saved copy.

So for that stretch, the Veil you had just finished paying for did not work. Your colony still appeared on the galaxy map, on the target board, in the war room, on the contact board and in your enemies' guild Watch Array to players who should no longer have been able to see it. Its coordinates were still printed on the public leaderboard row, on the public wanted board and in the global activity feed.

The same delay ran the other way for Sensors. A Sensors level that finished while you were offline did not extend your reach until something else saved one of your colonies: a slot you should have been able to see stayed drawn as empty, a target stayed off your board, a launch was refused for a colony you could now find.

Both sides of every visibility check are now read at the moment you ask. A level that has finished counts from the second it finishes — the Veil that hides you, and the Sensors that see through someone else's.

The address of a veiled colony stays off the public boards (#2892)

The public leaderboard, the wanted board and the activity feed each pick a colony with no Veil before printing its coordinates. They were reading the same stale copy, so a player whose first Veil level had just completed still had their planet published. They now read the current level, so a Veil counts against those three the moment it is finished.

Also

Nothing in the Veil or Sensors numbers changed: the levels, their strengths and the threshold between them are exactly as they were. What changed is when a finished level starts counting.

1.118.0 — 2026-09-06

New

• The best spy report now shows the fleets standing on the target, not just the ones that left it (#2899). The deepest espionage tier sold "fleets away from home" and delivered exactly that: the missions that departed the colony you probed. A fleet that arrived there — lent by an ally, or moved in by the owner from another of their own colonies — fought beside the defender and appeared in no report, at any tier, at any price. A commander could pay the most expensive Discipline in the game, read the report, and meet a wall that was not in it. Top-tier reports now carry a Garrisons on station section: the hulls, and how long each watch runs.

  This is a deliberate balance change, and it cuts one way. Joint defence keeps its ambush value against everyone below the deepest tier; against a commander who has paid for it, the ambush is over. Hiding a fleet on a friendly colony is no longer free — it now costs the attacker a probe, a round trip and a detection roll to undo, which is what the rest of the intel layer already charges for knowing how a fight will go.

Fixed

• The report no longer implies a colony is undefended when it simply never looked (#2899). On a target with nothing flying, the top tier printed a flat "everything is home" — a categorical negative, read as "no surprises", over colonies with hundreds of hulls parked on them. In the reports on record that sentence went out on 40 of the 53 top-tier probes ever filed against a garrisoned colony, with a median of 389 unreported hulls standing on the target and a worst case of 9,000. The fleets-away line now speaks only about what left that colony, which is all it ever measured.

• Every report now says which half it is showing (#2899). A report that did not read the garrisons — every older report, and every tier below the deepest — says so, instead of leaving a silence that reads as an empty planet. A probe that did look and found the colony clear says that instead, which is a different sentence and an answer you paid for.

• The launch forecast stops quietly leaving those hulls out (#2899). It now names them and states that it does not fight them, so the recommendation and the report card no longer describe two different battles. It says the forecast may be optimistic and points you at each garrison's watch — it does not pretend to know by how much, because a lent fleet shoots with its own commander's research and mastery, which no probe reads.

• A garrison past its shift no longer reads as gone. A watch that has run out can still be standing — a guard can be posted for zero hours, and any garrison is held on station past its own bell while a battle is landing. The card would have printed a time earlier the same day, which reads as "they have left" about ships that are still shooting.

Also

• The Veil clause proposed on the issue — hiding a garrison whose sender holds an Eschaton Veil — is not in this release. There is no per-fleet cloak in the game today (cloaking is a property of a colony's buildings), so whether a garrison inherits its sender's Veil, its host's, or neither is a design question of its own rather than a wiring job.
• The report does not name who lent the fleet, only what is standing there. What an ally may see of a defence is still open (#2310).
• Synthetic empires read no more than before: the new field is additive and their intel reader takes named fields only, so agent behaviour is unchanged. Whether they should have to scout at all is open (#2933).

1.117.12 — 2026-09-06

Fixed

• An interception no longer fights with research you finished hours ago but never came back to collect (#2914). Your empire's research is banked lazily: a level finishes on schedule, but the number on your account is only brought up to date the next time something touches it. Until then a weapons, hull or shields level you had already paid for simply was not there. Every other battle in the game already corrects for this before the shooting starts — a colony assault, a raid, the season boss, the campaign. An interception did not, on either side: the picket owner and the convoy each fought at whatever was last written down, and because the two accounts go stale independently, the error did not even cancel out. Now both sides are brought up to date at the moment of contact, so a convoy escort that finished a shields level while its owner was asleep defends with it, and a picket fights with the guns it has actually bought.

  Right now this changes no battle you can see: interception is still running in shadow on the live server, where contacts are recorded but nothing is destroyed. This is the fix landing before the mechanic is switched on, not a change to a fight anyone has had. Measured on the live database while fixing it: 31 accounts were holding a finished-but-uncollected combat research level, the oldest 28 days old, and every single one was exactly one level short.

1.117.11 — 2026-09-06

Fixed

• A synthetic empire whose capital is under attack now brings its fleet home (#2928). Until now it never did. An agent could watch an overwhelming fleet close on its own capital and simply carry on with the attack it had launched elsewhere — not sometimes, not slowly, but never, for any agent, against any inbound. The reflex had been written, but it asked the server for a piece of information that endpoint has never sent, and the miss was swallowed silently instead of raising, so the whole behaviour sat dead behind it. Nothing in the game ever showed it: no event, no notice, no line anywhere. Agents now turn the fleet around when the garrison and the wall together cannot beat what is coming and the fleet can still get back before the blow lands — and, as for any player, it arrives only after flying the return leg.

  How often you will see it: not often. Agent capitals really are attacked constantly — 289 attacks landed on 67 of the 76 synthetic capitals in the last week alone — but agents attack mostly through guild raids, and a raid cannot be recalled by anyone, human or synthetic. Over that same week, the situation where an agent both had its capital threatened and had a recallable fleet of its own in the air came up three times. So this repairs a reflex that was broken rather than adding a new habit: if you besiege a synthetic capital you will usually find things exactly as they were, and occasionally you will now find the defenders home in time to meet you.

  When it does happen you see what you would see from any human defender: the withdrawal notice your panel already shows, and the fleet back at the capital.

• An agent that calls off a strike to defend its capital can now tell you so (#2928). A withdrawal is announced to the empire it was aimed at, so the target knows the fleet turned around — but the agent itself had no way to refer to why. Asked about the attack it had been talking up, it knew only that the fleet had come back, with nothing true to say about the reason. It can now say plainly that its own capital was facing more than it could hold. It does not claim the defence succeeded, because at the moment it turns around that is not yet known.

1.117.10 — 2026-09-06

Fixed

• The Wraith empires had quietly stopped closing their day (#2917). When all thirty crossed Command Center 20 their contract board grew from three cards to six, and the Daily Wrap — which asks for two thirds of the board — went from two cards to four. But a Wraith only ever completes three kinds of daily card, so from that moment none of them earned a single Wrap, and without Wraps there is no Weekly Chest. Nothing broke and nothing was logged; the board simply got longer while the part they could actually play stood still. This release does not change any board — it records the measurement, in the code and in the design document, so the fix is made deliberately rather than discovered a third time.

Changed

• The contract catalog now carries an honest note about what the machine empires can actually do (#2917, following #2895). #2895 stopped dealing "Investigate an anomaly" to synthetic empires because they can never finish it, and reasoned that the five cards left were five they could work. Measured over a fortnight on the live game, they are not: the Wraiths never complete "Strike the Pirate Beacon" or "Harvest a debris field", and the agent empires never complete "Harvest a debris field" either — three more cards that could never move, on top of the one #2895 found. The counts and the reasons are now written down next to the rule they belong to.

Also

• Nothing a player sees has changed in this release. No contract, target, reward or board size moved, for people or for machines. Human boards are dealt exactly as they were.
• The supply figures that issue #2572 is waiting to tune anomaly stock against were re-derived independently and hold — with one correction: every completion percentage on that board is a lower bound, because claiming a finished contract is a manual tap and an untapped card counts as a miss.

1.117.9 — 2026-09-06

Fixed

• Adding Novarchy to an iPhone home screen now gives you the crest (#2935). The site never told Safari which icon to use, so the tile on your home screen was whatever iOS could work out on its own — and the two addresses iPhones check by convention when they are not told both answered with the web page itself instead of a picture. The game now hands iOS a proper 180-pixel icon of the same star-and-plate crest the app uses, and the addresses that used to return a page now either return the icon or nothing at all. If you already added Novarchy to your home screen you will keep the old tile: iOS copies the icon at the moment you add it, so remove it and add it again to pick up the crest.

Also

• The browser-tab icon, the Android launcher icons and the manifest are unchanged; this only adds the path Apple devices use.

1.117.8 — 2026-09-06

Fixed

• The Activity feed stops spending half its space on "Fleet returned" rows that say nothing (#2916). Every spy probe wrote two lines into your fifty-line feed: the report you sent it for, and a dim "Fleet returned — spy mission complete: ships and cargo returned to the departure colony." A probe flies empty, so that second line could never tell you anything — there was no cargo to land and none to lose. Send fifteen probes in a session and thirty of your fifty rows were gone, fifteen of them blank, pushing your own espionage reports off the bottom of the list. Those empty lines are no longer shown, and the reports they were burying are. For the busiest scout in the galaxy the feed went from 18 espionage reports to 43 in the same fifty rows.

  Nothing else about the feed changes. A fleet that brings something home, or that loses part of its cargo to a full warehouse, still reports exactly as before — harvests, pickups, transports, raids and the founding convoy all keep their landing line, and a spy mission you recall still tells you it turned back. The empty rows are only hidden from the feed; they are still recorded.

1.117.7 — 2026-09-06

Changed

• A synthetic empire that turns to defence now recalls the strike it had been preparing, and says so (#2931). Measured on 2026-09-06: of seven agent attack operations with a spy report in hand, four belonged to empires whose plan had turned defensive, and they sat there for hours — the strike neither launched nor recalled, the agent silent about a target it had announced. From now on a defensive plan calls a mustered or scouted strike off out loud, and the empire's next scouting starts fresh when its plan turns again. What you will see: an agent that told you it was coming and then chose to hold the line will say it recalled the strike, instead of going quiet.

1.117.6 — 2026-09-06

Fixed

• The Proving Ground's bracket cards now say what the leader's number is (#2920). A bracket you are not in collapses to a small card, and it printed two numbers on the same right-hand edge, one line apart: the fleet size that opens the bracket — labelled, "from 100k power" — and the score of the empire leading it, labelled with nothing at all. Both are printed at the same size, in the same abbreviated style, so the second read as another power figure. A player in Armada saw "5.7M" beside the leader of Fleet, a bracket below his own, and reported the game as ranking its brackets wrongly.

  Every number on that card was already correct. Brackets are decided by military power; that 5.7M was boss hull destroyed, by an empire with a smaller navy — which is exactly the feat the brackets exist to reward, and the card was hiding it. The leader's figure now reads "5.7M of hull", the same words the panel already uses for your own best strike and in the board's own legend. Nothing was retuned: no bracket moved, no score changed, and no ranking is different.

1.117.5 — 2026-09-06

Fixed

• Synthetic empires now plan with the engines they have actually researched (#2915). An agent that had been buying Propulsion for weeks was still planning every pursuit as though it had none, so it declined contacts its fleet could comfortably have reached — silently, with no event and no diary row to show for it. At Propulsion 4 an agent's reach was about 38% short of the truth; a contact eleven to fifteen cells away was refused outright. Agents will now go out after contacts they used to let pass.

  ⚠️ What this does not yet change for you. Interception runs on this server in shadow: pickets are placed and contacts are computed, but nothing is destroyed and nothing is looted. So you will see synthetic empires send fleets out to meet traffic where they previously sat still — you will not yet lose a cargo to one. The honest value of this fix today is that the measurement shadow mode exists to take ("does anybody place a picket at all?") was being answered by seventy-five empires whose planner was stuck at Propulsion 0. In ten days they placed one picket between them. That number was never about the mechanic.

• Agents can spend a Chrono fragment on a research order (#2915). Every fragment an agent had ever spent went to a building or a hull, because the research lane was invisible to them — zero research spends by any agent in seven days, against 157 on buildings. The first one went out within six minutes of this deploy. Agents with fragments banked and long research running will start using them, so expect a one-off drawdown of their stocks.
• The shared galaxy directory stops picking a blind anchor (#2915). Where two agents had equally good sensors, the tie was meant to go to the better Astrometry — and instead went to whoever came first in a list, because the Astrometry read as zero for everyone.

Also

• Agents can now ask their guild for a hand on a research order, and will not withdraw the request while the order is still running. Both halves of that were broken in a way that cancelled each other out, so no player has seen an agent do either — it is new behaviour rather than a repair of something visible.
• Under the hood, all four bad reads had the same cause: the agent runner asked /research for keys the server has never sent, and the miss was swallowed by the same guard that protects against a failed read, so nothing ever raised and nothing was ever logged. Three of them were live and one was latent. The paths now live in one place, with a test that fails if any part of the runner reaches into that payload by hand again — and, separately, a test that fails if the server ever renames one of the keys it depends on.

1.117.4 — 2026-09-06

Fixed

• A ship can no longer mistake another ship's test verdict for its own (#2922). Several releases are prepared in parallel, and each one runs the full test suite before it is allowed out. Those runs wrote their reports into files named alike, so two of them could end up in the same file — and twice in one night a release read a neighbour's "all green" as its own, once from a report that was missing two of its checks entirely. Both times a person noticed. Now every run stamps its own identity on its report and on its verdict, and a report holding more than one run is refused outright instead of being read as a pass. Nothing about the game changes; what changes is that a release can only ship on a green it actually earned.

1.117.3 — 2026-09-06

Fixed

• The leaderboard was telling everyone how big everyone else's fleet was (#2910). Opening the board quietly downloaded, for every player on the server, the full composition of their Power — how much came from buildings, research, tree and territory, and how much from ships. None of it was ever shown: the board has five columns and not one of them was any of those numbers. Knowing the size of a rival's fleet is what a spy probe is for, and in fact nothing in the game sells that figure for a whole empire at any price. It is no longer handed out. The board still publishes the Gate, which is the number it prints, colours the bands with and ranks on — that one is public by design, because both sides of an attack have to be able to read it.

Changed

• Nothing moves on screen: the same columns, the same numbers, the same order. The Gate column, the podium, the phone cards and the band colouring are untouched, and your own empire sheet still breaks your Power down into Structural and Military exactly as before — that has always come from your own empire, not from the board. The leaderboard response is about a third smaller, which the phone will feel more than the desktop.

Also

• The design document that specifies the board was amended to match, rather than left contradicting the code.

1.117.2 — 2026-09-06

Fixed

AI empires can name a wanted player again, without being told where to find them (#2882).

Yesterday's fix went one step too far. Making each AI empire see only what its own sensors reach was right — but it also took the bounty board away from them, and a bounty is not a secret. The game has always been explicit about this: a wanted player wholly hidden under a Veil still shows up, by name and by price, on the Wanted tab and on the leaderboard, for everyone. Only the address is withheld.

For a few hours the AI empires had it the other way round: a veiled Warlord that every player could see red-badged at the top of the board was a name none of the 75 would say, and could not answer a question about. That is the kind of small, always-in-the-same-direction gap that makes a machine recognisable, which is the one thing they are built not to be.

Now they follow the same rule you do. They can name a wanted commander and quote the price on their head; they are told where to find them only if their own sensor grid actually reaches that colony. And when it does not, they do not guess — an empire that cannot locate a target does not put it on a target list, so nothing it says about a hiding player is ever an invented coordinate.

Also

The machinery that gives each empire its own view is now held in place by tests that fail if it is ever unhooked — it was working, and nothing would have noticed if a future change quietly disconnected it.

1.117.1 — 2026-09-06

Fixed

• A watch now chases at the speed you have actually finished paying for (#2893). A picket priced its pursuit — and a hand-flown chase priced its rendezvous — from a Propulsion level the game had not written back yet. Research finishes the moment its timer runs out, but your empire's record is only rewritten when something else happens to touch it, so a level you completed while offline was simply not counted. It cost you two different ways, and one of them was silent: a standing watch could find no interception where one was there to be made, and it said nothing at all — no report, no diary entry, no error, just a flight crossing your ring untouched. A chase launched by hand was the louder half, refused at the door as unsolvable, or booked to a rendezvous later than your fleet needed, frozen in for the whole flight. Both now read the level you finished, and the Doctrine sails keystones with it.

  Nobody has lost an interception to this yet: watches are still in their measuring trials and none was standing when this was found. But 125 accounts were carrying a finished-but-unwritten Propulsion level at the time — one of them for 36 days — and every one of them would have flown a slower picket than they bought. Correcting the price can only ever give a fleet back speed it had already earned; no watch becomes worse, and nothing needed repairing in anyone's empire.

Also

• Two neighbouring defects found in the same sweep were filed rather than folded in, because each changes something different and deserves its own line: an interception simulates both fleets from combat research that has not been written back either (#2914), and the synthetic empires plan every pursuit as though they had no Propulsion at all (#2915).

1.117.0 — 2026-09-06

New

• Synthetic empires now strike a target they have scouted without waiting for the dice (#2906). Measured on 2026-09-06: eight agent attack operations sat with a spy report in hand for up to nine hours, and four with a force half-gathered for up to sixty-five, while agents struck humans twenty-seven times a day against four hundred and twenty the other way. Every five minutes the agent rolled twice for whether to act at all on an attack it had already decided, and when both rolls passed it re-drew the target among the neighbours of the one it had scouted. From now on an operation past its scouting is a standing intent: the fleet leaves minutes to an hour after the report, on that empire's own rhythm rather than on the dice, against the colony the probe looked at, and only if that colony still passes every check the agent always made — its own power band, its reach, the enemy's fleet, its refusal and defeat memories, its hours. What you will see: agent fleets that follow their probe within the hour, on the world the probe visited.

1.116.5 — 2026-09-06

Fixed

The anomaly contract is no longer dealt to empires that cannot complete it (#2895). "Investigate an anomaly" was being handed out to the AI empires and to the Wraiths as well as to you — and neither of them has been able to finish it since August, when their anomaly expeditions were taken away for exactly this reason: to leave the anomalies to real players. The card still went out to them every day and simply sat there, unfinished, for ever. Over the whole season the machines were dealt it 1,389 times and completed it five, all five before that change landed.

Nothing changes on your own board: you are dealt the same cards, with the same odds, as before. What changes is that the week's anomalies are now contended only by the players who can actually go and claim one.

Also

The measurement behind the open work on anomaly supply was corrected while fixing this. The earlier count identified machine empires by one email domain and so counted the thirty Wraith empires as human players. For the week of 31 August the real split is 470 cards to machines, none of them completable, against 227 to people, of which 27 were completed — so two thirds of the demand on the anomaly stock was never reachable, not 45%, and the human completion rate is 11.9% rather than the 7.1% previously published. The scarcity is real; it is about half as severe as the raw board figures suggested.

1.116.4 — 2026-09-06

Fixed

• The bounty warning is back on the launch screen (#2900). Attacking someone far below your Power puts Dark Matter on your head, and the launch screen has a line that tells you how much before you commit. Since 22 August you often never saw it. The screen shows one warning next to the Launch button and folds the rest behind the "n notes" chip, and a warning added that day — the one saying the launch leaves your own colony with no defences — was taking the slot. On the strike where both apply, which is the ordinary one (you hit someone much weaker, and you send everything to do it), the price was the line that got folded away. Players have been launching without seeing what it would cost them for two weeks. The price now takes the slot. Nothing about the bounty itself changed: the same number, shown instead of hidden.

• Nothing was lost to make room. The "no defences at home" warning still stops the launch with the same "Leave … without its own defences?" question, printed in full, before the fleet goes — it always did, which is why it can give up the line. It is also still there under the notes chip. The Picket and Contact sheets are untouched.

1.116.3 — 2026-09-06

Fixed

A colony you hide under a Veil is now hidden from the AI empires that cannot actually see it (#2882).

The 75 AI empires read the galaxy through a shared map, and that map was fetched using a single one of them as the eye. The server does its job — it hides a veiled colony from any viewer whose Deep Sensor Grid is too weak to find it — but it was answering the question "what can that one empire see?", and the answer was then handed to all 75.

So an AI empire with no sensor network at all could see, target, scout, besiege and name in chat a colony it had no way of detecting, as long as the one empire doing the fetching could see it. Before this change that was happening 4,103 times over: 72 of the 75 AI empires were being shown at least one veiled colony beyond their own reach, across 167 colonies — 49 of them belonging to players who had paid for the Veil precisely so this would not happen.

Now each empire narrows that shared map to what its own sensors reach, and the eye doing the fetching is the strongest grid on the roster rather than whichever empire happened to be first in a list. Your Veil is worth exactly what the game says it is worth, against every AI empire individually.

Also

The map's eye now says who it is and what it can see, once per refresh, so this can never quietly become true again without somebody noticing.

Two limits are known and filed rather than left implied: an AI empire can still be handed a siege target it cannot personally see (#2905 — narrowing that list turned out to retire sieges that were still live, and closing it safely needs its own change), and the anomaly half of the same map still depends on one empire's Astrometrics (#2904 — measured at 0 in production right now).

1.116.2 — 2026-09-06

Fixed

• The picket now counts the Sensors level you finished while the fleet was still flying out (#2858). Sensors levels settle in quietly: for a while after an order completes, the colony still reports the old level. Two parts of the picket system believed that old number.

  The launch screen sized your allowance from it, so a player who had just finished the Sensors level that buys their next watch was told their pickets were full — and the same attempt would go through later, once anything else happened to touch that colony. And the watch itself froze the old number in when it took station, where nothing ever corrected it: a level finished during the outbound leg was lost for that watch permanently, which is exactly the case the picket was meant to honour.

  Both now read the level you have actually finished. The watch is still fixed at the moment it takes station — a level completed afterwards still does not upgrade a picket already standing.

Also

• Nothing became visible to anyone that was not visible before: neither of these readings is compared against another player's Veil. The five map and war-room surfaces named in the same report were deliberately left alone, because correcting only the observer's side of a Veil comparison can expose a colony that correcting both sides keeps hidden — that one needs its own pass.

1.116.1 — 2026-09-06

Fixed

The Emergency crate no longer tells you nothing attacked you (#2859).

A colony of yours was attacked. You watched the battle resolve and read the report. Then the guild store refused you the Emergency crate with "No attack on one of your colonies in the last 48 hours."

The rule behind the refusal was right — a token attack has not opened the crate window since #2041 — but the sentence never mentioned it, so it named a condition you could see was false. One commander concluded from it that the crate needs a defence you lose. It does not, and it never has: the crate has never cared whether you won.

The refusal now says no attack counted, names all three exclusions — your own attacks, garrisons you lent to somebody else's colony, and token attacks — and states plainly that whether you won or lost the defence is not part of it.

The crate's row in the store now spells out what makes an attack a token, in the terms the game actually measures: a fleet that brought too little against the ships and emplacements that met it — or met none at all — and that looted nothing and destroyed no ship of the defence in the field. It also names what is not weighed there: the Shield bubble, the Fortress and the Cannon. So if your bubble absorbed an assault with nothing else on the field, the row now tells you why the window stayed shut.

Nothing about which attacks open the crate has changed. Only what the game tells you when one does not.

Reported by Spacca-Mazza, whose refusal is the one that started this; QQQQ supplied the citation that settled how much of the rule could be said out loud.

1.116.0 — 2026-09-06

New

• Synthetic empires now put a Shielded Silo over every warehouse worth one (#2885). Measured on 2026-09-06: the Silo, the one building that keeps a fifth to a third of a warehouse out of a raider's hold, stood at a median of zero on the agents' colonies holding 150,000 or more, because sixty-one of the seventy-six agents had never researched its node — it was nominated only while a capital was saving, and their research queues were full of other things — while humans took 10.7 million from agent warehouses in a day. From now on an agent empire holding 150,000 in any colony nominates the Silo's node to its research whenever it is within six purchases, and every colony holding that much, outposts included, puts the Silo up its build list to a useful level that differs from colony to colony, as long as its power grid can carry the next level — a colony in deficit buys power first. What you will see: Shielded Silos rising on agent outposts and capitals, and raids on them finding less in the warehouse.

1.115.2 — 2026-09-06

Fixed

• "Investigate an anomaly" is no longer dealt when there is nothing to investigate (#2868). Anomalies are surveyed once a week and only run down from there, so from midweek the daily board was handing out a card the galaxy could not supply — while the Galaxy map was already telling the same commander, on the same day, "No anomalies remain in the core sectors this cycle". Measured over the week of 31/08: the whole reachable stock of 119 anomalies was gone by Sunday, and the board dealt that card 690 times; of the 382 that went to human commanders, 27 were completed — 7.1%, against 24.8% for the next-worst daily card and 82.8% for the best.

  The board now asks the sky instead of your research tree. The card appears only while at least one anomaly your Astrometrics can actually resolve is still out there — unclaimed, and not already settled on. The Event Sector counts too, but only while a season is running, since that is the only time you can fly there.

  What that means at the board: on a day with nothing left to find you get a different daily card in that slot, or — if nothing else is within your reach either — a slightly shorter hand. The Daily Wrap scales with the hand it is dealt, so it never gets harder, and it is never left empty: if withholding the card would leave you with no contracts at all, you are dealt it anyway. A card you cannot finish is a bad card; a board with nothing on it is worse.

  One honest limit: the check runs when the hand is dealt at 05:00 UTC. If somebody else claims the last anomaly at 06:00, the card you were already given stays on your board for the day — hands are never re-rolled once dealt.

Also

• How many anomalies a week should hold is a separate question and is not touched here: it is tracked in #2572.

1.115.1 — 2026-09-06

Fixed

• A campaign clear that fits no longer says your warehouse is full (#2871). Clearing a level could show "Warehouse full: Ferrolite 0 did not fit. Raise your Warehouse or spend before the next clear." on the results card — directly under a Banked line listing the whole reward, on a colony holding a small fraction of its ceiling. The banked line was the true one: every unit arrived, every time. The refusal was rounding noise in how the receipt was worked out, and it appeared or not depending on where your stock happened to sit, which is why the same colony could see it on one clear and not on the next. The card now says only what happened. A real full warehouse still gets a real warning, with the real number.

Also

• The reveal's reward lines no longer name an amount they would print as "0" — a leftover too small to show is no longer announced as a refusal. The same fix covers the back-pay line.
• Read-only check of every stored campaign receipt: 139 clears across 44 players carried this false refusal, and none of them lost anything — the resources were in the warehouse the whole time. Nothing needed to be paid back. The 25 genuine refusals in the same data were real full warehouses and are unchanged.

1.115.0 — 2026-09-06

New

• Synthetic empires now keep their treasury at the capital (#2881). Measured over the day to 2026-09-06 09:00 UTC: 45 of the 64 human strikes on agent colonies landed on outposts and took 9.07 million there, 201,000 a strike against 86,000 at the capitals; the agents' 188 outposts held 58 million while their capitals held 237,000 each, because the freight line moved only a colony's dominant resource and a balanced warehouse never left. From now on an outpost's stock above its keep, 150,000 of a resource or the price of its next Command Centre rung, goes home to the capital on a four-hour clock when nothing more urgent is on the line (never past a ceiling of 450,000: a grown outpost saves for its own rungs from its own income), the capital sends an outpost only what it lacks under a 150,000 floor, never the treasury, and the freighter fleet is sized for those runs. What you will see: agent outposts holding a working stock instead of a hoard, more convoys heading home, and less loot waiting behind thirty hulls.

1.114.0 — 2026-09-06

New

• Synthetic empires now scout the target they chose without waiting a day (#2879). Measured on 2026-09-06: 27 of the 75 agents sat in the reconnaissance phase of a strike, every one of them having spent the day's six spy probes in an early burst of scouting, so the target they had chosen waited up to a day for its probe — and the phase after it, the fleet, waited with it; only six agents were gathering and three were ready. From now on a chosen target gets its probe outside the scouting cap, two a day of its own, sent at once instead of on a die roll. What you will see: an agent that picked your colony sends its probe within the hour, and the fleet follows its reconnaissance instead of a day-old picture.

1.113.0 — 2026-09-06

New

• Synthetic empires now climb their Command Centre (#2877). Measured on 2026-09-06: the agents' capitals sat at Command Centre 13 against the active humans' 18, and since every other building is capped at that level their mines sat there too; 72 of 76 could not pay the next rung while only four to five hours of income away from it, because the household budget switched itself off for the life of an attack operation, the shipyard's four Bombers a tick out-ate the Xenogas income, and the build pass bought lesser rungs from the money it was saving. From now on, when the next rung is within a day of the capital's own income, the budget holds through a gather (only a siege thins it), a gathering shipyard buys the strike's own hull or waits, the build pass keeps its own reserve for the rung and buys the Silo that roofs it, and the freight line brings the Xenogas the Command Centre waits for home from the outposts before anything else. What you will see: agent capitals whose Command Centre, mines and shipyards climb a level at a time, freight convoys bringing Xenogas home, and gathers that fill more slowly while a capital saves.

1.112.0 — 2026-09-06

New

• Synthetic empires now turn a full warehouse into a fleet (#2873). Measured on 2026-09-06: the agents held 74.5 million in their warehouses, a median of 416,000 per empire, while their fleets sat at a median of 430 combat hulls against the active humans' 2,222; 23 of 76 capital shipyards were idle with a hull affordable, because the shipyard's caps were the fixed tables of a starving-economy simulator, and a scientist's mix was the probe alone. From now on the shipyard's per-type and per-order caps grow with the purse of the world it stands on, one multiple per 100,000 of Ferrolite and Xenogas, and a scientist or a diplomat keeps Monitors and Corvettes at home instead of a million behind thirty probes. Below 100,000 nothing changes; the simulator and the Wraiths keep the old caps. What you will see: bigger agent fleets in the docks, sized to what each empire actually holds, and less loot waiting in their warehouses.

1.111.0 — 2026-09-06

New

• Synthetic empires now garrison their outposts (#2870). Forty of the fifty raids humans flew against agent colonies in sixteen hours landed on outposts, thirty-six of them on an empty dock over a full warehouse, and 7.5 of the 8.5 million taken came from there; the agents' developed outposts all had a Shipyard and most had no combat ship at all, while human outposts of the same size hold thousands. From now on every agent outpost with a Shipyard arms the world it stands on from its own warehouse: Monitors, Corvettes and Interceptors, one per Command Centre level plus one per 25,000 in stock once the warehouse holds 100,000, up to the commander's fleet cap, and those hulls stay home instead of leaving on the next sortie. A raid on an agent outpost now meets a fleet.

1.110.2 — 2026-09-06

Changed

• Synthetic empires now build the freighters their hoards need (#2867). The freight line shipped this morning moved a colony's hoard on the freighters at hand, and the freighters at hand were the bottleneck: a median of three per colony, 2,500 a hold, so a three-million hoard was forty-plus round trips. From now on an agent's home shipyard keeps a cargo fleet sized to what the line has to move, one freighter per 150,000 of hoard with somewhere to go, up to twenty-four, and counts the freighters away on a fetch as owned instead of ordering them twice. Nothing changes for an empire without a hoard.

1.110.0 — 2026-09-06

New

• Synthetic empires now run freighters between their own worlds (#2863). Until now an agent could sit on three million of Ferrolite in one colony while a neighbouring colony of the same empire starved for it, and raiders found those warehouses full and undefended: in sixteen hours humans took 8.5 million from agent colonies, the hangar empty at impact in 36 of 50 strikes. From now on an agent moves the resource one colony is drowning in to the own colony that is short of it and can spend it, the receiving world sends its freighters to fetch it, or the hoarding world delivers with its own, one convoy at a time, until the two are in balance, and never to or from a world with a fleet inbound. Nothing is sold and nothing is built by this rule; a fleet of freighters sized to the hoard is the next step.

1.108.0 — 2026-09-06

New

• The system list now shows which planets you have already probed (#2841). Open a system in list view and every slot where one of your spy probes found a commander carries your own report: the espionage mark and how long ago that probe landed — 🛰️ 3h ago. A slot you have no report on carries nothing, and a line at the foot of the table explains the mark.

  It came from a player. TartarusMkII, in universe chat, had just found how quickly you can send probes from the list view and then hit the wall: "I cant see that Ive already scanned a planet from this view" — and asked whether there was any way to record what he had found in the game, or whether he should keep the coordinates in a note outside it. The answer, until now, was outside it.

  The odd part is that the game already knew. The planet's own card has remembered your whole scouting history for days: open one slot and it tells you about your report however old it is. The list — the screen built for working through many slots quickly — was reading from your activity feed instead, which holds the fifty most recent things that happened to you plus a handful of rescued reports. So the screen you use one planet at a time remembered everything, and the screen you use to get through twenty remembered about three. Probes were being spent twice on the same coordinates.

  Both screens now read the same history. Your scouting is in the game: open a system and the slots you have already worked are marked, with their age.

Also

• A report has a date on it, and nothing else. The mark says when your probe landed and leaves the judgement to you. It deliberately does not expire, fade or grey out with age, because nothing on the server expires either — a report you took a week ago still pierces the Veil and still unlocks the launches it always did, so a screen that made it look spent would only have talked you into paying for a second probe you did not need.

  There is one thing that genuinely voids a report, and the game already had a sentence for it: a report from the Event Sector taken before the current season describes worlds the end-of-season sweep has emptied and repopulated since. Those marks say · pre-season, in the same words the planet's card and the mission composer have always used.

• The mark is only ever your own report, never a live reading of the slot. It shows what your probe found when it landed, and nothing whatever about a coordinate you have not probed yourself — a planet nobody has sent a probe at looks exactly the same to you as one your neighbour scouted this morning. The Commander column is untouched: a free planet and a colony under an Eschaton Veil still read identically there, and the mark appears on occupied and empty slots alike, so its presence never says anything about what is standing on the slot.

• What this release does not do, and the report asked for it: there is still no curated hit list — no saved shortlist of scouted opponents with your own notes against each. That was raised in the same report and it carries a real design question with it, which is whether a hit list only ever helps the attacker while the defender gets no equivalent. That question is still open, so it was left alone rather than answered quietly. What changed is the reason it was asked: the knowledge a probe buys you is now kept, and shown where you next decide what to do with it.

• One kind of probe is deliberately not marked, and the wording on screen is careful to say so: a probe that lands and finds nobody there files a different kind of record, one that carries no coordinates to match against, so it leaves no mark. On the live galaxy that is 162 probes against 20,880 reports of a commander found.

1.107.19 — 2026-09-06

Fixed

Your screens no longer disagree about an incoming attack (#2846).

The Sensor Grid is what pierces the Nightfall Protocol: at level 8 anywhere in your empire, a fleet flying under it is supposed to show on your warning surfaces from the moment it launches. But a Sensors level takes a short while to be written back to your colony after it finishes, and for that stretch only some screens counted it.

So a commander who had just finished Sensors 8 could see the attack on the Incoming panel and, in the same minute on the same colony, be told by the build advisor that their early warning was showing nothing — with that reassurance attached to the advice to spend on economy rather than defence. The guild door badge stayed dark over a guildmate whose entry on the threat board behind that door was already showing the attack. And the alert and phone push that should have fired at launch were held back while the panel counted the fleet down: the fleet arrived on a silent phone.

All of these now read your Sensors at the same instant, so they say the same thing.

Changed

A defender whose Sensors 8 has just finished can no longer be launched at from the dark (#2846). They were always meant to see a Nightfall attack from launch — that is what the level is bought for. Now the alert and the push arrive with the panel, instead of the panel alone. Nothing became visible that you could not already see somewhere else; the screens that were quiet simply caught up with the ones that were not.

1.107.18 — 2026-09-06

Fixed

• Synthetic empires now send a force sized to win, instead of massing for days (#2854). Before, an agent that had picked a target built the smallest fleet that could break its walls, judged that it would lose, and kept "gathering" for a bigger fleet it already had: on 2026-09-06 eleven of the sixteen agent operations in that state could have won outright with the ships in their hangar, six of them with certainty. Now the agent tries a larger force, doubling it step by step up to everything it can spare, and launches with the first size its own estimate says will win. An agent that really is outgunned still waits, as before.

Also

• Synthetic empires stop farming the Pirate Beacon once they have outgrown it (#2835, live since 2026-09-06 02:37 UTC with fc07625; its release was held by the deploy check that #2845 fixed). Over the previous week the agents struck the Beacon 776 times for a median of 62 resources a strike, 241 of those strikes taking nothing, and lost 229 of the fights. An agent whose Power gate is past the tutorial bar now leaves the Beacon alone and keeps its fleet for colonies worth the sortie, unless the day's contract card asks for one strike; a young empire still learns to fight there, with the same daily cap as before.

1.107.17 — 2026-09-06

Fixed

• The fleets-in-transit board and the mission advisor each stopped doing the same work once per reader (#2818). Both keep a short-lived shared answer — the contact board for 15 seconds, the advisor for a minute — and both used to mark that answer fresh only once it had finished being built. So everyone who asked during those few hundred milliseconds started their own identical rebuild: eight Galaxy tabs opening together, or a wave of synthetic empires waking on the same tick, each rebuilt the whole board. Now the first one builds it and the rest wait for the same answer. Nothing on screen changes; the server simply stops repeating itself.

  Measured read-only against the live database, before the change and again against the deployed build. Eight simultaneous readers of the contact board cost 2.04 seconds and a 215 ms worst pause before; they now cost 262–290 ms and an 81 ms pause — the same as a single reader, which is the whole point. Four readers went from 912–938 ms to 246–280 ms. A single reader is unchanged at 222–275 ms, and that is the control: one caller never duplicated anything, so the fact that it did not move is what shows the saving at four and eight is the duplication going away rather than a quieter galaxy.

  It was wasted effort and database connections, not a freeze anybody watched. This is the last and smallest of the four caches that had the same gap.

  For the mission advisor the saving is a double-click or a retry rather than a crowd, because that answer is private to the player who asked for it: it is computed from their own colonies, research and ship stars. That privacy is now pinned by a test that fails if anyone ever makes the advisor's cache shareable between two players.

Also

• Corrected a stale note in the sovereignty code that claimed the galaxy map and the home routes read their control data through it. They do not, and neither of the two readers it named has any caller in the shipped game.

1.107.16 — 2026-09-06

Fixed

• A session left open on an account that has since been removed now ends promptly (#2845, closes #2844). Before, a tab or an installed app still holding the key of an account the abandoned-accounts sweep had removed kept polling the server until the key expired, and on the server every one of those polls was logged as if presence had stopped being written for everyone, which is also what turned every deploy check red for an hour on 2026-09-06. Now the server recognises the missing account on the first poll, refuses that key from the next request on with the same answer the session-recovery door already gives, and logs one quiet line an hour with the count. A live account sees nothing new.

1.107.15 — 2026-09-06

Fixed

• The synthetic empires start striking on their own again (#2807). Since 01/09 the agents had almost stopped attacking as individuals: solo strikes fell from around 300 a day to between 13 and 24, and stayed there for five days. Nobody caught it, because guild raids kept running at full volume and the two were being added into one number — the collapse only became visible when raids stopped too, and the person who said so was a player, QQQQ, reading his own battle log.

  We broke it. A change of ours on 01/09 let an agent open a scouting job on a colony it had not yet probed — sensible on its own, and it fixed a real deadlock — without noticing that an open job took over the whole of that agent's attack. From then on the agent looked at that one colony and nothing else, and attacked nothing at all for as long as the job lived, with targets it could have taken sitting in the same view. The same change also gave the strategists a much better list of targets, so nearly every agent picked up one of these jobs within the hour; one half of it armed the other. Deciding to look at a colony now costs an agent nothing else: it sends its probe and still takes a target it can see, in the same turn.

  What we can and cannot claim yet. The clearest sign it is working is inside the agents: the jobs that had been stuck now move, and for the first time there are agents holding a scouted target with a force ready to fly. The strike count is a weaker signal so far — the first half-hour after the deploy is consistent with recovery, not proof of it, being one half-hour in the quietest part of the night on a service that had just restarted. In the same half-hour window the three broken nights produced nothing at all.

  This does not bring back the old figure of ~300 a day, and it should not. Almost all of those were strikes flung at colonies nobody had scouted, and closing those was a deliberate choice on 04/09, after we measured that 91% of the attacks agents lost had never been probed. What comes back is the strike an agent makes when it knows what it is flying into.

Also

• The raid fix from the day before held. Guild raids, which stopped dead on 05/09 (#2800), restarted and are still running. They are not yet back to their old volume — the first hour brought seven raid contingents against a historical fourteen to twenty — so that one is recovering rather than recovered, and is worth another look.
• An agent that has told another player in chat that a fleet is coming, or that has spent days gathering hulls for one target, now keeps that commitment: neither is mistaken any more for a passing glance at a colony. An agent also no longer talks about scouting a colony that has since changed hands, and no longer opens a scouting job on the pirate Beacon it is already attacking.
• Still to do (#2808): the counter that would have caught both collapses on the day. Two things it has to get right, or it will not. It must separate solo strikes from raid contingents — adding them together is precisely what hid this for four days. And it must separate scouted strikes from unscouted ones: the old healthy number was 99% unscouted strikes, so an alarm set on the raw total would be demanding we bring back the very behaviour we removed on purpose, and would have gone off on 04/09 when we did.

1.107.14 — 2026-09-06

Fixed

• The build advisor no longer tells you there are no threats nearby (#2822). The advisor's commentary used to open with lines like "production is strong and stable, with no threats nearby, so it's a good time to push economic upgrades" — a verdict on your safety, offered as the reason to spend on economy instead of defence. The advisor cannot make that judgement. The only danger signal it has is the inbound scan for that one colony: attacks already launched at it and already detected. It knows nothing about ships still in dock, nothing about a neighbour's strength, nothing about attacks aimed at your other colonies, and it has no notion of "nearby" at all.

  It now says what it actually knows — that nothing is showing on that colony's scans — and says plainly that a clear scan is not safety. When a fleet is detected, the wording is unchanged; that claim was always honest.

  Nothing about the game itself moved: what counts as a threat, and what a defence level is worth, are exactly what they were. Only the sentence changed. And it is the same sentence either way — a colony with a fleet hidden under the Nightfall Protocol reads word for word like a colony with an empty sky, so the new wording tells nobody anything about what is hiding.

  Reported by TartarusMkII, who asked how the game had judged that there were no threats nearby rather than assuming it had.

1.107.13 — 2026-09-06

Fixed

• The tech tree no longer quotes a price your warehouses cannot hold without saying so (#2817). Research is paid out of a stockpile, and a stockpile has a ceiling. The tree showed the price and, under it in red, how much was still missing — and never mentioned that the number could be larger than anything you are able to store. Tibeh01 reported it on the Tier 9 War Crucible: it asks for 2.67B Ferrolite, his warehouses top out at 472,632,000, and the screen let him save toward it anyway. A node priced above what your storage can ever hold now says so, names the resources that are walled, and points at the Warehouse: "Lava Hole tops out at 472M per resource — the Ferrolite, Xenogas and Data gaps will not close by waiting. Raise the Warehouse."

  The line is careful about what it claims. It appears only where a gap really cannot close, so it stays quiet on a resource you can already pay for, on an order already running, and on any price a free Requisition Surge could still bank for you. It names only the legs that are actually blocked — Dark Matter has no storage cap at all, and the other resources usually fill on their own. And it quotes the same warehouse figure the Colony tab shows, printed in full digits whenever compacting it would make it look identical to the price it is calling out of reach.

  Nothing costs more or less than it did, and no warehouse holds more or less: the ceiling was always there, it just never appeared on the screen that asked you to save for it. 31 players currently have at least one unlocked node priced above their ceiling. For every one of them the wall is real but not permanent — the tree contains no dead content, and raising the Warehouse clears it.

1.107.12 — 2026-09-06

Fixed

• The galaxy map is built once a minute for everyone, instead of once for every reader who happens to arrive while it is being built (#2816). The political map is assembled from an edition — one scan of every colony in the galaxy, plus the guild, sovereignty, relic and outpost reads around it — that the server keeps for a minute and shares between all players. The note saying "this edition is fresh" was only written once that work had finished, so for its whole duration every player opening the map started another identical scan of the galaxy. Eight players landing together did the same work eight times over.

  Nobody was looking at a broken screen because of this, and it should not be dressed up as a rescue: what it cost was database work and connection-pool time, not a freeze. No page went blank and no request failed. Measured against the live database with 1763 colonies, at eight players arriving together it went from 3.0 seconds to 0.4 and the longest moment the server spent unable to answer anything else went from about a third of a second to under a tenth. The cost no longer grows with the number of players looking.

  This one mattered more than its two predecessors because the map is the most-polled screen in the game and its refresh interval is exactly the server's own, so nearly every poll landed on a rebuild — and because the sovereignty sweep marks the map stale every five minutes, which made every player miss at the same instant rather than by accident. That scheduled pile-up is now a single rebuild that everyone shares.

  The same fault has now been fixed in four places (#2790, #2802, #2816); this was the last large one. The map's contents are unchanged — every reader still sees exactly what their own Sensors, guild and research entitle them to see, which was the part that had to be proved before anything could be shared at all.

1.107.11 — 2026-09-06

Fixed

• Beating a derelict no longer keeps it awake for the next attacker (#2821). Attacking an abandoned empire is meant to earn no infamy — a corpse cannot defend, so the power gap is maximal and hitting one was the cheapest reputation in the game. The rule that decided whether an empire counted as abandoned looked at the last write to its colonies, and that timestamp is moved by whoever attacks it. So striking a derelict reset the clock that was about to stop paying for striking it, and a target hit more often than once a week never stopped paying at all. Silence is now measured by when the owner was last here, which nobody but they can change. Sixty-six empires whose owners had been gone a week or more — every one of them being farmed, between them absorbing 850 attacks in the last month — stop rewarding their attackers, on the target board, on the planet sheet, in the solo resolver and in guild raids alike.

Changed

• The game now says "silent" about the owner, not the planet. The planet sheet used to read "nobody has answered here in days" over a colony a rival guild might have razed yesterday, and the activity feed said the same of an empire it was showing you a battle on. Both now say what they actually mean: its owner has not signed in.
• Settings explains the seven-day stage. The leave blurb listed what happens after 14 days and after 30, but never mentioned that from day 7 an absent empire stops costing its attackers anything — which is the stage a player would actually notice first. It says so now.

Also

• The ruin and the account removal were deliberately left exactly where they were: an empire that has only just fallen quiet still has a fresh colony clock, so nothing is newly halted, decayed or deleted. Verified on the live galaxy before and after the deploy — the "in ruin" and "removable" counts do not move at all, while "silent" goes from 250 to 316.
• The same defect in the account removal is written up rather than fixed (#2830): abandoned guest accounts are un-deletable for the same reason, 23 of them holding a starting slot in the cradle, but deleting them is irreversible and rests on a threshold that is still an open question (#2558).

1.107.10 — 2026-09-06

Fixed

• The AI empires start fighting again (#2800). For a day they did not. Guild raids by synthetic empires went from about a hundred a day to none at all, and because most AI attacks arrive as a raid contingent rather than a lone strike, that was almost every attack they make. A player noticed and told us before we noticed ourselves: QQQQ, in universe chat, reported his log showed one AI attack in twenty-four hours and it was against the pirates. He was right.

  It was our change that did it. A day earlier we had given the raid two new rules at once: only call the guild onto a colony you have actually scouted, and do not call it onto a fight your own contingent would lose flying alone. Each is sensible on its own. Together they closed the door completely — the second rule only switches on once the first one is satisfied, and the first one asked for scouting reports of colonies nobody had ever had reason to scout. And the second rule could only ever say no: a raid exists precisely for the colony too strong for one fleet, so asking whether one fleet could take it alone is asking whether the raid is pointless.

  The lone-fleet rule is gone. What decides a raid now is the forecast for the coalition that would actually arrive, which is the question that matches the verb. Scouting still comes first — a colony you have seen always outranks one you have not — but it is a preference now rather than a locked door, and an unscouted target still has to clear the strength, garrison and defence checks before anyone is called. Those are the checks that stop a guild throwing itself at a wall, and none of them needs a probe to have landed first.

Also

• Two things this release does not fix, both found while measuring it and both filed: solo AI strikes (the ones a single player is most likely to see in their own log) fell sharply on 01/09 from a different cause and are still low (#2807); and nothing on our side watches for an AI behaviour disappearing, which is why a player got there first (#2808).

1.107.9 — 2026-09-06

Changed

• Synthetic empires strike what they can beat and stop besieging what they cannot (#2814). Over the past week the agents launched 515 of their 759 strikes on humans at a defending fleet worth at least four times their own, and lost every one; at two to one or better they won 139 of 146. The losers were war and league targets, which the fleet ceiling exempted because it read the leaderboard, a proxy. A spy report is not a proxy: from tonight a scouted fleet the empire could not launch against (its deployable share over its safety margin, about 0.6 of its combat hangar for a hawk and 0.3 for a timid one) is never a target, whatever the war says, and an enemy colony the probe finds empty is the one struck. The agent says so too: its reality sheet carries the strike it called off and why. An operation whose target turns out to be beyond reach ends at once instead of holding the agent's scan on that colony for two days; an operation that has not launched in three days is dropped; and an agent assembling a strike against emplacements now builds the bombers that break them instead of waiting with a hangar of Interceptors. Yesterday the agents launched 26 attacks all day against 420 the day before, because 65 of 75 were waiting on exactly such walls.

1.107.8 — 2026-09-06

Fixed

• The galaxy target board is built once a minute, not once per reader (#2802). The list of every colony you could attack is assembled from a single sweep of the galaxy that the server keeps for a minute and shares between everyone looking at it. The note saying "this sweep is current" was only written once the sweep had finished — so while one was running, every player who opened the board was told the shared copy was out of date and set off a fresh sweep of their own. Eight players arriving together did the same work eight times.

  Nobody saw this as a broken screen, and it is worth being plain about that: the board always showed the right colonies, and it is not what caused the ten-second freeze fixed a few releases ago. What it wasted was the server's own time and its connections to the database — three seconds of duplicated work for eight readers, on a machine that also has everyone else's turn to run.

  A reader arriving while a sweep is running now waits for that one instead of starting another. Measured against the live galaxy at 1757 colonies: eight simultaneous readers went from just over three seconds to four tenths of a second, and the longest the server spent unable to answer anything else dropped from about a third of a second to under a tenth. The board itself is unchanged — same colonies, same order, same numbers.

  What each reader sees is still worked out for them alone: what your Sensors can resolve through an Eschaton Veil, your own Gate band, the bounties, the ledger of grudges owed to you and the pacts that refuse a strike are all still decided per player, after the shared sweep. Only the sweep is shared, never a point of view — and the test suite now holds two different players to one sweep and checks they still get their own answers.

Also

• The same missing shape was found in four more places while fixing this one. The largest, the political map's own edition, is filed as #2816 with its measurements; the remainder is filed as #2818. Neither is fixed by this release.

1.107.7 — 2026-09-06

Fixed

• The chat no longer decides your history ended just because a page came back short (#2794). Scrolling back through a conversation could hit a wall that was not there: the "↑ Earlier messages" button stopped being drawn, scrolling up loaded nothing, and jumping to a pinned message gave up — while the messages were still sitting on the server the whole time. It happened because the chat judged whether older messages existed by counting how many arrived, and a page is only ever a window of at most fifty messages you are allowed to read. When someone's account was being deleted at the moment you loaded, their messages dropped out of that window and a perfectly ordinary page arrived at forty, which the chat read as "this is the whole conversation". It stayed that way until you switched channel or reloaded, and nothing told you. Now only an empty page ends the history, so the past stays reachable.

Also

• A channel with fewer messages than a page — a new guild, most whisper threads — now offers "↑ Earlier messages" once. Tapping it finds nothing and the button goes away. That is deliberate: it costs one request to be sure, and the alternative was hiding conversations that really existed with no way for you to tell.

1.107.6 — 2026-09-06

Fixed

Upgrading a building past the point where it stops helping now asks first (#2804)

A few buildings stop improving well before level 30 — the Colonial Outpost after level 1, the Trade Portal after 16, the Autonomous Factory and the Orbital Arsenal after 1 — while their energy drain keeps climbing with every level. The game already said so, on the button and beside it. What it did not do was pause: one tap on Upgrade and the order was placed, and since buildings cannot be demolished or downgraded, a mis-tap was permanent.

Now that tap opens a question instead. It names the building and the level, repeats what has stopped improving, shows exactly how much more energy the level will draw for as long as the building stands, and says plainly that it cannot be undone. Confirm and the upgrade goes ahead as before — the level still earns colony points and structural power, and the Outpost's artwork still advances at levels 10 and 20, so this is a purchase some players make on purpose and it stays available. Cancel and nothing is spent.

The question is asked wherever the upgrade can be started: the building sheet, the build menu, the Colony tab's rows and the advisor's cards.

Reported by T04st.

Also

Colonies that are already above one of those levels are left exactly as they are — this change prevents the next mistake, it does not undo an earlier one. Being able to demolish or downgrade a building is a separate, still-open request (#2626).

1.107.5 — 2026-09-06

Changed

• A farmed synthetic capital now buys a garrison, keeps it home, and keeps saving (#2796). Measured on 2026-09-05, 48 of 316 synthetic empires had three or more battles at home in two days. Under the household budget shipped in v1.106.12 such a capital spent its whole construction reserve and handed the shipyard every unit of income, and the aggressive personas spent it on Titans and Scythes that left on the next sortie. Now, from two battles at the capital in two days, the reserve frees about half of what the capital holds and no more, the shipyard buys Monitors, Corvettes and Interceptors up to a garrison sized on the Command Centre, those hulls stay home while the siege lasts, research and supply runs keep the calm reserve, and the agent's own reality sheet says its home is under siege. Battles at an ally's world or at a distant outpost no longer count as a siege of the capital. Also carried in from the previous release's review: the defence nomination counts research purchases when choosing the nearest rung.

1.107.4 — 2026-09-05

Fixed

The game no longer freezes for ten seconds at a time (#2790). Every so often the whole server would stop answering — not one screen, all of them at once — and anything you had in flight at that moment simply failed. Opening the galaxy map returned an error, the chat refused to load, a fleet order came back as a red banner. It cleared up on its own after a few seconds, which is exactly why it was hard to pin down: by the time you retried, it worked.

The cause was one screen asking the same question eighteen times. The galaxy view checks the Power rating of the owner of every planet slot it draws, and each of those eighteen checks was starting its own full recount of every empire in the game instead of waiting for the one already running. The recount is not interruptible, so eighteen of them ran back to back while the server could do nothing else — long enough for anything already open to give up and fail.

Now the first check does the work and the other seventeen wait for its answer. Measured against the live database, before and after: with eighteen at once, the server used to lock up for 1.9 seconds and take 8.9 seconds to answer; it now locks up for 1 millisecond and answers in 0.56 seconds. With thirty-six at once the freeze was 5.2 seconds; it is now 1.7 milliseconds. The cost no longer grows with how many people are looking.

Also

The operator panel can now see a freeze while it is happening. The admin dashboard has a new "API event loop" card showing how long the server has gone without responding — the worst pause in the last minute, the worst since the last restart, and when it happened — and the server writes a warning to its log whenever it stops for a second or more. This condition used to be visible only by inference, reconstructed after the fact from failed requests in a log that is wiped on every deploy. It is now a number anybody can read.

Not covered by this change: the galaxy target list has a milder version of the same duplicated-work pattern. It was measured at 216 milliseconds rather than seconds, so it is wasted database effort rather than a freeze, and it is left for its own fix.

1.107.3 — 2026-09-05

Fixed

• Opening the chat no longer fails when the server is busy (#2788). The chat panel could answer with a generic error instead of the conversation — three times inside one second on 05/09, on the screen the game is most read on. The messages were never the problem: the page was being fetched under a ten-second deadline that the server had to meet while it was busy doing something else entirely, and when it missed the deadline the whole page was thrown away. The chat now reads without that deadline, so a busy moment costs the panel a pause instead of an error, and the pinned-orders bar is read the same way. If a player's account is deleted at the exact instant you are reading, their message quietly leaves the page — which it was about to do anyway — instead of taking the page down with it.

Also

• The stall that made the chat miss its deadline is a separate defect and is not fixed here: it is filed as #2790. Until it is, the galaxy map can still fail the same way — the chat is now immune to it, the map is not.

1.107.2 — 2026-09-05

Changed

• Synthetic empires now put a roof over their savings and a wall in front of a farmed capital (#2785). Measured on 2026-09-05, active empires older than two weeks: 46 of 66 human capitals held the Shielded Silo node, 50 the Orbital Fortress, 51 the Orbital Cannon; of 75 agents, 14, 0 and 13. Human strikes on agent capitals won 55 of 58 in a day. Before, an agent capital saving for its next building was saving for the raider. Now, while the capital holds a construction reserve, its research pass is nominated the nearest missing rung toward the Shielded Silo (the Discipline level it lacks, or the prerequisite node, or the Silo itself once available), and the Silo goes up its build list up to a useful level of five or six. When the capital is hit at home twice in two days, the same happens for the Orbital Fortress and the Orbital Cannon, whichever is closer to hand and within six Discipline levels. A nomination reorders research, it does not decide it: the strategist's own research priority still comes first, and an unaffordable rung waits its turn.

1.107.1 — 2026-09-05

Fixed

• An abandoned empire under regular attack could sit out the ruin entirely, and a returning player could be told twice what it cost them (#2762). An empire left silent for two weeks falls into the ruin, which halts its production and takes part of its garrison for every further day of silence. The game measures that silence by when the colony was last touched, and the ruin deliberately writes to a colony without disturbing that mark — otherwise it would keep resetting the very silence it is counting.

  The trouble was that the same mark is also how every other write to a colony notices that something changed underneath it. A battle that read the defending colony just before a ruin round, and finished its calculation just after, still looked current — and saved the older picture over the top. Measured on the real code before changing it: with an attack landing on each round, a derelict's garrison was never worn down at all, holding at 1000 ships where the same colony left alone fell to 662 over five days, with its ruin clock pushed back to where it started. The empires most likely to be attacked every few hours are exactly the abandoned ones, so the rule was weakest precisely where it was meant to bite.

  The same overwrite could also put back a notice that had already been sent. A player coming home is told what the ruin destroyed and the record is then cleared; an attack landing at the wrong moment restored it, and the next round sent a second, identical message — and put the production halt back on an empire that had just been restored.

  Colony writes now check the colony's own contents as well as the clock, so anything that changed in between is noticed and the write is redone against what is really there. Nothing about how long an empire has been quiet is measured differently, and the ruin's rate and thresholds are untouched.

• A razed empire could be frozen with ships the ruin had already destroyed (#2762). When a colony is taken off the map by a Forced Exodus, the empire is frozen into an offer the owner re-founds from. That snapshot was taken with the same blind spot as above, so a razing that straddled a ruin round could hand the player back a garrison that no longer existed, along with a loss notice they had already received. It is the one place where nothing later corrected the mistake, because the colony the snapshot came from was gone. It is now guarded like every other colony write.

Also

• Whether any player was affected before this cannot be established from the data, and this entry does not claim nobody was. The ruin keeps only its current position and no history of what it charged, and the server logs that would show it do not survive a deploy. That blindness is itself an open defect (#2758). What is certain is that the message the ruin sends had never yet gone out to anyone, so no duplicate can have reached a player.
• The fix covers every battle, raid, arrival, market trade and colony build. A further set of colony writes still guards the old way; they are narrower windows and are tracked separately (#2782, #2786).

1.107.0 — 2026-09-05

New

Novarchy installs on Android (#2778)

Open novarchy.com in Chrome on an Android phone and the browser menu now offers Install app. The game lands on your home screen as a launcher with the Novarchy icon and name, and opens full-screen — no address bar, no browser chrome, the whole display for the galaxy.

Before this, Android had no install path at all. The game shipped a web manifest and the seven icon files it needs, served them correctly, and had done so for months — but no line in the page ever pointed at the manifest, so no browser ever fetched it. On top of that the manifest itself was missing the name, the start address and the display mode that make a site installable. Both halves are fixed: the page links the manifest, and the manifest is complete.

If you are on iPhone, you already had this and were never told: Add to Home Screen in Safari has quietly produced a standalone, browser-chrome-free Novarchy app for months, because iOS reaches the home screen through tags the page already carried. Nothing there breaks. One thing improves — iOS has also read the manifest since version 16.4, and the page declares no icon of its own for that purpose, so a home-screen icon added from now on should be the real Novarchy crest rather than a snapshot of the page.

The home-screen entry starts the game at the site root and keeps the whole site inside the app, so links to the privacy and support pages stay in the window instead of throwing you out into a browser tab.

What was tested, and what was not. Every published Chrome installability criterion was checked against what the live site actually serves, using real Chromium driving the same manifest parser the browser uses to decide: it now finds the linked manifest, parses it with no errors, and all eight criteria pass where seven of eight failed before. That the install entry appears on a physical Android handset follows from those criteria but was not observed on a device — there was none to test on.

One thing deliberately not claimed: Android will not pop up an unprompted "install this app" banner. That banner needs a service worker doing work this game intentionally does not do. Installing is there when you go looking for it in the menu.

Also

The launch splash colour now matches the colour the page itself paints first, so opening the installed app no longer flashes between two slightly different blacks.

1.106.12 — 2026-09-05

Changed

• Agent capitals now save for their next Command Centre (#2775). Before: an agent capital that had every building at the Command Centre's cap spent each hour's income on a pair of hulls the moment the ship lane freed, and never accumulated the price of the next upgrade; human capitals of the same age sat eleven Command Centre levels higher. Now: the capital keeps a construction reserve for the building it wants next, and the shipyard, research, the guild donation and the outpost supply run spend only what lies above it plus a share of the income between two spends. Under threat the reserve thins and the hangar's share grows; while an attack operation is gathering the shipyard sees the whole purse. Expect agent capitals to climb again, and their hangars to fill more slowly while they do.

1.106.11 — 2026-09-05

Fixed

• A quiet season no longer means an invisible one (#2766). When a season ends, the game hands out placements and end-of-season rewards. Until now it only looked at guilds that had actually put points on the board: a guild that finished the season without scoring was not ranked at all, took no place, counted towards nobody's total, and its members received nothing. On today's galaxy that was 8 of the 9 guilds — and the season would have been recorded forever as a one-guild season, with its winner shown as "#1 of 1".

  Now every guild that played the season is ranked, and the number beside a placement is the real size of the field. Guilds that had a quiet season still take their place and their members still collect the end-of-season reward.

  No season has ever closed yet, so nobody has lost anything — this lands before the first one.

• The podium still has to be won. Because everyone is ranked now, a season where only one guild scored still has a second and a third place. Those places do not carry the podium rewards or titles unless the guild actually scored: finishing above someone else because nobody scored is not the same as winning, and it is not paid as though it were.

• Who counts as being in the season. A guild is in the field if it was already there when the season opened, or if it put points on the board during it — so a guild founded in the final days does not collect a full season's reward, while a guild founded mid-season that went on to compete still places properly. Guilds made up entirely of the Wraiths are ranked the way they already are on every other board: not at all.

1.106.10 — 2026-09-05

Fixed

• A garrison built around a Destroyer could not be beaten by any single attacker, however large a fleet you brought (#2769). The flagship at its damage floor absorbed every shot aimed at your fleet, and while it did, whatever was parked beside it was never fired on at all — so the attacker lost two thirds of its ships to the retreat rule while the defence lost nothing. The same trick worked on the attack: a flagship and one escort could take a colony off any fleet in the game. The stalemate rule written to end exactly this could never fire, because the defence was never "reduced to" the flagship alone while the sheltered escort lived.

  Your Destroyer is exactly as indestructible as it was: nothing you can face alone will ever kill it, and it still takes a coalition of two empires eight rounds to bring one down. What it no longer does is make everything standing next to it immortal too. Once its floor has been spent, the enemy's guns move on to targets they can still hit — so those fights now end, as a draw or as a flagship withdrawing alive, instead of grinding for thousands of rounds towards a result the attacker could never change.

  Reported by Spacca-Mazza, who built the counter-example after this desk had twice told him the scenario was a draw. It was not, and he was right.

Changed

• Battles involving a lone Destroyer or Annihilator against a single opponent will resolve differently, and usually far faster. Measured over 2,000 battles before and after: fights with no legendary hull are completely unchanged, and so are fights against a real coalition of two or more empires — the mechanic that arms this is the flagship's own damage floor, and a coalition switches that floor off. Where results do move, the average fight falls from about 7,700 rounds to under 200, and no change anywhere creates a win the flagship's owner did not have before. Across 800 such battles the flagship survived every single one.

Also

• Fleet forecasts inside an already-installed phone app are computed on the device, so until the app updates they may still predict the old outcome for these fights. Every battle that actually resolves is decided by the server and is correct immediately.

1.106.9 — 2026-09-05

Changed

• Agent empires now develop the worlds they settle (#2764). Before: an agent's outpost grew on one Extractor's income and nothing ever came from the capital, so at two weeks it held a handful of building levels where a human's held hundreds, and an outpost with a shipyard kept no ships. Now: every few hours the capital sends its freighters to the outpost that has waited longest, with a balanced load sized to what the outpost can hold; the shipyard builds a freighter for each outpost that is behind; an outpost whose build lanes are busy keeps a small garrison of its own; and a young outpost builds production first whatever the capital's plan says. Expect agent outposts to catch up with their human neighbours over the coming weeks, and their supply convoys to show up in space.

1.106.8 — 2026-09-05

Fixed

• Coming back to a smaller fleet is no longer a mystery (#2757). An empire left silent for two weeks falls into the ruin, and a ruined empire loses part of its fleet for every further day of silence. The game never said so — no message, no report, no line of text anywhere. You came back after three weeks, found ships missing, and the only conclusion available was that somebody had attacked you, so you went looking for a battle report that does not exist. The game now keeps a record of what the ruin destroys and sends you a message when you return: what was lost, how many of each, and plainly that no battle and no other player took them. It also says what the ruin did not touch — buildings, defences and stored resources are never part of it — and points at the declared leave, which holds the ruin off for up to three weeks if you know you will be away.

  The message goes only to empires that actually lost ships. If the ruin took nothing from you, you hear nothing.

Also

• Nothing about the ruin itself has changed. No threshold, no rate, nobody new becomes eligible. The mechanic has worked this way since August; what is new is only that the game admits to it.

• Losses from before this release cannot be itemised, and are not guessed at. The only thing the game recorded until now was a running billing position, and that mark is also written when the ruin destroyed nothing at all — so it cannot tell a player who lost a fleet from one who lost nothing. Rather than send a vague message to people who may have lost nothing, players whose ships were destroyed before today are told nothing. On the live galaxy that is 157 commanders carrying the old mark, of whom only five still have a fleet left for the ruin to take. Being able to count what the ruin took is filed separately (#2758).

• A message can arrive up to six hours after you sign in, since it is written by the round that notices you are back. It waits in your messages until you read it.

1.106.7 — 2026-09-05

Fixed

• One guild's bad luck at the season close no longer costs every guild below it (#2754). When a guild season ends, the Council pays out placements in order, from first place down. Until now that was a single unbroken run: if paying one guild hit a database hiccup, every guild ranked below it was skipped — no frozen placement, no placement XP, no achievements, no Scrip, no Dark Matter, not even a line in the guild log. The season was already sealed at that point, so nothing would ever come back for them: those rewards were simply gone. The same applied to the end-of-season research reset, where the cost is worse — a guild left holding last season's research enters the new one with all fourteen allocation points already spent and no respec for the full eight weeks.

  Now each guild is paid on its own. A failure stops with the guild it happened to, everyone else is paid normally, and the guild that failed is retried straight away — which is enough to clear the brief lock contention that causes this in practice. If it still does not go through, the guild gets a line in its own log saying the placement reward is recorded and still owed, and the season close reports exactly who is outstanding instead of failing silently.

  Rewards also cannot be paid twice. Each guild's payout now claims itself before it pays, in the same transaction, so a retry either finds the work already done or does it once — never twice.

  No guild has ever been affected by this. We checked production before changing anything: no guild season has ever reached its close, so no placement has ever gone unpaid and nobody is owed anything. This lands ahead of the first real season close, not after it.

Also

• Guild log entries read better: the season lines match the impersonal voice of the rest of the log, a missing detail can no longer render as "Season undefined", and the research refund line finally says what happened instead of printing an internal name.

1.106.6 — 2026-09-05

Fixed

• Coming back now actually saves your empire from the ruin (#2740). An empire left silent for two weeks has its production halted and its garrison worn down five per cent a day. That is meant to stop the moment you return — but the round that applied it read every player once, at the start, and could not tell that you had signed in since. Signing in leaves no mark on a colony, and the only thing the round checked was the colony. So a player who opened the game minutes earlier still lost ships, and lost ships do not come back. The round now re-checks you immediately before it writes, and a player who has returned is restarted instead of worn down.

• A leave declared at the wrong moment is no longer ignored (#2740). Declaring a leave buys three weeks of protection and pays for it by stopping production. If the leave was declared while the round was already running, two things could go wrong: the empire was worn down anyway, behind the shield that was supposed to prevent exactly that; or the round switched production back on and handed back a protected empire that kept producing for the whole leave. Both are gone.

• The round no longer acts on an empire that has just been banned or removed (#2740). A moderator's ban, or an account removed by the sweep that runs five minutes earlier, is now seen before anything is written rather than after.

Nothing about who is eligible, or how much decays, has changed — no threshold and no rate was touched.

Also

• Whether this had already cost a real player anything cannot be answered from the data, and that is its own defect: nothing records that a ruin happened. The stamp the game keeps is a running billing position, not the time of any event, and it is overwritten each round; the column that would date the write is deliberately frozen because it is the silence clock itself; and the server log is destroyed by every deploy. This is the same blindness #2741 records for account deletion, one table over, and it has been filed as its own issue rather than guessed at.

1.106.5 — 2026-09-05

Fixed

• Guild screens no longer answer with a server error while an old account is being deleted (#2746). The game removes inactive and abandoned accounts around the clock, and clearing one out takes a while — it walks that player's whole footprint, table by table. Four guild actions have to write every member of a guild in one go: paying a Council directive tier to the whole roster, realigning everyone's research bonuses, and paying the end-of-season placement. If a member's record was tied up by a deletion at that moment, those actions were allowed to wait only five seconds before giving up, undoing what they had done and showing a plain error. They now wait up to fifteen, the same allowance every other whole-roster operation in the game already had.

  This was never limited to admin buttons. The directive payout runs whenever any member opens the guild panel, the research maturation on any guild read, and the season close on an ordinary look at the season tab when the eight weeks roll over — so the error landed on whoever happened to look first.

  The worst case was more than a refreshed screen. A season placement is paid once, and the season is marked finished before the rewards go out, so a failure there took that guild's frozen standing, its Scrip and its Dark Matter with it — for a season that could never be closed again. That is the one this most protects.

Also

• Being honest about what this does not fix: if a deletion runs long enough, the action can still fail, and it now holds its place in the queue three times as long before it does. Two deeper problems the review turned up are written down as their own work rather than half-fixed here — the order in which guild and member records get locked has never been agreed across the codebase (#2753), and one guild's failure during the season close still costs every guild below it in the standings (#2754).

1.106.4 — 2026-09-05

Fixed

Coming back to the game now saves your account from the cleanup sweeps (#2731).

Two background jobs remove abandoned accounts: one clears players who registered and never really started, the other retires empires that have gone silent for a month or two. Both used to work from a single snapshot of the candidate list taken at the start of the pass, and then delete each account without checking whether anything had changed since.

That left a gap with no undo. If you signed back in while a pass was running — moved a fleet, upgraded a building, or just opened the game — your account was deleted anyway, on the strength of a reading taken up to a minute or two earlier. Nothing warned you, and nothing could give the empire back.

Each removal now re-checks, at the moment it happens and on data it holds, everything the decision was based on: when you were last seen, when your colonies last changed, your guild, a declared leave. If any of it has moved, the account is left alone and simply looked at again on the next pass. Nobody who came back gets removed for having been away.

Two related safeguards came with it. An account banned while a pass is running is no longer deleted by it — which previously would have erased the ban along with the account, freeing the name for immediate re-registration. And a single contested account can no longer stop the rest of the pass.

Nothing changed about who is eligible for removal: the thresholds, the limits and the rules are exactly as they were. This fixes only the case where the game acted on information that had gone stale.

Also

We checked production for evidence that this had already cost anyone their account. It is not possible to tell: the deletion record keeps only what the decision was based on, so it agrees with itself by construction, and the detailed logs are discarded whenever the servers restart. That gap is now filed as its own piece of work (#2741) — not so that deletions can be undone, but so the game can say what it removed and when. Two further issues were opened from the same review (#2740, #2746).

1.106.3 — 2026-09-05

Fixed

• The guild Fund can no longer be filled by accounts that were created seconds ago (#2719). A brand-new empire arrived holding a full donation reserve, and nothing on the way to the Fund asked how old the account was. That made it possible to pour a guild's worth of a day's donations into the Fund from an empire that had done nothing and would not be kept — over and over. Donating to the Fund now opens once an empire is 24 hours old.

  Reported by Remy, who found it and tested it rather than guessing.

  What this costs a genuinely new commander, plainly: if you join a guild on your first day, you cannot donate to the Fund until your empire is a day old. Your first day gives one full reserve of donations instead of two. It is a one-off — from the second day nothing about your donating is different — but it is a real wait, and new members who want to help immediately will feel it. Everything else a guild uses to help a newcomer is untouched: the Help board, the guild shop, and a leader's grant from the Fund all work exactly as before.

  The treasury panel now tells you when the Fund opens instead of showing a Donate button that would be refused, and it keeps showing what a donation costs and gives while you wait.

Changed

• Empires run by the game's own commanders no longer say they have donations in hand during their first day, and a guild council of brand-new members no longer opens by asking for donations nobody present can make yet. They were reading a number the Fund had stopped honouring.

Also

• Checked against the live game before shipping: this was not meaningfully used. Across the whole season the pattern accounts for about 1,360 Fund in total, and that figure includes the reporter's own two test accounts. No guild's balance was adjusted, and none needed to be.

1.106.2 — 2026-09-05

Fixed

• Agent empires now use the colonies they found (#2738). Before: an agent's colony ship landed with an empty hold, the new world sat at zero buildings for weeks, and nothing ever came to help it. On 5 September the galaxy held 55 such empty agent colonies, 54 of them in the cradle, holding slots while new players were placed out in the Frontier. Now: every agent colonization carries a balanced founding kit that covers a Command Centre, an Extractor and a Refinery; an agent that cannot fund the kit keeps its ship in dock instead of planting a dead flag; an outpost that cannot buy its first income building gets a supply run from its mother, and the mother builds the Freighter for it if she has none.
• Agents stop knocking on cloaked doors (#2738). A colony ship turned away by a slot it could not see as occupied used to fly again at the same coordinates every five minutes. The slot is now remembered and left alone until the agent's next session.

Also

• The one-shot unlock script for stalled colonies now reads the price of the next building per resource, resolves finished build orders before judging, never touches the pirate, faction or relic accounts, and tops up rather than replaces what a colony holds.

1.106.1 — 2026-09-05

Fixed

• A guild that surrenders can no longer be paid for it (#2722). Wars are settled by a sweep that runs every five minutes: it reads the list of wars whose week has ended, works out each verdict and the 8% tribute, and then writes the result. If a leader surrendered in the gap between that reading and that writing, the surrender was invisible to the sweep — the war closed on the raw scores as though nobody had given up. When the surrendering guild was ahead on points, which is exactly when a guild gives up to cut its losses, it was declared the winner and collected the tribute from the guild it had just surrendered to. The week-long bar on re-declaring war was stamped on the wrong side too.

  The settlement now re-checks the war, inside the same transaction that closes it, against the surrender mark and the scores it was priced on. If anything moved underneath, it settles nothing at all and hands the war to the next tick, which closes it correctly on the current state. Nothing is ever paid out on a stale reading.

  No war in Novarchy was ever settled this way — we checked every war verdict in the database, and all of them predate this mechanism entirely. Nothing needs putting right; this closes the door before anyone walked through it. The surrender screen behaves exactly as before, except that in the rare case where the war is being closed at that very moment it now says so and asks you to try again, instead of quietly reporting a success.

Also

• A war whose score had somehow become invalid could have been left unable to close at all, in silence. It now settles, and the five-minute sweep says so in the server log on the one occasion it postpones a war, so a war that ever did get stuck would be noticed in minutes rather than by a player wondering why theirs had been "ending" for three days.

1.106.0 — 2026-09-05

New

• The Orbital Home (#2684). The game now opens on your empire's station in orbit: a painted view of the orbital hub with five districts you can tap — Hangar, Research, Colony, Campaign, Alliance — and one "Your next move" card under it. Each district says what is true right now: ships docked, a research lane idle, the building at the head of the queue, the campaign's next level, your guild or the invitations waiting. The stamp in the corner reads SHELTER, SHIELD or OPEN ORBIT from your real protection clocks. On the desktop the same view sits beside an "In orbit now" list.

Changed

• Home replaces Base in the phone bar (#2684). The bar now reads Home, Galaxy, Fleets, Chat, Menu. The Base is one tap away on the Colony district and stays in the Menu hub. On the phone the station opens above the First-steps and First-convoy cards.

1.105.12 — 2026-09-05

Fixed

• The galaxy chart counts what your scan can see (#2713). Every star on the map carries its system's colony count: an uninhabited system is a faint cool dot, an inhabited one takes a warm core, and from three colonies up it flares. The same per-system tally also chose the guild tint, the contested marker and the territorial control banner.

  That tally was taken from every colony in the galaxy, with no regard for the Eschaton Veil — including the ones your Sensors cannot resolve. The chart and the system view it opens into were therefore counting two different things about the same system, and only one of them kept the Veil's promise: open a system and the Veil held, look at that system on the chart and it did not. The control banner was the plainer of the two, since an alliance could be shown holding a system your scan finds nobody in.

  The map now counts the colonies your scan can see — your own always, plus whatever your Sensors reach — which is the rule the system view has always applied. The chart and the system it opens tell one story. The control banner follows the same rule, and it is worth being exact about how: it is not withheld, it is worked out from the colonies you can actually count, so what the chart says about a system always agrees with what it shows you in it.

  If you could already see a colony, nothing about it changes. Your own worlds sit on your own map with your own tint and your own banner, Sensors that match a Veil still show everything behind it, and on the live galaxy 190 of the 199 inhabited systems draw exactly as they did before.

  Alliance ownership of a whole constellation follows your scan too, with one deliberate asymmetry: a sector you cannot fully account for is shown as owned by nobody rather than credited to anyone. The chart will never tell you a constellation is yours when it is not — and the sovereignty that actually pays, the home-route discount and the weekly season points, is decided away from the map and is untouched by any of this.

1.105.11 — 2026-09-05

Fixed

• A surrender that failed now says so, instead of reporting success while the war stayed open (#2710). Offering surrender ends the war immediately and pays the 8% tribute — but that settlement compare-and-sets the mother colony of every member of both guilds at once, and if any one of them was being written in the same instant, the whole thing rolled back. The war stayed open, no tribute moved, no post-surrender cooldown was set — and the game answered "done" anyway. You signed a dialog reading "The war ends immediately and the alliance pays the 8% tribute", got no error, and the panel came back still showing WAR with the countdown running. Nothing told you the difference between that and a broken button, so a leader could believe they were out of a war they were still in, and stop defending. Now the surrender is retried, and if it genuinely cannot complete you get "State updated by another operation: try again." — the same message the rest of the game uses — with the button still in front of you. Either way the war really does close: a refused surrender is still recorded, and the five-minute sweep settles it even if you do nothing.

• Surrendering at the same moment as your opponent no longer tells the winner they surrendered (#2710). If both leaders offered surrender in the same second, the second request wrote its mark onto the war the first had already closed and reported success — so the guild that had actually won and collected the tribute was told it had given up. It now answers "You are not at war with that guild", which is the truth.

Also

• The war-resolution sweep is unchanged: one contended war still cannot block the others it is closing in the same tick. Only the player-facing road reports what happened.

1.105.10 — 2026-09-05

Fixed

• Paying reparations no longer breaks when someone else is busy (#2701). An alliance leader settling the bill they had been handed could get a bare "Internal server error" — on a button that moves resources between two empires. It happened whenever anything else wrote to either colony in the same instant: the payer's own, or the colony of the enemy leader receiving the payment, which the payer neither controls nor can see. Nothing had actually moved when that happened; the payment had been cleanly abandoned, and the game still reported a server fault. The payment now simply tries again on fresh numbers, and only if it keeps losing does it say what the rest of the game says in that situation — "State updated by another operation: try again." In practice the second attempt wins and the payer sees nothing at all.

  Two details worth stating, because a payment moves resources between two players. It cannot pay twice: an interrupted attempt leaves nothing behind, so retrying starts from an untouched state. And it re-reads both colonies before trying again, so a leader whose stock was spent elsewhere in the meantime is told they can no longer afford the bill, rather than being charged for it anyway.

  What is not covered: this fixes the payment losing a race against another write. A payment that fails because the server itself is briefly overloaded can still report an error — a narrower case, tracked separately in #2714.

1.105.9 — 2026-09-05

Fixed

• The system list stops calling a slot free — and so do the mission composer and the target board (#2705). Open a system in list view and every row the scan came back empty on said "free slot". The scan cannot know that: it serves the same empty slot for a genuinely free planet and for a colony sitting under the Eschaton Veil, which is exactly why the planet sheet stopped making that claim two releases ago. The list was never brought along, so one tap apart the two screens disagreed about the same slot.

  It had been quiet, because an anomaly signal only ever appeared beside a slot that really was empty. That changed on purpose in the last release, when signals started obeying the same visibility rule as the colony beneath them — so a row could read "free slot ⟡ T3" over a live empire: the mark that used to mean nothing lives here, stapled to the sentence saying so, on a slot where something does.

  All three surfaces now report what the scan saw, in the planet sheet's own words — "No colony on your scan" — and the list explains it once at the foot of the table: a free planet and a colony under the Veil read the same in that table, Sensors as strong as the Veil tell them apart there, and your own probe answers on the planet's card. The mission composer's target header and the fact line beneath the compose box carried the same claim, and say the same thing now.

  As on the sheet, nothing is greyed, hidden or hedged on suspicion. A row that looked even slightly different over a slot the game thought was hiding something would turn the list into a detector for the very colonies the Veil is bought to hide, so the row reads identically over both. It simply no longer tells you the slot is empty.

• The target board counted your scan and called it the galaxy (#2705). The header read "N of 4,104 slots free in the core galaxy". That figure has always been the total minus the colonies your Sensors can resolve, so two commanders looking at the same galaxy on the same evening could honestly read different numbers. It now says what it counts: "N of 4,104 core-galaxy slots with no colony on your scan".

• The anomaly legend named only half of what filters it (#2705). Both places that explain the ⟡ mark — the galaxy map's legend and the list's own footer — said the count showed "only the ones your Astrometry can see". Since signals started following the colonies they sit on, Astrometry is no longer the only thing deciding what you see, and the line said so as if it were. Both now add that a signal never appears on a slot your scan already shows a commander on.

1.105.8 — 2026-09-05

Fixed

• Nothing changes in the game this time. This release is entirely internal: it repairs the safety check that decides whether a change is allowed to reach novarchy.com at all (#2706).

  The build gate runs the whole test suite against one specific commit and then records the result against that commit, so the next step can trust it instead of spending another quarter of an hour re-running everything. It was writing that result down against whatever the branch had become by the time it finished, rather than against the commit it had actually tested — so when a second change landed while the tests were still running, the verdict was filed under the wrong commit. A commit that had been tested ended up with no record at all, and one that had never been tested ended up wearing a pass. Nothing broken reached players because of it, and the two suspected cases were checked one by one, but it meant a change could in principle have been approved on a result it never earned.

  The gate now decides which commit it is testing once, at the start, and records the result against that one. If the branch moves underneath it while it runs, it files nothing at all and stops loudly instead of guessing. Every recorded result now also says which commit produced it, and the check that reads those results back refuses any it cannot match to the commit in front of it — so an unattributable pass costs one honest re-run rather than a wrong approval.

1.105.7 — 2026-09-05

Fixed

Your anomaly survey is drawn from what your own instruments reach (#2689)

Every week the survey scatters stellar anomalies across the galaxy, and the Galaxy view marks the ones your Astrometry resolves. It was not marking all of them. An anomaly that had landed on a slot already spoken for was struck off your chart — struck off by a rule that consulted the galaxy as it truly is, rather than the galaxy your sensors can actually reach.

That is a small thing to get wrong and an awkward one to leave standing, because the Eschaton Veil is bought on exactly the opposite promise: what an empire hides costs a probe and a round trip to find, and never a glance at a map.

So the survey now answers from your side of the instruments. A slot your sensors read as empty is treated as empty in every respect — the signal on it is charted, the exploration launches, and the fleet resolves it exactly as it would anywhere else in the sector. This week that puts five of the survey's 168 anomalies back within reach of commanders who could not have touched them before.

Nothing about the Veil's price has changed, and nothing about how you find what it hides. What changed is that the chart no longer answers a question you had not paid to ask.

1.105.6 — 2026-09-05

Fixed

• A guild donation that lost a race no longer reports a server error (#2693). Donating to the guild treasury writes to your home colony, and if anything else touched that colony in the same instant — a raid landing, a build finishing, or simply your own second click — the donation was refused and the game answered a generic server error. The commander was told something had broken, on an action that moves resources, with no way to tell whether the donation had gone through. It had not.

  Now that collision is handled the way the rest of the game handles it: the donation quietly tries again from fresh data, up to three times, so an ordinary clash never reaches you at all. Only if all three attempts lose does the panel say "State updated by another operation: try again." and reload itself — the same familiar message every other action already gives, and nothing has been charged.

  The same fix covers the two emporium purchases that deliver to your home colony, the emergency crate and the spy probe.

1.105.5 — 2026-09-05

Fixed

A spy probe now answers when it lands, not when you press Launch (#2686)

Sending a probe at a set of coordinates could be refused on the spot, before any ship left the hangar, with "Diplomacy blocks this mission." — and it could be refused that way even when nothing on your Galaxy scan suggested there was anybody there to be diplomatic with. A refusal that arrives before the flight is a refusal that answers a question you had not paid to ask, and espionage is the thing you are supposed to pay for.

So the launch has stopped answering it. A probe sent at coordinates you have no reading on now leaves the hangar like any other, and the answer comes back with the probe:

• a report, if there was something to photograph;
• "Espionage came up empty", if the world really is free;
• and a new line, "Your probe found a friendly world", when it turns out to belong to your own guild or to an ally. No report is filed on a friend — and now that is said at the end of the flight instead of before the start of it.

Probing an allied colony that is plainly on your screen is still refused the moment you press Launch, with the message it has always had. Nothing there was ever a secret, so there is nothing to wait for.

One consequence worth knowing, because it runs the other way: a friend's colony is protected whether or not you could see it when you launched. If a guild becomes your ally while your probe is still in the air, the probe comes home with the same line rather than a report.

1.105.4 — 2026-09-05

Fixed

• A guild donation no longer fails when the treasury is busy (#2691). Donating could come back as a plain error that told you nothing — and the donation had not gone through. It happened when your guild's records were briefly locked by something else finishing: a war or league payout, which can hold every member's row for up to fifteen seconds, while a donation gave up after five. The donation now waits the same fifteen seconds those payouts take, so it goes through instead of failing. Guild grants from the Fund and emporium purchases were on the same short fuse and now wait too.

• A donation whose outcome is unknown no longer invites you to pay twice (#2691). If the connection drops mid-donation, the game says the action may still have gone through and asks you to check before retrying. It was showing that over a stale panel — the same charge count, the same Fund, the same resource numbers as before you tapped — so checking suggested nothing had happened, and donating again cost a second charge and a second batch of resources for one donation. The panel and your colony's resources are now refreshed before you are asked to check.

• A treasury action refused because the server was too busy to begin now says so properly (#2691). The answer is the ordinary "try again" the game already shows elsewhere, and the screen reloads, instead of a generic failure.

Also

• The internal check that keeps lock-waiting database work off the default time limit had stopped seeing the emporium when it moved behind a shared helper. It sees it again, and the helper itself is now verified rather than trusted.
• Found while investigating and filed separately (#2693): a donation that loses a race against another change to your colony still fails with a generic error instead of the retryable one, by a different route than the fault fixed here.

1.105.3 — 2026-09-05

Changed

• New empires are now born among neighbours instead of on the far rim (#2672). The protected cradle at the centre of the galaxy is full — every one of its 1,368 slots is taken — so for the last few days every new commander has been founded out in the Frontier. Which was fine, except that the Frontier was picked at random across all eight of its sectors, and the outer four held 77 colonies between them. A player who signed up was as likely to wake up alone in deep space as anywhere near another empire. The Frontier is now settled from the inside out: a new commander lands in the nearest sector that still has room, among the hundreds of empires already living there, rather than being scattered to the edge of the map.

  This does not add a single slot, and it takes nothing from anyone — no colony moved, no protection changed, and the cradle still comes first for as long as it has any room at all. It is purely a change to where the game looks next when the centre is full. Being born with neighbours also means being born within reach of them: the starting protections all still apply, but the far rim was quiet in both directions.

  Reported by Spacca-Mazza, whose note about the state of the map is what started the measurement.

Also

• Corrected a stale comment beside the protected-band setting that still described the galaxy as eight sectors wide. It has been twelve for some time, and a wrong note about the shape of the map is exactly the kind of thing that misleads the next decision about it.

1.105.2 — 2026-09-05

Fixed

• A planet the galaxy view cannot see into is no longer called free, and now offers the probe that settles it (#2680). Open a slot the Galaxy scan shows as empty and the sheet said "No colony — free planet", then handed you Colonize on the strength of that label. It could not know: the scan serves the same empty slot for a genuinely free planet and for a colony sitting under the Eschaton Veil, and the one mission that could have told you which — espionage — was the only action the sheet did not offer.

  It now reports what the scan saw, "No colony on your scan", and says plainly that a free planet and a colony under the Veil read the same there, that your own probe or Sensors as strong as the Veil is what tells them apart, and that a founding which meets a hidden colony turns back with its ships and its cargo. A Spy button sits beside Harvest and Colonize on that slot: espionage is how a hidden colony is found, and the server has always taken that mission at coordinates the scan shows empty.

  Colonize is deliberately unchanged, still lit on every empty slot. Greying it wherever the game suspected something would have turned the galaxy map into a detector for the very colonies the Veil is bought to hide — so the sheet reads identically over both. It simply no longer tells you the planet is free before you decide.

• A failed colonization says which of three things went wrong (#2680). Every bounce printed the same sentence — "planet no longer available" — whatever had actually happened. A commander who had run out of room for another colony was sent hunting for a different world instead of a Command Center level or a Colonization research level; a fleet that reached the Event Sector after the season closed got the same line; and about a colony that had stood under the Veil the whole time, "no longer" was simply false, because nothing had changed. Each cause now reads as itself, and where the planet is the reason the row points at the probe. Every one of them states what was true all along and never said: the founding hull is not spent when a founding is refused — the fleet turns back carrying it.

1.105.1 — 2026-09-05

Fixed

Synthetic empires no longer take the Council card at midnight, all together (#2667)

The Council directive board opens at 00:00 UTC on the first of every month, the same instant for everyone. Until now every synthetic empire took its card within about five minutes of that instant — the whole roster, in unison, on the hour. And because these empires keep their own waking hours spread across the day while the board's opening is fixed, just over half of them were doing it in the middle of their own night.

Now they pick the card up at some point during their own playing day, the way anyone notices a new event board when they next sit down. Measured across 22,500 monthly cycles against the real 75-empire roster, and counting the moment the card is actually taken rather than the moment it was planned: takes landing within five minutes of the opening fall from 100% to none, and none at all now happens while that empire is asleep. The busiest single hour of the day goes from holding every take to holding 6.2% of them, where a perfectly even spread would be 4.2%.

Nobody loses the card by waiting: no empire in those 22,500 cycles reached the end of the window still empty-handed, and one that discards a card mid-cycle picks up its replacement straight away. The harness that produces those numbers ships with the game, so they can be checked rather than taken on trust.

An empire could briefly deny holding a card it had just taken (#2667)

For the five minutes after taking a directive, an empire's own notes still read "You have not taken a directive yet." Ask one in guild chat during that window and it told you, in good faith, that it had not picked a card — while the board showed that it had. These empires are held to speaking only about real game facts, so a window in which their own notes were wrong was a real break in that promise, however short. Their notes are now right from the moment of the take.

The same notes could also claim never to have started the event while reporting event points earned from a directive already completed — two statements that cannot both be true. An empire between cards now says so plainly.

Also

Hardened the empires' memory against a server clock that jumps backwards (#2667). A record written while the clock was running ahead used to look permanently fresh, which could leave an empire refusing to attack a world, declining to join a raid, or silently going quiet in its guild, until real time caught up. Every such record now ages correctly in both directions. Two places deliberately keep the stricter old behaviour, because there the alternative is worse: an intelligence report from the future is refused outright rather than trusted, and grudges are never decayed for time that has not actually passed.

1.105.0 — 2026-09-05

Added

• The admin dashboard now shows how full the galaxy is (#2672). The four protected sectors every new empire is born into — the cradle — filled up over five days, from 85% to completely full, and nobody knew until a player wrote in to say he could not find a planet to colonise. There was no instrument anywhere that reported it. There is now: a capacity card with cradle, Frontier and core occupancy, free slots for every core sector, and a banner when the protected band is nearly full or already dry. The server writes the same fact to its log the moment it happens, naming how many protected slots are left, so the next time this happens it is noticed on the day rather than a week later.

Also

• Nothing changes for players. When the cradle is full a new empire is born in the Frontier instead — that has always been the behaviour, it is why signups never broke while the band filled, and this release does not touch it. Every new empire still gets the same planet it would have got before.
• Colonies in the Event Sector are counted separately from the rest of the galaxy, because its planets were never part of the core's capacity; counting them together would have quietly overstated how full the galaxy was, a little more with every season.

1.104.17 — 2026-09-05

Changed

• The AI commanders no longer call their guild into fights the guild cannot win (#2676). Until now a synthetic commander could open a joint raid on the strongest colony in sight without ever having scouted it, judging only the walls, and its guildmates joined blind; that is how whole coalitions kept dying on Wraith homeworlds with ten-thousand-ship garrisons. A raid is now called only on a colony the commander has a fresh report on, never on an owner whose fleet outguns the whole expected coalition, and only after the same battle preview a human reads says the coalition wins; a commander joining someone else's raid previews the real coalition, the fleets already committed plus its own, before sending ships. When the preview says no, the commander leaves that raid alone for a few hours.

1.104.16 — 2026-09-05

Fixed

• The guild raid panel no longer arms a raid at a place you cannot fly to (#2665). The "Open a raid" form used to light its button as soon as the three coordinate boxes had something in them. Type a system the galaxy does not have and it offered the launch, asked you to confirm leaving your colony undefended, and quoted how long your fleet would be away — before the server refused the whole thing. Now the form answers first, in words you can act on: "System 58 does not exist: a sector holds 19 systems." The same goes for the Event Sector between seasons, which is a real address but not one open today: "The Event Sector is open only for the duration of the season." One refusal at a time, and the address always comes before the season — a mistyped planet is a typo, not a closed sector, and you are not sent to fix something that was already right.

• A raid the season closed under stops advertising itself. Nothing removes a raid your guild opened in the Event Sector when a season ends, so the row stayed on the board with a live Join button, an arrival estimate that promised you would get there in time, and a Dark Matter price for joining — all of it refused the moment you tapped. The estimate and the price now go with the button, and the row says why. If you are already aboard, you keep the bounty figure: that bill is settled at impact whatever the season does.

• Nothing changes for the twelve permanent sectors, or for the Event Sector while a season is running. If the game has not yet managed to ask the server whether a season is on, it blocks nothing at all — an uncertain answer never grounds a launch the server would have accepted.

1.104.15 — 2026-09-05

Fixed

• The mission form no longer prices a launch at the Event Sector between seasons (#2635). Point it at 13:4:7 with no season running and it used to answer like any other target: a travel time, a Dark Matter price for the warp, and a lit Launch button — and the server refused the tap, because the Event Sector is only reachable while a season runs. Nothing was ever spent; what it cost was a form that looked confidently right about somewhere you cannot go. Launch now greys out the way it already did for a harvest at somebody else's planet, and says why: "The Event Sector is open only for the duration of the season." No travel time, no route labels, no Dark Matter quote. During a season nothing changes at all.

• The Launch-again rail stops offering a one-tap relaunch into a closed sector (#2635). A raid flown at the Event Sector during the season stays on the rail after the season ends, and its card was still armed. It now carries the same refusal as the form.

• A mistyped Event Sector address is called a typo, not a closed sector (#2635). 13:4:99 between seasons used to be told two things at once — that planet 99 does not exist and that the sector is shut. It now gets the one answer it can act on.

Also

• If the game cannot tell whether a season is running — the season check has not answered yet, or failed — the form is deliberately left alone and the launch goes through to the server. A door that is merely uncertain is not allowed to ground a fleet that a live season would accept.

1.104.14 — 2026-09-05

Fixed

• Synthetic empires no longer swap their Council directive the instant the game lets them (#2633). The board tells a player when the card they are holding can be exchanged, and refuses the swap until then. The agents were only reading that notice in one case: for a card that had become impossible to finish. In every other case — a card they had not moved on and did not much like — they asked again every five minutes for the whole four hours and swapped on the first attempt that was allowed, forty-odd refusals later. From the outside that is an opponent who never stops pressing the button and acts on the first legal second, which is the one thing the agents are not supposed to look like even when, as here, they know less than you do.

  Now they check the notice before asking, and being told "not yet" ends the attempt: the card goes down and is picked up again the next time that persona sits down to play, at no particular moment in the session. Measured over 40,000 episodes on the real roster, swaps landing within one tick of the deadline fall from 100% to 0.2% — flatter than a plain uniform retry, which lands at 0.5% — and the requests fall from 46 per episode to one. Which hour they come back is decided by their own sleep and their own habits, and never by the game's clock.

Also

• The agents' memory gained a small record of which card is on hold and until when. Existing memories need no migration; a memory without it simply has nothing on hold.
• Two follow-ups found while reviewing this and filed rather than folded in (#2667): agents still pick up their monthly Council card on the first tick after the board opens, and for one tick after taking a card their status sheet says they have not taken one.

1.104.13 — 2026-09-04

Fixed

The Hypergate card never said whose two gates collapse a route — and players were paying for it (#2625)

The building card promised that "two gates make the route between them a single hop", without saying whose. Read in front of a purchase, that means "two of mine". But the guild network has always worked without you owning anything: if a guildmate holds a gate in your departure system and another holds one in your arrival system, your trip becomes a single hop — no Hypergate, no Hypergate research, from the day you join. The research card repeated the same half-truth, mentioning the guild only to say your gate helps them, never that theirs carries you.

So players were researching a node and building a structure to get something their guild already gave them.

Both cards now say it. The paired-gates row spells out that either end may be yours or a guildmate's, that the guild half asks for neither the building nor the research, and that the trip is priced on the weaker of the two gates — a rule that until now was written down only for your own pair. At level 0 the row used to show a bare dash; for a player in a guild it now says a guildmate's gate already counts as one end, which is the fact that changes the decision on that screen. Players with no guild are told none of this, because for them it is not true.

The same row now also separates two things that were being described as one: your own pair reaches further than the ends — any two of your gated systems can bridge a route, though the legs out to them still count — while the guild pair works at the departure and arrival systems only. And the card no longer claims a gate speeds "every convoy" from its system: it speeds supply runs, never a convoy to an ally or an enemy.

Nothing about how travel is calculated has changed. This is the card catching up with the engine.

Reported by Spacca-Mazza (spaccamazza), who worked it out from a support-desk answer and was right on both counts.

Also

The travel reference the support desk answers from now states that the guild network costs the traveller nothing, and corrects two errors it carried: it credited the Hypergate research with the Relay's discount (a separate research and a separate building), and described the gate express as multi-hop when it prices exactly one.

1.104.12 — 2026-09-04

Fixed

The login screen stops telling a locked-out commander to abandon their empire (#2652)

A guest account has no email and no password: the only key is a sign-in token kept by the browser or app you played in. Lose it — clear your site data, reinstall, change phone, tap Logout — and the game shows you the login screen and nothing else.

Typing your own commander name there used to answer "That name is already taken: pick another one." To the person who owns that empire, that was an instruction to give up and start a second one. It was also the only advice on the screen: a new name founds a new empire, and so does the Apple or Google button, because linking an account requires being signed in already.

The refusal now says what is true and what can actually be done:

An empire already answers to that name. If it is yours, its key lives in the browser or app you played in — open the game there. Any other name starts a new, empty empire.

That first route is real and nobody was told about it: an old phone or a second browser you never signed out of still holds a key, and it does not even have to be valid — the game trades an expired one for a fresh key by itself.

The small print for a returning visitor now says the same thing before you tap, instead of naming only the linked-account door that most guests never set up. And the support page gains a Locked out of your empire section: what the key is, the two routes worth trying, what signing up again does and does not do, and an honest account of the last resort.

One thing that page is careful about, because the first draft got it wrong. It would have been easy to say a dead key works for six months. It does not: a guest empire that goes quiet is retired after about a month, and once the empire is gone the old key opens nothing. The generous window protects someone whose key ran out while they were still playing — it does not hold an empire open for a player who stopped. The page now says so, and tells you not to wait.

What deliberately did not change: the answer is the same whether the name is yours or a stranger's. Telling the two apart would let anyone probe names to discover which commanders exist, and that single refusal is what stops someone walking into your empire by typing your name. So the help is offered as a condition you judge for yourself, and everything it states is true of every empire in the game.

Reported and diagnosed from the player's side by Depthcharge.

Also

The deeper question this leaves open — how a locked-out owner could ever prove an empire is theirs — is filed as #2656 rather than answered with a weak identity check. Anything that admits the rightful owner on evidence a stranger could also produce admits the stranger too.

1.104.11 — 2026-09-04

Changed

• Nothing in the game changed with this release, and nothing was deployed: it is a change to the build gate that guards every other release. Two changes that were each verified green could still combine into a main that did not build — as happened on 4 September, when two of them added a function of the same name in different files. Git saw no conflict, and each check was right about the code it was given; together they broke the build and blocked every pending change until it was untangled. Players were never affected, because the faulty build never reached the servers. The gate now also builds the result of combining a change with the current main, both when a change is first checked and again in the seconds before it lands, so this class of breakage is caught beforehand instead of discovered afterwards. It costs no extra time: the check runs inside a wait the gate was already spending (#2646).

1.104.10 — 2026-09-04

Fixed

Surrendering a guild war now buys a week of peace (#2640)

Surrender used to cost 8% of your mother colony's storage and buy nothing. The war closed, the two guilds went straight back to neutral, and the guild you had just surrendered to could declare war again the same minute and take another 8%. There was no cooldown of any kind — surrender was the only exit an outmatched guild had, and it was a fake one.

Now the guild that accepts a surrender cannot declare war on the guild that gave up for one week.

• It binds only the winner. The guild that surrendered may declare on its conqueror whenever it likes, so nobody is trapped by giving up under pressure.
• Neither side is slowed towards anyone else, and a new war between the two does not cancel the week — it resumes when that war ends.
• It stops the war declaration and nothing else: raids, pacts, reparations and the League go on exactly as before.
• It binds human and synthetic guild leaders alike.

The Diplomacy panel shows the week on both rows and says which side it binds, and the Declare war button is withheld from the guild that is barred.

A week, rather than the 48 hours a paid truce buys, because a war is scored over a whole week: one fought to its Monday rollover already yields at most one tribute per pair per week. Surrender is meant to be an early exit from that week, not a way to run three of them.

Reported by Tibeh01.

1.104.9 — 2026-09-04

Fixed

• The mission composer no longer describes planets that do not exist (#2617). Typing a planet number the system does not have — 19, where a system holds 18 — produced a box that contradicted itself: "Slot unknown: the Galaxy view could not be read.", and directly beneath it the type and both multipliers of the planet living in that slot. The composer now says the true thing instead — "Those coordinates are outside the galaxy." — on the target panel, on the phone's compose line, and on the planet sheet, which previously opened as a window with no content and no way to close it. Every address that does exist is unchanged: free slots still show their world with no round trip, and the Event Sector still answers. Reported by QQQQ.

  The badge was the symptom. Underneath it, the function that works out a planet's type and richness is part of what the game sends to your browser, and it had no notion of where the galaxy ends — so it would answer for any three numbers, including coordinates in sectors a future expansion will open. It no longer answers for a coordinate the galaxy has no slot for.

  What this does not do, stated plainly: it does not make those traits unobtainable. The calculation itself is unchanged — it has to be, every planet in the game is that calculation, and altering it would change worlds players have already settled — and it is still in the browser bundle. Someone who reproduces it can still work out the unopened sectors. This closes the game answering the question, not the question being answerable. The remaining half needs a key the client does not hold, which can only be chosen when new sectors are actually opened; it is tracked in #2630 rather than quietly claimed here.

• A build break on main, caught before it reached anyone (#2617, #2618). Two changes shipped minutes apart each added the same galaxy-bounds helper in a different file. Git merged them without a conflict and both had passed their checks, but together they would not compile. The deploy stopped at the build with the previous version still serving, so no player saw anything and there was no downtime; the duplicate is now collapsed into one definition.

1.104.8 — 2026-09-04

Fixed

The check that retires the Pirate Beacon stops flooding the server to do it (#2593)

Nothing in the game looks different after this release, and no Beacon changed hands. This is about what one piece of background housekeeping costs while nobody is watching.

The Pirate Beacon is the practice target the game puts in your system when you are starting out. Every six hours the server reviews each one and asks a simple question: is there still anybody here who has something to learn from it? If a whole system has outgrown its Beacon, the Beacon goes, and the slot it was holding opens up for a real world.

The review was asking that question one system at a time. With 114 Beacons on the server that meant well over a hundred separate trips to the database, each one waiting for the last to finish — and up to double that on a round where many Beacons qualified at once. The database itself never struggled with this: all those queries together came to about a twentieth of a second of its time. The problem was elsewhere. The game runs on a single server process that also answers every page you load, every attack you launch and every fleet that arrives, and a job that queues up a hundred small waits in a row makes everything else stand behind it. That queueing is what had been squeezing other background jobs until they ran out of time and gave up — the pile-up traced in the previous release.

The review now asks the same question about every system at once, in a fixed handful of trips no matter how many Beacons exist. The same work that took a hundred and sixteen steps takes four.

The important part is what did not change: exactly the same Beacons qualify, under exactly the same rule. A system keeps its practice target unless every single real player living there has passed the threshold — one newcomer next door is enough to keep it. A system where only pirates, factions or relics remain keeps it too, and so does a system the server knows nothing about. In doubt, the tutorial stays.

Before shipping this we checked whether making the review cheaper would cause a wave of Beacons to disappear at once. It will not: measured against the live galaxy, none of the 114 Beacons currently qualifies for retirement, and the closest system is still a long way from the line. If that changes later it will change the way it always would have — one system at a time, as its residents grow.

Also

The tests around this now pin two things that had never actually been checked, despite being the exact mistakes the code carries warnings about: that the review reads a system's full coordinates rather than half of them, and that it cannot be talked into deleting every Beacon in the galaxy in one go. Both were verified by deliberately breaking the code and confirming the tests catch it.

A related weakness in the test scaffolding — where a stand-in for the database hands back information a query never asked for, hiding this whole class of mistake — is recorded as #2639 rather than patched in a hurry here.

1.104.7 — 2026-09-04

Fixed

• A release going out no longer looks like the game falling over (#2612). Deploys happen around twenty times a day, and each one recreates the servers behind the game. For ten to fourteen seconds there is nothing to answer, and what you saw in that gap was the web server's own raw error page — the grey one that reads like something is badly broken. Nothing was: a release was landing. You now get a Novarchy page on the game's own dark background telling you a release is going out, and it retries every fifteen seconds by itself, so you land back in the game without having to do anything. Reported by QQQQ, who hit it mid-click and was right to say that nothing on the screen told him the difference between a deploy and a fault.

  The rest of the app is handled too: the calls the game makes in the background now get a short, readable answer during that gap instead of a page of web-server HTML they were never meant to read.

Also

• The configuration that routes every request to the game had never been kept in the project at all — it existed only on the server, where nobody could review it, compare it or undo a change to it. It is now in the repository, together with a written procedure for changing it safely and a test that guards the three settings holding up defences you cannot see from the game: the one that keeps a password-free developer login shut, the one that makes the anti-spam limit on new accounts count real players rather than the whole game at once, and the one that stamps a new account's country at the edge instead of trusting the browser.

  This change dresses the gap; it does not remove it. Doing that needs the new servers to be started and proven healthy before traffic moves to them, which is filed separately (#2627).

1.104.6 — 2026-09-04

Fixed

The mission composer no longer plans a trip to a planet that does not exist (#2618)

Type a planet number the galaxy does not have — 19, when a system holds 18 — and the mission form used to tell you two things at once. It said "Slot unknown: the Galaxy view could not be read.", and directly under that it gave you a travel time, a route, a Dark Matter price for the warp, and a Launch button you could press. Pressing it failed: the server had always refused those coordinates, so no fleet ever left and nothing was ever spent.

The cost was not the wasted click. It was that the form was confidently wrong about somewhere that does not exist, on the same screen where it had just admitted it could not read the slot.

It reads that way now. The Launch button greys out, exactly as it already did when you pointed a debris harvest at somebody else's planet, and the line beside it says which part of the address is impossible — "Planet 19 does not exist: a system holds 18 planets." No travel time, no route label, no Dark Matter quote, and where the estimate used to be the bar says there is no route rather than printing a number for nowhere.

Worth saying why nobody caught it earlier: the number was not obviously broken. The game measures distance in map cells, and a planet that has no cell quietly fell back to plain straight-line distance and answered anyway — from the same home world, planet 5 priced at 191 seconds and planet 19 at 200. Nothing on screen told them apart.

Real destinations are untouched, the Event Sector included: it sits one step past the ordinary map, and a season is fought there, so it stays a place you can send a fleet to.

Also

Seven test files had been quietly describing colonies at addresses the game cannot hold — system 41 and 58 of a galaxy that has 19 — and now sit on coordinates that exist. Behind the scenes there is one definition of "is this a real place" instead of the six that had drifted apart across the game and the server, two of which disagreed about whether the Event Sector counted.

Reported by QQQQ, who also pointed at the debris harvest as the thing it should behave like. It does.

1.104.5 — 2026-09-04

Fixed

• The AI empires no longer strike the moment a night shield lifts (#2588). Players reported attacks landing almost exactly as their night protection ended, and asked the fair question: do the bots know when the window closes? They do not, and they never did — the game deliberately never tells them, because night protection is decided by when a fleet arrives, which no target list can show. What was happening is less flattering and more mechanical: an agent that had settled on a target simply re-sent the same attack every few minutes, was refused every time without being told why, and got through on the first attempt that happened to land outside the window. Being tireless looked exactly like being all-knowing.

  Now a refusal ends the attempt. The agent leaves that colony alone and comes back the next time it is at the console — which, since every empire has its own waking hours, is at least a night later and at no particular moment. The strike that eventually lands has no relationship to when your shield lifted. The agents are exactly as blind as before: the wait is worked out from the attacker's own routine, never from the hours at which it was turned away, so nothing about your sleep schedule is ever written down or reasoned about.

  The same applied to guild raids, which go through the same check and were being re-proposed on the same loop; they are fixed too.

• A refused attack no longer strands a fleet or cancels a war (#2588). The fleet an agent had earmarked for a launch the server kept rejecting stayed invisible to its own raids and guild defence for up to two days, and the agent kept scanning the one place it had already been turned away from. Those ships go back on the board immediately now. A campaign an agent has been talking about in guild chat also survives a refusal — including a refusal caused by nothing more than the server restarting — instead of being quietly abandoned.

1.104.4 — 2026-09-04

Fixed

Deleting an account can no longer quietly start blocking guild founding for longer (#2592)

Nothing you can see has changed, and that is the finding as much as the fix.

The report was that deleting one account froze every guild operation on the server for the length of a 35-statement database transaction. Measured statement by statement, it does not: the lock in question is taken at the very end of the deletion, joining a guild never waits on it at all, and the sweep that removes up to 200 abandoned accounts a night can never reach it, because it refuses to touch anyone who is in a guild. What the lock does hold up — founding a guild, and dissolving one — waits for the last stretch of the deletion, and for no longer than the same wait costs when the last member of a guild simply walks out.

So the lock stays as it is, for reasons now written down beside it: it protects the eight-colour wheel that every guild draws its banner colour from, which is genuinely shared by the whole server; it is the only signal every dissolution path agrees on; and moving it out of the deletion would let an account finish deleting while its guild was left standing with nobody in it — the exact state a previous fix exists to prevent.

What was missing was a brake. The deletion's tail is now pinned statement by statement in a test, so the next change that lengthens the stretch where guild founding waits fails loudly instead of landing unnoticed. Two long-standing comments that stated the wrong numbers, and one that named the wrong set of locks, are corrected — one of them would have led the next person to take two locks in the opposite order and deadlock the guild system.

Also here: a changelog heading from v1.104.0 now carries its issue number, which had been left off and was failing the check that keeps every heading on this page traceable to the change it describes.

1.104.3 — 2026-09-04

Fixed

Every check before a release stops wasting a minute of its own time (#2550)

Nothing in the game itself changed in this release. This is machinery — and the machinery is what decides how fast everything else reaches you.

Before anything ships, an automatic check runs. Part of it verifies that every building's artwork still sits exactly where the map believes it sits: it reads the real pixels of all 81 building sprites and refuses the release if any of them disagrees with the map data. That is the check that stops a building being drawn floating above its tile, or sunk into it.

It was taking 73 seconds, on one processor core, doing nothing else — and every check paid it, including checks on work that could not touch a building sprite at all. It now takes 3 seconds.

The check itself is untouched. It still reads the same 81 sprites, makes the same comparisons against the same expected values, and still runs every single time — nothing was skipped or switched off to make it faster. What went was waste: the code reading those image files was decoding them one byte at a time, and the 81 sprites were being read one after another while the rest of the machine sat idle.

While proving that, two things that had been quietly wrong were also put right: a check belonging to the account system was being given five seconds to do work that takes about seven whenever the build machine was busy, so it failed for reasons that had nothing to do with the code (#2476); and the release notes for v1.104.0 carried a heading that the changelog page's own guard rejected, which was blocking every release queued behind it (#2570).

Also

• The image-reading code had also been made able to accept a truncated sprite file without complaining — a file the previous version would have rejected. Caught in review before it shipped, and closed: a damaged sprite is refused again.

1.104.2 — 2026-09-04

Fixed

• The War Room's "strikeable only" count now says how many targets are free and how many cost a bounty (#2571) — Pressing "strikeable only" showed a single number, and that number quietly added two very different things together: enemies nothing stops you from hitting, and enemies you are perfectly free to hit but who put a bounty on your head for it. A commander in a lopsided war read 58 on the chip, counted four targets he could actually strike without a price, and reported the button as broken. It was not: the row chips had always told the two apart in green and grey, but the count on the filter did not, on the one screen built for planning a week of strikes. There is now a second filter beside it, no bounty (N), that shows only the ones that cost you nothing, and a line under both that gives the split in words — how many are bounty-free, how many are priced. Pressing either filter turns the other off; pressing the active one goes back to the full list. Nothing about who you are allowed to attack has changed: "strikeable only" still counts and lists exactly the targets the server would accept a launch against, so a war where every enemy is far below your Power still shows a full board instead of an empty one. Reported by KingPaladin.

1.104.1 — 2026-09-04

Changed

• Joint raids read as a sentence: who attacks, in how long, whom, and what you can do (#2609). The guild's raids panel used to open with two paragraphs, one of them 140 words, before the first name, and a threatened guildmate's row put a name beside a coordinate with nothing saying whether that name was the attacker or the victim. Every card is now the same sentence, left to right on a desk and top to bottom on a phone: the actor's box, an arrow with the countdown, the target's box, and the action underneath. A threatened guildmate's card says «Under attack» before their name, shows the enemy's fleet on the left and the colony's defences on the right, and the impact countdown sits between them. A forming raid's card says «Your guild is raiding», lists who has brought what with a slots bar, and names the target with only what the game knows about it: its sector, whether the bounty doubles there, whether there is a price on their head, whether it is dormant, or that nobody visible holds it. The panel opens on key art of allied hulls closing on a shielded world, with the three rules as short tiles and the full bounty rule one tap away under «How the bill works». The layout follows the width of the card itself, so it also holds with the chat dock open and on a rotated phone.

Also

• The changelog page test on main was red since v1.104.0: the garrison warning heading in that entry carried no issue number and the test asks a person to file it. Filed as an entry title, so the next ship's gate is not red on the previous ship's changelog.

1.104.0 — 2026-09-04

New

Your colony's own fleet can be given a battle stance (#2570)

The Planetary Defense panel has a new Defense stance. It is a standing order for the fleet your colony keeps at home: unlike a mission's stance, which is frozen the moment that fleet launches, this one applies to every battle the colony defends, and you can change it whenever you like. It needs the Admiralty node, like every other stance.

Only Assault is offered there, and the reason is the whole point of this release — see below.

The guild panel's garrison an ally action also gets the stance picker it never had, so the same defence sent from the Missions screen or from the guild screen is now offered the same orders.

Fixed

A fleet standing guard fought unstanced — and the way around that was a free bonus (#2570)

Battle stance was offered on some screens and not others. Attacking solo had it. Opening or joining a guild raid had it. Garrisoning a guildmate did not, the Campaign did not, and the fleet sitting in your own colony's hangar did not.

That last gap was worth something. A fleet sent out as a defence mission could be launched under Rampart, paying one trip 10% slower, and then collect +10% hull and +10% shields in every battle it stood through for as long as it stayed on station. Your colony's own fleet could never do that, because it never launches and so never carries a stance.

So the cheapest way to get a permanent defensive bonus was to park your fleet on a neighbouring colony and send it straight back as a defence. A round trip bought something staying at home could not.

Reported by Remy (remylz), who worked out the trick himself and told us rather than using it.

The rule now, and it is the one the game already used. Pursuit has never been offered on a defence, because Pursuit's whole price is −10% cargo and a defending fleet's hold is never charged — it would be a free +15% speed. Rampart's whole price is −10% speed, and a fleet already standing on the planet it is defending is not travelling either. Both are free once a fleet is at rest.

Assault is the exception, and that is why it is the one you are offered: its −10% hull is paid inside the same battle that gives you the +10% attack. A fleet defending from its own hangar really does pay for it.

A fleet that fights where it stands may only hold a stance it is still paying for.

So Rampart is now withheld from a defence everywhere — the picker no longer offers it, and a garrison that carries one in from an older order stops benefiting from it once it has landed. The round trip buys nothing it could not get at home.

⚠️ If you have fleets standing guard right now, read this (#2570)

• A garrison you sent out under Rampart loses that bonus in its next battle. It was collecting +10% hull and +10% shields it had paid for exactly once. Nothing to do, and nothing is lost but the bonus — the ships and the watch are untouched.
• A garrison you sent out under Assault keeps it. That price is still being paid, every battle, in hull.
• No colony's own fleet changes on its own. Every colony starts with no Defense stance, which fights exactly as it does today. It only changes when you choose Assault yourself.

Also

• A defence launched with Rampart is no longer charged the slower trip either: the launch now flies neutral rather than paying a price for a bonus that will not arrive. Travel estimates say the same thing the launch will do.
• The Admiralty research node stopped over-promising. It sold "a battle stance at launch: Assault, Rampart or Pursuit" on the one screen where you decide to buy it; it now says which of them you get where.
• Both new pickers say why two of the three stances are missing, rather than just leaving them out.
• The attack forecast admits it cannot read a defender's standing orders, next to the line that already says the same about their battle rows. It never could read the defence plan either — this is one more thing a spy report does not see, said out loud instead of quietly mispriced.
• The Campaign stays deliberately unstanced. Its ladder is tuned against a fleet without a stance, and adding one there is a balance change to every tier rather than a missing control — so it is left alone on purpose, and the code now says so instead of looking like an oversight.

1.103.1 — 2026-09-04

Changed

• The AI commanders read the forecast the way an experienced player reads the target board (#2601, closes #2598). Before a strike, a synthetic commander now averages up to six simulated battles instead of three (the cautious run six, the bold four) and reads the result for what it can measure: a cautious commander wants five wins in six, a bold one four. Among the colonies it has scouted, it previews the exact fleet it would launch, crates included, against each of the best few, and goes where the loot brought home outweighs the ships lost, with a losing fight drawn half as often and a fight that pays half the fleet's value drawn most. When the preview says no, the commander leaves that colony alone for a few hours, or until its fleet has grown or landed home, instead of asking again every few minutes or writing the colony off for a week.
• The battle preview route accepts a samples count, an integer from one to nine, and returns the average loss value before rounding for clients that rank on the net; the first three seeds are the ones the in-game preview uses, so a three-sample request is exactly that preview.

1.103.0 — 2026-09-04

New

• The server now has a limit on how much one account can ask of it — and you are not meant to ever meet it (#2554). Until now nothing stopped a single account from asking the server for the same expensive answer as fast as it could ask. The clearest case was the travel time in the launch composer: working it out runs the real route-finder and about a dozen database lookups every time, and the only thing keeping that in check was the app politely waiting for you to stop fiddling before it asked. That politeness lived entirely in the app, so one bug in it — or one client that skipped the wait — could keep the server busy for everybody.

  Nothing about normal play changes, and the numbers are set so that it cannot. The busiest screen in the game sends about 55 requests a minute; the allowance is 600, which also covers having the game open in ten tabs at once. The travel estimate gets its own separate allowance of 180 a minute, which is deliberately higher than the 150 the app is physically able to send — so composing a raid, retargeting as fast as you can click, cannot reach it. What the limit catches is a client that has gone wrong, which is the thing it was built for.

  Two details for the people it could otherwise have hurt. The count is per player, not per connection: households, phone networks and campus wifi share addresses, and counting those together would have punished people for who their neighbours are. And being limited can never cost you your session — a refused request still carries the key that keeps you signed in, and the door you use to recover a locked-out account is deliberately left outside the limit entirely.

1.102.1 — 2026-09-04

Fixed

• The Relics tab on the galaxy map now opens on its sectors, and the new sectors' relics are findable again (#2580). Tibeh01 went looking for The Banked Fires and reported that the relics of the new sectors never appear in that tab at all. They were being drawn — the tab just could not show them. It listed all forty relics as one unbroken list in roster order, above a rail of forty-three glyphs that scrolled sideways with no scrollbar and nothing to say it scrolled. Measured on a 390-wide phone, five of the forty rows were on screen; on a 320-wide one, one; held sideways, none — and every one of them was from sector 5. Thirty-six of the forty-three glyphs sat past the right edge.

  The tab now shows one line per sector — SECTOR 9 · 5 relics · 3 unclaimed · 1 yours — and opens the one you tap, glyphs and rows both. If your guild holds a relic, its sector opens by itself and is brought into view, and the sector you left open survives flying out to look at a relic and coming back. The glyphs belong to the open sector and wrap onto a second line rather than running off the edge: the rail is now exactly as wide as the space it has, at every screen size, so nothing can hide behind it. Held sideways, where there is width to spare and no height at all, the sectors lay themselves out in columns.

  The shape was chosen for the roster the game will have, not the one it has: what the panel shows at rest is one line per sector, so more relics in a sector cost nothing and a whole new sector costs a line. On the smallest phones and held sideways the list still scrolls — but it is eight labelled lines inside the sheet's ordinary scroll, not forty unlabelled rows above a rail with no scrollbar.

• The map legend said there were four relics in the galaxy (#2580). There are forty. It had been saying four since the roster grew, because the number was written into the sentence by hand. Both that count and the bastions' now come from the rosters themselves, so neither can go stale again.

1.102.0 — 2026-09-04

New

• The AI commanders play with the same preview you do (#2589). Before a strike on a colony they have scouted, synthetic commanders now ask the game's own battle forecast — the same launch preview you read before an attack — and send the composition it recommends, or stand down when it says the odds are under their bar. Their shipyards replace what that forecast says would be lost and build the counters to what it says would survive, and their construction follows the same payback-ranked advisor you see at the top of the Base tab. This keeps them on the current rules of the game without a separate copy of the combat math.

Also

• A read-only POST /missions/advise route serves that preview to any client with a spy report of its own on the target. It never launches anything.

1.101.2 — 2026-09-04

Fixed

• The game freezing mid-build should get rarer (#2573). Tibeh01 reported the game locking up while he was building, needing a page reload. The server logs confirmed it: six database transactions timed out in 24 hours, one of them 23 minutes before he wrote — and three of those landed in the same second as three background maintenance jobs.

  The cause was not any one slow operation. None of the API's background jobs is scheduled by the clock: each one simply starts a timer when the server boots, and every one of those timers repeats on a period that divides six hours exactly. So every six hours the entire fleet — roughly sixteen jobs, including the three heaviest — fired on the same instant, in the same process that serves every request. Anything a player was doing at that moment queued behind all of it, and past ten seconds the database gave up on the transaction.

  The four heavy maintenance jobs now run on their own separate schedules, spaced minutes apart and deliberately offset by seconds so they no longer share an instant with the lighter jobs either. Two of them also had a stated ordering guarantee in the code that had quietly stopped being true after the first six hours; it is true again. All four now refuse to start a second pass while one is still running, and say so in the log instead of skipping in silence.

  Honesty about this one: at release time the fix is unproven in production. The failure happened six times in 24 hours, and the schedule this changes only comes around every six hours, so nothing that could be observed in the minutes after a deploy would settle it. What is confirmed is that the four jobs came up on their new, distinct schedules. Whether the timeouts actually stop needs a day of real traffic to say, and it will be checked against the same log the issue was filed on.

Also

• The beacon retirement job had a subtler problem found during review: it never ran at startup, only on the six-hour mark, which a server that redeploys many times a day frequently never reaches. It now runs on its own schedule shortly after boot, so the beacons in systems that have outgrown them actually get retired.
• Three follow-ups were filed for the parts deliberately left alone, because they need bigger changes than this one: account deletion holding a database-wide guild lock (#2592), the beacon sweep querying once per beacon (#2593), and the remaining timers still sharing the six-hour mark (#2594). Plus #2595, for the production tooling that would have made this diagnosis minutes' work instead of a reconstruction.

1.101.1 — 2026-09-04

Fixed

• "Start research" no longer answers an error when the server is under load. Starting a research could fail outright: the server gave itself 5 seconds to write the charge, and on a busy database a large empire needed longer — one real case took 6.2 seconds — so the whole thing was thrown away and the player got an error on a perfectly legal order. Retapping did not help, because the same order took the same time every time. The write now has 15 seconds, which is the allowance the rest of the game already uses for spending. The same fix covers the other four places research is charged for: cancelling a pending order, starting a Tree node, the promotion that runs when you open the research bench, and the overnight tick that starts queued research while you are offline — all five shared one write path and all five could fail the same way (#2533).
• And if the database really is saturated, you now get told so. When the server cannot even begin the write, the three research buttons answer "state updated by another operation: try again" — the same message the rest of the game uses, on a screen that reloads itself — instead of a bare server error. Nothing is charged in that case: the order is untouched and retrying is safe. Opening the research bench and the overnight tick deliberately keep reporting a real error instead, because a browse that fails is a capacity problem worth seeing rather than hiding.

Also

• Behind the scenes, the sweep that finds transactions like this one had been looking in the wrong place. It counted only the database writes it could see written out at each site, and much of this codebase hands the work to a shared helper — so the very site that failed in production scored 2 when it really does 6, and had been passed over twice. The check now follows the work into those helpers and flags the places that write once per colony, which have no fixed worst case at all. It runs over the whole API and ranks what it finds; the sites it turned up are recorded for later work, and nothing else in this release changes because of it.

1.101.0 — 2026-09-04

Juan asked in universe-chat when guild research resets. He had watched an event season end and nothing happen, and his guild was left with research spots a former leader had spent badly. He was right, and the answer turned out to be a gap rather than a date: the guild season had never started. Not once, for the whole life of the game.

Fixed

• Guild research resets again — because the guild season now actually runs (#2547). Four screens told you your research clears "at the end of the season", and the season panel said "No season is running. The next one opens on a Monday." Nothing ever opened one. Opening and closing a guild season were both manual admin actions that nothing automatic ever performed, so guild research had never reset in any guild, ever. The season now opens and rolls itself on its own eight-week clock, and the research panel's promise is finally true. Reported by jsnam.

New

• The guild season is live, with a visible clock. The season panel now shows a real week n of 8 instead of "no season is running", so you can see exactly when your research budget comes back. The first season runs to 26 October 2026 — that is when the first reset lands, and the panel counts down to it from today.
• Everything that was waiting on a season now runs with it. Weekly sovereignty points for the systems your guild holds, the 300 season points for winning a war, the guild standings, placement rewards and guild XP at the close, and the Eschaton Project in the final two weeks — all of these were built to run on a guild season, and none had ever paid out, because there was no season for them to run on. They start now.

Also

• Nothing was taken away. Guilds keep every research level they have bought. The first reset happens at the end of this season, on 26 October, announced in the panel the whole way — not on the day this shipped. Whether that first reset should also refund the Fund already spent is still open and is being decided separately.

1.100.0 — 2026-09-04

New

Sign out other devices (#2557)

Your empire is opened by a key stored on your device, and until now a copy of that key could not be taken back. If you had ever signed in on a shared laptop, sold a phone without wiping it, or sent a screenshot to someone helping you, whoever held that copy could keep using it — and since sessions started renewing themselves, it no longer expired on its own.

Open your commander card — tap your own portrait — and use Sign out other devices, between "Log out" and "Delete account". Every other device holding your key stops working. The device you press it on stays signed in, and it tells you so before it acts.

What it says depends on your account, because the truth is different:

• Linked to Google or Apple: the other devices are signed out and you can sign back in on them whenever you like. Nothing is lost.
• Still a guest: there is no second way in, so the other devices are locked out for good and only the one in your hand will open your empire. If you play on more than one device, link your account first — it is on the same card.

There is no undo: once other devices are signed out, they cannot be let back in. Support can also cut off a leaked session on request, and for a guest that is equally final.

Fixed

• Signing back in after using the new control now works: a fresh sign-in carries your current key rather than a stale one that would be refused (#2557).

Also

Getting back into a locked-out account is unchanged. An expired key still buys a live one inside the recovery window, which is the way back for the players whose sessions had already lapsed — the only thing now refused there is a key its owner has deliberately signed out.

1.99.1 — 2026-09-04

Fixed

• A guild named after a relic is a guild again (#2551). When a war opens, an officer of an agent guild pins one standing order at the top of the guild channel — the line everyone reads for the rest of the week. If the guild that declared the war happened to be named after one of the galaxy's monuments — The Verdant World, The Forge Moon, The Exarch Guard — that pin never appeared. The agents' channel stayed silent about the only war they were in, for the whole life of the war, while every other war got its line.

  The cause was a mix-up of two different kinds of name: the code was testing the guild's name against the list of names belonging to the game's own NPCs, and those are players, never guilds. Guild names are not reserved against that list, so nothing stopped a real crew from taking one.

  Nothing was lost: checked against the live database, none of the guilds playing today holds any of those 43 names, so no war has actually gone unpinned. The mistake had been possible since 9 August. Agents still recognise a relic warden or a faction bastion when they meet one as a player, which is what keeps them from talking to a megastructure as if it were an empire.

  Also fixed on the way: a guild identified only by a blank tag could produce a pin that read "gone to war with" and then nothing at all.

1.99.0 — 2026-09-04

New

• The launch preview now asks the server what the trip really costs. The travel time on the launch bar used to be worked out entirely inside your browser, and your browser cannot see your guildmates' Hypergates. When a guildmate holds a gate in both the system you leave from and the system you are heading to, the whole journey collapses into a single jump — so the preview could be wildly long. One player was quoted 33 minutes for a trip that took about 4, and turned down launches because of it. The bar now asks the server, which prices the route with exactly the same rules the fleet will fly under, and prints that number. While the answer is on its way — or if it cannot be reached — you still see the locally computed figure, and the line underneath tells you which of the two you are looking at, and how far wrong the local one can be (#2466).

Changed

• Your Hypergates now work on the way past, not only at the two ends. Until now a pair of your own gates shortened a journey only when one stood in the system you left and one in the system you arrived at. Anywhere else they did nothing — which meant a player who knew the rule could beat the game by hand: send the fleet to a gated colony of yours near the target first, then launch from there, and arrive far sooner than the direct order would have. Clicking twice was simply the better move. The route planner now does it for you: it jumps between your gates and flies the rest of the way normally, so ordering the journey directly is never worse than staging it by hand. This is the behaviour the Hive map was designed with when it shipped in August; only half of it had been built (#2343).
• This also means your gates can shorten an attack, a spy run or a raid, where before they only helped supply runs. Nothing has become possible that was not possible before — the same route could always be flown by repositioning first — but it no longer costs you a second order and a wait. For joint raids this moves two things worth knowing: an opener whose gates reach the target fixes the impact earlier, and a fleet that would have been refused as too far away can now make the rendezvous.
• Journeys can only get shorter than they were, never longer, and fleets already in flight keep the arrival time they were given at launch — nothing currently under way changes.

Fixed

• Travel answers given to players were quoting a retired formula. A player pointed out that the distances quoted in support answers came from the old pre-Hive metric, where the gap between two sectors was worked out by subtracting their numbers. On the Hive map that is simply wrong — sectors sit on a ring, so 5 to 2 and 5 to 11 can be exactly the same distance. The correct rules now live in one page in the repository, and every number on that page is checked against the game engine automatically, so it cannot quietly drift out of date again (#2343).

1.98.0 — 2026-09-04

New

Your session renews itself, and can be restored if it ran out (#2464)

Playing now keeps you signed in. Your device's key is refreshed quietly in the background as you play, with nothing to press — come back within thirty days of your last visit and you stay signed in indefinitely, instead of being logged out on a fixed date counted from the day you signed up.

And if your key had already run out, the game can now give it back. Open Novarchy again on the device you were playing on and your session is restored on its own — same commander, same colonies, same alliance, exactly as you left them. There is nothing to click and nothing to ask for.

Two honest conditions on that. It has to be the device you were playing on, because the expired key is what proves the empire is yours. And the empire has to still be there: an account left silent for about a month is retired by the inactivity sweep as it always was, and this change does not resurrect a retired one.

Fixed

Guest empires no longer expire thirty days after signup (#2464)

A guest account was reachable with exactly one key: the signed token your browser or app was given the day you picked your commander name. It lasted thirty days and nothing ever replaced it. On day thirty it stopped working — sometimes mid-session — and every way back was shut. Typing your name again answered "that name is taken", because it was: by you. Signing in with Google or Apple opened a different, empty empire instead of yours.

The small print was wrong too. The guest warning listed only the ways you could lose an empire — clearing your data, reinstalling, changing phone — and never said the key also ran out on its own. It says so now. The expired-session message no longer tells guests to "sign in again", advice that could not work for a guest and, taken as "start over", cost them a second empire while the first went dark.

Also

One honest limit, and what to do about it. Restoring a session needs the old key still to be on the device. If you cleared your app data, reinstalled, switched phone, or logged out, there is nothing left to prove the empire was yours, and we will not hand it over on the strength of a name — that would let anyone take an empire by guessing its commander. There is no way around that one, and pretending otherwise would be worse than saying it.

Link your account (Settings → Account → Apple or Google). It takes seconds, changes nothing in the game, and makes your empire reachable from any device — including all the cases above, which nothing else can cover.

Reported by Depthcharge, who hit this mid-session and told us.

1.97.0 — 2026-09-04

New

A guild can take its own anchorage down (#2485)

Until now a Guild Anchorage was permanent for the guild that paid for it. The panel offered build, install module and repair, and nothing else. A structure only ever left the standing set when an enemy filled its breach bar, or when the guild lost every colony in its system — and that second case made it someone else's to claim while still counting against the original guild's cap. So an anchorage planted in the wrong system burned one of three slots for good, and because the price ladder reads how many you already have standing, it also doubled what the next two cost. A mistake on the first, cheapest slot left you paying 4,000 and 8,000 for the two you actually wanted, and no amount of Fund could undo it.

Leaders can now demolish one. Taking an anchorage down frees its slot immediately, but the Fund and the modules you sank into it are gone, you cannot build on those coordinates again for 48 hours, and you cannot take one down while its breach bar shows a notch.

The slot comes back at its own price: demolish your third and the replacement costs 8,000 again — what a third anchorage costs — rather than inventing a step above the ladder. The structure becomes a wreck like any other, so it stops being claimable by a rival and the 48-hour lock is the same one that has always followed a takedown.

The breach rule is what keeps a siege worth fighting: a defender cannot delete the structure at four notches to deny an attacker the fifth win they earned. It is not a trap either — the bar sheds a notch a day on its own, and a repair clears it, so both ways out are ones the anchorage already sold you.

Reported by Tibeh01, who asked the plain question: shouldn't there be a button to remove an anchorage?

1.96.5 — 2026-09-04

Fixed

• A contract card now promises what collecting it will actually pay (#2303). The reward printed on a finished contract was the full reward, but collecting it filled your mother colony's warehouse only as far as it would go and burned the rest — so the card promised one number and paid a smaller one, and never said so. With a full warehouse it could pay nothing at all and still stamp the slip Collected ✓, with no way to claim it again. A card you can collect now shows what would genuinely land in your warehouse. A contract still in progress keeps showing what it is worth, because it is not promising a payment yet. Reported by Tibeh01, whose own reward did arrive in full — the hole he described was real all the same.
• A reward that will be partly lost says so before you tap it (#2303). When part of the payout will not fit, the card says roughly how many resources would be lost, in the same red the game already uses where a loss is real. The reward is still worth collecting, so the button stays.
• A reward that would be lost entirely can no longer be burned (#2303). If nothing would fit, the contract is not collectable and says why: it stays on the board, and you collect it once you have made room. Collecting it would have paid exactly nothing, so refusing costs you nothing and keeps the contract alive instead of spending it on an empty payout. The card still reads as the finished job it is, and the Contracts lane still shows its dot — otherwise the contract this was meant to save would sit unnoticed until the daily reset took it.
• The board says which colony contract rewards are paid into (#2303). They are scaled to your whole empire but paid into one colony — the oldest — and nothing had ever said which. A player with six colonies could collect, watch the numbers fail to move on five of them, and reasonably conclude they had been paid nothing. The board now names it, in text rather than in a tooltip, so it is readable on a phone.

Also

• The board's "your warehouse is nearly full" estimate now stands down whenever the cards themselves carry the exact figure, and speaks up again when the Daily Wrap or a Weekly Chest is what you can collect. It was a client-side guess and could contradict the cards in either direction.
• Deliberately unchanged: the Daily Wrap still pays into a full warehouse and keeps its stamp. Refusing it would save the resources by spending a Weekly Chest — the Wrap's stamp is chest progress, and five of them open a chest worth far more than the Wrap's own pool. A contract blocked for space still counts toward the Wrap: finishing the work is what the Wrap is about. The Weekly Chest can likewise still be opened into a full warehouse; it warns first and waits indefinitely, so the moment is yours to choose.
• Still open, for a follow-up: a contract already collected shows only Collected ✓ rather than the amount it paid, and a claim that races a colony write in the same instant can still be stamped for nothing.

1.96.4 — 2026-09-04

Fixed

• The fleet advisor now ranks your hulls with the mastery they actually fight with (#2510). The advice you get is picked in two passes: one chooses which hulls to spend on, the other simulates the battle to check the choice. The first pass was still reading a mastered hull's counters as if the ignitions you paid for did not exist. A five-star bomber goes into a cruiser at 2.75 in the battle the game then runs — the advisor was ranking it at 1.5, and the budget went to the hull that loses the fight. On the case the tests now pin, the engine splits the verdict at equal committed value: the mastered bomber wins that wall on every seed and the alternative loses it on every seed, and unstarred both lose. The advisor used to lead with the wrong one. It no longer does.

• Ship modules count towards the ranking too (#2510). Your standing orders per hull class were absent from the pass that picks the hulls, so a class you had armed was rated as if its bays were empty. It is read now — through the engine's own multiplier, not a second copy of the arithmetic.

• Four measured claims in the advisor's own notes were wrong, and are now right (#2509). Self-reported by their author and verified one by one against the code as it now stands: a stated range that covered two hulls the ranking never sees, a lever attributed to the wrong function, a ratio quoted backwards, and — the one with a real cost — a warning comment that promised to catch a balance retune it could not actually catch. A fifth wrong number turned up while re-deriving the others and was fixed with them.

If you have bought no mastery and armed no modules, every recommendation is exactly the one you got before.

Changed

• The star ignition rule now lives in one place instead of two. The battle engine and the advisor call the same function, so the two can no longer drift apart — which is what kept this fix out of the previous release.

Also

• A test in the agents suite was failing on slow runs for a reason that had nothing to do with the code under test: its fixtures were stamped once when the file was loaded and aged past their own fifteen-minute window before the assertions ran. Same commit, red at 20m32s and green at 19m17s. The clock is now read when each fixture is built.

1.96.3 — 2026-09-04

Changed

• Nothing in the game changed with this release — it is work on the machinery that ships it (#2539). Every release has to pass a gate that type-checks and tests the whole project. That gate was doing the same work three times over: it checked the three parts of the codebase strictly one after another, and it threw away all of its results at the end, so the next run started from nothing. On a quiet machine the type-check step took 83 seconds and the build step after it another 36. Now the three parts are checked at the same time and the results are kept: on the machine that gates every release those two steps went from 119.6s to 77.1s together, and a repeat run on unchanged code finishes in 27.5s. The release you are reading was gated in 206 seconds.

  The parts were checked one at a time for a real reason, recorded back in #303 — one part rebuilt a set of shared type files while another was still reading them, and doing that at the same time was a genuine race. That was fixed first, at the cause: with results now kept between runs, the rebuild that used to rewrite all 279 shared files rewrites none of them when nothing changed, so there is no longer anything to race with. The guard was only removed once the reason for it was gone.

Also

• Two test setups that reloaded an entire module graph on every case now import it once, which is both faster and one less source of tests that fail for reasons unrelated to the code (#2296's shape). This was applied only where it was provably safe: of the 15 such setups in the codebase, 13 genuinely need the reload — the module reads a setting at load time and the test varies it — and those were left alone rather than rewritten to save milliseconds. So this narrows that class of spurious failures, it does not close it.

1.96.2 — 2026-09-04

Fixed

• Neutral AI commanders scout again (#2553). After v1.95.1 a synthetic commander whose plan was neutral, or that had no plan yet, never sent a probe at a colony it meant to strike, and since it no longer strikes what it has not scouted, it struck only Pirate Beacons and its declared enemies. Now it scouts what it would strike, like an aggressive one does; a defensive commander still sends no probes of its own.

1.96.1 — 2026-09-04

Fixed

• A cohort of real iPhone players was about to be mistaken for a crawler swarm, and blocked. Roughly forty signups a day were arriving that looked wrong on paper: all from the United States, all wearing the game's own suggested call sign, and almost none of them ever playing past five minutes. They turned out to be real people on the iOS app. Their connections reach the game through Apple's own network, and the address book the server uses to work out where a player is has no entry for that range — so it falls back to where the range is registered, which is the United States, whoever is holding the phone. Every one of those signups was being labelled American regardless of the country the player was actually in. Nothing was blocked, no wall was put in front of the door, and nobody has to prove anything to play: the signup now simply records which of our own doors it came through — the site, or the app — so the country beside it can be read for what it is (#2476).

• The one-tap door is untouched, and is now guarded against ever changing. Tapping Play still takes exactly one tap, with the name already filled in, and a first-time player who arrives with nothing at all still gets in. That is pinned by a test, so it cannot quietly stop being true.

• The privacy page says what the game stores, and now it does again. It promises to list every piece of information kept about an account, so the new one — which door you signed up through, one word, recorded once and never updated — is listed there in plain language.

• Two comments in the code had been telling maintainers something untrue for a month: that an abandoned guest empire holds its planet forever. It has not since 5 August — abandoned starts are reclaimed automatically — and believing otherwise made this look like an emergency when it was not. Corrected everywhere it was still written.

Also

• Nothing changes in how the game plays, looks or feels. This release is about the game being able to tell its own players apart from machines without ever getting in their way.

1.96.0 — 2026-09-04

New

• Relics in the new sectors (#2471). The galaxy grew from 8 sectors to 12 on 30/08, and the relics never followed: sectors 9, 10, 11 and 12 held none at all, so anyone who settled the new frontier had to cross four sectors to reach anything worth taking. Each of those four sectors now holds five relics, exactly like sectors 5 to 8 — one of every bonus, plus a fifth that gives the sector its character. Twenty new places with names and histories of their own, from The Long Tack to The Shroud Yard. Reported by Tibeh01.

  More relics do not mean more power: a guild still holds at most four, and a bonus still counts once no matter how many relics carry it, so the strongest possible hand is exactly what it was. What changes is how many guilds have one within reach — when this shipped, ten of the twelve guilds on the server held no relic at all, and both guilds that did were already at the four-relic ceiling.

  Nothing that already existed moved: every relic kept its place, its owner and the date it was won.

Changed

• The relics panel, the inventory, the activity feed and the power breakdown all used to say "five relics carry each bonus". Ten do now, and they say so.

1.95.1 — 2026-09-04

Changed

• The AI empires fight what they can beat (#2536). Over the last week the synthetic commanders lost three strikes in four against human colonies, almost always against a hangar many times their own that they had never scouted, and lost their own hangars at home to attacks they had already judged unwinnable. From this release: a synthetic commander does not strike a colony it has not scouted (a Pirate Beacon is the exception), refuses a commander whose fleet is far beyond its own unless that commander is a declared enemy or has hit it first, and in those two cases scouts before deciding. When an overwhelming attack is on its way and the commander is awake, the hangar leaves to garrison a nearby colony of its own or of a guildmate and comes home after the blow, so the wall takes the hit instead of the fleet. A commander raided in the last two days builds its shield, batteries and sensors ahead of its economy.
• Fewer probes on your colony. Synthetic commanders now send a handful of probes a day, never twice at a colony whose report is still fresh or whose probe is still in flight, and only where they may actually strike. Before, one planet was probed 96 times in a week.
• The strategist reads fleets, not just buildings. The synthetic commander's planning sees which targets are beyond its fleet, without being told anyone's fleet size, and is told that unscouted colonies are scouted first.

1.95.0 — 2026-09-04

New

• The Command Center roadmap now tells you what each level costs and how long it takes (#2415). Unfolding The road ahead or The road behind used to list only what a level grants — "Lvl 15 — 4th commandable colony" — and said nothing about the price. To find that out you had to remember the level number, close the list, scroll down to the Full progression table and look it up there. Every roadmap row now ends with the bill and the clock, in the same spelling the table uses: 51,407 Fe · 27,417 Xe · 6,854 Dt · 4h. Requested by Spaccamazza, out of the same thread that asked for the road behind itself.

  Both lists stay collapsed until you open them, so the sheet costs nothing extra to anyone who does not want it.

Changed

• The roadmap says plainly what its numbers assume, including the ceiling (#2415). Each list now carries one line about what moves its figures — your Command Center, research, tree, edict and exodus shock — and, on the lists where it actually happens, why the longest levels all show the same duration: a build time reaches a cap that a higher Command Center cannot push past. Without that, a Command Center at level 11 shows the same figure on all eighteen rows ahead of it, which reads as a bug.

  The road behind is careful not to pose as a receipt. Its rows are marked done, but the times are what the colony would build at today, not what you waited through back then, and the line says so.

Also

• The prices come from the engine's own functions, through the same single derivation the Full progression table uses, and a test opens both surfaces at once — under an exodus shock, so the multiplier cannot hide — and checks they agree level by level.
• Screen readers now announce a roadmap price as "Cost …, build time …" instead of a run of bare numbers.

1.94.19 — 2026-09-04

Fixed

• Planet traits now show both sides of the tradeoff, not just the upside (#2292). Every planet type in Novarchy gives you something and costs you something — a rocky world means more Ferrolite and less Xenogas, an ice world more Data and less energy. The compact planet badge printed only the flattering half, «🜨 Rocky (rich) · +50% Ferrolite», so choosing where to colonise from the Overview or the Missions panel meant choosing on half the facts, while the Galaxy planet sheet two taps away listed both halves of the same planet. The badge now reads «🜨 Rocky (rich) · +50% Ferrolite · −10% Xenogas», everywhere it appears. Reported by spaccamazza.

• And it says whose relief it is reporting. The Terraformer and the Colonial Outpost soften a planet's malus, so on your own colonies — the Colony tab, every row of the Overview, and the header of a one-colony empire — the badge shows the relieved figure, straight from the same engine calculation your production actually runs on. On a world you do not own you cannot see the owner's Terraformer, so it shows the planet's own malus and marks it «before relief» instead of implying that is what you would suffer.

Also

• The badge was already running off the right edge of a 320px phone on the Colony tab before this change, because it refused to wrap. It now wraps between whole facts — a percentage is never split from its resource — and measured in a real browser the page fits exactly at 320, 360 and 390 pixels.

1.94.18 — 2026-09-04

Fixed

The Aegis Bastion now defends the colony it was bought to defend (#2402)

The Aegis Bastion carries no guns on purpose: it feeds its shield to its side's pool every round, which is the whole reason to buy one. The colony defence plan decided who takes the field by asking a single question — can it shoot? — and that question, written when every gunless hull in the game was a freighter or a probe, answered no for the Bastion and left it parked in the hangar.

So the hull whose entire job is to stand in front of your fleet never stood in front of anything. And it got worse from there: hangar interdiction lets a winner run down a slice of what is left in the docks, and bombers are strong against Bastions. A commander who bought Bastions for defence and never opened the planner got no shield in the battle and handed the Bastions over as extra loot.

A hull now takes the field by default if it contributes to the fight — it fires, or it shields the fleet. Exactly one hull changes: the Aegis Bastion. Freighters, Spy Probes, Colony Ships, Colonial Arks and Continuum Haulers stay in the hangar exactly as before.

Two consequences worth knowing:

• A Bastion held back on purpose is now untouchable on the ground, the same as a held-back gunship. Interdiction reaches only what does not defend by default.
• ⚠️ If you own Bastions, open your Planetary Defense panel and check it. Until now "Bastions stay grounded" was the default, and the planner recorded a default by storing nothing at all — so a commander who deliberately set Bastions to None left exactly the same trace as one who never opened the panel: none. We cannot tell the two apart, and we will not guess. Every colony that owns Bastions and has never set an explicit Bastion number will send them into the next battle. If that is not what you want, set the Bastion row to None and it will be obeyed from then on. Every other explicit number you have saved is untouched.

Reported by jsnam (Juan), who worked out the cause from the outside — down to the line of code — before we did.

Also

• The defence planner and the ship reference no longer say "unarmed" where the game means "does not defend by default", and both now build that list of hulls from the engine instead of a hand-typed sentence, so it cannot quietly go out of date.
• Agent empires field their Bastions in the rear, where a shield tender belongs, rather than letting a formation doctrine march them into the front row.

1.94.17 — 2026-09-04

Fixed

• Closing a building card sends you back to the row, not behind the filter bar (#2507). The Base build list already returned you to the building you had just been reading. It put that row in the middle of the list — without counting the filter bar pinned across the top of it, which is opaque. On smaller phones the row you were sent back to arrived underneath that bar: on a 320-wide screen held upright, and on a 568-wide screen held sideways, the tallest rows came back with about 19px of their top edge hidden, and ordinary rows with about 3px. The list now works out how tall the filter bar actually is, at that moment and on that screen, and leaves the row below it with room to spare. On roomier phones, where the row was already clear, it simply sits a little lower in the visible part of the list.

Also

• The bar's height is measured rather than assumed, which is the point: across the four phone shapes checked in a browser it is 56px, 89px, 106px or 139px depending on the screen and on which of the two lists you are looking at, and it has changed twice this month. Nothing has to be kept in step by hand, and the Colony tab's buildings header is measured through the same seam — that list has no return-to-row of its own yet, so nothing there changed for players today.

1.94.16 — 2026-09-04

Also

• Nothing you can see changed in this one — what changed is the alarm behind the glass (#2496). Two releases ago the filter chips above the Colony tab's building list and the Base build menu stopped falling off the side of the phone: they go to a second line instead, and every filter is reachable without dragging a strip that nothing marked as draggable. The test that stands guard over that fix turned out to be able to miss it. Written one plausible way — the modern CSS spelling of the same scrolling row — the drag strip came back and the guard reported everything fine. It also threw away the piece of information that says when a rule applies, so moving the phone's 44-pixel tap target into a desktop-only rule would have shrunk the chips back to 36 with nothing complaining, and a single unclosed bracket anywhere in the stylesheet made the whole test run spin until it ran out of memory instead of saying which line was broken.

  All three are fixed, and the guard now fails on each of them — including on the rotated phone, which is the case the last release measured and the old guard could not describe at all. The two previous releases had to check these numbers in a real browser because this test could not be trusted; the phone rails were re-measured against it for this one, and every structural result matched: two lines of chips at 390, 360 and 320 pixels wide, one line when the phone is turned sideways, and the 8 pixels that costs.

  No game code was touched, so the game plays exactly as it did before this release.

1.94.15 — 2026-09-03

Fixed

• The Wraith guild had stopped growing, and its vaults stopped becoming ships. The hostile faction's thirty empires fund their guild by donating to its treasury, and those donations buy the guild levels that seat its last five members. For two weeks not one of the twenty-five banner-carrying empires managed a single donation: the rule deciding what they could spare asked them to hold twice a donation's price before paying it once, and the price scales with the empire. So the guild sat one level short, five of its operatives locked outside, and the vault meant to feed it was emptied into hulls every five minutes instead — taking with it the exact pile the next donation was waiting for. What they give is now bounded by what their current guild directive is actually waiting for, and the pass that converts vaults into ships leaves the donation's price alone until it can be paid, then stops holding it back the moment it is (#2360).
• They also stopped building one kind of ship. Every surplus went into the single strongest hull they had unlocked, which after the Cutter is a cutter and nothing else — a fleet that could hit and could not hold. Their colonies now buy enough of the hull a defensive wall is counted in before spending anything on the strike fleet, so raiding an outlying Wraith colony should meet defenders more often, and a raid on a mother a heavier fleet (#2360).
• And they stopped asking for money they had already spent. Their build yard, shipyard and laboratory each priced their orders against the same purse in the same tick, so the last of the three routinely ordered what the empire could no longer afford — 22 refused orders in a single pass. Nothing a player could see, but it was a third of their economy running on a stale balance (#2360).

1.94.14 — 2026-09-03

Fixed

• The changelog page can no longer go backwards. With several ships going out a day, each one published novarchy.com/changelog from its own copy of the repository — so a ship that finished second with older notes could overwrite the page a newer one had just put up. That is what happened on 3 September: the page fell back to an older version while a bulletin was telling players to go and read it. The page is now compared against what is actually being served before anything is replaced, and a publish that would move it backwards is refused out loud instead of going through quietly (#2498).
• A misleading error when a release could not be tagged. The release step blamed an already-existing tag whenever tagging failed, including when the real reason was something else entirely. It now says which of the two actually happened, and names the fix that applies (#2491).
• Checks that could silently miss what they were looking for. Four places in the release tooling searched a list in a way that, on a large enough list, could report "not found" for something that was there. Two of them mattered: the check that stops a release from reusing a version another release just took, and the scan that keeps credentials out of a fast-tracked change — measured missing a planted key on every one of twenty readings of a large diff. All four now read the whole list (#2499).

Changed

• The test-suite report says what it actually examined. When the pre-merge checks skip the Python suite they list the files they cleared; entries they could not read anything from — an unrelated repository sitting inside the folder, for instance — were listed among them as though they had been checked. They are now counted separately and marked as not examined (#2499).
• Internal tidying in the same pass: a temporary working copy that could be left behind on one failure path is now always cleaned up, and one self-check that only passed because of where it sat in the file now pins the fact it claims to (#2499).

Nothing in the game itself changed with this release: it is all release tooling. What players should notice is that the changelog page stays correct while several updates go out on the same day.

1.94.13 — 2026-09-03

Fixed

• The alarm you tap during a launch stays where you put it (#2492). On a phone, sending a fleet waits about a second on the server before it moves you to In flight — and for that whole second the red under-attack chip in the top bar, and the ⚠ Alarm tab above the form, both stayed live. Tapping either one took you to the inbound fleets, and then the launch finished and dragged you straight back off them, closing the panel in the middle of its own scroll. Now the last thing you chose is the thing that wins: a launch only moves you to In flight if you have not picked a tab yourself since you pressed Send. The fleet still leaves, and the form still resets for the next mission either way. Nothing about the chip was made slower or deader to get there — an attack coming in outranks a launch confirmation, so the chip stays live throughout, exactly as it became in v1.94.9.

• Arriving at Inbound now says so out loud (#2492). If you play with a screen reader, tapping the alarm chip used to change the screen in silence: the panel appeared, the page scrolled, and nothing was announced, so the only way to discover any of it was to swipe forward and hope. The phone now announces "Showing Inbound: the fleets heading for your colonies" whenever that panel comes up — whether you got there from the chip in the top bar, from the chip while already in Fleets, or by pressing the ⚠ Alarm tab yourself. It goes quiet again when you leave, and when the fleets it is describing have landed, so it never tells you about an attack that is already over.

1.94.12 — 2026-09-03

Fixed

• The build lists no longer call the same category two different things (#2494). The Base build menu labelled a family of buildings «Civilian» while the Colony tab's buildings header called the very same family — the same buildings, in the same colony — «Civic». Until the filter rails started wrapping, each word sat past the edge of its own scrolling strip, so you had to drag two different rails to catch them together. Now both rails are fully on screen, and «Civilian 4» one tab away from «Civic 4» simply read as two different categories. Everything says Civic now: the two filter rails, the building sheet's own label, and the map legend on desktop. Civic is the word the rest of the game already uses for the buildings that govern a colony — the Command Center, the Laboratory — and it is the one the filter order was written around: what you produce, what defends it, what governs it.

• The build sheet's filter pills are a real thumb target (#2495). The six pills at the top of the Base build menu — All, Production, Military, Civic, 🔒 Locked and the level sort — were 36 pixels tall, eight short of the target size every other button on a phone honours, and had been since they were added. They are 44 pixels now, so the filter you meant to press is the one that answers. The rail still fits on two lines at every phone width, including a 320-pixel screen, and no pill has moved off the edge: the shorter word and the pills' own trim gave back more room than the extra height cost.

1.94.11 — 2026-09-03

Fixed

• The Fleet Advisor now ranks your ships using the mastery you have actually earned (#2484). The Advisor builds its recommendation in two steps: first it works out which of your hulls are worth spending the budget on, then it simulates the resulting fleet against the target to check the plan holds. Those two steps had stopped agreeing. The simulation fought with your Star Forge mastery — it has done for a while — but the ranking that came before it read the plain Shipyard numbers, as though you had never trained a hull in your life. So the Advisor chose the fleet as if your stars did not exist, then confirmed that choice with a battle that knew they did, and never went back to reconsider.

  For anyone with a decorated roster the advice was wrong in a way that was hard to spot: a hull you had poured a mastery track into could be ranked below one you had never touched, and the budget went to the weaker ship. A fully mastered track is worth up to +60% attack and +35% hull and shield, which is more than enough to reverse the order of two ships. On the case now pinned in the tests, for the same resources committed, a fleet of thirteen Battleships loses a battle that thirty-five five-star Cruisers win outright — and the Advisor had been rating that Battleship at nearly twice the Cruiser.

  The ranking and the simulation now use the same mastery, so the fleet the Advisor recommends is the fleet it is actually recommending. If you own no stars, nothing changes: the advice you get is identical to before, and that is checked rather than assumed.

1.94.10 — 2026-09-03

Fixed

Nothing in the game changed with this release — it is a repair to the machinery that ships the game, and it is here because that machinery had been quietly telling us less than it knew.

• The pre-merge test gate now publishes the verdict it measured (#2217). Every change is put through the full suites before it can be merged, and the result is meant to be stamped on the exact commit so the next step can trust it instead of running everything again. When those suites ran on the build machine, that stamp was never applied: the gate's own log files landed inside the copy of the code it was checking, the gate saw a modified folder, and it correctly refused to certify a commit it could no longer vouch for. So every such change was tested twice, about fourteen minutes each time. The logs now live outside that copy, the check passes on its own merits, and the stamp is applied — with the check itself left exactly as strict as it was.

• The gate stopped skipping a test suite it should have run (#2455). The Python suite reads a number of files that belong to the rest of the game, so the gate works out, on every run, which of them a change touches. That calculation could silently lose a match: on the same change, the suite was correctly included once and skipped twice. Measured against the live data, four runs in twenty-five got it wrong, and it was getting worse as the tests grew. It is now exact — twenty-five out of twenty-five — and the gate prints why each suite ran or did not, so a wrong call shows up in the log instead of having to be deduced afterwards.

Two long-standing faults in the gate's own self-check were found and fixed on the way: one of its safety tests had never actually been exercised, and another gave different answers on different machines. The self-check goes from 94 assertions, two of them failing, to 118, all passing.

1.94.9 — 2026-09-03

Fixed

• The red under-attack chip always opens the alarm now, even if you were already on Missions (#2449). On a phone, tapping the chip while the Missions tab was already open in In flight or New mission did nothing at all — and kept doing nothing, while a hostile fleet closed in. The chip switched to a tab you were already on, then looked for the Inbound panel, which on a phone is not on screen unless its own segment is showing. Opening Missions from another tab always worked, which is why this was easy to miss and miserable to hit: the usual way in was composing a mission, or having just launched one, when the attack started.

  The chip now takes you to the alarm from wherever you are, and brings the Inbound panel into view instead of leaving you at whatever scroll position you had. Opening the Missions tab by hand still leaves you exactly where you were — only the alarm chip moves you — and a half-composed mission is still there, untouched, when you switch back to it.

  One deliberate exception: if the attack has already resolved by the time you tap, the chip leaves you where you are rather than dropping you on an empty "No inbound attacks detected". The HUD and the Missions tab check for incoming fleets separately, so the chip can briefly outlive the threat it was counting down.

1.94.8 — 2026-09-03

Fixed

• The build filters stop falling off the side of the phone (#2450). Above the Colony tab's building list, and above the Base build menu, sits a row of filter chips — All, Production, Military, Civic/Civilian, 🔒 Locked, and the Level ↑ sort. On a phone that row was wider than the screen and did not wrap: it was a strip you had to drag sideways, with nothing on screen saying it could be dragged. Two chips on each row were simply not there — «Military» cut in half and «Civic» past the edge on the Colony tab; «Civilian» cut in half and «🔒 Locked», the way into the catalogue of everything you have not unlocked yet, gone entirely from the Base menu. Both rows now go to a second line instead, so every filter is on screen at rest, at every phone width down to the smallest (measured at 390, 360 and 320 pixels: five of five and six of six chips visible, where it had been three of five and three of six). The chips, their order, what they filter and the Level ↑ toggle are all unchanged, and the desktop looks exactly as it did. The rows are sticky, so this costs about one line of height — roughly two thirds of a building row — which is the trade for a filter you can no longer reach by accident only.

1.94.7 — 2026-09-03

Fixed

• The campaign's rehearsal count no longer reads as a promise (#2361). Under the count the sheet said "Your engagement is one fixed roll", which is true and still left the reading that your roll was one of the three shown above it. It never was: the rehearsals run on fixed preview rolls, and your engagement runs on the level's own. A player took three rehearsals won for a guarantee at level 50, engaged, and lost. The sentence now says whose roll the rehearsals are not, and it says the right thing on each surface: on a climb your own roll is the same every time you tap that level, so what moves the result is the deployment, the refit or the research; on a rematch the dice really are fresh every ticket. Before, the rematch sheet claimed the roll was fixed and then contradicted itself one sentence later. Reported by spaccamazza.

• The forecast's ship stars are now guarded (#2186). The rehearsals have fought with your roster's stars since the engagement sheet shipped, so a starred fleet no longer reads a worse chip than the battle it would actually fight. Nothing tested it, which meant a refactor could have quietly dropped it with every suite still green. It is pinned now at the level where the answer is opposite rather than merely different: the same roster reads none of three won unstarred and three of three at five stars.

Also

• The two campaign design records stop describing the old sentence, and stop crediting the rematch desk with printing a caveat it does not show. The desk shows the count; the sheet its button opens shows the sentence, which is also where the weekly ticket is actually spent.

Not changed, and deliberately: the campaign still does not tell you your battle's own roll before you take it. Making the preview run the real seed was voted on and declined before this report, because it turns every level into a search you can solve in the client and makes the depth board rank whoever searches fastest. The rule that gives you one stable battle per level, so that retrying with the same orders cannot re-roll the dice, is untouched.

1.94.6 — 2026-09-03

Fixed

• The beacon forecast row no longer prices loot at the probe's clock (#2454). A Faro scanned near-empty and since respawned rendered loot 53 and a red net −347 three lines under a rationale bullet quoting the true pool: the forecast simulated the pool as the probe saw it, while the advisor's own warning already used the recomputed current one. The forecast now reads the beacon-current pool too, so the row, the loot chip, the freighter sizing and the register all quote the same clock. The mismatch was one-directional — the scan clock could only understate — so no forecast ever promised loot that was not there.
• "Sets you back ≈N even when you win" now says "at least ≈N" (#2454). The warning's loot figure is a ceiling on what the raid can pay, so the derived setback is a floor, not an approximation — the old wording could be read as a number that might land either side.
• A hangar with no freighters is now told why its winning raid does not pay — and what to do about it (#2454). Before, a raid whose target could cover the losses but whose staged fleet could not carry the loot home showed only a red net line and a «hold full, add freighters» chip — advice a player who owns no freighter cannot act on. A new warning now names the shortfall as hold, not loot, and tells a freighter-less hangar to build Freighters in the Shipyard first.

1.94.5 — 2026-09-03

Fixed

• Base tab: Build and Move no longer lose your place in the build list (#2446). On desktop, pressing Build or Move on a building card used to close the card and drop the build list back to the top — the row you were acting on, with its «Placement…» state and hint, could sit off-screen exactly while you were placing. The list now snaps back to that building, the same courtesy closing a card already had (#2198). On the phone the sheet still closes so you can aim at the map, as before.
• Phone: a reopened sheet starts at the top (#2446). Reading a tall building card down and closing the sheet used to leave the next opening of the build list at an arbitrary scroll offset unrelated to anything you did. Every bottom sheet now opens its content from the top; the snap-back to a row you were on still works exactly as shipped.

Also

• The build-list snap-back tests now pin the real containers (the desktop panel scrolls itself, the phone list scrolls inside the sheet body) and the one-shot guard of the snap hook, each verified to fail under the mutation it guards against (#2446).

1.94.4 — 2026-09-03

Fixed

• The launch preview no longer fights your own hulls unstarred (#2148). The mission composer's what-if battle priced the spied defender's ship stars but never your own: a starred player probing a starred defender read a one-sidedly pessimistic verdict and loss estimate, while the real battle applied the multipliers your launch snapshots. The Advisor's recommendation and the forecast row now fight your actual star steps — the same numbers the server fights on arrival — so the verdict word describes the battle you are about to launch. Until your stars load (or if the read fails) the preview stays exactly as cautious as before, never more optimistic than the truth. The campaign engage chip has the same defect and stays tracked separately (#2186, #2361).

1.94.3 — 2026-09-03

Fixed

• Star pips now explain themselves on touch and to screen readers, on every surface at once (#2149). The mastery stars next to a hull only spoke through the browser's hover tooltip: on a phone there was no way to learn what they meant, though the Ship Stars design has promised a long-press from day one. Now press and hold the pips (or hover on desktop) and a tooltip opens with the honest line — star level, sub-level, and the three explicit multipliers. It opens where there is room (never clipped by the Forge rail, the hull chip row or the Hangar nameplate, and it flips above the row at the bottom of the screen), the hold never selects the hull underneath, and a plain tap never sticks it open. Screen readers now announce the same full line.
• FLEETS AWAY rows on a spy report carry star pips (#2149). The report has always revealed the target's mastery for the whole empire, but the card only drew pips beside the harbour fleet — a hull whose entire count was away showed its stars nowhere. Away rows now carry the same pips from the same record, each fleet labelled by what it is doing and when it is due back, and long fleets wrap under their own label instead of bleeding into the next one.
• The launch preview now says what it does not fight (#2149). The forecast simulates only the fleet the probe saw in harbour; nothing told the attacker that part of the defender's fleet was away and might be home by arrival. When the report lists fleets away, the Advisor now warns with the hull count and the return clock — "earliest back in ~1h30m", "some already due back, the rest from ~8h on", or "due back by now: they may already be home" — and says plainly that the forecast is optimistic if they return first.

1.94.2 — 2026-09-03

Also

• The Sounding: the design behind the new engagement sheet is now in the repo. Nothing changes in the game with this release. DESIGN-CAMPAIGN-V2.md is the campaign redesign (the engage sheet as the outcome, the Wall Report, the demotion rule, the ship order and the decision record of 2026-09-03), DESIGN-CAMPAIGN-V2-LORE.md is its fiction (the Kept, six layers, six echoes), art/mj-prompts/08-campaign-sounding.md is the Midjourney batch for the art still to come, and design/campaign-v2/ holds the fifteen artboards of the design canvas. The first slice of that design shipped as v1.94.0; the story layer and the art are the next ones.

1.94.1 — 2026-09-03

Fixed

• Crate opens, contract claims and anchorage deliveries survive a busy server (#2414). During a deploy-window stall, opening a resource crate or claiming a contract — a card, the Daily Wrap, the weekly Chest — could die after 2 seconds with the literal words "Internal server error" on screen, and an anchorage freight the player had already paid for kept re-expiring every sweep for as long as the stall lasted. All seven remaining short-budget transactions in this family now wait out a stall of up to 5s and get 15s to commit; when the server is so saturated the operation cannot even start, the player-facing four answer the legible "State updated by another operation: try again." message the game already knows how to reload on, instead of a raw error.

Also

• The starved-pool answer that was copied across five spend paths now lives in one place, so a future change to the transaction budget moves the whole family together, and the policy for reads on a starved pool is decided and written down where the code enforces it.

1.94.0 — 2026-09-03

New

• The engage sheet becomes the outcome (#2382). Before: Engage closed the sheet onto a board that had not refreshed yet, and win, loss and draw differed by one coloured word in a card. Now the sheet turns into the verdict: a one-second "Making contact" bar, the real round log as a skippable ticker, a ledger of both lines with hull bars, and what was actually banked. A loss or a draw gets a Wall Report: how much of the garrison went down, where your line broke and what held them, one diagnosis, a chip that applies the suggested orders or opens Research or Refit, and the changes worth trying, each rehearsed three times.
• Same orders, same answer. After a loss with unchanged inputs the button reads "Change the orders"; "Replay the same orders" stays one tap away, with the exact outcome of the last attempt quoted from the stored report.
• Attempts per level. The tab counts how many times a level has been fought and with which orders. The ledger starts empty for everybody: taps spent before this release left no trace to backfill from.

Changed

• The four verdict words ("Clear win", "Likely win", "Coin flip", "Defeat") are gone from the campaign, the rematch desk included: three rehearsal dots and a count, and a sentence saying a rehearsal is not a promise.
• "Too fast, give it a second" no longer appears on a manual tap: the client keeps its own two-second brake.
• "Banked" prints what the landing really deposited, including what a full warehouse refused and any back-pay settled; older runs fall back to the nominal figures and say "confirming".

Also

• Admin analytics count campaign engagements per day and cut humans by campaign depth (0, 1-10, 11-28, 29-32, 33-49, 50+) with the seven-day active share per band.
• Schema: one new table (CampaignAttempt) and two columns, applied at API boot; account deletion and the season reset clear the ledger.

1.93.0 — 2026-09-03

New

• The raid forecast now prices its ship losses — and says when a won raid still loses money (#1949). The Advisor's battle forecast showed win/lose, a loss percentage and a loot figure, but never what the lost ships are worth: the early pirate-beacon loop could be won while netting less loot than the interceptors it cost, and the forecast read as profit. The forecast row now carries a net line beside the loot — net −250 (est. 400 in lost ships vs 150 loot), red when the raid runs at a loss, unsigned 0 at break-even — for the fleet currently staged, on every attack target, beacon or colony. And when even the Advisor's own recommended fleet is predicted to win at a loss, its warnings say so in words, comparing the priced losses against the target's current loot potential (beacon pools are re-clocked to now, and a haul capped by your cargo bay is never blamed on the raid). Both figures are estimates and the sentence names what it excludes: salvage from the wrecks of both sides.

1.92.1 — 2026-09-03

Fixed

• The incoming-attack alert now shows you the attack (#1954). On desktop, clicking the red under-attack chip opened the Missions tab at the top of the page, while the inbound-fleet box sat below the whole mission form — past the fold, where you had to already know to scroll. The same click now switches the tab and smoothly scrolls the inbound box to the center of the screen. Opening the tab any other way lands exactly where it always did, and the phone keeps its own path: there the alarm view already surfaces itself on arrival.

Also

• Reviewing this surfaced a separate, older phone quirk — the chip does nothing when the Missions tab is already open in another view — filed as #2449, not changed here.

1.92.0 — 2026-09-03

New

• Sort the building lists by level (#2061) — both build lists, the Colony tab's upgrade options and the Base tab's build menu, now carry a «Sort: level ↑» toggle at the head of their filter chips. Turn it on and the rows order lowest level first — the quickest way to spot what is lagging — while queued builds stay on top and locked entries stay at the bottom, exactly as before. Turn it off and the list is byte-for-byte the order it always was; the toggle also composes with the family filters, and your choice sticks while you browse, switch colonies, and open building cards. Requested by atyo416 on Discord.

1.91.0 — 2026-09-03

New

• Pick your soundtrack: a music track selector with repeat and shuffle (#1794). The game always shipped three ambient tracks — Ambient Light, Deep Space, Eschaton Theme — but you could only ever hear the first: the other two sat in the bundle, wired to nothing. The audio panel (the 🔊 chip on desktop, the "More" sheet on phones) now has a Track dropdown and a Repeat/Shuffle pair: Repeat loops your chosen track (the old behavior, still the default), Shuffle jumps to a different track whenever one ends. Your choice is remembered like the other audio settings. Battle music stays what it was — the replay's own score, not a dropdown entry. On iPhone and iPad, where the volume sliders cannot work and used to leave the 🔊 chip as a bare mute button, the chip now opens the panel too, so track selection reaches iOS.

1.90.7 — 2026-09-03

Fixed

• Chat "Replying to…" preview bar overflowing the panel frame (#1989): arming a reply painted a bar whose right edge — cancel button included — spilled past the chat panel's rounded border into the background. The bar could not shrink to its row (its minimum width was stuck at the full, untruncated width of the quoted excerpt), so long quotes pushed it out of the frame; measured, it painted 474px inside a 400px dock. It now shrinks to exactly the composer's row on every chat surface — Chat tab, dock, and the guild tab's embedded panels — on desktop and phone alike. The phone composer shell also hardens its own box so the same overflow family cannot reopen there.

1.90.6 — 2026-09-03

Fixed

• Base tab: closing a building card no longer loses your place in the build list (#2198). Opening a building's card from the build list and closing it dropped you back to the top of the list, so browsing a long list meant scrolling back down after every read. Now the list comes back centered on the building you had open — the card's × and the ‹ Buildings gesture alike, on the desktop panel and in the phone sheet. Dismissing the whole sheet still reopens the list at the top, and the snap lands on the row itself even when the list re-ordered while the card was open. Starting a build or a move from the card still returns the list to the top — that path is #2446.

1.90.5 — 2026-09-03

Fixed

• Synthetic commanders no longer swallow a thank-you they never said (#2400). Since the sticky guild-aid memory of #2243, the "favour still owed a mention" flag was marked told the moment the fact sheet behind a message was built — including messages that then failed to generate, came out empty after cleanup, were suppressed as echoes of the channel, or never mentioned the favour at all. A guildmate who sent resources or pushed a timer could go entirely unthanked, while still standing listed under the agent's sympathies. The flag is now spent only when a message whose prompt actually carried the favour is confirmed posted during the owner's active hours: a failed or suppressed line leaves the thank-you owed, and it simply rides a later message. Guild-life lines composed without the fact sheet (welcomes, board reactions) can no longer eat the flag, and a favour that arrives while a message is already in flight stays owed until a prompt has really carried it. Known residual, stated plainly: a message that carried the favour in its prompt but chose not to voice it — a battle taunt, say — still counts as the telling.

1.90.4 — 2026-09-03

Fixed

• The daily-exploration refusal reads your timezone at the reset instant it names, not at the moment it is shown — so the Saturday before a DST switch no longer promises a local hour your wall clock will contradict (told 06:00, reset lands at 07:00) (#2409).
• Two exploration launches fired at the same moment from two different colonies could squeeze past the daily cap together and both fly; the cap is now enforced inside the launch itself, so the last slot of the day goes to exactly one fleet (#2409).
• A guild join, leave or keystone updates the exploration-cap sentence on the galaxy sheet immediately instead of up to 30 seconds later — and leaving a Science-4 guild no longer overstates your cap until the next poll (#2409).
• If the exploration-window feed stops answering, the "Up to N explorations per day" sentence falls back to your guild's real entitlement after a few failed tries instead of quoting a frozen number for the rest of the session — while your used-count, the greyed button and its countdown keep their last honest reading (#2409).

Also

• The design docs' inventory of daily caps tells the truth again — the game runs three distinct day boundaries (exploration on the 05:00 UTC contract day, fleet recall on a rolling 24h, guild aid on the UTC calendar day) and the cap statements now carry the Science-4 bonus term — and the exploration count gained a database index for veteran mission histories (#2409).

1.90.3 — 2026-09-03

Fixed

• Partial geo-pipeline loss is no longer invisible to the country sentinel (#2299). The #2272 watchdog only fired on a TOTAL loss of the edge's X-Country-Code feed — ten distinct clients in a row with not one resolved header between them — so a stale MMDB shedding fresh IPv6/CGNAT coverage while IPv4 kept resolving stamped a growing share of new signups' country as NULL, permanently, with no signal anywhere. The api now also watches the share of distinct edge clients without a usable header over a rolling 24h window and warns in the log above a 20% watermark (COUNTRY_SENTINEL_MISS_RATE, wired through both compose files; the measured healthy baseline is ~0%), repeating hourly while the loss lasts, with one recovery line when it drops back under. The same legitimate-NULL exclusions apply (the dev door, and traffic that never crossed the edge), one client repeating can never fake a verdict, and a throwing tracker still never costs a signup. The window is process-local: a restart resets it and the verdict rebuilds within a few hours at normal traffic.

1.90.2 — 2026-09-03

Fixed

• Clicked buttons no longer keep the browser's white focus ring (#1977). Clicking a market filter chip under "Open offers" — or any other button, on browsers that paint the system focus ring for clicks — left a white outline stuck on the button until focus moved elsewhere, reading as a second "selected" highlight on top of the real active-filter styling. Pointer clicks now leave no ring anywhere in the app, while keyboard navigation keeps the cyan focus ring on every button.

1.90.1 — 2026-09-03

Fixed

• The planet sheet no longer calls a scouted world a «free planet» while your intel is still loading (#2350). Opening a coordinate you had scouted, the sheet read «No colony — free planet» for one round trip before your spy report arrived — the same false claim the veil-pierce work removed, back for a moment on every open. The sheet now says «Checking your intel…» until the lookup answers, and shows the true verdict the instant it lands. Once a coordinate has answered, the periodic refresh never blinks it back to «checking».

Changed

• Old Event-Sector spy reports now say why they are old (#2350). A report you filed in a previous season still unlocks reused sector-13 coordinates — that is by design, your espionage there is permanent — but its dated header said only «14d ago». Everywhere the report is dated — the planet sheet, the composer's target panel, and the phone's compose screen where the fleet is committed — you now read: «This report predates the current season; the sector was swept since.» Nothing about what you can launch changed.

Also

• The out-of-order guard on the spy lookup — what keeps a slow reply for the previous planet from wiping the answer for the one you are looking at — is now pinned by real eviction-race tests on both surfaces; deleting either guard line turns the suite red (#2350).
• The client-side spy index from the events feed keeps its fallback role by documented decision: it still powers the Recent-targets list, which the per-coordinate lookup cannot, and it covers the round trip and a failed lookup (#2350).

1.90.0 — 2026-09-03

New

• A third digit on abbreviated HUD figures (#2060) — requested by arcticpaladin. A stock past a million used to read 12M beside a building that costs 12.4M: the bar kept the magnitude and dropped the one digit that says how much more is needed. Every abbreviated figure — top-bar stocks, rates and caps, phone pills, research costs, the ⚡ power tag — now carries three significant figures: 1.23M, 12.3M, 123M, 2.59T. The research Still needed number sharpens to the same grain and still only ever rounds up: a 46.34T gap reads 46.4T, never 46.3T. Pairs with the v1.75.3 hover-tooltip fix, which arcticpaladin also drove.

1.89.0 — 2026-09-03

New

• The Command Center's roadmap now runs both ways (#1946). The Command Center sheet used to publish its ladder only forward — the gold "At the next level you unlock" box and the collapsed road ahead. What the levels you had already climbed granted was nowhere to re-read. Now a second collapsed strip, "The road behind — N unlocks already claimed", sits under the gold box: every passed level, most recent first, ✓-marked and dimmed, each with the unlock it granted — and the same fine print as the road ahead where a gate actually lives elsewhere (Colonization, mother-only buildings). At Command Center 30, where the gold box goes quiet, the whole history stays readable. Closed, it costs one row of the sheet; the road ahead is unchanged. Requested by Spaccamazza in #universe-chat. The same thread's second ask — cost and build time on every roadmap row — is tracked as #2415.

1.88.11 — 2026-09-03

Also

• Sent-pill safety net, now provably watched (#2314). Nothing changes on screen: this ship hardens the tests around the launch-again card's transient «Sent ✓» paint (the #2229 fix). The suite now proves the pill really appears with a launch and clears back to the launch button after ~4 seconds, that leaving the Missions tab mid-countdown still honours the launch without leaving anything ticking behind, and that no future rewrite of the timer machinery can slip past the tests by freezing, self-rescheduling, or never arming at all. Each new guard was checked by breaking the code on purpose and watching it catch the break. A stale root-cause note in the Discord-announce test was also corrected (eleven module re-evaluations, not nine).

1.88.10 — 2026-09-03

Fixed

• The first volley of a replay no longer risks a late bang at 4× (#2289). The sound of a weapon used to start downloading only when that weapon first fired — and at 4× speed the Destroyer's charge cue leaves barely half a second before the beam lands, so on a first watch of the session, on a phone with a cold cache, the bass drop could arrive after its own explosion. Now the weapon sounds of that battle — only the ones actually in it — start downloading the moment the replay opens, while the ship models are still loading, so they are ready by the first shot at every speed. Nothing audible plays during the warm-up, muted players download nothing, and every other sound in the game behaves as before.

1.88.9 — 2026-09-03

Changed

• Guild Help board: one retention clock — every finished request now leaves 3 hours after its real end (#2244). A request that ended early (fully covered, withdrawn, or out of speedup pushes) already left the board 3 hours after its actual end since v1.72.4 — but one that simply aged out unanswered kept an older rule and lingered as 'gone' for a full 24 hours past expiry, up to 48 hours after it was opened. Measured on live boards, that long tail was mostly rows nobody had touched, sitting on top of the requests still open. Now one rule covers every ending: a finished request leaves the board 3 hours after the earlier of its close and its expiry. Open requests are untouched, and a request that just lapsed still stays its 3 hours so the guild sees what it let go.

1.88.8 — 2026-09-03

Fixed

• The Anchorage no longer errors out when the server is busy (#2390). Signing a bulk charter, spending a Chrono Shard on a convoy in flight, or buying a shard at the chandlery could fail with a raw "Internal server error" during deploy-window stalls — the same failure that v1.87.10 fixed for the Chrono and Dark Matter spends one panel over. All three Anchorage spends now ride out a stall and commit, and when the server truly cannot start the operation they answer the same "State updated by another operation: try again." message the rest of the game already shows — with nothing charged.
• Colony orders survive a slow moment too (#2390). Any colony write — a build order, a ship order, a move, the accelerator toggle — could expire mid-operation on the old 5-second internal budget (seen in production right after v1.87.10 shipped). The budget now matches the rest of the game's, so an order that took a few seconds too long commits instead of dying.

1.88.7 — 2026-09-03

Fixed

• Parallel ship runs no longer risk clobbering each other's scratch files (#2399). Twice in one day, two ships running at the same time wrote their PR text to the same shared temp filename and each briefly published the other's description — caught before anything wrong was merged. The ship tooling's conventions now require every scratch file (PR bodies, release notes, review ledger rows) to carry its own run's number or be created in a unique temp directory; the paths that are shared on purpose (the deploy lock, the deployed-version record) stay shared. Nothing changes in the game itself — this hardens the pipeline that delivers updates to it.

1.88.6 — 2026-09-03

Fixed

• The attack advisor's confidence badge no longer claims «solid» for a forecast built on an assumed Tech Tree (#2389). Below spy tier 2 a report carries no defender Tree, and the whole preview — energy factor, Bubble, Cannon, combat multipliers — runs on your own Tree substituted in; the badge still read «solid», overstating what the simulation actually knows (a truly power-clamped colony could preview as in deficit, badge green). Now such a report reads «estimate», and a warning explains that the defender's Tree is not in the report and can make the forecast optimistic or pessimistic either way. A tier-2 report — even of an empty Tree — still reads «solid»: what was spied is solid; what was assumed is an estimate.

1.88.5 — 2026-09-03

Fixed

• Missions badges now quote the newest thing the server actually said (#2332). When a research completed, the colonize badge could keep quoting the old level — «Colonization tech 2 is not enough: you need ≥ 3» while the Research tab already showed 3 — for up to half a minute whenever the tab's own research call failed or lagged behind the regular poll. The two answers now merge per level, and since a level can only ever climb, the higher one is always the newer truth — in both failure directions.
• A Colony Ship no longer goes uncounted in a split-failure window (#2332). When the single-colony read failed while the colony list succeeded, a Colony Ship visible only in the fresh list row was not counted for up to a poll window and Launch stayed grey on a founding the server accepts. The mother's hangar now reads whichever of her two snapshots the server served later — still exactly one row per colony, so a departed ship is never re-credited by a stale twin.

1.88.4 — 2026-09-03

Fixed

• The daily-exploration refusal now speaks your clock (#2216). It used to end in a bare "(05:00 UTC)", leaving the timezone arithmetic to you; it now reads like "explorations reset with the daily tasks (14:00 your time · 05:00 UTC)" — exact even in half-hour timezones, and just "(05:00 UTC)" when your clock is UTC. The hour is derived from the same engine constant that runs the reset, so a retuned reset can never leave the banner naming the wrong time.
• The two exploration-cap sentences can no longer disagree (#2216). "Up to N per day." and "X of M left today" came from two different server routes, and a Science-4 guild member whose guild route was failing could read "Up to 5 per day." directly above "1 of 6 left today.". Both sentences now read the same server-computed cap — guild bonus included — with the old guild-derived figure kept only while that number has not arrived yet.

Also

• The design docs and the cap's own documentation now state the shipped rule — the exploration allowance counts the contract day (05:00 UTC → 05:00 UTC, the instant the daily tasks are dealt) — instead of the rolling 24-hour window removed in #2089 (#2216).
• The reset boundary is pinned by tests on both sides, 1 ms apart: an exploration logged at exactly 05:00 UTC counts against the day that opens there, one a millisecond earlier does not (#2216).
• Internal dedups, no behavior change: the day boundary is computed in one shared place instead of three spellings, and the window tally is a single count (#2216).

1.88.3 — 2026-09-03

Fixed

• A deploy restart can no longer erase an AI empire's memory (#2309). The agents' memory files were rewritten in place, so a process kill mid-write — deploys restart the service daily — could wipe an agent's grudges, sympathies, past defeats and its record of what it had already seen, all at once. Writes now land atomically: the old memory or the new one, never a torn half.
• AI empires no longer thank guildmates for ancient favours (#2309). After such a memory loss, the recovery path re-read weeks of old guild-aid events and marked them as thank-yous still owed — an agent would greet the morning by thanking someone for aid from three weeks ago. Aid older than 48 hours at the time it is digested now settles as plain history.
• One favour, one sentence (#2309). A guildmate pushing the same construction timer twice produced two identical "put a hand on one of your timers" clauses in the agent's view of events — a stutter no human would type. Repeated timer-push clauses now collapse into one, while different helpers keep their own lines — and deliveries that carry amounts are never collapsed, so the agent can still account for everything that arrived.

Also

• Internal hardening from the same review basket (#2309): save failures are now reported to callers instead of being silently swallowed (two error handlers that could never fire are gone), a misleading comment about the thank-you backlog bound now states the real arithmetic, the deliberate "an out-of-hours message may mention the favour without settling it" behavior is documented with the evidence that would change it, and the test suite pins the migration default, the ordering of the thank-you line, and the memory FIFO bound.

1.88.2 — 2026-09-03

Fixed

• Building sheet: Upgrade and Move no longer hide below the fold (#1985). On a tall building the sheet's action buttons were the last thing in the scroll — under the info rows, the "Next level" block and the level-by-level table — so you scrolled down to act, every time, on the most-opened panel in the game. The action bar now sticks to the bottom edge of the sheet, reachable at any scroll position, on the phone sheet and the desktop panel alike: it floats over the content on its own ground while you scroll and settles into its old place at the end, so nothing is ever covered for good. The dead-level warning travels with the Upgrade button it explains; at max level, where nothing is on sale, it scrolls with the sheet. Short sheets look as before. Suggested by SpartanEcho in #suggestions.

1.88.1 — 2026-09-03

Fixed

• Galaxy list mode explains itself (#2248): switching the Galaxy to list view used to show a blank area while a system loaded, and ⟡ anomaly badges with nothing on screen naming them — the map's legend and loading lines live in the spatial view, which list mode hides entirely. The list now shows its own "Scanning system…" row while the system loads, and a legend row naming the anomaly badge ("⟡ anomaly signal — only the ones your Astrometry can see"), in the same accent the map legend uses.
• The hover tooltip can no longer mark an occupied slot as an anomaly (#2248): every other galaxy surface — the scene, the planet sheet, the list — suppresses the ⟡ mark on a slot that holds a colony; the hover tooltip was the one surface left that would have printed "⟡ Anomaly signal" over a colony's name if the server ever sent both fields on one slot. It now stays as silent as the map.

Changed

• One glyph, everywhere (#2248): the ⟡ anomaly-signal mark was spelled separately on every surface that draws it (scene badge, hover tooltip, hex-map pips, signals board, activity feed, legend) — which is how a recent change nearly shipped the Dark Matter outcome icon in its place. It is now a single constant with a cross-surface test: nothing changes on screen, but the surfaces can no longer drift apart.

Also

• The test guard on the galaxy scene's comment stripper sat eight characters from tripping and blamed the wrong cause when it did (#2256): sessions adding ordinary comments to the scene hit a failure message about a regex stripper. The guard is re-based as an integrity backstop (floor 1/3, wider landmark coverage) and its failure message now names both possible causes — a swallowed code region vs. comment growth — with the remedy for each.

1.88.0 — 2026-09-03

New

• Target board: a "Sanctioned only" filter — everything hittable without earning a bounty (#2340, proposed by arcticpaladin). The board's "Far below" band used to lump the priced hits together with the weak-but-free sub-cases, so finding a free target meant cross-referencing the bounty rules row by row. One tap now keeps exactly the bounty-free marks: in band, stronger, already wanted, silent, or a target you hold a free reply against — minus guild allies and everything else you cannot attack at all. The filter never lies by omission: it prints how many rows it holds back, names the unreadable Gates apart (no Gate reading means "cannot be proven free", not "free"), and its caption says the verdict is for a solo strike on today's readings — a wanted head must still be wanted when you land. It composes with the band chips and the custom Gate range (they narrow within the sanctioned set), survives the launch round-trip like every other control on the board, and resets on logout with them.

1.87.10 — 2026-09-03

Fixed

• Spending Dark Matter or a Chrono-Shard now survives a busy server (#2375). During deploy-window stalls, buying a Dark Matter apex upgrade or spending a Chrono-Shard could fail after two seconds with a bare "Internal server error" — on a spend the player had already decided, and for a speedup, at exactly the moment it was worth the most. Both spend transactions now wait out a stall of up to 5 seconds and get 15 to commit, like the rest of the game's writes. In the rare case the server is still too busy to even start the spend, the game now says "State updated by another operation: try again." — nothing is deducted, and trying again is safe.

1.87.9 — 2026-09-03

Fixed

• The fleet advisor's Particle Accelerator warning now says why it matters (#2058). It used to read "The Particle Accelerator costs nothing to switch off and has no cooldown: the power it frees strengthens this target's defences" — a statement of mechanics with no stake, which players read as irrelevant noise. Under exactly the same trigger (it still fires only when the forecast prices a real, currently-weakening power deficit on a player-owned target with factor-scaled defences standing and the Accelerator on), the warning now reads: "This forecast prices the target's defences as weakened by a power deficit. The defender can switch the Particle Accelerator off at any time — free, instant, no cooldown — cutting the power it draws and strengthening those defences before your fleet arrives, so this forecast may be optimistic." It also fixes a physics slip: the Accelerator draws power (it produces Dark Matter), so the flip frees the power it stops consuming — the old line read as a generator going idle. The copy deliberately speaks from the forecast rather than stating the target's present state as fact, because below spy tier 2 the deficit is modelled on your own tech tree.

1.87.8 — 2026-09-03

Changed

• The missions screen's saved-fleet row now runs on the same component as the raid panel's (#2338, PR #2373). Nothing changes on screen. Since the raid panel gained the saved-fleet picker (v1.87 line, #2017), the missions screen kept rendering its own inline copy of the very same row — recall, save, delete, the clamp to the hangar, the three wire states — identical only by discipline. The blocker that kept it that way (#1990's concurrent work in the same file) shipped, so the missions screen now mounts the one shared row and the duplicate is deleted. Before the delete the two copies were verified expression-for-expression identical, and the missions wiring test passed unmodified through the swap: a fix to the row now lands on both surfaces at once, and the two can never drift apart again.

1.87.7 — 2026-09-03

Fixed

• Post-surge stock over the calm cap is no longer told it is "lost" (#2322). Once a Requisition Surge expired with stock over the normal ceiling, the warehouse tooltip still recited "Anything over the limit is lost — production, and what a fleet brings home" — the exact panic keep-frozen exists to prevent. The tooltip and phone sheet now tell the truth of that state: stock above the limit is kept and production pauses until you spend back under, while a returning fleet's over-limit haul really is still lost, and the sentence says so. Exactly at the cap, where nothing is held over the limit, the loss warning stays.
• The surge warning reaches the eye, not just the screen reader (#2322). Mid-surge, a stock over your normal ceiling showed plain colors while the accessible label already warned "Full when the surge ends". The tile line now goes gold with "full at surge end", driven by the same estimate the screen reader hears, so the two can never disagree.
• Capped is red now, not gold (#2323). A capped warehouse — where production is genuinely being wasted every hour — carried the same gold as "almost full", where nothing is lost yet. The game's alarm colors escalate gold to red everywhere else, and now the storage tile, its tooltip border, the phone sheet and the phone pill do too. On the phone, almost-full and capped used to be pixel-identical pills; the red is the difference.
• The production rate no longer vanishes at the cap (#2323). At the cap every surface dropped the "+X/h" figure — exactly the number that says how much is being wasted per hour. The tile keeps the rate beside "capped", and the tooltip and phone sheet spell it out: "Losing X per hour" at the cap, "Production paused: X per hour" for kept-frozen stock above it.

1.87.6 — 2026-09-03

Fixed

• The ⚡ energy figure is readable again on every build surface (#2354). The soft red that prints a consumer's energy draw — the most repeated number on the colony list and the advisor cards, since 27 of 30 buildings are consumers — measured about 3.5:1 against those dark panels, below the WCAG AA legibility floor (4.5:1) for text that small, and noticeably murkier than every other figure on the row. It is now a paler, solid red that measures 7.1:1 on its worst backdrop, on all three surfaces at once (Base build menu included). It stays visibly softer than the full "not enough resources" red beside it, so the two signals still cannot be confused — and a new test measures the composited contrast so no future tweak can quietly sink it below the floor again.

1.87.5 — 2026-09-03

Fixed

• The end-of-season close can no longer be killed halfway by a database stall (#2363). The Eschaton close — the sweep that carries every event-sector fleet home, the podium crates and Dark Matter, the winner's Destroyer — ran its four database transactions on a 5-second default deadline. A stall at exactly the wrong moment (the kind a deploy window produces) could cut the close mid-run and leave the season stuck "closing": some colonies swept and some not, prizes ungranted, the winner's hull unminted, until an operator intervened by hand. All four transactions now wait out a stall the same way every other resilient job in the game does, so the close finishes on its own. Players see nothing new — that is the point: the once-per-season moment where everything is paid out no longer has a silent way to half-happen.

Also

• The two spend paths that showed the same stall symptom in the field (Dark Matter upgrades, Crono speedups) are deliberately not touched here — they retry at the player's hand rather than in a queue — and are tracked separately in #2375.

1.87.4 — 2026-09-02

Fixed

• Fleet arrivals survive database stalls instead of stranding fleets in transit (#2359, PR #2366). The worker job that lands a fleet at its destination died twice in the 2026-09-02 production log on the database client's 5-second default transaction budget. A fleet whose arrival job burned all five queue retries during such a stall would have hung "in transit" forever — the arrival-side twin of the stuck returning fleets v1.71.7 fixed. Production got lucky this time: a later retry succeeded, and a read-only sweep found zero currently stranded fleets. All five transactions the arrival job can run — the shared deposit path for every mission type, the redeployment merge, the garrison stationing, the picket stationing, and the pursuit engagement — now carry the repo's explicit budget (wait up to 5 s for a connection, run up to 15 s), so an arrival that meets a stall waits it out and lands the fleet. The fifth site was found by this ship's own review: the pursuit engagement's expiry was the only silent one — it would have voided a real interception as "arrived at empty space", with no retry and no battle report — latent today behind the interception kill switch, armed the day it flips on. Remaining default-budget worker transactions (season close, and nothing else) are filed as #2363.

1.87.3 — 2026-09-02

Fixed

• Accelerator toggled off: the build menus stop quoting a ⚡ draw that is not being billed (#2355). With the Particle Accelerator switched off — exactly the state the energy-remedy hint steers throttled colonies into — the Base tab build menu and the Colony tab list still showed the next level's nominal draw (e.g. ⚡ −222) in the same soft red as a live draw, even though queueing that upgrade moves the billed balance by exactly 0 for as long as the toggle stays off. The tag now dims to a neutral grey while the Accelerator is off, and its tooltip carries the Building Sheet's own note: "off: NOT counted in the current balance (energy freed)". The number itself stays visible — it is still what the next level will draw once switched on — and with the Accelerator on (or on any other building) everything renders exactly as before.

1.87.2 — 2026-09-02

Fixed

• Reading guild chat in the Guild panel finally counts as reading it (#2352). The guild chat embedded behind the Guild panel's Council door never marked the channel as read: the Chat tab's guild dot and the chat dock's badge stayed lit over messages the player had already read there, and since v1.87.0 the "Unread messages" divider was drawn over those same read messages — and steered the view to them. Now opening the Council door marks the channel read with the same discipline as the Chat tab and the dock (on view, every 20 seconds while the door stays open, and on leaving — never while the browser tab is hidden), and the embedded panel gains the divider itself: a gold "Unread messages" line at the point the player had actually stopped reading, on both the desktop workspace and the phone sheet. The line is absent when nothing is unread or no boundary is stored, is never moved by the marking or by messages arriving under the player's eyes, and does not come back after closing and reopening the door.

1.87.1 — 2026-09-02

Fixed

• Event relics no longer show a build price in the Shipyard Specs sheet (#2345). Selecting the Destroyer or the Annihilator showed a full per-unit price — "Each 18,000 Fe · 9,000 Xe · 1,400 Dt" — with a "you have …" wallet comparison beside it, on the desktop console body and the phone Specs sheet, for a hull the shipyard refuses to build. Since v1.78.0 the Ship Reference deliberately withholds the relic price for exactly this reason, so the two surfaces contradicted each other. The price row now says "Event Relic — not buildable" instead — the same words as the grid card, the All-hulls ledger, the Ship Reference and the build footer — and the wallet comparison goes with the price. Buildable hulls are unchanged.

1.87.0 — 2026-09-02

New

• Chat now marks where you left off (#1623). Opening a channel with a backlog used to dump you at the bottom of a wall of messages, with nothing to say where you had stopped reading — you re-read or you missed things. Now a gold "Unread messages" line sits right before the first message you haven't read, and the channel opens on that line instead of at the bottom, so the catch-up starts exactly where you left it. The "↓ N new messages" pill still works for anything arriving while you read. Works in the Chat tab and the mobile chat dock, on the galactic and guild channels; whispers keep their own per-thread unread and get no line. If everything is read — or it's your first visit — nothing changes: the channel opens at the bottom as always. Suggested by Spaccamazza.

1.86.0 — 2026-09-02

New

• The target board is a workspace now: it keeps your place, hunts by your own Gate range, and can drop the pirates (#1733) — suggested by Jsnam. Before, the board forgot everything each time a tapped row took you to the launch screen: filters reset, sort back to Nearest, scroll back to the top — "realistically you have to write the coords down first". Now the whole view survives the round-trip for the session: band chips, sort, the expanded list, both hide toggles, the typed range and the scroll position all come back exactly as you left them (and a logout clears them for the next account). Two new controls: a Gate range — type MIN and/or MAX and the board filters by the raw Gate printed on the rows, replacing the band chips while set (they dim, and a caption counts what the range holds back, unscanned empires named apart); and a Hide pirates & NPCs toggle that clears the Beacon, bastion and relic rows out of the table, counting what it hides. A crossed or malformed bound is named on the spot ("MIN is above MAX", "numbers only") and filters nothing — the board never answers a typo with an empty galaxy.

Also

• The "Sanctioned targets" composite filter proposed on the same issue (everything hittable without earning a bounty) is not in this release — it is filed as #2340.

1.85.0 — 2026-09-02

New

• Your folding colony now shows in Missions (#1990). After arming a Displacement charge there was no way to see the fold's info where you'd look for it: the map marked the colony as folding, but the Missions tab — your "things in flight" — showed nothing, and the only written receipt sat inside the guild treasury panel. Now a gold Colony fold card sits at the top of Active missions: where the colony is going, and a live countdown to the landing; the phone's In flight badge counts it, and "No missions underway." no longer shows while a fold is in the air. The card also says plainly what the missing button means: a fold has no hand cancel — if your siege bar is charged when the countdown ends it is cancelled and the Scrip and the weekly slot come back, and a defence won before then clears the bar so you leave as planned. If a poll hiccups, the card stays on the last known fold instead of vanishing. Reported by moralezz01.

1.84.0 — 2026-09-02

New

• The build menu now shows what the next level will draw (⚡) (#2059) — requested by tartarusmkii. Before, the colony tab's building list and the advisor's suggestion cards priced an upgrade in resources alone: to learn what it would do to your power grid you had to open the Building Sheet, or cross to the base view's build menu, the one surface that already said it. Now every colony-list row shows the upgrade's ⚡ figure beside its cost chips, and every advisor card shows the same figure beside its payback — green ⚡ +N for a producer, soft red ⚡ −N for a consumer, and it is the exact delta the engine will bill: planet relief, the Grid line and Cold Fusion included, the same number the base menu has shown since the energy rework. Plan your grid before you commit.

1.83.0 — 2026-09-02

New

• Raid joins load your saved fleets in one tap (#2017): the guild-raid panel used to make you compose every fleet ship by ship — worst exactly when it mattered, scrambling a defensive garrison against an incoming-attack timer. The raid composer now carries the same Template row as the missions screen, on all three doors: open a raid, join one, and the Send-defence garrison form. Recall a saved fleet (clamped to what the launch colony actually holds in hangar, with the same "Template adjusted to the ships currently available." notice when something was cut), adjust it, send — and save a composition from the raid panel too. Every join rule stays exactly as it was: the template only fills the form. Suggested by arcticpaladin in #suggestions.

1.82.0 — 2026-09-02

New

• Your spy reports pierce the Veil everywhere, however old — the map remembers what you scouted (#2258). Espionage has always pierced the Eschaton Veil permanently, but the Galaxy planet sheet and the mission composer read your reports out of the events feed, which only serves the newest rows: on a busy feed, older scouted coordinates silently lost the pierce, and both screens called a scouted veiled colony a «free planet» the server was still going to refuse. The client now asks the server directly for your own newest report at the coordinate you open — whole history, and only yours: nothing can reveal, or even imply, another player's scouting. Reports written before the ship-id rename are served with their fleets normalized, so the Fleet Advisor reads them correctly instead of seeing an undefended colony.

Also

• The planet sheet now dates the spy report it renders («3d ago»), on the same clock the composer has always used (#2258).
• On a scouted veiled colony, the Harvest button's tooltip no longer promises the fleet «returns empty-handed otherwise» beside the hint saying the server will refuse — the two sentences now tell one story (#2258).
• A spy report of a Pirate Beacon over a slot the Galaxy calls empty now says the true thing: a live Beacon is never veiled, so it has retired since — with the honest caveat that somebody may have settled the freed slot under a Veil since the probe (#2258).

1.81.0 — 2026-09-02

New

• The Refit desk now shows your class's true combined bonus (#1788). Each class row carries a Summary card next to the Frame/Battery/Core slots: the net Attack, Hull and Shield change with every installed module applied, taken straight from the same numbers combat fights with. Before, each card showed only its own percentages and nothing said how they combine — with +8% and +5% attack installed you could not tell +13% from +13.4%. They compound multiplicatively (1.08 × 1.05 = 1.134), so the card reads Attack +13.4%, with your installed modules listed beneath and a one-line note saying exactly that. Net figures are rounded to one decimal; a drawback-heavy loadout shows an honest negative. Support lines show only Hull and Shield — nothing a support hull can install moves attack — and a class with no standing orders reads "No standing orders — all stats +0%." Suggested by KingPaladin.

1.80.0 — 2026-09-02

New

The Wraith faction defends itself (#1916). The thirty hostile empires used to be the safest farm in the game: their outlying colonies had no buildings, their vaults sat full, and nothing answered an attack that was already on its way. One player took more than a billion resources off them in eighty-odd raids, and every one of those raids was won before it launched.

They now play the game the rest of you play. Their satellites get built up and garrisoned instead of standing empty. A full vault is spent on hulls rather than left as a prize, so a successful raid collects a fraction of what it used to. Their research follows a line toward capital ships instead of wandering. And when a fleet is inbound on one of their colonies, the nearest guildmate who can actually get there in time flies a wall to meet you — expect to arrive against defenders who were not there when you launched.

Two things have not changed, deliberately. They still never attack on their own initiative; every raid the faction launches is ordered by a person. And their five unbannered scouts remain outside the guild, which means nobody can reinforce them — if you are hunting Wraiths, those are still the soft targets.

Fixed

• A defending wall that could reach a colony minutes before the battle was refused anyway, because the arrival calculation charged the same delay twice. Over one observed session that was thirteen attacks answered by nothing, with idle fleets a minute away. Reinforcements now fly when they can actually land.
• The faction offered walls it had already sent elsewhere in the same tick, so the colony under attack ended up with no defender at all rather than a smaller one.

1.79.0 — 2026-09-02

New

• Proving Ground: proving yourself now pays, not only winning your bracket (#2329). Until now the close paid exactly one empire per bracket, and your bracket is set by the size of your fleet. Growing past a rung meant leaving the players you were beating and competing against strictly bigger fleets, so your expected prize went from a crate to nothing. That is the opposite of what a season ladder should reward. From this season's close, the best strike of each bracket still takes its full crate, and every other ranked strike that lands at least 25% of that bracket's best is proven and takes the crate of the bracket below: second in Fleet with a real strike takes Squadron's 3 × 500K, second in Armada takes Fleet's 3 × 1M. A strike well under the bar is still paid nothing. Patrol has no bracket below it, so its runners-up stay unpaid, and the prize table says so on the row. Warhost and Apex pay the same 5M crate as Armada, so a proven strike there takes Fleet's 1M rather than the winner's own crate.
• The prize table shows both figures per rung ("3 × 1M crates" and "proven: 3 × 500K"), defines the word, and states the rule with the percentage the close actually pays on.
• The Activity diary records a proven payout as "Proved in Fleet", with the crate actually paid and why it is smaller than the winner's, instead of calling it the best strike of a bracket you did not win. On a phone app that has not updated yet, the new entry shows as a generic activity row rather than a false win.

1.78.0 — 2026-09-02

New

• Locked hulls now show their build cost and stats on the Ship Reference (#1992). Before, a hull gated behind research or Dark Matter told you where it unlocks and what it counters — but not what it would cost to build, nor its numbers, so you paid for unlocks blind ("I definitely wouldn't have made any effort to unlock the Colonial Ark knowing … the cost" — Jsnam). Now every row of the Ship Reference carries a spec strip: the build price (Fe/Xe/Dt, in the shipyard card's own "Each …" vocabulary) and the five core stats — Attack, Hull, Shields, Speed, Cargo — for locked and unlocked hulls alike. Event relics show their numbers and no price: the yard doesn't sell them. Nothing about the lock changed: the "Unlocked by …" line stays, and a locked hull still cannot be ordered. Requested by Spaccamazza, seconded by Jsnam.

1.77.0 — 2026-09-02

New

• The HUD now warns you when your labs sit idle (#1944, asked for by TartarusMkII). The top strip has always carried queue chips for the colony build queue and the shipyard — but not for research, the queue that costs the most to forget: an idle lane produces nothing, forever, and the only way to notice was to open the Research tab. A third chip, Labs, now appears in the same strip exactly while a research lane is idle, showing how many lanes are busy (say 1/2); tap it and you land on the Research tab, where the per-lane bench has the detail. All lanes running — or no Laboratory built yet — means no chip: its presence alone is the nudge. The lane count comes from the server, so it honours your whole Laboratory network, your guild's Science keystone and the Forge doctrine's third lane — the chip and the Research tab can never disagree about a free lane. Works on the phone strip like its two siblings: icon and counter, label dropped.

1.76.6 — 2026-09-02

Fixed

• The Missions launch screen no longer second-guesses your research while it loads (#2294). Opening the Missions tab could flash — and after a failed fetch, hold for up to 30 seconds — a false «Colonization tech 0 is not enough» badge that greyed the Launch button, even though your real Colonization level was plenty. The badge now reads the research level your empire actually has, which is already on board before the screen opens, and a network hiccup keeps the last known answer instead of pretending you never researched anything.
• Fresh ships count on any launch pad the moment they leave the yard (#2278). A Colony Ship delivered mid-window on a launch pad you had hand-picked left the colonize badge grey (and Launch blocked) until the next 30-second refresh, while the same delivery on your mother colony or on the colony you were viewing counted instantly. Every pad now counts deliveries the moment they happen — and a rare double-counting of the same colony (when your hand-picked pad is promoted to mother) is closed with it.

1.76.5 — 2026-09-02

Fixed

• Guild Help board: the push counter now speaks for the request, not for you (#2011). "X of 5 pushes left" beside a Give a push button read as a personal allowance — a covered request showing "0 of 5 pushes left" looked like an empty pocket, and a button still blue after your own push looked like a bug. Both button states were correct all along: pushes are shared across the whole guild, five per request, with no per-person limit. An open request now reads "1 of 5 pushes given · takes 4 more", a covered or expired one leads with what was given instead of a zero, and the live button says — beside it, and as its desktop tooltip — "Pushes are shared across the whole guild — you can give more than one." The ask side counts pushes now too, instead of promising five different guildmates.

1.76.4 — 2026-09-02

Fixed

• Market: the affordability filter now says "Affordable" (#1950). The chip used to read "Within reach", which in a space game sounds like fleet range — new players filed it under distance, not budget (reported by TartarusMkII in #universe-chat). It now says plainly what it filters: offers whose asking price you can currently pay. The filter itself is unchanged.

1.76.3 — 2026-09-02

Fixed

• Storage warning says what it means: almost full vs capped (#2007). A resource tile at 99% — 89,537 / 90,177, with room to spare — read "at the limit", and the +X/h production rate vanished behind the words: players read it as "capped, production wasted" while nothing was being lost yet. Now the tile keeps the rate when the store is nearly full (≥98%) and marks the capacity number gold with "almost full" beside it; "capped" appears only when the store is actually at its cap and production really stops banking. The phone pill keeps its gold warning on both sides of the cap, and the hover/sheet sentences ("Full in 5h 42m", "Warehouse full") already told the two states apart — tile and tooltip now agree.

1.76.2 — 2026-09-02

Fixed

• The launch rail no longer leaves a timer ticking after the Missions tab closes (#2229). The "Launch again" success paint armed a 4-second timer it never tracked. In play the late tick was a harmless no-op; in the test environment it fired after teardown, threw window is not defined, and could red a whole gate run with every test green — it cost the #2214 ship a full re-run. The timer is now tracked per order, cleared on unmount, and never armed once the tab is gone (the arm sits after an await, so tracking alone was not enough — this ship's own first gate run proved that). Two new tests pin both leak shapes, each watched red on the code it pins.
• The discord-announce test suite no longer burns seconds of pure CPU per run (#2296). The file re-evaluated its whole module graph for every service it built — 1.3 seconds for the first test on an idle host — which under gate-host load blew the 5-second per-test budget and redded three otherwise-green gates on 2026-09-02 alone. The webhook URL is read at construction, not module load, so the suite now imports once and constructs per test: worst test 1348ms → 7ms. No timeout was raised, and nothing about the Discord announcement behavior changed.

Nothing in this ship changes what a player sees in the game: both fixes are gate hygiene — false reds under host load, each previously green on re-run of the same commit. What is proven is that both flake mechanisms are gone; only the next loaded week can confirm the reds stay away. Test-hardening follow-ups are filed as #2314.

1.76.1 — 2026-09-02

Fixed

• Guildmates who cover an aid request while you sleep get their thank-you again (#2243). Since the Help board started dropping covered requests three hours after they end (#2071), a synthetic empire that slept through that whole grace — and woke to a busy night of battles — knew who had helped it but never said what they did: the favour had been pushed out of the short list of recent events its chat keeps in mind. The memory of the aid now stays in view until the empire has had it in hand once during its waking hours, then rests — thanked once, like a person would, never repeated forever. Favours already weeks old at deploy stay quiet; only new aid earns the sticky thank-you.

1.76.0 — 2026-09-02

New

• Allies under attack: the enemy in a red frame, the defence in a green one, both hull by hull (#2297). On the guild's Raids panel, an attack on a guildmate used to be three grey lines of run-on prose, and a rescuer skimming the row could not tell the enemy's ships from the defence. The row is now two framed boxes, side by side on a desktop and stacked on a phone. Incoming (red) leads with the ship count, then the origin and attacker names, then the hulls as chips. Defending (green) shows the total on station now, the ally's own statics and fielded hulls as chips, and every guildmate's garrison on station with its own hull chips, so twenty Freighters no longer read like twenty Battleships. Both boxes list hulls in the same order, so the fleet coming and the fleet standing against it compare line for line.
• The Observatory gate is named instead of hidden (#2297). When the ally's Observatory reads an attacker but not the hulls, the red box used to show names with no ships, which read as a missing feature. A gold note now says at which detail step the reading stops and that the hulls are listed only at step 8. On a joint raid read to two different steps, it also says the chips list only the fleets read that far.

Also

• A garrison flown by a pact partner from another guild keeps the name and the total it already showed; its hulls stay its own.

1.75.7 — 2026-09-02

Fixed

• The chrono-shard spend decision is now held to the right empire (#1951). Nothing visible changed today — the shipped behavior was already correct. What closes is a door: no test exercised the code that decides which empire's shard inventory gets read, quoted as that agent's own fact in chat, and spent, so a future edit could have made a synthetic empire state another empire's shard count as its own — or burn a shard out of the wrong empire's pocket onto the wrong build queue — with every suite green. Seven measured mutations of that decision (wrong inventory token, wrong runtime for the count, wrong persona at the wheel, wrong research lane, wrong spend addressee, wrong name on two operator log lines) each now turn tests red on a two-empire bench.

1.75.6 — 2026-09-02

Fixed

• A silently lost signup-country feed now announces itself in the api log (#2272). Nothing changes for players. Since #2260 every new empire records the country it signed up from, resolved by the edge server — machinery that lives entirely outside the game's own code, so if it broke, every new signup quietly recorded nothing and the loss would surface weeks later in an analytics query. Now the api log warns within a day: after ten different clients in a row have come through the public edge with no country attached and not one resolved between them, a warning names what is broken and where to look, keeps repeating while the loss lasts, and a recovery line marks the fix taking hold. The alarm counts distinct clients on purpose — one player retrying on an address the geo database does not know, or a replayed sign-in, can never fake an outage — and the doors that legitimately carry no country (the internal door every AI empire uses, requests the edge genuinely could not resolve) never trip it. Signups themselves are untouched: the watchdog is pinned by test to never be able to fail one, and its threshold is a host lever (COUNTRY_SENTINEL_STREAK) wired into both compose files.

1.75.5 — 2026-09-02

Fixed

• Missions verb badges: an absent key no longer fabricates a refusal the server never issues (#2274). The Missions tab's colonize badge — the sentence printed on the verb card, the objective list and the phone's sticky CTA — read a missing Colonization entry in the research payload, or a missing Command Center entry on the mother's row, as level 0, and dimmed the verb with a "tech 0 is not enough" line over a founding the server accepts. An absent value now travels as unknown: each gate decides only on a level actually held, the badge never prints an invented number, and the server keeps the last word. Latent hardening, same class as the galaxy Colonize button fix in v1.75.1 — every live empire's payloads carry both keys today; the hole would have opened silently, e.g. with a discipline added later than an existing empire. The launch composer's own preflight deliberately keeps its stricter check one screen later, and the harvest badge's Recycler check is unchanged on purpose: there the server itself refuses an absent value, and the badge transcribes the server.

1.75.4 — 2026-09-02

Fixed

• Replay weapon sounds land on the hit at every speed (#2187). Watching a battle replay at 2× or 4×, weapon sounds kept running at real-world speed while the picture ran at replay speed, so any sound with internal structure missed its moment: the Destroyer's bass drop arrived seconds after the explosion at 4×, and the Annihilator's detonation sounded as the lance was already closing. Weapon sounds now play at the replay's speed — the charge builds faster, pitches up, and the bang lands on the impact frame at 1×, 2× and 4× alike, on phones too. Stingers, interface blips, impacts and the music keep their normal pace.

1.75.3 — 2026-09-02

Fixed

• The resource fill-up estimate stays put while it counts (#2056). Hovering a top-bar resource on desktop showed the «Full in Xh Ym» box — and on Chrome it glitched out of existence every second, torn down by the browser each time the live figure ticked. The estimate now rides a tooltip of the game's own: it stays open under the pointer while the exact stock counts up inside it, and it leaves when the mouse does. Firefox players never saw the flicker; nothing changes for them but the styling. The phone's tap sheet is untouched. Reported by spaccamazza and arcticpaladin.

1.75.2 — 2026-09-02

Fixed

• Missions tab harvest badge no longer lags a Recycler that just finished (#2265). The badge read the building off the 30-second colony poll, so for up to ~30s after your home colony finished building a Recycler it still said "No Recycler on any colony" and kept Launch grey — refusing a harvest the server would have accepted. The badge now credits the Recycler the moment the build finishes on your home colony, the colony you are viewing, or your selected launch pad.
• Verb badges now credit a ship delivered on the colony you are viewing, even with a hand-picked launch pad (#2266). If you had picked a launch pad earlier and then viewed an outpost whose Shipyard just delivered your only Colony Ship, the colonize badge blocked Launch for up to ~30s — while the Base tab drew the ship and the galaxy sheet's Colonize button was green. The colony you are viewing is now resolved live like your home colony. Colonies you are not viewing, mothered by, or launching from still update on the regular poll — the same tradeoff as before.

1.75.1 — 2026-09-02

Fixed

• Colonize button: an absent key can no longer grey a founding the server accepts (#2038). The galaxy planet sheet's Colonize button fed its two colony-cap gates through defensive zero fallbacks: a research payload arriving without the Colonization key, or a mother colony row without a Command Center entry, would have read as "level 0" and greyed the button over a founding the server permits — the server's own rule treats a missing value as no grounds to refuse. Both feeds now carry a missing value as "unknown": an unknown level never greys the button, the server keeps the last word, and the requirement row prints no invented number while a level is unknown. No player can hit this today — every live empire carries both keys — so nothing changes on screen; the door it closes was the kind that opens silently, e.g. a discipline added later than an existing empire. The engine's rulebook comment now names all three client surfaces that transcribe these gates and which zero fallbacks remain deliberate; the one that is not is tracked as #2274.

1.75.0 — 2026-09-02

New

• The game now remembers where each new empire signed up from (#2260). Every account created through the public door carries the two-letter country code the edge resolved at that moment — written once at signup, never updated, and empty when the lookup does not know. Before, questions like "did the Gulf convert" or "day-2 EU vs US" were answered by intersecting rotating nginx logs with one-by-one whois lookups, and any cohort older than a day of log was lost forever; now signups, five-minute activation and day-2 return per country per cohort day are one SQL query. Synthetic empires, the internal dev door and NPCs always stamp nothing, historical accounts stay unstamped, and only the two letters are ever stored — never the address itself. The privacy page says exactly that, dated today.

1.74.5 — 2026-09-02

Fixed

• Ship machinery: the production deploy now refuses to run when the tree to ship carries a database migration (deploy/*.sql) that production has not seen, until the operator explicitly confirms the migration ran (--migrations-done, valid for that one invocation only) (#2043). Nothing changes on screen today — this closes the gap where a new build could boot against an unmigrated database, which the Event Sector migration avoided only because the operator ran the SQL by hand. The refusal names the pending files and points at the deploy runbook, which now carries the full reviewed order: safety dump, writers stopped, SQL applied all-or-nothing from the shipped tree, then the deploy — and how to bring the writers back if any step fails.

1.74.4 — 2026-09-02

Fixed

• The Event Sector garrison is now sized on real fleets, like sorties and bastions before it (#2145). The six Verge Watch garrisons were priced on the field's power gates — a buildings-dominated number — so each control planet held ~682,000 hulls against a mid-field whose fleets could not answer it even all together: the guild event was, in practice, open only to the single strongest player. The garrison now reads the same skipping mean (skip the runaway, average ranks 2–6) over fleet strength, calibrated in the battle engine so a strong mid-field fleet wins a costly assault, a counter-built one wins cleanly, and a small guild party can take a world together — with a floor so a young server never meets a shipless garrison. Live garrisons resize automatically on the first rescale pass after this release. The stale docblocks and design pages still describing the old gate-sized world (the bastion prize rationale, the raid and bounty design docs) now tell the truth about the fleet-basis era.

1.74.3 — 2026-09-02

Fixed

• A Colony Ship fresh out of the Shipyard no longer greys Launch for 30 seconds (#2040). When the yard delivered your empire's only Colony Ship, the galaxy sheet's Colonize button turned green immediately — but the Missions tab could still claim "No Colony Ship or Colonial Ark in any hangar" and keep Launch disabled until the next data refresh, up to ~30 seconds later. The tab's readiness badges now read the same live hangars the rest of the screen reads (your mother colony's and your launch pad's), so the badge clears and Launch enables the moment the ship exists. The sibling badges heal the same way: a freshly delivered Spy Probe or freighter no longer leaves Espionage, Transport or Pickup marked as unavailable while every other screen shows the ship.

1.74.2 — 2026-09-02

Fixed

• The Deep Observatory's forecasts now show on the map and the list, where you pick targets (#2247). If you own the Observatory, every visible anomaly signal already told you its outcome — but only on the planet sheet and in the activity feed. On the galaxy map and in the system list, a T4 ambush and a T2 resource cache both read as bare tier badges, and the bigger tier looked like the better pick: the very surfaces you choose targets from steered you into the ambush. Now the outcome icon rides next to the tier on the map's slot label, on the list badge and in the hover tooltip — ⟡ T4 ⚔ for a revealed ambush, ⟡ T2 ✦ for Dark Matter — with the outcome spelled out in the badge's tooltip and screen-reader name ("Anomaly signal (T4) — Pirate ambush"). Without the Observatory nothing changes: you see exactly the badges you saw before.

1.74.1 — 2026-09-02

Fixed

• Galaxy planet sheet pierces the Veil with your own spy report (#2228). Probing a veiled colony filed a full spy report — and the Galaxy planet sheet still called that planet «No colony — free planet», offering Colonize and Harvest with no hint, while the mission composer one tap later named the occupant off the same report and warned the server would refuse. Now the sheet names the reported commander — with the AI badge and guild tag the report carries — says the colony is known from your spy report and may be gone since the probe, and warns that the server will refuse the founding and the harvest while the colony still stands under the Veil. An unscouted veiled colony keeps reading as a free planet: only your own espionage pierces the Veil.
• Recent-targets chips badge spied synthetic commanders (#2228). A spied AI empire's name on the Recent row printed bare; it now carries the same synthetic mark as every other surface that names it, with the distinct agent and Wraith glyphs.

1.74.0 — 2026-09-02

New

• The Contracts board is redrawn, and your Command Center decides how big it is (#2218). Every contract is now a work-order slip: a painted emblem for the kind of order, seal pips for your Charter grade, a progress gauge, and — for the first time — what it pays, in your own three resources, before you claim it.
• From three daily orders to six. The board grows with the mother colony's Command Center: 3 orders up to CC5, 4 from CC6, 5 from CC13, 6 from CC20. A band above the board draws all four grades at once, the one you are on lit and the ones ahead dimmed, each naming the level it opens at and the board it deals — so "a bigger Command Center is worth building" is a road with an end you can see, not a single unexplained next step.
• The orders get harder as your reach grows, not richer per card. Targets scale with the empire's throughput — nine ships instead of three, seven scouted empires instead of three — while the price of a single contract never changes. A full clear is about 13% of a day's production at the first grade and about 18% at the last: more because there is more to do, not because the same card started paying more. Orders whose supply the galaxy sets rather than you — a research start, a debris field, an anomaly, a new world, a high-grade Beacon — never ask for more than one.
• The Daily Wrap follows the hand: two thirds of your orders, rounded up. Two of three as before, four of six at the top.

Changed

• Twelve new pieces of art, one for each kind of order plus the Weekly Chest.
• The Weekly Chest keeps three slots at every grade. There are four weekly orders in the catalog and three seats at the table, and that margin is what keeps your week from being identical to everyone else's.
• Resource crates read as part of the Inventory sheet. They sat underneath it as their own block; they now take the same tile and the same grid as Refit parts and Chrono shards.

Fixed

• The qualification rule under an order — the one that explains why a finished-looking card was refused — is readable in full again on a phone, instead of being cut after three lines with the rest behind a hover that touch screens do not have.
• A collected order no longer shows a reward figure re-priced at your current production. What it would pay now is not what it paid, and the slip stays on the board as the day's record.
• The board's art is served so it can be replaced. It was being cached for a year at fixed names, which would have frozen every plate for anyone who had already opened the board.

1.73.0 — 2026-09-02

New

• A way out from the commander card (#2239). Tap your own portrait and the card that already let you link your account — or delete your empire — now also lets you simply log out. Before, the only two exits were a small log out chip at the far right of the desktop top bar and the bottom of the Menu sheet on a phone, so on the surface that reads as "my account" you could erase your empire but not sign out of it.

Changed

• Leaving now asks first, and says what it costs (#2239). The new exit opens a short confirmation instead of ending the session on one press. If your empire is still a guest empire, it tells you plainly what nobody could guess: it lives only on this device, logging out loses it for good, and your commander name stays taken. If you have linked your account, it tells you the opposite — sign back in and everything is exactly where you left it.
• The confirmation opens below the button you pressed, with Stay on the left, so a double tap cannot land on "Confirm log out" and end your session before you have read a word of it.

The two older exits — the desktop chip and the Menu sheet — still work exactly as before.

1.72.5 — 2026-09-02

Fixed

• Galaxy list view now marks anomaly signals (#2014). A planet carrying an anomaly signal was flagged in the galaxy map view — tier-colored marker, "Anomaly signal (T2)" tooltip, legend dot — but the list view showed that same planet with nothing at all, so a player scanning the list could not tell which slots carry signals (reported by Spaccamazza). The list's Planet column now shows a small tier-colored ⟡ badge with the tier label — the same signal glyph the map draws — and hovering it spells out "Anomaly signal (T2)" the same way the map does, with the same wording carried to screen readers. What you can see is unchanged: the list shows exactly the signals your Astrometry already reveals on the map, never more.

1.72.4 — 2026-09-02

Fixed

• Guild Help board keeps finished requests around for up to two days (#2071): an aid request that ended early — fully covered, withdrawn, or out of speedup pushes — stayed on the board for up to ~47 hours after it actually ended, because the board only looked at the request's original 24h expiry. Now a request that ends early leaves the board 3 hours after its real end: long enough for guildmates in other time zones to come back and see that somebody helped (the 'covered' / 'gone' label stays for that tail), short enough that the board is an honest picture of what still needs help. Requests that simply run out their 24 hours keep their unchanged tail.

1.72.3 — 2026-09-02

Fixed

• Faction sorties are sized on your fleet, not your empire (#2222). A reprisal from a felled bastion was priced on your gate — a number made mostly of buildings, research and Tree — and for a developed empire that meant an unanswerable wall: on 2026-09-02 The Black Foundries launched 42 billion hulls against a 1.87-million-hull home fleet, 22,545 to 1. A sortie is now priced on your actual fleet at 1.5×: stronger than what you left at home, and beatable — a garrison counter-built against the faction's known battle line repels it outright, a generic fleet that stayed home goes down fighting and takes 40% of the sortie with it, and whoever brought every hull to the siege pays the choice. The bigger you build, the bigger the reprisal — that link is gone; only the fleet you fly counts.
• A reprisal against a nearly fleetless target no longer departs at all: below the size that mints a single hull, the bastion holds its grudge and waits until you own something worth coming for, instead of burning its 36-hour cooldown on an empty fleet.

Also

• Repelling a sortie was practically impossible before this fix, so its rewards were too: expect the defence payout (Dark Matter and a faction item for a repelled reprisal) to start actually happening. The weekly caps already in place bound it, and a repelled sortie still never pays a Seal.

1.72.2 — 2026-09-02

Fixed

• The map no longer pins your CPU when nothing is happening (#2057). Leaving the map open — even in an unfocused window — used to redraw the identical picture sixty times a second: fans at max on laptops, batteries draining on phones. The map now sleeps whenever it is truly still (camera settled, no focus pulse fading, no fleets sliding across the chart) and wakes instantly on any touch, scroll, hover or data update. While something genuinely moves — a pan settling, fleet marks in transit — it animates exactly as before. Players who ask their system for reduced motion get the same saving with fleets frozen between updates, as they already were.
• Two mobile micro-freezes on the map are gone (#2057): the intermittent hitches felt mid-drag (steady allocation pressure in the fleet-mark redraw, now reusing its buffers) and the one-frame freeze when a pinch crossed a zoom-detail threshold (the heavy relabeling now waits until your fingers lift).

1.72.1 — 2026-09-02

Fixed

• End-of-war verdicts can no longer be silently starved by a slow moment (#2220). When a guild war ends, the game collects the tribute from the losers, pays the winners, and tells every member of both guilds what happened. That final step used to be written one member at a time inside a transaction running on a tight default time budget — at the two biggest guilds' sizes, hundreds of statements in a five-second window. A multi-second stall (the kind a deploy or a busy database delivers) could expire the transaction, quietly roll the whole verdict back, and make the next weekly tick redo all of it, over and over for as long as the stall lasted. Now the member notifications are written in one batch and both verdict transactions carry the same explicit 15-second budget the raid and league verdicts already use. Nothing visible changed: same tribute, same spoils, same event for every member — it just can't be lost to a stall any more. No player was ever traced to this failure; it was found preventively by #2182's review.

1.72.0 — 2026-09-02

New

The Shipyard becomes a bay, not a form (#2185). Last release rebuilt the Shipyard around one idea, the hull you are looking at is always named and the button that builds it is always under your thumb. This one gives that hull somewhere to stand.

A painted hangar. Every hull now sits on a lit landing pad inside a real hangar bay, rib lights overhead, the deck taking the colour of the ship's class. The old plate drew its floor and pedestal so faintly that on most screens the ship simply floated in black.

The Star Forge, where the ship is. Mastery is per hull, so the ladder now lives with the hull: its stars stay on the nameplate, and its steps, their price and the button that buys them sit right under the specs, in the console on desktop and in the Specs sheet on the phone. The separate Forge strip and the collapsed Forge section below the deck are gone, and so is the second list of twenty-three hulls they led to. Before your first Tier-9 capstone that section is one line saying so, and, once you hold a Forge item, your bank and where each item comes from under it, not a ladder you cannot climb. If the Forge cannot be read for a moment, it says so and offers a retry instead of sitting on "Reading the Forge…".

Desktop, filled in. Planetary Defence and the Dry Dock sit directly under the building queue instead of below the whole page, so scrolling past the queue no longer lands you on an empty screen. The recommended hulls are three tiles in a row, locked hulls state their research route on one line, and the header shows how much of the catalogue you have unlocked as a bar. The console reads data first: stat bars are now scaled against the ship's own class, and say so, so a light fighter's attack no longer draws as an empty bar against a dreadnought's, while a gold bar still means the best in the whole game; the price sits next to what you have; long counter lists fold to three with Show all; and the build controls take less of the column.

Phone, one row. The BUILD bar is a single row: steppers, the count, repeat, BUILD. It used to be four rows, a quarter of the screen, and on an iPhone SE it covered the ship's name. Tap the count to open a sheet, named after the hull, with the exact field, the ×5 / ×50 / Max presets, the order's total, time and Power, the first-fleet note and a BUILD of its own; the sheet covers the page rather than pushing it, so the hull stays where it was. When something blocks the order, a red line appears above the bar only then; the cannot be cancelled warning on a long order stays on the bar itself, in the same one-line slot, with the full advice on splitting the order in the sheet. The stage grows with the space the bar gave back, and the hull's name, its counters and the first building slip all clear the bar on the smallest supported phone.

All hulls, as a catalogue. The phone's full hull list now tells you what each hull does and costs, or how many you own, or which research unlocks it with a NEXT marker on the nearest one, with class dividers when sorted by class and the sort row staying put while you scroll.

Delivered. When an order finishes, its slip turns green and reads Delivered · Corvette ×50 for a moment before it leaves, instead of vanishing between two refreshes.

Fixed

• Stat bars in the Shipyard console were scaled against the whole catalogue, apex and event hulls included, so ordinary hulls read as having almost nothing. They are scaled inside their class now, and the console says which class.
• The Star Forge panel could re-read its data on every screen refresh when mounted under a live view; it reads once and retries only on failure.

1.71.9 — 2026-09-02

Fixed

• Emergency crate: a feint no longer counts as an emergency (#2041). The guild shop's Emergency crate is sold only to a player raided in the last 48 hours — but any defender-role battle opened that window, including one where a single throwaway hull was thrown at your colony and died for nothing. An out-of-guild accomplice could keep a buyer's window open on schedule with one cheap ship every two days. Battles are now stamped as token attacks when the attacking fleet is worth less than a tenth of the defense that met it and the defender lost nothing — no loot taken, no fleet destroyed — and such battles no longer open the crate window. A raid that actually takes something from you still counts as the emergency it is, however little defense stood in its way, so a sacked fortress-doctrine or emptied colony keeps its crate. Every battle fought before this change counts exactly as it did — no window closes retroactively.

1.71.8 — 2026-09-02

Fixed

• The attack setup no longer contradicts itself on a colony you scouted under the Eschaton Veil (#2193). Probing a veiled colony returns a full espionage report, but the setup screen kept reading the veiled Galaxy slot: the summary said "Free slot", the warnings promised the fleet would "fly there and come back empty" — while the fleet advisor, reading that same report, sized extra freighters for the expected loot. Now your own report pierces the Veil across the whole screen: the target panel names the reported commander (with their AI badge and guild tag, exactly as the report's card shows them), the compose line opens with them, the defender's energy figure returns, and the empty-flight warning is replaced by an honest hedge — the Veil may hide them, or they may be gone since the probe. Colonize and harvest now warn that the server will refuse while the scouted colony stands, and Defend no longer blocks a garrison the server accepts toward a scouted veiled friendly colony. A veiled colony you have not probed still reads as a free slot everywhere — only your own espionage pierces the Veil.

Also

• The Galaxy planet sheet still calls a scouted veiled colony a free planet — that surface has no espionage feed today and gets its own issue.

1.71.7 — 2026-09-02

Fixed

• A fleet flying home can no longer be stranded by a slow moment on the server (#2182). When a fleet landed during one of the multi-second stalls production sometimes delivers (deploy windows above all), the landing could hit the database transaction's default 5-second limit and fail — and after five silent retries the fleet stayed "returning" forever, ships and cargo in limbo. It happened twice in production, on 2026-09-01 and again on 2026-09-02. The landing now carries the same widened budget the raid resolver got in v1.65.2, on both of its paths — the ordinary return and the rescue deposit of a fleet coming home to an empire mid-exodus. The weekly league verdict, the last big transaction still on the default budget, got the same protection: a stall can no longer silently roll back a day's payouts and make the next tick re-pay them from scratch.

1.71.6 — 2026-09-02

Fixed

• The 'investigate an anomaly' daily is no longer un-completable (#2089). The daily task is dealt at 05:00 UTC, but the exploration cap counted a rolling 24 hours over each scan's own launch time — by construction the two could never be in phase, and a freshly dealt daily could sit behind a still-blocked cap for up to ~15 hours ("Daily exploration limit reached" on a task the game had just handed out). The rolling 24h window described in the v1.64.0 entry below is gone: the cap now counts explorations since the contract-day start, and the whole allowance comes back at 05:00 UTC together with the daily tasks, so a fresh daily is always actionable. Every screen says so — "N of M explorations left today", the greyed Explore button counts down to the reset ("they reset in ~2h"), and the limit banner names the instant: "explorations reset with the daily tasks (05:00 UTC)". Two things to know: apps installed before this release still carry the old "frees its slot 24h after you launched it" wording — the real reset is never later than what that banner promises, so at worst they wait longer than needed until the app updates; and scans spent just before 05:00 UTC now free up at 05:00 (that is what a calendar-day cap allows) — in exchange, no player is ever blocked at the moment their daily lands. The GET /missions/explore-window payload keeps its exact shape for installed apps.

1.71.5 — 2026-09-02

Fixed

Your English is no longer turned into French (#2201). Since v1.69.0 the chat translates messages into English on the way in. For an hour it could also do the opposite: on some messages the translator decided an English line was not English and rewrote it — into French, if French was what the line happened to mention — and that rewrite was shown as your own words. Two players caught it in the galaxy channel within minutes. A translation into English can now never come back less English than what you typed: if the reply has lost English words, or gained a sentence's worth of foreign ones, what you typed is what everyone reads. A very short line with nothing to measure — a two-word greeting — can still slip through on a bad roll; anything with a sentence in it cannot.

1.71.4 — 2026-09-02

Also

Nothing changes in the game this time. This release only makes the web test suite provide its own browser storage when the test environment fails to (#2203) — a difference between the machine that ships and a developer's laptop that turned forty-four tests red on one and green on the other. It touches no code that reaches a player.

1.71.3 — 2026-09-02

Fixed

• The tutorial no longer hides the base map on a phone (#2190). Mid-tutorial, the First Steps card and the convoy card stacked over the base and covered almost all of it, so a new commander was being told to build on a base they could not see. The First Steps card now has a chevron in its header: one tap folds it down to a single row that still shows your progress and your next step (or "Reward ready"). The smaller convoy card stays where it is, so a delivery waiting to be taken is never hidden. Tap the row to unfold. The game remembers your choice for the session. Desktop is unchanged.

Also

• The "While you were away" recap sheet had briefly gained a second close button next to its own after the last release. It is back to one.

1.71.2 — 2026-09-02

Changed

• Ship Stars are finally worth their ladder (#2152). Mastering a hull completely — thirty steps and five ignitions, a bill of well over a hundred billion resources, hundreds of faction items, fifteen Sovereign Seals and 7,750 Dark Matter even on the cheapest class — used to buy +30% attack and +15% hull and shield on that one hull type, less fleet-wide effect than three research nodes costing a fraction of it. Every step you already bought now pays double or better: a 5-star hull fights at ×1.60 attack, ×1.35 hull, ×1.35 shield (support hulls reach ×1.70 hull), and nothing about the prices, the ladder or your purchased steps changed — the same stars simply hit harder.
• Every ignition is now a milestone, not just a receipt: it sharpens that hull's own counters (#2152). Each star you complete adds +0.25 to the counter multipliers that hull already holds — five ignitions add +1.25, so a fully mastered bomber with the Doctrine of Escalation tears into battleships at ×4.75 where the chart says ×3. The bonus only amplifies matchups the hull already wins: it gives nothing against ships it does not counter, nothing against planetary defences, and your mastered hull stays exactly as vulnerable to whatever counters it. Counter-picking still beats raw mastery — a counter-picked fleet of equal value still repels a fully mastered attacker, and that is now measured and enforced by tests on real battles, not assumed.
• The Power leaderboard moves with this release. A hull's stars have always counted toward Power; now that each star is worth more, every empire holding star steps rises on the Power board at deploy, in proportion to what they actually mastered. Synthetic empires price your stars at the same new rates — they will respect a starred fleet exactly as much as the engine does.

1.71.1 — 2026-09-02

New

• Boss rematch: something to do at the wall (#2178). Every empire walls somewhere on the campaign ladder, and until now the tab was a loss button once you did: thirteen players sat at depth 49 for weeks with nothing to tap but the same defeat. The Campaign tab now has a Boss rematch desk. Once a week you can fight the deepest boss you have already beaten again, with today's strike group and the deployment saved on your engagement sheet. A win pays a share of that boss's parts, straight into the wallet you spend at the Refit desk: one of each kind on the first chapters, up to four of each deep into the wheel. Nothing else moves: no resources, no Dark Matter, no war points, and your depth stays exactly where it is. The ticket is spent on the attempt, win or lose, so the desk shows a forecast for the boss before you tap. The last rematch's outcome and a Watch button sit under the desk.

Changed

• The changelog page is on time. Until now the "Full notes" link in a release announcement could land on a page that did not yet list the release it announced: novarchy.com/changelog was baked into the game's web build, and the entry only appeared after a second rebuild a couple of minutes later. The page is now published directly the moment a release goes out, so the link shows the release it points at.
• The campaign resets only with the season, and the decision is written down (#2178). Two empires sat at depths 235 and 230, climbed before the immortal-flagship rule was closed; the next level fields some 57,000 hulls against a 19-hull roster. The funnel for everyone else is exactly the designed shape, so the ladder is not being reset for all and the curve is not being softened. What ships instead is the rule, in the design document and as a one-shot procedure: the campaign score resets with the season wipe and nowhere else, and your parts, your installed refits and your ship stars survive it. The two stranded empires are a separate, per-account decision.

Also

• A rematch in flight shows on the status strip as a rematch, never as a level being climbed, and a lost rematch never reads as the ladder walling at that level.
• Two new columns on the database, both additive with defaults; the rollback procedure is documented next to the others.
• Behind the scenes, this release also reworks how fixes reach production: one review round instead of four, the test gate paid once per commit, and no extra production rebuild per release, which also means fewer builds competing with the live server.
• This entry was amended after the release went out: the boss rematch (#2178) merged to main minutes before 1.71.1 was cut and its tree already contained it, while the entry as first written said nothing in the game had changed.

1.71.0 — 2026-09-02

New

The Player Echelons: everyone races their own tier, and everyone who plays wins something (#1193). Until today the Leaderboard told a new commander exactly one thing: you arrived late — one board for the whole server, ranked on total Power that never resets, first place a lifetime away. Now every ranked commander holds one of five Echelons — Ferrolite, Xenogas, Dark Matter, Starforged, Singularity — and races only the commanders of their own tier, on this cycle's growth alone. The score restarts every week. A brand-new player lands in Ferrolite with a race that is winnable from day one; the veterans fight each other at the summit.

Your Echelon opens the Leaderboard tab. Your tier's standings, ranked by growth this week (on equal growth the smaller empire ranks higher), the countdown to the weekly close, what closing at your current rank would pay, and your last verdict — promoted, relegated, or held — with the receipt of what it actually paid. The full server board sits right below, unchanged, now with each empire's tier on its row.

At the weekly close the top 2 with real growth rise a tier and the last 2 can drop one — with honest guards: an empire whose Gate climbs past the tier above's midfield rises whatever its rank, an empire whose Gate is above its own tier's midfield never drops, nobody drops out of Ferrolite, and nothing rises out of Singularity. Rewards go by your position inside your tier: a share of your own warehouse capacity (12% for first, tapering to 1% for every active racer — substantial for a small empire, pointless to farm with a throwaway account), plus Scrip and Dark Matter in the top three tiers. A still week pays nothing; accounts in their first 48 hours race but are not paid yet.

Synthetic empires race in the tiers too — the AI badge marks them, they keep the field alive, and they take no purse: every prize stays in the playerbase.

1.70.0 — 2026-09-02

New

• The Annihilator has its own gun, and its own voice. In battle replays it used to fire the Destroyer's beam and play the Destroyer's sound. Now it fires a lance: a gather at the muzzle, a twisting pair of violet strands inside a wide sheath that snaps shut instead of fading out, recoil rings kicking back off the barrel, and a hit that flares and throws real light onto the ship it struck. It has a new sound to go with it — a short charge, then the bang — and the shot is timed so the bang lands on the impact. Every other ship fires exactly as before.

1.69.0 — 2026-09-02

New

Every message in the galaxy and guild channels now reads in English, whoever wrote it (#2163). Until now, a message typed in Italian stayed in Italian for everyone else — unless the admin wrote it, in which case it was translated on the way in. That translation now applies to every player: write in your own language, and the channel stores and shows the English. What you typed in English is stored exactly as you typed it, typos and all — the layer only ever changes a message that was not English to begin with.

Private whispers are not translated, and never leave the server. A whisper is two people who already understand each other, so there is nothing for translation to buy and something real for it to cost. The privacy page says so in as many words.

Changed

The privacy page now names the AI service the game uses, and what it sees. It was incomplete before this release — not only for translation: the name check at registration, the AI empires' own replies (which carry the recent chat around them as context), and the build advisor already went through the same service, and the page's "that is the whole list" did not include it. It does now, along with what is not sent: no email, no account id, no name alongside your message, and mentions of other players replaced by placeholders first.

Fixed

A threat written in Italian still counts as a threat. The AI empires read the stored message to decide whether they have been threatened, and half of what they look for was Italian — which translation would have made unreachable. The English renderings are recognised now, and the anchor is at a word boundary, so "your warehouse is full" and "I'm afraid you're right" no longer read as declarations of war either.

An AI empire that gets no reply from the server when it whispers no longer says the same thing twice.

1.68.0 — 2026-09-02

New

• Every pop-up now has a close button (#2176). On a phone, the server-message box opened close to full screen, and the only way out was to hit the sliver of dark backdrop left around it. Every pop-up in the game now carries a ✕ in its top-right corner, sized for a thumb, and it stays put while you scroll a long list underneath it. Tapping outside and pressing Escape still work exactly as before. The Android back button still leaves the game rather than closing the pop-up; that is a separate, larger change.

1.67.0 — 2026-09-01

New

• While you were away — a return recap (#2021). Coming back after 8+ hours no longer drops you on the colony screen with zero acknowledgment that time passed. One compact sheet now opens first — a centered card on desktop, a bottom sheet on your phone — with the night's story: the net resource change across your colonies (signed — loot taken from you shows as the loss it was), every building, research and ship order that finished while you were gone, attacks and espionage you suffered with a working View report link on each battle, convoys and fleets that docked, and the next thing about to complete — or an honest "your build queues are idle" nudge when nothing is running. If hostiles are already inbound, the sheet leads with that. A warehouse that filled up overnight is called out by colony name ("Ferrolite storage full at Tycho Outpost — production paused there") instead of masquerading as a night of zero production. One tap dismisses it, it never shows on same-day resumes, and there is nothing to claim on it — it orients you, it does not pay you. Eight quiet hours show nothing at all: the colony screen is better than an empty sheet. The recap is built from data the game already loads at boot, so signing in is exactly as fast as before.

1.66.5 — 2026-09-01

Fixed

• The Annihilator is black again. The last release made it visible by painting it grey, which fixed the wrong thing — it bought readability by throwing away the ship. It is now exactly the hull that was drawn: black, with the plating, the panel lines and the violet glow at the eye and the prongs all readable. What changed instead is the light in the Shipyard bay, which is now a proper showroom rather than a dim room, so a mirror-finish hull has something to reflect — and every other ship in the bay reads a little better for it. The stray white sparkle along its edges is gone too: that was damage from compressing the model, and it now ships at full detail. Nothing about the ship itself changed: same stats, same cost.

Changed

• Battle replays draw one ship per squadron, with its name plate on. A squadron used to appear as a wedge of hulls that grew with its size — and at replay distance those hulls piled on top of one another and read as wreckage. Each squadron is now a single ship, and its plate (Interceptor ×2147) stays up while the squadron has anything left, so the size of a force is still on screen without tapping. Losses show on the bar beneath, as before; the ship blows apart when the last of them goes.

1.66.4 — 2026-09-01

Fixed

• The Campaign strike group now includes the Annihilator (#2166). Nobody can win one yet — the Eschaton event still grants the Destroyer and nothing grants this hull — so this is the fix landing before the ship does, not a flagship handed back to anyone tonight.

  The tower strike group is built automatically from the hull types your empire has unlocked, plus whichever legendary you actually hold, and it fielded every other hull that qualifies — your Destroyer included — while skipping the Annihilator no matter who you were. Each legendary now boards for the empire that owns that hull, and for nobody else.

  Worth knowing for the day it does matter, because it is not simply an upgrade: a strike group carrying two legendaries loses the Spectral Bulwark's damage cap. That cap normally stops a lone legendary losing more than an eighth of its hull in any single round, and it only applies to a fleet fielding exactly one of them — so a group with both takes its punishment undiminished, in exchange for the extra guns. (The Bulwark's other half, the rule that a flagship cannot be destroyed outright, is not active in the tower and has not been since the tower started fighting you as a coalition.) On the deep endgame build we bench, the two-legendary group still reaches one level deeper than the same roster with only the Destroyer in it. A gain, just not a free one.

Also

• Nothing else was missing from the Campaign roster. Every armed hull you have unlocked boards, the Aegis Bastion and Continuum Hauler continue to board on their shields, and freighters, probes, colony ships and Arks continue to stay out of the fight — all unchanged, and stated here because "the roster skips a hull" deserved an answer wider than the hull that was reported.

1.66.3 — 2026-09-01

Fixed

• Updates no longer freeze the live game while they install (#2158). Every deploy used to build the new version at full speed on the same machine that serves the game, and switched to it mid-build: during those minutes the game could stall for tens of seconds — actions timing out and screens refusing to load for whoever was playing right then, several times a day. Builds now run first, held below the running game in priority, while the current version keeps serving; the switch happens afterwards, in seconds, on a machine that is no longer busy building. If stalls are still seen during updates, the next step — preparing updates on a separate machine entirely — is already named in the deploy notes.

1.66.2 — 2026-09-01

Fixed

• A guild can no longer be double-charged when a purchase times out (#2138). When a synthetic empire's purchase hit a network timeout or a dropped reply — a rolling restart at exactly the wrong moment — the client quietly sent the request again, and if the server had already honored the first one, the guild paid twice for one item: two Scrip bills for one shield. The same exposure sat on every action that spends — building, ordering ships, starting research, launching missions and raids, spending crono fragments, Dark Matter upgrades, guild donations, grants, aid, outposts, Eschaton contributions, and market trades. Every one of those now sends exactly one request: a failure surfaces as a failure instead of a hidden second purchase, and a guard keeps the rule enforced for every future endpoint too.

1.66.1 — 2026-09-01

Fixed

• The Annihilator is visible again in the Shipyard. Its card showed a black outline with a few white glints on it — you could tell something was there, and nothing more. The hull is now lit like every other ship in the catalogue: plating, panel lines, the pink glow at the prongs and at the eye. It is still the darkest capital ship in the game, which is the point of it; it is no longer a hole in the card. Nothing about the ship itself changed — same stats, same cost, same everything you build it for.

1.66.0 — 2026-09-01

New

The Shipyard becomes the Guided Hangar (#2144). The old shipyard asked you to read it: a wall of cards on desktop, a phone page you scrolled through before you knew what you were looking at. It has been rebuilt around one idea — the ship you are looking at is always named, always on screen, and the button that builds it is always under your thumb.

One place to stand. The hull now stands on a lit hangar plate — a floor, a pedestal, rib lights in the colour of its class — with its name, its step on the rarity ladder, its class and its position in the catalogue written on the plate itself. On a phone the plate is a swipeable carousel of every hull you can see, locked ones included, so the catalogue is something you flick through rather than something you scroll past. The name can no longer be covered by the build bar: that was the first thing the redesign was built to fix.

For you. A short strip at the top proposes two or three hulls you can actually afford right now, one per class, so a new commander has somewhere to start instead of twenty-three equally loud cards. The picks hold still while you are looking at them — they do not reshuffle under your finger as resources tick — and they refresh when you switch colony, so an outpost is never guided by the home world's treasury.

Building, on the thing being built. BUILD is pinned: on desktop at the foot of a console that scrolls independently, on a phone in a bar above the tab strip. It always reads the same way — how many, and how long — and when something blocks the order, the reason is on its own line above the button (Short 12,000 Fe) instead of hidden inside the label. Press and hold the steppers to ramp through large numbers, tap Max to fill the yard, and tap ↺ to repeat your last order for that colony.

The queue, and the shard that shortens it. Building slips are full-width rows that cannot clip, idle slips say they are idle, and Speed up sits on the slip itself: tapping one opens the shard picker aimed at that exact order, with the order named above it and Orders are committed stated before you spend, not after. Your shard inventory is on the deck header, and the way back to the Shard Chandlery is never more than a line away — including when you have none.

Everything else is one bar away. Defence, the Dry Dock and the Star Forge sit below the deck as three collapsed bars that all behave the same way, and the Forge now shows the selected hull's stars right under the stage with a door straight to its ladder. Ship Reference, the full hull ledger with sorting, and the per-hull Specs sheet — stats, counters, the 3D model, and what the order costs — are all reachable without leaving the deck.

Long waits read like waits. A queue that used to quote 1323h 34m 43s now says 55d 3h, and above an hour the seconds are dropped everywhere they were noise.

Fixed

• The Speed up sheet can no longer aim a shard at the wrong order. Two identical orders in the queue, or one order finishing while the sheet was open, could re-point a paid shard at a lane you never tapped. The sheet now resolves the lane exactly the way the server does, so what it shows is what gets cut; an order that has genuinely left the queue says so instead.
• Spending a shard no longer makes the panel claim the order has finished when it has hours to run, and you can spend a second shard on the same order without closing and reopening.

1.65.2 — 2026-09-01

Fixed

• Raids no longer risk re-fighting their whole battle when a deploy stalls the server (#2130). Joint-raid resolution ran on the database library's default 5-second transaction budget — the largest such transaction in the game. Production evidence from today shows the server suffers multi-second stalls during deploy windows (the same class that produced this afternoon's guild-aid error), and a stall landing on a resolving raid would silently roll the whole battle back and re-run it from scratch, over and over, delaying loot, points and fleet returns with nothing visible but worker logs. Both raid-arrival paths — the battle resolution and the empty-target abort — now carry an explicit 15-second budget sized to survive those stalls, and tests pin it so it cannot quietly fall back to the default. Nothing changes in how battles are fought or paid out.

Also

• The investigation acquitted raids of today's 18:43 UTC connection-pool burst and convicted the deploy's on-box image builds instead — filed as #2158 for a proper fix.

1.65.1 — 2026-09-01

Fixed

• Farming PvE targets no longer pays PvP rates — the 50% haircut now reaches every PvE-owned world (#2099). Beating the neutral garrison on an event control planet, a relic warden, or a Wraith faction bastion used to pay the attacker season and weekly-war points at full player-versus-player rates: a maximum-value battle against the garrison wall paid 100 points of destruction value, same as against a real player, off a target that retunes itself and never fights back in the standings. Now those battles take the same 50% anti-farm haircut the Pirate Beacon has always taken, on both boards, win or lose, solo or raid — that 100-point battle pays 50. What the haircut does NOT touch: loot and victory bonuses stay whole (a loot-heavy garrison win is trimmed by a few percent, not halved — the halving bites the repelled and low-loot battles that were the farm), player-versus-player battles pay exactly what they paid before, and the human garrisons a holding guild stations on a control planet or relic keep every share they earn for repelling an attack. The event-control code claimed this haircut "already applied"; now it does, and the docs say since when. Completes the cleanup #2073 (v1.54.12) started on the same standings from the defender side.

1.65.0 — 2026-09-01

New

The Annihilator — a second legendary warship (#2124). Until now the Destroyer stood alone at the top of the catalogue: the one hull nobody can build, and the one that comes back from the dry dock instead of dying. It now has a bigger sister.

The Annihilator is stronger on every measure that decides a fight — half again the Destroyer's attack, hull and shields — and it fights differently. Every other ship in the game picks one enemy group and pours its whole salvo into it. The Annihilator fires that same full salvo at its target and strikes two more groups besides, at half strength. It is never a trade: whatever the enemy brings, it hits at least as hard as a single-target ship would, and against a spread-out fleet it hits far harder. A defender can no longer park one fat stack in front of it and soak the round.

It is slower than the Destroyer, and deliberately so: it is a far heavier ship, and the Destroyer keeps its title as the fastest thing in the fleet.

Like the Destroyer, it can never be built in the shipyard, no research opens it, and losing it does not lose it — the wreck is towed home and returns for half its cost and twelve hours in the dock. The dry dock now tracks each legendary separately, names the hull on every countdown and every button, and prices each repair on the ship that actually went down.

Not awarded yet. The hull is finished — art, model, combat rules, dry dock — but no Event grants it and no research unlocks it, and its card says exactly that rather than promising a prize the season does not hand out.

Changed

The activity feed, the Holdings sheet and the Eschaton sweep all used to say "Destroyer" wherever a legendary was involved, because there was only one. They now name the hull that was actually lost, repaired or rescued.

Also

The dry dock, the "no weak spot" rule, the debris exemption and the Star Forge's price tier used to name the Destroyer directly. They now describe legendary hulls as a class, so nothing about the ship that cannot die depends on remembering to update five separate places.

1.64.1 — 2026-09-01

Fixed

• The research card can no longer disagree with its own prices (#2116). On an uncapped tech line's sheet, the note saying where the price flattens — "all three flat by lvl N" / "Dark Matter rises 100 each level" — was computed by your browser from the tables baked into the page, while every price beside it came from the server. After a balance retune, a tab left open showed server prices next to a plateau note the retune had made stale, and the one line whose whole job is to be believed contradicted the chips under it. The server now states the plateau level itself, right beside the prices computed from it; a page loaded before this release keeps working exactly as it did.

1.64.0 — 2026-09-01

New

• The Explore button now knows the exploration cap (#2117). Explorations are limited to 5 per rolling 24 hours (6 with your guild's Science-4 keystone), but the button on an anomaly's sheet never knew it: at the cap it stayed lit, the tap opened the mission composer, and the attempt died with no error and no mission. Now the game tells you where you stand:
  • at 0 left the Explore button greys, with the reason readable right on the sheet — "No explorations left (5/5 in the last 24h) — the next slot frees in ~2h 10m". The countdown is the server's own instant, because the window is rolling, not a calendar day: nothing resets at midnight, each launch frees its slot 24 hours after you made it;
  • the anomaly panel and the mission composer show your remaining count — "3 of 5 explorations left in the last 24h";
  • the composer refuses a sixth exploration before launch with the same sentence, instead of a late server error;
  • if the count is not known yet (or your cached count has visibly expired), nothing greys: the button never withholds a launch the server would accept, and the server's own refusal banner — now also reworded to explain the rolling window — remains the safety net.

Also

• The exploration-cap tally moved into one shared computation used by both the launch check and the new count, so the button and the server can never disagree about the same refusal.

1.63.0 — 2026-09-01

New

• The spy report shows the defender's star mastery (#2115). A probe has always brought back the defender's Ship Stars beside their fleet — the screen just never showed them. The spy card now renders the star pips beside each starred hull in the FLEET section, at the same reveal as the fleet itself: unstarred targets, Ghost-fogged reports and old reports look exactly as before.

Fixed

• Synthetic empires no longer fight a battle model that cannot see Ship Stars (#2115). Since stars entered the engine, every battle has multiplied a hull's attack, hull and shield by its owner's mastery — up to x1.30 / x1.15 / x1.15 at five stars — but the synthetic empires' raid forecasts, force sizing and strategic planning priced every navy as unstarred. A starred defender would watch them read a "favourable" ratio, sortie, and lose, with no signal to correct. They now price a spied defender exactly as the engine will fight them, using only what the probe legitimately revealed: their own fleets stay unstarred (they never buy stars), and where a report is fogged they see the same fog you do.
• The launch preview prices the spied defender's stars (#2115). The what-if battle behind the mission composer had the defender-stars seam wired and never fed: the preview fought a five-star wall as if it were bare. The probe's star record now reaches the forecast, so the verdict describes the battle the server will actually resolve. Your own hulls' stars are not in the preview yet — that half needs its own wiring and is tracked separately.

1.62.5 — 2026-09-01

Fixed

• Faction bastions are fightable again: their garrisons now follow the strongest FLEET, not the strongest empire's total power (#2127). The three bastions had grown into walls of 22,434,296 hulls each — sized on the top empire's gate, a number that at endgame is over 90% buildings, research and Tree, and converted entirely into ships. No fleet in the game was priced for that fight, and the bastions stood unfelled. They are now sized on the server's strongest fleets (the same measure the relic garrisons have used since July), and within a minute of this deploy every live bastion re-tuned itself from 22.4M hulls to 2,002,945 — a hard siege for the top fleet, a real target for a coalition, and exactly what the wall was supposed to be. The prize for felling one is unchanged: three 4-hour Chronoshards per qualifying participant, 48 Dark Matter at the bulwark step, and the full stockpile.

Also

• Bastion garrisons can now shrink as well as grow between re-tunes: they track the strongest fleet in both directions, on the same slow six-hour rhythm as before.
• One review note for the record: felling a bastion no longer carries the weekly war board's underdog bonus (that ratio compares total-power figures the bastion no longer inflates). Season standings never paid points on bastions, and the fixed Chronoshard prize is untouched.

1.62.4 — 2026-09-01

Fixed

• Faction bastions are fightable again: their garrisons now follow the strongest FLEET, not the strongest empire's total power (#2127). The three bastions had grown into walls of 22,434,296 hulls each — sized on the top empire's gate, a number that at endgame is over 90% buildings, research and Tree, and converted entirely into ships. No fleet in the game was priced for that fight, and the bastions stood unfelled. They are now sized on the server's strongest fleets (the same measure the relic garrisons have used since July), and within a minute of this deploy every live bastion re-tuned itself from 22.4M hulls to 2,002,945 — a hard siege for the top fleet, a real target for a coalition, and exactly what the wall was supposed to be. The prize for felling one is unchanged: three 4-hour Chronoshards per qualifying participant, 48 Dark Matter at the bulwark step, and the full stockpile.

Also

• Bastion garrisons can now shrink as well as grow between re-tunes: they track the strongest fleet in both directions, on the same slow six-hour rhythm as before.
• One review note for the record: felling a bastion no longer carries the weekly war board's underdog bonus (that ratio compares total-power figures the bastion no longer inflates). Season standings never paid points on bastions, and the fixed Chronoshard prize is untouched.

1.62.3 — 2026-09-01

Fixed

• A refused launch now says so right at the Send button (#1952). On desktop, when the server turned an attack away — the target under night protection, novice cover, a pact, an active shield, or any other refusal — the only notice was a banner at the top of the Missions tab. The Send button sits at the bottom of a tall form, so if you were scrolled down at the button the click looked like it did nothing: no fleet left, nothing changed, and the reason was off-screen. The refusal now also appears on the launch bar itself, beside the button you pressed, with the reason spelled out. It clears when the launch goes through, retires when you change the mission, target or launch pad, and carries a Dismiss control. When the connection drops and the game cannot know whether the fleet left, the line says exactly that — in its own tone, never claiming a refusal — so you know to check In flight before pressing Send again. The top banner still works as before, and phones keep their bottom-pinned notice, which was never affected.

1.62.2 — 2026-09-01

Fixed

• Guild "Investigate N anomalies" directive now counts dark-matter and ambush outcomes (#2128). The Council directive only advanced when an anomaly turned out to be salvage or resources: a dark-matter find or a pirate ambush resolved the anomaly, paid its reward and advanced your personal contract — but left the guild directive stuck. Since each anomaly's outcome is fixed in advance, a run of dark-matter or ambush anomalies made the directive impossible to complete no matter how many scans you flew (as spartanecho found: three explorations, one counted). Now every real anomaly investigation counts — salvage, resources, dark matter, and ambushes won or lost — exactly matching what the personal contract counts. Lost claim races and already-colonized slots still count for neither.

1.62.1 — 2026-09-01

Fixed

• Guild actions, fleet recalls, chat pins and Star Forge purchases no longer fail when the server hiccups (#2121). A transaction was measured in production dying at 5.78 seconds against a 5-second internal time budget nobody had chosen, during a momentary server stall — any purchase, guild join, pin or recall in flight at such a moment could come back as an error even though nothing was wrong with it. Every action that waits its turn on a per-player or per-guild lock now carries an explicit, wider budget, so a brief stall or a queue behind another commander's action completes instead of failing. When one of these actions does fail, it still fails safe: nothing is charged and nothing is half-granted — a guard test now keeps every such action on the explicit budget for good.

1.62.0 — 2026-09-01

New

The last Italian voice in the channel now speaks English (#2122). Novarchy's chat has been English for a while: the interface, the Registry chips, every synthetic empire. One account was still typing Italian into it, and everyone else was reading a language the rest of the channel does not use. From this release that account's messages arrive in English like everybody else's.

What you see is simply a message in the same language as the channel around it. What sits behind it: the translation happens when the message is written, not when it is read, so the galaxy channel, the guild channel, a whisper and the iOS app all show the same sentence. There is no second version of a message anywhere.

The people, battles and events cited in a message are untouched by it. A chip pointing at a raid still points at that raid, with the name it had when it was tagged, because the references are lifted out of the sentence before the translation and put back exactly as they were. If any of them would not survive the round trip, the message goes out as it was typed instead: nothing is worth a citation that quietly loses what it was citing.

And if the translation is not available at all, for any reason, the message is still sent. It just goes out in the language it was written in, the way it does today.

1.61.0 — 2026-09-01

New

• The Shipyard is now the Hangar Deck (#2113). Before, the Shipyard was a grid of 22 small cards with the ship's details in a side panel (or, on the phone, in a sheet you had to open card by card). Now the hull you are looking at is the centrepiece.
  • On the phone: swipe through your hulls on a full-width stage, class by class (Light · Medium · Heavy · Support, or All). Under the stage, a nameplate with the ship's rarity, class and role, and a signal strip that summarises its stats at a glance — tap it for the full Specs sheet. The build controls stay under your thumb in a bar that sticks to the bottom of the screen: quantity presets, a stepper you can press and hold (it ramps to ×10 and ×100 on its own), and one BUILD button whose face quotes the build time and the cost.
  • On desktop: a three-zone bay — the class-grouped hull list on the left, the ship large on a stage in the centre with its 3D model, and the full stats and build console on the right.
  • Locked hulls stay in view as dimmed silhouettes with the research node that unlocks them, and a NEXT tag marks the nearest one in each class. A hulls unlocked meter in the header (e.g. "8 / 21") shows how far along the catalogue you are.
  • The queue shows its idle slips: a free build slot is named ("Slip 2 — idle") instead of being invisible.
  • Repeat your last order: after you place an order, a gold chip remembers it per colony; tapping it prefills the same ship and count (clamped to what you can afford now).
  • When you cannot build, the reason is written above the button — not enough resources (with the exact shortfall), queue full, locked, no shipyard yet — rather than being buried in the button's label.
  • Your stationed fleet is listed under the queue on the phone, one chip per hull with its count; tapping one opens that hull's Specs.

Changed

• The ship's 3D model now lives on the desktop stage and in the phone Specs sheet, no longer inside the phone's hull carousel — a rotating model under the finger was fighting the swipe.
• The Specs sheet on the phone keeps its build block pinned at the bottom, as before the redesign, so the allowance and "orders cannot be cancelled" notes sit next to the button they qualify.

Also

• The ship reference button, defence planner and dry dock are all still there — the reference button moved into the Shipyard header, and the defence planner and dry dock sit below the hangar. The old "stationed fleet" block became the chip row on the phone and the count badges on the desktop hull list.
• Nothing changes in how ships are priced, timed or queued: this is a new face on the same shipyard.

1.60.0 — 2026-09-01

New

The endgame research lines stop being a wall (#2083). Doctrine of Annihilation, Beyond the Horizon, Production Singularity and Abyss are the four lines with no level cap, and until now their price doubled every level — which is not an endless line, it is a wall a few levels further out than it looks. Measured on the largest empire the rules allow, one level cost about 36 days of total production at lvl 20, five months at 22, two years at 25 and forty-seven years at 30.

Now each resource stops rising at its own ceiling — roughly 60 days of a fully built empire's production — and the three stop one after another, so a line is completely flat somewhere between lvl 22 and lvl 24 depending on which one it is. No price goes up anywhere: every level costs the lower of the old price and the new ceiling, never more than it cost before.

Once all three resources are flat, each further level costs Dark Matter — 100 for the first, 200 for the next, 300 after that, and so on without end. This is a new cost on levels that previously charged none, and it rises by 100 each time rather than doubling. It is what paces the line now that the resource price has stopped growing.

The node sheet says both halves: below the flattening it names the level where the prices go flat, and past it the Dark Matter chip carries the running figure with the level the ramp started from.

Changed

Planning a Concord capstone now needs the Dark Matter in hand. The four tier-9 capstones — War Crucible, Resource Concord, Industrial Confluence, Veilsight — cost 1,000 Dark Matter each. Until now you could queue one while short and pay when a lane freed; the button now refuses, reading "Not enough Dark Matter".

This reaches players who will never see an endgame plateau, so it is worth the plain reason: when the queue reaches an order it cannot pay for in Dark Matter, it does not skip it — it stops, and everything behind it stops too, Disciplines included, in a currency mines do not produce. Meeting that wall on the button is better than meeting it on a frozen bench.

Also

A very deep empire's Power score now counts what its endgame levels actually cost rather than what the old doubling price would have charged. No current standing moves.

1.59.0 — 2026-09-01

New

• The Star Forge is on screen (#2118). Open the Shipyard and it is there: all 22 hulls, each with a 30-step mastery ladder drawn as a grid you can read at a glance — five stars across, six sub-levels down, every cell marked with the thing that step buys. Attack, hull and shield take turns in a fixed cycle, so you can see three steps ahead without opening anything, and the sixth cell of each star wears two marks because it is two things: the ignition, and a shield step. On a support hull with no guns the attack cells show as hull, because that is where the step actually goes — the picture never promises what the engine will not pay.
• Every price, before you pay it. Pick a hull and the card itemises the next step: what it costs in resources, faction items, Dark Matter and Sovereign Seals, each as have / need, with every leg you are short on called out at once rather than one refusal at a time. The button carries the bill, so what your thumb reaches and what it spends are never in two different places — and on a phone every purchase asks before it spends.
• The weekly meters tell you when, not just how much. Each faction family and the Seals show what is left this week and when the next credit ages back in — the caps roll drop by drop, so a full meter says "2d 6h", never a Monday that never comes.
• The Forge is visible before it opens. Below the Tier-9 capstone that unlocks it you still see the whole ladder, the prices and your wallet — with a line under each item saying which faction's bastions drop it. The items bank long before the door opens, and the screen that explains what they are for should not be the one thing you cannot see.
• Take a Destroyer to five stars and you may name it — the one ship in the game that carries a name of your choosing.

Also

• Your mastery shows as a star row wherever a hull is listed, down to a 24-pixel line on a ship card, and the tooltip is honest: three separate numbers for attack, hull and shield, never one blended figure.
• The cost card reads your mother colony's warehouse, because that is the colony the Forge charges — and it says so, so the number never has to be reconciled against the readout at the top of the screen.
• A purchase that times out and is retried cannot buy a second step: the request names the step it believes it is buying, and the server refuses anything else.

1.58.0 — 2026-09-01

New

• The AI empires learned to play (#2093). The silent machines got a real brain. Their strategist now runs on the strongest model we have, reasons before it commits, and — more importantly — finally sees the game: targets near their own space and in their fair-fight band (instead of everyone staring at the same corner of sector 1), their own unlocked ships, the spy reports they scouted, and their whole empire rather than one colony.

  On the ground that means: they defend now — fleets come home when their colony is about to be hit, garrisons scale with the actual threat, and the wall is counted the way battles count it. They attack like they mean it — sized forces with a home reserve instead of all-in blind strikes, no more repeating an attack that already lost, and their war fleets carry freighters, so a defence you lose can now cost you a real share of your warehouse, not pocket change. They stop wasting money on buildings that do nothing (see below) and research according to their character instead of one shared script.

  Fair play is unchanged: newcomer protection, the power band and the bounty rules are all enforced by the server exactly as before. What changed is that the machines now play well inside those rules — expect the leaderboard to feel it.

• The Factory and the Orbital Arsenal now tell you the truth on their sheets. Each opens its shipyard queue slot at level 1, and no level above adds anything but energy drain. The sheet, the build lists and the Advisor now say so — the same gold warning the Colonial Outpost already carries. Nothing is capped and nothing you built is taken away: levels climbed still count for points, power and the building's look.

Also

• A bounty on someone's head is now genuinely dangerous: every AI empire in range sees it. If that head is yours, expect visitors.
• The AI guilds still keep their silence — no war declarations, no surrender notes. Wars involving them end at the weekly rollover as always.
• If you write to an AI empire, your message still sits unread rather than being ignored: nothing is consumed, nothing answers.

1.57.0 — 2026-09-01

New

Tier 10 of the tech tree: four capstones, built to be overpowered (#2068)

The tree no longer ends at Tier 9. Each branch now crowns at a Tier 10 capstone — 10B resources plus 2,000 Dark Matter, unlocked by the branch's Tier 9 capstone and its endless line at level 8:

• Doctrine of Escalation (War): your naval counters above 1x gain +0.5 (capped at 3.5, never lowering what Sundering already reaches), and your Batteries and Fortresses fire and shield at x10 — hull unchanged, the Cannon excluded, season score unchanged. Battle reports and your raid forecast know the node.
• Boundless Reserve (Frontier): every Warehouse ceiling x1.5 (x3 with Deep Vaults), plus the Requisition Surge — a button on the Research tab that triples every colony's storage ceiling for 24 hours, once every 5 days. Deliveries, crate opens, prizes and reparations all land against the raised roof, and every screen reads the same ceiling the server holds — including the "Full in…" clock, which now knows when a surge ends.
• Accretion Engine (Forge): Particle Accelerator Dark Matter output x3 at double the energy draw. A tier leg refills in days, not weeks.
• Nightfall Protocol (Spectre): your attacks reach the defender's screens only in the last stretch of the flight — ten minutes, or 15% of the flight if longer — always tagged, so the victim knows why the warning came late. A Deep Sensor Grid at level 8 anywhere in the defender's empire sees them from launch, phone push included. Flights of ten minutes or less were never hidden and still warn at launch.

A storage rule changed on purpose

Stock above a ceiling is never destroyed any more — not by a shrinking cap, not by a surge that ended, not on the next tick. Production pauses on the excess until you spend back under, and the interface now says "production is paused", never "lost", where nothing is lost.

Also

• Recalling an attack closes the defender's alarm exactly when an alarm was ever shown: a feint under the Nightfall Protocol leaves no trace, and a countdown you were watching never vanishes without its closing line (#2068).
• The Anchorage prices every charter at the room its convoy will find on landing, so a surge can no longer sell a convoy that would burn at a calm doorstep hours after the window closes (#2068).
• Synthetic empires learned the new machines' arithmetic: an agent that buys the Accretion Engine keeps it lit exactly when a human would (#2068).

1.56.0 — 2026-09-01

New

• The Star Forge accepts its first purchases (#2111). The faction items and Sovereign Seals you have been banking since last release can now be spent: every one of the 22 hulls has a 30-step mastery track — 5 stars of 6 sub-levels — and each step buys exactly one thing, permanently. Attack, hull and shield take turns, so a step never quietly inflates everything at once, and nothing you buy can ever be refunded, uninstalled or lost. The Forge opens with any Tier-9 capstone; below that you can already see the whole ladder and its prices, because the items bank long before the door opens.
• Every price is itemised before you pay. Resources come out of your mother colony, faction items and Seals out of your wallet, Dark Matter on the star ignitions — and if you are short, the refusal names every leg you are short on and by how much, in one answer, instead of sending you back four times. Each weekly cap now also tells you when headroom returns: the ledgers roll credit-by-credit as old drops age out, so the meter shows a real time and not a Monday that never comes.
• The sixth sub-level of every star is an ignition — it costs Dark Matter and a Sovereign Seal, and it is the one that lights the star. Take a Destroyer all the way to 5 stars and you may name it: the one ship in the game that carries a name of your choosing.

Also

• The Forge's purchases show up in your Activity feed under Economy, with the full bill — what the step cost in resources, items, Dark Matter and Seals — so you can find where a wallet went without reconstructing it.
• A retried purchase can never buy a second step: if the request arrives twice, the second one is refused rather than charging you for a step you did not order.

1.55.0 — 2026-09-01

New

• The faction bastions now drop faction items (#2107). Felling an outpost, stronghold, citadel or bulwark of the Silent Choir, the Black Foundries or the Exarch Guard pays every qualifying raider Choir Resonators, Foundry Ingots or Exarch Sigils — fixed counts by bastion size (1/2/4/7), no dice, full share per qualifying head on a joint raid. Repelling a faction sortie pays the defender a smaller guaranteed haul: the solo channel. Your feed shows every haul next to the Dark Matter line, including exactly which weekly ledger clipped you when one is full (45 items per family, rolling 7 days).
• Sovereign Seals begin to drop. Bulwarks pay one guaranteed; citadels roll a fair, seeded 25% — the same result no matter when the battle resolves, and never worse than one Seal per four citadel fellings: after three Seal-less citadels the fourth pays, guaranteed. Seals are capped at 4 per rolling week, across every source.
• These items and Seals are the currency of the Star Forge, the end-game per-hull mastery ladder coming in the next releases. There is nothing to spend them on yet — bank them: guild raids on the weekly bulwark are how wallets fill fastest, and everything you collect now is waiting when the Forge opens.

1.54.16 — 2026-09-01

Fixed

• The Missions tab now counts a pending exodus toward the colony cap, like the server does (#2037). With an exodus offer pending and the colony cap exactly filled by it, the Colonize verb badge stayed lit and the launch panel showed no blocker — then the launch died on a raw "maximum colonies reached" error that sent you to upgrade a Command Center that was not the problem. Now the badge grays and the launch panel blocks before the launch is spent, on the same sum the server refuses: colonies owned, colonizations in flight, and the pending exodus. When the offer is what holds the slot, the blocker says so — the research line reads "… + your pending exodus" and the Command Center line reads "(your pending exodus holds one)", the same wording the galaxy planet sheet already uses. With no offer pending, nothing changes.

1.54.15 — 2026-09-01

Changed

• Nothing changes in the game this time either. Second rung of the Ship Stars ladder: every battle door and every Power reading now carries a hull's mastery record the way it already carries refits — an attacking fleet will fight with the stars it left under, a defender with the stars it holds at the moment of impact, and the NPC factions, relics, pirates and the season boss with none, ever. Nobody holds a single star yet — the Star Forge that sells them is still to come — so every battle fought today resolves exactly as it did yesterday, and the leaderboard moves by nothing.

Also

• When stars exist, battle reports and spy intel will show them honestly: each side's report row carries what that side actually fought with, and a spy probe reveals a defender's mastery only at the same clearance that reveals their fleet — a target hidden behind a Ghost Manifest keeps its mastery fogged too.
• Joint operations are ready for mixed company: in a raid or a shared defence, every ally's contingent fights under its own record — a starred ally lends you their strength, not their paperwork.

1.54.14 — 2026-09-01

Fixed

• Synthetic empires no longer show two colonies both named "Prime Colony" (#1938). Three agent empires — r.costa, rmendes, prahman — each displayed two identically-named "Prime Colony" colonies, which read as a machine deliberately obfuscating its map. The cause was a login window during a Forced Exodus: when an empire's only colony was razed, the empire waited in exodus limbo with zero colonies, a login during that window seeded a fresh home, and the exodus settle then brought the razed mother back — two default-named homes. The home seed now recognises an empire in exodus as in transit rather than homeless, so the duplicate can no longer be minted (for humans as well as agents). A boot-time repair renamed the three existing duplicates to their coordinate names ("Outpost 2:14:4", "Outpost 1:7:8", "Outpost 5:11:16") — nothing was deleted, and the real home of each empire keeps the "Prime Colony" name. Verified in production after the deploy: zero empires carry duplicate home names.

1.54.13 — 2026-09-01

Fixed

• Battle reports for Tier 9 empires understated the multipliers the battle actually used (#2069). The report sheet stores each commander's loadout reduced to the tech-tree nodes that fought, and that list was never told about the Tier 9 capstones: the War Crucible (attack ×1.1) and Veilsight (+10% shields on defense) decided real clashes but were stripped from the stored report, so a T9 owner's card recomputed weaker attack and shield multipliers than the battle resolved with, and the capstone chips never appeared. Reports written from now on carry both nodes, with each chip grouped beside the multiplier it explains (reports stored before this fix were reduced at write time and stay as they are). A census guard now derives the list of battle-relevant nodes from the combat engine itself, so a future tier's combat capstone cannot be forgotten silently again.

1.54.12 — 2026-09-01

Fixed

• The Season Leaderboard ranked a PvE faction among the players (#2073). "The Verge Watch" — the neutral garrison holding the contested event worlds — sat at rank 3 of the live season board with 547 points, pushing every player below it down one rank and one prize rung. PvE entities (the pirate beacon, factions, relic wardens) no longer appear on the season board, the season podium, or the guild event podium's member ranking: ranks and prize rungs slide up to the real players. LLM empires still rank — they are competitors — and the Wraiths stay excluded exactly as before.
• PvE entities no longer earn season or war-week points at all (#2073). The garrison had genuinely earned those points: the battle reward only knew to skip the pirate beacon, so factions and relic wardens collected points on every attack they repelled. Both battle channels now refuse every PvE kind — while the members of the guild holding a contested world or relic keep every point and directive they earn defending it, exactly as before.
• The end-of-season Destroyer can no longer be minted for a PvE bastion (#2073). A faction finishing on top of the raw score would have received the season's unique hull; now any PvE entity on top withholds it from everybody — the signature of a bug, and the server now says so in its log — matching the rule that already existed for the pirate beacon.

No prize was ever paid to a PvE account: the current season is still open, and this closes the gap before its first payout. Had the season closed as it stood, the faction would have taken the rank-3 crates from the player who reported the bug.

1.54.11 — 2026-09-01

Fixed

• The Concord ledger no longer prints fourteen digits per resource (#2082). A research bill paid by eleven colonies filled four lines with figures like 45,588,937,919,675 Fe · 45,588,937,919,675 Xe · 45,588,937,919,675 Dt from Lantea; …. The ledger now speaks the same K/M/B/T shorthand the research costs and the Concord treasury adopted this week: 46T Fe · 46T Xe · 46T Dt from Lantea; …. Figures below a million keep every digit, exactly as before.

Changed

• The «Exact figures» switch now sits in the ledger too, and gives back every digit. It was a treasury-only control; on a phone that put it hundreds of pixels above the bills it was supposed to open up, and a big bill drains the treasury — so the switch used to disappear at the very moment you wanted it. There is now one in the ledger's own heading, and pressing either one expands both. On a computer, hovering a shortened row still shows the full figures.

1.54.10 — 2026-09-01

Fixed

• War Room: "strikeable only" no longer empties the board in a lopsided war (#2081) — Pressing the War Room's "strikeable only" button while your side's Power towered over the enemy's made every target vanish behind "their empires are all under protection" — even though the unfiltered view attacked those same empires without complaint. The filter still treated the Power band as a wall, but since early August it is a bounty price, not a refusal. Bounty targets now stay in the filtered list, the button's count and the "Targets (N)" tab count agree with it, and the empty message only appears when everyone really is refused (novice protection or shields). The rows themselves stopped contradicting the toggle too: a bounty target's Attack button is now full-strength red like any legal strike, showing the price in prose instead of the "You can still probe it." consolation reserved for targets the server actually refuses.

1.54.9 — 2026-09-01

Changed

• Nothing changes in the game this time. This release opens the engine seam for Ship Stars, the end-game per-hull mastery ladder: battles and the Power score can now read a hull's stars — but nobody has any, nothing writes them yet, and every battle fought today resolves exactly as it did yesterday, verified down to the byte in the tests. When the Star Forge opens in a coming release, the strength you buy there will flow through combat, the Power score, battle reports and the fleet advisor's forecast through the seam this release installs.

Also

• Battle reports are ready to show star pips on both sides once stars exist — publicly, on purpose: counters are the counterplay to stars, and picking a counter requires seeing what you are countering.
• The fleet advisor's what-if battle and its forecast row will price stars identically, on both the attacking and the defending side, so the preview never describes a different battle from the one the server will fight.

1.54.8 — 2026-09-01

Fixed

• Research-focused AI empires now actually fund research (#1948). Player Aglet Green spotted it from the leaderboard: the synthetic empires excelled at shipbuilding and construction but consistently under-invested in research — even the ones whose strategic plan declared a research focus. That focus only nudged them to build more laboratories, while actual research spending always came from whatever the construction queues had left over. Now a research-focused synthetic empire pays for research first each turn and holds back part of its capital treasury — never more than half of what remains — to save toward the next technology it cannot yet afford. Its outposts keep their own purses untouched, it keeps building from the remainder, and every other strategic focus behaves exactly as it did before.

1.54.7 — 2026-09-01

Changed

• The campaign's deepest-climbers board only lists ranked empires (#2077). The "deepest climbers" list on the Campaign scene — and its ticker — used to be the one board that filtered on its own rules: the operator's account sat at #1 on a depth from an earlier era of the tower, and a deep-climbing Wraith or the announcer would have been listed too. The board now subtracts exactly the identities every other standing subtracts, so the twenty rows are all empires you can actually race: ranks renumber from 1 with no gap, and your own depth in the ticker still reads from your campaign, not from the list.

1.54.6 — 2026-09-01

Fixed

• The weekly League's dissolution payouts are back in agreement with the League's own prices. When the League purse was halved earlier today so the sport would stop out-paying the season, three checks covering what happens when a guild dissolves mid-fixture were left quoting the old, larger figures. Nothing you were paid was wrong — the purse itself was already correct — but the safety net that watches those payouts had stopped watching. It is watching again.

Also

• Groundwork for Ship Stars, the end-game per-hull mastery ladder announced in the design published a few releases ago. This release ships the arithmetic — how far each of the 30 steps moves a hull's attack, hull and shield, and what each step costs across the five currencies — plus the icons the Forge will be drawn with. Nothing in the game reads it yet: no ship is stronger, no price has changed, and the Star Forge is not open. It is the foundation the Forge is built on, landing ahead of the screen so the screen has something true to stand on.

1.54.5 — 2026-09-01

Fixed

• A raid kill is no longer bragged as everyone's private solo feat (#1943). When a guild raid tore a chrono shard off a bigger empire, every participant's reward carried the coalition's power ratio with nothing saying the kill was joint — so each eligible synthetic empire could boast "tore a 4-hour shard off an empire 3× my power" about the same battle, and your activity diary told you the same solo story. The shard now records how many empires struck together: the chat boast only goes out for a kill the shard can prove was solo (a genuine solo underdog kill brags exactly as before), and the diary reads "torn from an empire 3.00× your raid's combined power — 6 of you struck together" for joint kills. With agent chat currently muted server-wide, the chat half closes the hole behind that switch rather than changing anything you see today.
• A felled faction bastion no longer reads as a battle against nobody (#1943). The diary used to render each bastion shard as "torn from an empire 0.00× your power" — a battle that never happened. It now says what it is: the prize for a felled faction bastion, exempt from the weekly shard pool in both directions.

1.54.4 — 2026-09-01

Changed

• The operator's account leaves every standing (#2070). The account "Alex" — the game's operator, whose economy and fleet are granted server-side and had reached about 13 trillion points — no longer appears on the leaderboard, the weekly war scoreboard or the season scoreboard, and collects none of their prizes. Every rank below that row moves up one: the player who was second on points is now first, the rank-1 share of the weekly war goes to a player who actually earned it, and the end-of-season Destroyer slides to the first ranked empire. The account had been out of the standings from 10 to 21 August, was let back in, and is now out for good — the wanted board still lists it, so the bounty on its head remains collectable, and the difficulty of relics and faction bastions was never sized on it in either direction.

Also

• The three League fixture tests that had been red since this morning's purse retune now assert the halved payouts the retune deliberately set; nothing changes in the game, the suite just stopped disagreeing with a decision already shipped.

1.54.3 — 2026-09-01

Fixed

• The weekly League no longer out-pays the season (#1971). When the League went weekly, its rewards stayed sized for a round every two weeks: a member of a guild that won every fixture banked 720 Scrip and 32 Dark Matter over a season — 1.8x the season's 400-Scrip first prize — and eight won matchdays paid 200 guild XP, double the ceiling the progression was tuned to, making sport the fastest road to guild levels and the member seats they buy. A match win now pays 45 Scrip and 2 Dark Matter per member (was 90 and 4), a draw 17 Scrip and 1 Dark Matter (was 35 and 1), a loss 7 Scrip (was 15), and a won matchday is worth 12 guild XP (was 25) — so a perfect League season lands back under the season's first prize on every axis, and the season stays the big prize it was designed to be. Everything already paid out stays yours: only fixtures resolved from now on pay the new purse. If your installed app still says a match win pays "up to 4 DM" on the Anchorage's Dark Matter list, that text catches up with your next app update — the payout change itself is already live.

1.54.2 — 2026-09-01

Fixed

• A guild dissolving mid-week no longer erases its opponent's League fixture (#1970). When a guild disbanded — its last member leaving, an account deleted, or the empty-guild reaper — every League fixture of that pair vanished with it: the surviving opponent read "Your guild drew a bye" about a fixture they were drawn into and may have been winning, the week's verdict was never paid (no purse, no season points, no guild XP, no log line, no verdict event), and already-finished rounds dropped out of their standings and history. Now the fixtures stay. Finished rounds keep counting, and a live fixture is settled within minutes of the dissolution on the scores as they stand, so the survivor collects the win, draw or loss the scoreboard actually shows. The dissolved guild itself leaves the table instead of holding a rank as a nameless "—" row, and synthetic players no longer risk citing that placeholder as if it were a guild's name in chat.

1.54.1 — 2026-09-01

Changed

• The AI empires have stopped talking (#2047). Players told us the machines in chat had become annoying, so they are now silent — everywhere. No opening lines in the galaxy channel, no replies when you mention them, no guild councils, no galactic gazette, and no answers to private messages. Chat now carries only what people write.

  They have not left the game. The AI empires still build, colonise, attack, trade, join guilds and climb the leaderboard exactly as before, and they still read what you write and remember it. They simply do not answer.

  A private message you send one is not lost — it stays unread on their side rather than being quietly swallowed, so if the voices ever come back, it gets an answer then.

• Kepler Observatory has gone quiet too. The weekly War Week bulletin that announced the podium in the galaxy channel is off with the rest. Your prizes are untouched: the podium still pays out, and the paid podium still shows on the War tab — only the chat announcement is gone.

• Guilds led by AI empires will not declare war or surrender for now. Those decisions were taken in the guild council, and a council that cannot speak should not be able to declare a war nobody can read about — you would find out from the numbers. Wars already under way still end as they always did, at the weekly rollover.

Also

• If the silence turns out to be too much, none of this is a deletion: the whole chat machine is intact and comes back with a single switch, no new build required.

1.54.0 — 2026-09-01

New

• "Still needed" on the tree node sheet (#2036). A node you cannot afford now says how much more it wants — per resource, and per Dark Matter when the node charges it — instead of leaving you to subtract two fourteen-digit figures. The line only ever rounds up, so banking what it says always clears the red, and it disappears on an order you already paid for (a running order never claims its owner is short — its chips read neutral too, Dark Matter included).
• An "Exact figures" toggle on the Concord treasury strip (#2036). The summed wallet's chips compact past a million; the toggle stacks them and shows every digit, on any device. It only appears when a figure actually lost digits, and desktop hovers carry the exact figure on every compacted chip.

Changed

• Research costs read as K/M/B/T past a million (#2036). Doctrine of Annihilation lvl 32 used to quote 46,707,609,302,782 Fe — fourteen digits nobody reads as anything but "about 47T". Tree node prices, Discipline next-level prices, the Concord treasury chips and the bench's Dark Matter sentence now compact past a million; the red/white affordability tint is still computed on the exact figures. The Concord ledger deliberately stays exact: a bill is subtracted against the warehouse, so it keeps every digit — as do trades, loot and the warehouse table.

Fixed

• A shortfall could understate above ten billion (#2036). The card footer's "how much more" figure rounded a 46.3T gap down to −46T — the one number contracted to only ever overstate. Its ladder now reaches the T tier.
• A Dark Matter debt reads as what it is (#2036). Bounties claimed on you can take the balance negative; the research wallet used to clamp that to 0, so the sheet asked for less Dark Matter than the purchase actually needed — banking what it said left the refusal standing. The balance now reaches the sheet signed, the gap arithmetic accounts for the debt, and a debt no longer paints phantom "needs 0 Dark Matter" walls on orders that charge none.

1.53.5 — 2026-09-01

Changed

• Nothing changes in the game this time. This release ships a design document for a feature that has not been built yet — Ship Stars, an end-game mastery ladder for individual hull types — plus the art it will use and one caching rule. No prices moved, no ship got stronger, and nothing in your empire behaves differently today.

• The Star Forge header and three item icons are now in the build, unused. When the feature ships they will be the faces of the currencies you spend on it; for now they are files nothing reads.

• A caching fix, invisible until it matters. Art files under /assets/ui/ were being served with a one-year "never check again" instruction, which is right for backdrops nobody redraws and wrong for anything that gets touched up later — a corrected image would simply never have reached you. The Star Forge header now revalidates hourly, like ship and item art already does. The two large space backdrops keep their year, because they do not change.

Also

• The design itself was re-priced against the Tier 9 that actually shipped. It had been written when a Tier 9 capstone cost 1B resources and 400 Dark Matter — a price that lasted about an hour before being retuned to 5B and 1,000. Everything in the ladder was anchored to the old number, so a system meant to be end-game was costed at a fifth of it. When it is built, a fully-mastered hull will cost roughly twice the entire four-node Tier 9 tier in Dark Matter, on top of its resource bill.

1.53.4 — 2026-09-01

Fixed

The guild Emergency crate now requires an emergency (#1980). The store row always said "get back up after a raid" — and never asked for one. Any guild member could buy it four times a week: 30 Scrip (three donations, costing about 1.8% of your Warehouse ceiling in resources) bought back 12% of that ceiling in fresh resources — about 6.7 times what the donations cost, minted from nothing, growing with your empire. At the reported ~38.7M warehouse that was ~34M free resources a week for one day of donating.

From this release the crate is on sale only in the 48 hours after an attack on one of your colonies — you must have been the defender in a battle there. Raids, solo attacks, Wraith and faction attacks all count; your own attacks, and garrisons you lent to a friend's colony, do not. The weekly limit drops from four to two. Price and payout are untouched: a raided player gets exactly the per-crate relief they always got, and a refused purchase costs nothing — no Scrip, no weekly slot. The store row states the rule, and the row, the refusal message and the server all derive their "48 hours" from one constant.

Two transition notes, said here so nobody files them as bugs: if you already bought three or four crates this week, the counter reads "3/2" or "4/2" until Monday 00:00 UTC — the button is correctly disabled, nothing is broken, it heals with the new week. And apps installed before this release show the refusal as a raw code (no_recent_raid) until they update; the rule is spelled out in the row's description just above it.

The synthetic empires now use the guild store — and mean it. A wiring defect had kept every agent out of the emporium since guilds launched: no agent had ever bought a single item there. It is fixed with manners. An agent that just lost a defense of one of its own colonies buys the Emergency crate when its stores are empty; it buys the 8-hour Peace shield only when its plan is defensive — and then genuinely holds fire on every door (solo attacks, raids opened, raids joined) while the shield lasts, though it still grinds its Pirate Beacon, which the shield permits. Their purchases appear in the guild log like anyone else's.

1.53.3 — 2026-09-01

Fixed

• The AI badge now appears everywhere names are written as text, not just where they stand alone (#1957). Since the game stopped hiding its machines, every synthetic empire carries a glyph next to its name — but surfaces that joined names into sentences quietly lost it, so a machine there read exactly like a person. Now the badge also rides:
  • the incoming-attack card: each attacker in "attackers: …" carries its own glyph, joint raids included;
  • the threat board: the threatened guildmate's row, the "On station" garrison list and the intel line under each threat;
  • the raid call: every contingent in the participant list;
  • the guild registry: every sentence naming a member or a rival — "X kicked Y", grants, aid, relic captures and losses;
  • the activity feed's frozen sentences: market receipts ("Sold to …", "Bought from …"), the spy-probe echo, the bounty collection notice, all three open-space interception rows, the "your colony was razed by …" notice, the relic chronicle and the guild-aid receipt.
• Rows written before this release keep their exact wording and stay unmarked — the feed never guesses. New events carry the badge from now on; live boards (rosters, threat board, registry) show it immediately for everyone.
• A person whose name contains a machine's name can no longer inherit its badge — the glyph lands only on the machine's own name, never inside or beside a human's.

Also

• The emails these badges are derived from never leave the server: every board and every event payload carries only the badge itself.

1.53.2 — 2026-09-01

Fixed

Big treasury figures no longer print over the panel next to them (#2033). On the Research tab, the Fe / Xe / Dt boxes under your Laboratory network showed every digit of what you hold. Late in a season that stopped fitting: under the Resource Concord, where the figure is every colony's stockpile added together, a number like 45,821,940,447,365 is wider than the box that holds it, and it was drawn straight across the Concord ledger beside it — two panels of text on top of each other.

Those three figures are now shortened past a million, exactly as the counters at the top of the screen already are: 46T, 1.2B, and every digit still printed below a million (999,999). On desktop the exact number is one hover away on the box itself.

One thing this does not do yet: a phone has no hover, so there the exact total is not reachable from that panel. Nothing else changed — what a research level costs is still printed to the unit in the panel where you order it, and whether you can afford it is still worked out from your real numbers, never from the shortened ones on screen.

1.53.1 — 2026-09-01

Fixed

• The galaxy Colonize button no longer lies about whether you can colonize (#1955, #1995). It counted colony ships in your mother colony only, so a Colony Ship or Colonial Ark stationed at an outpost left the button grey — even though the same colonization launched fine from the Missions tab, because the game accepts a founding ship from whichever colony you launch it from. In the other direction, the button ignored the checklist right under it: with the Command Center colony cap reached, or your Colonization research short, it stayed green and the click dead-ended at the launch panel. The button now counts founding ships across your whole empire — including one your Shipyard finished seconds ago — and greys out whenever a requirement is known to be unmet, while a requirement whose data has not loaded yet never disables it: green stays a forecast the launch panel verifies.
• The button now says where to launch from and why you are blocked. When your founding ship is parked away from the colony you are viewing, a note under the button tells you to pick its colony under "Launch from" instead of letting the launch panel refuse a mission the button just promised. When a pending exodus is what fills your colony cap, the requirement rows and the tooltip name it — instead of pointing you at a Command Center level that would unblock nothing — and a grey button's tooltip names the same requirement the game refuses on first.

1.53.0 — 2026-09-01

New

• The galaxy has grown to 12 sectors (#1937 — it was 8). Four new frontier sectors, 9 through 12, sit on a new outer ring: 76 more star systems, 1,368 more planet slots to colonize. Nobody moved to make room — every existing colony, sector and travel time kept its place; the new space simply appeared beyond the old rim, and the map, target lists and sector pickers grew with it.

Changed

• The Event Sector is now sector 13 (it was 9, which has become an ordinary frontier sector). The season's contested zone moved one step past the new edge, and everything in it moved along: event colonies, control planets and their held-since clocks, fleets already in flight, even the battle debris. Chat coordinate links, the season panel and the "Go to the Event Sector" button all point at 13.
• The galaxy map opens one step further out on most screens — twelve sectors are simply more world to fit. Control rings, fleet labels and coordinates appear one pinch in, as they always did at galaxy zoom.
• Routes from the inner sectors to the Event Sector are longer than last season (it sits past sector 12 now); the 2-hour travel ceiling still caps every route, and the new sectors 9-12 are the closest launch points.

1.52.1 — 2026-09-01

Changed

• Tier 9 prices were retuned on launch day. A Concord capstone now costs 5 billion resources plus 1,000 Dark Matter (it shipped this evening at 1 billion + 400, and both measurable end-game empires bought instantly — the whole tier was cleared within the hour). The tier now totals 20B resources and 4,000 Dark Matter; tiers 10 and 11 re-curve to 10B and 20B. At this scale a single capstone costs most of the second-richest empire in the game, which makes the Resource Concord's summed wallet the practical way to pay — exactly the fantasy the tier was built around. The seven purchases already made at launch prices keep them — nothing is clawed back, and orders that were paid and in flight complete at the price they paid.

1.52.0 — 2026-09-01

New

• Being a Recruit now means something — and so does demoting someone to it (#1529). The two lowest guild ranks used to carry identical powers: none. A player reported the obvious symptom: "Why is there a button to demote members and it does nothing?" Now a full Member holds the one thing a Recruit lacks — asking the guild's Help board for supplies or for a hand with a timer. A Recruit can still give aid, donate to the Fund, buy from the emporium, chat and take directives; the ask side opens on its own when their 7-day probation ends. Demoting a Member to Recruit restarts that week and closes their ask side, so the ▼ button finally does what it looks like it does — and it now asks for confirmation that spells out the cost, on desktop and phone alike.
• The Help board explains itself to a Recruit instead of refusing them after the tap. The ask form is replaced by a plain sentence saying when it opens; the counters that cannot count for them show a dash; a brand-new joiner is told the 48-hour countdown opens the giving side, not the ask side; a demoted member with a request still on the board is warned that withdrawing it will not free a new slot; and the «Ask the guild» buttons on the Speed up panels disappear for recruits instead of answering an error.
• Synthetic empires obey the same wall. A recruit agent reads the closed ask side from the board itself and stays quiet instead of promising requests it cannot post, its reality sheet says so in as many words, and bot leaders are told what a demotion now costs before they hand one out — most of their roster reviews end with no demotion at all.

Also

• If your installed app predates this release, a Recruit tapping the old ask form will see the raw code aid_rank_too_low until the app updates; the current version shows the full sentence.

1.51.0 — 2026-08-31

New

• A knock when your First Convoy docks (#2022). Your First Convoy has always arrived exactly 24 hours after you founded your empire — but if you were not in the game at that moment, nothing told you. Now, if you have allowed notifications, one soft push arrives the moment it lands: "Your convoy has docked at [colony]. The cargo is waiting." Tapping it opens your Base, where the Take delivery button is — on iOS and in the browser alike. One push per convoy, never more; it respects your quiet hours, the daily notification cap, and your opt-out, and nothing is sent while you are already in the game or after you have taken delivery. Android will inherit it automatically when Android push support lands (#1206).

1.50.0 — 2026-08-31

New

• A knock when your First Convoy docks (#2022). Your First Convoy has always arrived exactly 24 hours after you founded your empire — but if you were not in the game at that moment, nothing told you. Now, if you have allowed notifications, one soft push arrives the moment it lands: "Your convoy has docked at [colony]. The cargo is waiting." Tapping it opens your Base, where the Take delivery button is — on iOS and in the browser alike. One push per convoy, never more; it respects your quiet hours, the daily notification cap, and your opt-out, and nothing is sent while you are already in the game or after you have taken delivery. Android will inherit it automatically when Android push support lands (#1206).

1.49.0 — 2026-08-31

New

• Tier 9 research: the Concord (#2009). The tech tree's ninth tier is open — four capstones, one per branch, and each is priced in BOTH currencies: 1 billion resources plus 400 Dark Matter, charged together. The tier's frontier is Resource Concord: once you own it, every research bill — Disciplines, tree nodes, the endless endgame lines — is paid from the summed stockpiles of ALL your colonies, your mother colony first, with a Concord ledger on the Research tab showing exactly which colony paid what. War Crucible multiplies your fleets' attack by 10%, Industrial Confluence adds +10% production of all three resources on every colony, and Veilsight grants +10% shields on defense and makes your spy reports reveal the target's empire-wide summed stockpiles (marked clearly as not lootable from the probed colony). Entry per branch: its tier-8 capstone plus its endgame line at level 5 — Spectre, which never received a tier 8, opens through the Eschaton Veil instead.
• The research bench understands Dark Matter. A queued tier-9 order that is waiting on the Dark Matter half of its price now says so on its slot, with the figure; the node sheet shows the DM price beside the resource chips, tints it red when your balance falls short, and greys the button instead of letting you tap into a refusal.

Changed

• Research pricing above tier 8 was re-anchored. Tiers 9–11 were priced years of inflation ago at 2.16M/4.32M/8.64M — pocket change for the empires the tiers are for. They now cost 1B/2B/4B. Nobody had ever been able to buy at those tiers, so no live purchase is affected.
• Dark Matter accrual is banked a little earlier. Research-driven colony updates now bank your Accelerator's accrued Dark Matter at the moment they write the colony, instead of leaving it for the next collection point. Totals are computed the same way; if your Accelerator finished an upgrade while you were offline, the window before the upgrade is now credited at the correct pre-upgrade rate instead of occasionally rounding in your favour.

Fixed

• Colonies whose planet data predated the trait system no longer bank research-time production at neutral rates: every colony a research bill touches is now resolved on its real planet.
• A spy probe that landed while your own research order completed mid-flight now counts that research in the report it files.

1.48.4 — 2026-08-31

Fixed

• Mobile galaxy map: the survey timer is fully readable again (#1716) — on a phone, the "New survey in …" notice at the top of the galaxy map could slide partly behind the left toolbar, hiding its first letters ("ew survey in 1d 10h…"). The notice now stays inside the top band, and its phone wording was re-measured to fit: it reads "Survey in 3d 22h — need Astrometrics 8." and the like, leading with the countdown so the number you act on is always visible; on the widest phones the whole line fits, and on the narrowest a too-long line trims at its end with an ellipsis instead of vanishing under the toolbar. The same issue also proposed removing the GALAXY › SECTOR › FLOWER breadcrumb on phones; it stays, deliberately: it is the only thing on a phone that names the zoom level (which decides whether a tap zooms or opens), and removing it would not make the map any bigger — in portrait the map is fitted to the screen's width, not its height.

1.48.3 — 2026-08-31

Fixed

• A healthy deploy can no longer fail its own smoke check on a wrongly empty log window (#2006). Since v1.47.29 the deploy verdict reads only the api log lines written after the deploy started, and it refuses to bless an api that appears to have logged nothing — that guard stays exactly as it is. But the query that reads the window could transiently come back empty on a freshly restarted api even though the boot lines were there, turning a green deploy into a false red and stopping the release chain right after a healthy boot. The window is now cross-checked against the Docker daemon's own log read before the verdict: real boot lines are found and vote, while a boot that truly logged nothing — or noise pretending to be a boot line — still stops the ship. Nothing changes in the game itself; what changes is that fixes stop getting stuck behind a verdict that was wrong about a healthy server.

1.48.2 — 2026-08-31

Fixed

• "Capital hulls" now reads "Heavy hulls" (#1530) — the Nanoforge building sheet, the build advisor tip and the Nanoforge research card all described the discount using a ship class the game does not have. The classes you see everywhere else are Light / Medium / Heavy / Support, so all three surfaces now speak that vocabulary: the sheet's first row says "Heavy hulls", the advisor says "Heavy hulls come out N% sooner", and the research card says "full on heavy hulls (Battleship and up)". The mechanic itself is unchanged: the cut still scales with hull mass, full from the Battleship's mass up, so massive Support hulls like the Colonial Ark keep the full discount.

1.48.1 — 2026-08-31

Changed

• Two research nodes stop pretending to be sensor and spy techs (#1470). "Deep Sensorics" sounded like a fleet-detection tech and sat one shelf away from the real fleet sensor node "Deep Sensor Grid" — but its only effect is +4%/lvl data production, and the name had already sent at least one player down the wrong branch. It is now called "Deep Computing". "Covert Networks", the other +4%/lvl data node, read as espionage infrastructure and is now "Data Networks". Only the display names changed: levels you have already researched in either node are untouched, keep their effect, and simply show the new name. Exactly one node in the tree now carries a sensor name, and it is the one that actually unlocks the Sensor building.

1.48.0 — 2026-08-31

New

The Colonial Ark founds colonies — and becomes the new world's first Warehouse.

The Ark sits at the end of the Frontier branch, behind Colonisation 14 and three other nodes, and the tech tree has always described it as a super-colony ship. It could not found a colony. It was a freighter with a colonial name, and the shipyard card deliberately avoided the word "colonise" because printing it would have promised something the game refused to do.

It does it now. A colonisation launches on a Colony Ship or a Colonial Ark, and an Ark that founds a world is dismantled into it as a level 3 Warehouse — the new colony opens with a 20,480 ceiling per resource instead of 5,000.

That last part is the whole point. A new colony used to hold only 5,000 of each resource, so a large starting endowment was trimmed on arrival and the remainder was destroyed when the founding ship was the only one in the fleet. The hull whose one selling point was its hold punished you for filling it. One Ark now lands a colony, a full hold of supplies, and the warehouse to keep them in.

The launch panel warns you when an endowment would still overflow the new world's ceiling, and says what it will cost you.

Changed

The Colonial Ark is no longer the worst carrier in the game. Its hold goes from 4,000 to 14,000 and its hull from 6,000 to 10,000. Its price, speed and shields are untouched.

For anyone who already owns Arks: each one is now worth considerably more Power, so expect to move up the leaderboard at this release. Nothing was taken away from anyone.

The transport advisor will now offer Arks where it used to ignore them — they were ranked dead last among carriers and only ever picked as a final filler.

Fixed

• The fleet screens no longer tell you a colonisation needs a Colony Ship specifically. The requirement row, the launch check, the objective list and the error message all name both hulls, and the Settle button stops greying out when the only hull you own is an Ark.
• The raid picker no longer offers your Colonial Ark as loot freight. Like the Colony Ship, it is now marked as a hull to leave at home.
• The colonisation report no longer tells you to build a Warehouse on a colony that already has one.
• The picket rules now name the kinds of flight that pass a blockade rather than a list of ships, which was missing the Ark.

1.47.40 — 2026-08-31

Nothing changes in the game this time — this release is armor for your future ship art.

Also

• A re-cut ship image can no longer hide behind your browser's cache (#1997). Ship art files keep stable names, so when one is re-drawn the game has to move a cache version for the new art to reach players who already hold the old copy. Buildings have had a test guarding that step since #1054; ship art relied on someone remembering it (the Spy Probe re-cut in #1902 was saved only by memory). Now all 81 ship art files — hero cards, shipyard thumbnails, fleet billboards and glow maps — are fingerprinted next to that version: replace one without moving the version and the test suite goes red before the change can ship. If a future release re-cuts ship art and you keep seeing the old image, that is now a bug the tests would have caught, not a silent inevitability.

1.47.39 — 2026-08-31

Fixed

• Siege forces are sized for the whole static wall, not one group at a time (#1988). Planning an attack on a colony defended by both Defense Batteries and a Fortress, a synthetic empire could assemble a strike force strong enough to break either defense on its own — but not both in the same battle, since the engine makes the fleet fight through the two walls with the same rounds. The measured example: three Breachhead Monitors against Fortress 5 and Batteries 30 cleared the old check and sank with the defenses still standing. Task forces are now sized for the combined engagement (the same example composes five Breachheads, which win with hulls to spare), and a hangar that can only beat the defenses separately keeps building instead of sailing a fleet into a measured loss. Colonies behind a single static defense are attacked with exactly the same forces as before — and because forces are now sized honestly, some fortified targets that were previously composed too small and then vetoed by the operation forecast become genuinely attackable.

1.47.38 — 2026-08-31

Fixed

• The Spy Probe's card art no longer has a jagged white halo (#1902). The probe's hero art — and its billboard in battle replays — carried a bright fringe around the whole silhouette, worst along the solar panels and antennae, left over from rough background removal when the art was rendered from its 3D model. It stood out next to cleanly cut hulls like the Battleship, and worse on a high-DPI phone. The edge has been re-cut: contaminated edge pixels now take the ship's own color, and the stray specks scattered around the silhouette in the files are gone too (in the replay those were already below the renderer's visibility threshold — the halo was the visible defect). Same art, same silhouette, same dimensions; the art cache version was bumped so everyone gets the clean version. A new test reads the shipped images and fails on the halo signature, so a future rough re-cut cannot land silently.

1.47.37 — 2026-08-31

Fixed

• Synthetic empires now weigh a siege fleet by what it actually brings to the wall (#1965). Until now an AI empire scored its siege plans with a flat bonus for merely carrying any siege-capable hull: one bomber tucked into a cruiser wing counted the same as a full Breachhead wing, so empires that had built the right fleet kept declining sieges they would win — eleven Breachheads at a level-30 Fortress were appraised below their own launch threshold — while the battle engine resolves the Breachhead at four times its attack against static defences. The appraisal now reads the same counter tables the engine fights with, weighted by each hull's share of the fleet's firepower and each defence group's share of the colony's strength, and damped so a confident forecast still has to be earned. Fortified colonies start drawing the sieges they always invited: attacks that sat parked just under the old threshold launch after this update, and a lone bomber no longer flatters a whole wing.

Also

• The adversarial review of this change measured that the empires' appraisal remains deliberately conservative — every newly launched attack shape won its engagement in the review's engine model, and some winnable sieges are still declined — and filed the pre-existing, currently harmless task-force under-sizing it found against colonies fielding both Batteries and a Fortress as #1988.

1.47.36 — 2026-08-31

Fixed

• The allies-under-attack board no longer vanishes when a background refresh fails (#1840). The guild's "Allies under attack" card refreshes itself every few minutes over the slowest route in the panel. If that read failed — a server hiccup, or a phone connection timing out — the whole card disappeared in one frame: header, every row, and an open Send-defence picker with a fleet already sized, all gone until the next successful read up to three minutes later. Now a failed refresh keeps the board you were reading; the rows are at worst one read stale, and the next successful read still updates everything as before.

Also

• The same board's promises are now under test (#1840). Three behaviours that were correct but unguarded — which colony row claims a standing garrison, which row draws the open Send-defence picker, and the freeze that keeps rows from jumping under your finger mid-refresh — could each have quietly regressed with every test green: telling a rescuer an uncovered planet was already held, painting a live ship picker under the wrong colony's header, or moving a 150–200px block of the board mid-tap. New tests now go red for each of those regressions, verified mutation by mutation against the real code.

1.47.35 — 2026-08-31

Fixed

• Galaxy names stop vanishing when only their empty margin touches an empty caption slot (#1166). On the galaxy map, a planet name could disappear because the caption slot of a nameless disc — an empty orbit slot, or a Pirate Beacon whose own name had just lost a crowding contest — touched the invisible margin around the name's letters, with no letter anywhere near it. The declutter pass now looks at the letters themselves, so those names stay on screen: measured on an eighteen-slot system, about one name in five frames comes back. A name that actually sits in another disc's caption slot is still hidden, exactly as before.

1.47.34 — 2026-08-31

Fixed

The League can see how strong a guild is (#1973). It pairs you against guilds of similar strength, and strength was measured on seven things — your record, your standing, your war research, your economy, your science, how many members you have, and how powerful you are. The last one was never actually read. In practice that left member count doing almost all the work, so a guild of one commander with a fleet of millions counted as a close match for a guild of one commander with three ships.

Power is now on the table, and it is read the way it should be: a guild ten times stronger reads as stronger, not as off the scale. That matters because power spreads across an enormous range, and measuring it flat let the biggest empire in the game squash everyone else into the same reading — so two genuinely well-matched guilds looked no closer to each other than to somebody a hundred times smaller.

You will see it in Monday's fixtures. On the current field four of the five pairings change, the two guilds with almost nothing to their name are drawn against each other instead of against opponents they cannot touch, and the two best-matched guilds in the game are put together rather than split up.

1.47.33 — 2026-08-31

Fixed

• Rare "something went wrong" errors on the galaxy map and in chat under busy moments (#1974). Twice on August 31 a request that worked for everyone else failed with a server error: the api's database connection pool was silently capped at 25 — a number derived from the server's CPUs, not from how the game is actually played — and at peak moments every connection could be busy for longer than a request was willing to wait. The pool ceiling is now set explicitly at 60, 2.4x the level where it saturated, so the map and the chat keep answering through the same bursts. Nothing changes in how the game plays; the two failed requests' code paths were checked and were never at fault.

1.47.32 — 2026-08-31

Fixed

The AI badge now marks a synthetic empire everywhere its name appears, not just on the handful of screens that already had it.

Before, the glyph beside a name — the hollow hexagon for an agent, the solid delta for a Wraith — showed up in chat, on the leaderboard, on the wanted board and in the guild panels, and nowhere else. On the target picker, the market, the battle reports, the contact sheet, the season and league boards, the events feed, the campaign ticker, the guild aid and invite lists and the mission intel, a machine looked exactly like a person. The board built for choosing whom to attack was among the screens that said nothing at all, which made it the worst of them: it is the one place a player meets a Wraith in a combat context, and Wraiths are kept off the leaderboard by design, so there was no other surface to learn it from.

The report that opened this named two screens. It turned out to be forty-one, and the badge is on all of them now, in the same place every time — after the name, before the guild tag. On the target row that ordering also had to survive a narrow phone: the tag and the bounty give way first, so the badge is never the thing a long commander name pushes off the line.

Screen readers were losing it too. The battle toast and the market row build their spoken text as a string rather than from what is drawn, so a player listening heard a bare name while the player beside them saw the mark. Both now say it aloud.

Also

The floating names on the galaxy map itself still carry no badge, and that is deliberate rather than unfinished. Those labels are drawn into the map as images, not as text the badge can be placed beside, and the two glyphs that would fit are already spoken for on that same chart — the hollow hexagon is an anchorage and the solid triangle is a contact, so a third one there would read as a place, not as a machine. Opening any of them shows the badge in the panel that appears.

1.47.31 — 2026-08-31

Fixed

• Synthetic empires now actually build the Breachhead Monitor they already knew how to use (#1848). Since the siege-evaluation fixes, an AI empire could correctly judge that a Breachhead fleet cracks your bastion — and then never commission one: its build plans only knew the old bomber/titan doctrine, so raising your Fortress or Batteries actually lowered the Breachhead's place in its shipyard queue. Now an apex empire (two Dark Matter Doctrines at 15) facing a seriously fortified colony — Fortress 3+ or Batteries 4+ — keeps a Breachhead wing in its standing build mix and buys the Breachhead first when massing for an operation against any standing static defence, falling back to titans and bombers when it cannot afford one. The seat it takes in the four-hull build mix comes at a price: against capital fleets the battleship is the hull that drops out. Ordinary colonies (Fortress up to 2, Batteries up to 3) see no change, and empires without the apex unlock see no change anywhere.
• AI siege forces are sized by what actually breaks walls (#1848). A fleet assembled against a Fortress used to be filled in raw-attack order, shipping redundant titans ahead of the hull its own wall check rated highest; the fill order now weighs each hull by its real siege multiplier, so a fortress force can be a single Breachhead where a titan would have tagged along for nothing.
• AI attackers stopped over-paying your Cannon in their estimates (#1848). The engine has always charged a fleet that fields a Breachhead only half the Cannon's saturation cap; the AI's estimators did not know that, so they overestimated the toll on exactly the fleets built for sieges and sometimes declined bastions they could take. All their estimate frames — target selection, raid planning, force sizing, and the operation forecast — now apply the halved cap. In guild raids only the empire's own Breachheads earn it: an ally's hull still makes the real battle cheaper than the estimate, never dearer.

Also

• What was deliberately not changed: the flat siege bonus in the AI's win-probability forecast still cannot tell a Breachhead armada from a lone bomber. That re-tune has its own issue (#1965) with the measurements attached.

1.47.30 — 2026-08-31

Changed

The League now plays every week. It used to alternate — one week of fixtures, one week of nothing — so a guild that drew a bye was out of the competition for a fortnight. Every week is now a round, and the panel never points further than seven days ahead. Nothing already played moves: the weeks that used to be rest simply become rounds.

A guild of one is now in the draw. The League used to ask for two members before it would enter you at all, and three of the guilds playing right now have exactly one: for them the League did not exist, week after week, behind a message telling them to recruit first. That requirement is gone. The pairing already matches guilds by size, so a solo guild is drawn against another guild its own size.

The table keeps its eight weeks. It shows the last rounds played, and it was sized in rounds rather than in weeks — so a weekly calendar would have quietly halved it, and a result earned a month ago would have dropped off the standings and out of your own fixture history. It now spans eight rounds, which is eight weeks, which is a season. Every result already on the board stays on it.

The panel reads its own rhythm off the server. Two lines used to state the rules from memory — "the League plays every other week" and "Table — last 4 fixtures" — and both went stale the moment the rules moved. They are now told to you by the server, so the app in your pocket says what the League is actually doing rather than what it was doing on the day that build was made.

The two sentences that stated the old entry rule out loud are gone with it.

1.47.29 — 2026-08-31

Fixed

• Releases no longer get blocked by errors from before the release (#1961). Nothing changes in the game itself this time — this is about the machinery that ships it. Three consecutive web-only releases were reported as failed by one api error logged hours before any of them: the deploy's health check read the api container's whole recent log window, and a web-only deploy never restarts the api container, so the old error kept re-triggering the check (and then also broke the rebuild of the public changelog page after each release). The check now judges only the log lines produced since the deploy started; older faults are still printed, as observations. A real boot failure in a container the deploy did (re)start still blocks the release — the new checks fail closed on every edge: a garbage timestamp, an unreadable log window, or a freshly started api that logged nothing all stop the ship instead of passing it.

1.47.28 — 2026-08-31

Fixed

• A harvest's two diary lines land under one Activity chip (#1232). A harvest is a round trip: the fleet collects at the debris field, then flies home. Its "Salvage collected" line was filed under the Economy chip while the same trip's "Fleet returned" landed under Empires — so following your fleets under Empires showed trips ending that never began, and Economy showed collections that never came home (the "Nothing to salvage" line had the same split). Both harvest lines now file under Empires with the rest of the fleet log, exactly as pickups do since #1229: the cargo only becomes yours when the landing credits it, and the landing line is the one that warns you when a full warehouse ate part of it. One knock-on worth knowing: if your whole night was harvests, the Economy chip may now be empty — the salvage story lives under Empires, whole.

1.47.27 — 2026-08-31

Fixed

• The message button's envelope is centred on the phone (#1901). At rest, the envelope icon in the chat "message" button sat noticeably left of centre: the unread counter, though invisible at zero, still reserved its space beside the icon. The counter is now a small gold badge in the button's corner that appears only when you have unread whispers — the envelope sits dead centre, and nothing jumps when a whisper arrives. Off the phone nothing changes.

1.47.26 — 2026-08-31

Fixed

• Mobile header no longer wraps to an extra row when your colony name is long (#1914). On a phone, naming a colony anything longer than a few characters ("FarFarWest" instead of "Marte") pushed the envelope button and the colony picker onto a second header row, shoving the resource chips and the map below them down. The header top row now always stays one row: a name that does not fit is trimmed with an ellipsis inside the colony picker instead of growing it, and the envelope button keeps its full size and tap target at every screen width. On desktop nothing changes.

1.47.25 — 2026-08-31

Fixed

• An agent only brags about a chrono shard it actually tore off someone bigger. Chrono shards drop from six different places — beating a stronger empire, but also the weekly chest, a bounty, a bastion, a refund and the shop. The chat brag was gated on the shard's size alone, so any 1-hour or 4-hour shard could trigger it, and the line it produced ("off an empire 1.0 times your power") stated a victory that had not happened. The worst case was publicly refutable: a bounty pays out on damage share even to the side that LOST the battle, so the battle report said the opposite of the boast. The brag now requires the shard to have come from actually striking above your weight, and the sentence that claimed shards "only drop when you beat someone stronger" is gone — it described the class of shard, not that one, and no gate could make it true. (#1932)
• And the number in that line is now one the server can have produced. The ratio is printed truncated rather than rounded, so a shard earned at 2.97× prints "2.9" and stays inside its own tier instead of borrowing the next tier's threshold. A ratio the engine cannot mint at all is refused outright.
• A brag goes stale. After the runner is down for a few hours the tick digests the morning's shard as new, and the agent would announce at lunchtime a raid the battle report dates to breakfast, while the victim has been talking about it since. Brags now expire, and when one is suppressed the reason is written where an operator can read it — so a feed that renames a field looks like a broken feed, not like a quiet day.

Note: production currently runs with the galaxy channel in addressed-only mode, so agents do not speak unprompted there at all. This fix is correct and live, but latent: no chrono brag reaches the global channel today either way. It takes effect the moment that mode is turned off.

1.47.24 — 2026-08-30

Fixed

A Colonial Outpost offered its next level as the thing to do, right under the line saying not to.

The Outpost stops helping after level 1 — the malus relief is complete there, and every level above it only raises the energy drain. The sheet has said so in gold since #1422. What it also did was present Upgrade to lvl 2 in the cyan styling it uses for "this is the move", eight pixels below. Matt D. reported the contradiction.

The button now says what it is: Upgrade to lvl 2 (no further effect), in the sheet and in the Base-map build menu. The Colony tab's row already carried the sentence and is unchanged.

It is still a button, and that is deliberate. Those levels are not worthless — they feed colony points and structural power, and the Outpost's own artwork changes at 10 and again at 20. Hiding the control would have frozen every Outpost in the game at its first look forever, and quietly closed a legal way to spend a surplus. If you are levelling one for the sprite or the points, nothing is in your way; you are simply no longer being told it is the sensible next purchase.

The same rule reads from the engine rather than a list, so the Trade Portal picks the note up at level 16 — where its offer slots and its per-resource ceiling both stop growing — and not before (#1681).

1.47.23 — 2026-08-30

Fixed

• The empire total no longer dips while your cargo is in the air (#1895). The resource totals at the top of the Overview page counted only what is landed in colony stockpiles — so during a large transfer or transport between colonies, and on the return leg of a pickup, the headline dropped by the whole cargo amount for as long as the fleet flew, and checking whether you could afford a big purchase mid-transfer gave a misleadingly low answer. The Empire tiles now fold the hold of every fleet in flight back into the headline, so the number answers "how much do I own"; when something is aloft, an "incl. X in transit" line says where the missing share is. The storage bar below keeps tracking what actually sits in your warehouses, and a pickup's manifest — still in the destination colony's stockpile until the fleet arrives to collect it — is never counted twice.

1.47.22 — 2026-08-30

Fixed

• Mutual-counter chips now say who fires the return shot (#1920). When two hulls counter each other, the ship reference and the shipyard card used to compress the exchange into "×2.5 · you ×1.5" — which read as "it usually is ×2.5, but for you it's ×1.5", as if a personal modifier softened the incoming counter, and the "you"/"back" label meant a different thing on the strong and the weak side. The second number is the return fire, and it now says so with a verb: a threat on your "Weak against" list reads "Voidlance Cutter ×2.5 · takes ×1.5" — it hits you at ×2.5 and takes ×1.5 from your reply — while a target on your "Strong against" list reads "Orbital Scythe ×2.5 · deals ×1.5" — you hit it at ×2.5 and it deals ×1.5 back. The panel legend explains the wording, an even exchange still reads "×2 both ways" on both sides, and mutual pairs stay listed on the side they net out to.

1.47.21 — 2026-08-30

Fixed

• Chrono-shard boasts now sleep with their author (#1921). Synthetic empires keep human hours: they go quiet outside their persona's activity window, exactly like a player in another time zone. One speaking path had no clock — the boast an empire posts after tearing a big chrono shard (1h/4h) off someone stronger could land on the galaxy channel at 04:00 its local time, while every other spontaneous line (taunts, guild talk, razing proclamations, post-attack outbursts) already waited for daylight. The boast now respects the window like everything else. Nothing changes on the channel today: unprompted galaxy chatter is currently switched off entirely, so this closes the gap for the day that switch comes back — and the shard itself still drops and gets spent whenever the battle lands.

Also

• Pinned in tests, without changing behaviour: a winning attacker may still gloat at night (deliberate — it always could), while a defender hit in its sleep stays silent until it plausibly wakes up.

1.47.20 — 2026-08-30

Fixed

• One over-committed offer no longer locks your whole market (#1924). If you spent a resource after putting it on offer (goods stay in your warehouse until someone accepts), every new offer you tried to publish was refused with "your other open offers already promise these resources" — even offers that had nothing to do with the over-committed resource. A commander with an oversized Xenogas offer could not post a plain Data-for-Ferrolite trade. Now only the resources your new offer actually promises are checked: unrelated offers publish normally, while promising even one unit of the over-committed resource is still refused until you cancel or shrink the offending offer. The Publish button and the server finally agree — the button already did this arithmetic correctly.

1.47.19 — 2026-08-30

Fixed

• Opportunist empires stop buying Dark Matter Lances as their second Doctrine (#1187). Dark Matter Lances grant shield penetration, which works against ship shields only — the planetary Shield Bubble soaks damage before any ship shield does, and an opportunist raider's whole go/no-go is about breaking that Bubble. Its spend table still priced Lances high enough that the second Doctrine purchase of its season was the one node useless against the defence it raids by. Opportunists now spend like the raiders they are: Veil first (mobility), Forge second (production — more hulls is what actually cracks a Bubble), and Lances late as a genuine residual — still worth having against shield-bearing defending fleets, so they are deferred, not dropped. Aggressive empires keep Lances first: fleet combat is their game, and penetration genuinely pays there. Validated in the self-play harness: the first Lance moves from the 2nd Doctrine purchase to the 10th at every income rate, with every balance criterion staying green.

1.47.18 — 2026-08-30

Fixed

A test that was supposed to keep agents quiet at night wasn't actually checking anything.

Agents sleep. Each one has an activity window, and a commander who sees one joking in chat at four in the morning has learned something they shouldn't be able to learn — it is the single clearest tell that they aren't talking to a person. A test existed to guarantee a sleeping agent posts no league trash talk, and it passed for the wrong reason: a random roll refused the joke before the sleep check ever mattered, so the sleep check itself was covered by nothing. Removing it left the whole file green.

The test now forces the roll to yes, which leaves the activity window as the only thing that can keep the agent quiet — verified by removing it and watching the test fail, with the agent visibly posting mid-sleep.

Nothing about how agents behave changed here: they were already sleeping correctly. What changed is that the guarantee is now guarded, so a future edit that broke it would be caught instead of shipping green.

Two comments in the same file that credited the wrong mechanism for holding an agent quiet across repeated ticks were corrected against a measurement rather than a guess.

Closes #1242.

1.47.17 — 2026-08-30

Changed

The Praxis Interdictor is priced as the expendable hull it was always described as (#1826)

The Interdictor exists to strip the enemy's rear row of its sanctuary — it is the only hull in the game that reaches capitals parked behind a screen. It has always been described as a cheap picket you bring several of and expect to lose.

It never cost that. At 1400 Ferrolite / 900 Xenogas / 300 Dati it was 34% more expensive than a Monitor Frigate — a ship with over three times its hull, more shield, more attack, available far earlier, and one that counters the Interdictor outright. Four of the five hulls that beat it cost less than it did. Bringing "a few to lose" cost more than bringing the ship that kills them.

It now costs 600 / 400 / 150. Ten of them come to 11,500 instead of 26,000.

Two knock-on effects, because in this game a hull's price is also its tonnage:

• Interdictors build faster, and a new colony's discounted starter allowance covers more of them.
• They leave smaller wrecks. Debris is a fixed share of what a hull cost, so an Interdictor field is worth proportionally less to whoever recycles it.

The ship itself is unchanged — same hull, same shield, same guns, same speed, same counters, same unlock. Only the price was wrong.

If you already own Interdictors, your score will drop. Leaderboard points are a live valuation of what your fleet cost, so re-pricing the hull re-prices the ones you already have: around 2,900 points for a player holding 200 of them, and your guild's standing moves with you. Nothing was taken from you and no ship was lost — the same fleet is simply now worth what it now costs. There is no way to rebalance a price without this happening, so it is said here rather than left to be noticed.

Fixed

• The design document that specifies the apex hulls still listed the Interdictor's old price. It was the only one of the six that disagreed with the game (#1826).
• The synthetic empires keep their own copy of the ship price list, and nothing in the project compared it against the real one — all six apex hulls could quietly drift apart. That copy is now checked against the source on every test run, in both directions (#1826).

1.47.16 — 2026-08-30

Fixed

The night-window lock no longer tells you that you moved it when you didn't.

Spacca-Mazza: "When changing time for night protection, and then clicking the undo button (nice addition btw), it will gray out the selector and say 'Moved — available again in 24h.' even though i didn't move it."

Undo works exactly as intended and is unchanged: it puts back your previous window, and it deliberately doesn't refund the 24h. Refunding it would let the window be re-aimed onto whoever just came online, taken back nine minutes later, and re-aimed again — which is the whole reason the lock exists. So the cooldown standing after an undo is correct.

What was wrong was the sentence. The selector could only see when the lock lifts, never what you did — a move and an undo look identical to it, and always will. It guessed, and after an undo it guessed wrong, so the game appeared to be scolding you for pressing undo.

It now states the rule instead of accusing you of anything:

One move every 24 hours — available again in 7h.

Which is true whether you kept the new hour or took it back. The same sentence appears before you spend the move, when the selector is still free, so the price reads the same way on both sides of paying it. The refusal you can hit from a second device was worded a third way — "can only be moved once a day" — and now matches too; "once a day" was its own small trap, since moving at 23:30 and finding it locked at 00:10 is a new calendar day but the same cooldown.

Closes #1390.

1.47.15 — 2026-08-30

Fixed

Browser notifications actually switch on now, and the Dispatches panel no longer offers a button that cannot work (#1899)

Since Dispatches arrived, the panel has offered an Enable button for browser notifications — attacks and announcements while the tab is closed. Tapping it never worked: it said "Asking…" for a moment and then "Notifications are not switched on for this server yet." That sentence was true. The server had no key pair for sending browser pushes, so the button could not succeed for anyone, on any browser.

Two things changed:

• The server can send now. The key pair is in place and the API reports Web Push enabled. Anyone who had already tapped Enable and granted permission is subscribed from here on with nothing further to do; everyone else can tap Enable and get the browser's own prompt, as intended.
• The panel asks before it offers. It now checks whether the server can send before the button can be pressed. The row is on screen from the moment the panel opens, with the button reading "Checking…" and disabled until the answer lands — then it becomes Enable, or, on a server that cannot send, the row says so and offers nothing to tap. If the check itself fails — the API being unreachable is not the same as the server having no keys — the offer is still made, and the tap finds out.

The changelog entry for v1.47.12 no longer uses the name it was correcting

That entry quoted a player's question verbatim, including the wrong name for the Orbital Cannon that #1906 had just removed from the changelog. The quote is now a paraphrase: same question, the game's own name for the weapon.

1.47.14 — 2026-08-30

Fixed

• A trillion no longer reads as "2586B" (#1896). Wherever the game shortens a number — the phone's resource pills, the leaderboard, the Power figure under your name — the scale stopped at billions. An empire holding 2,586,278,951,586 Ferrolite got "2586B": four digits and a unit nobody counts in, longer to read than the thing it was replacing. It now says 2.6T.

Changed

• The top band on desktop shortens big figures too (#1896). A late-season HUD printed seventeen characters per resource, three of them side by side, and a Power line reading 1,791,388,233,546 PP · S 1,791,120,471,854 · M 267,761,692 · Gate 1,791,120,471,853. None of those numbers was wrong, but the band had to take a second row to hold them, and that row comes out of the map. It now reads 2.6T, +595M/h · 46T and 1.8T PP · S 1.8T · M 268M · Gate 1.8T — the same abbreviation your phone has used since the pills shipped.
• Below a million nothing changed at all. Every digit is still printed, so a young empire still compares its stock against a building cost exactly as before.
• The exact figures are still there. Hovering a resource tile now opens with your precise stock and warehouse cap in full, in front of the production estimate that was already there; the Power tile's hover keeps its exact Structural / Military / Gate split.
• Energy and Dark Matter still print in full, on purpose. The energy tile compares two numbers against each other, and shortening both sides would let a thin margin look like a tie; Dark Matter is counted to the decimal.

1.47.13 — 2026-08-30

Fixed

The era readout at the top of the Research screen no longer looks like a button.

Spaccamazza put it exactly: "What is it? it doesn't open if I click it." Two complaints in one sentence, and both were fair. The era sat in a rounded pill with a coloured border — the shape used all over the game for things you press — and it has never had anything behind it, because it is a status readout. And a lone uppercase HEGEMONY doesn't say what kind of thing it is, which is why the first question was what it even was.

It now reads Era: Hegemony — no border, no pill, and the word is introduced. The era keeps its colour, which is what tells one era from the next at a glance, and it is now in the same case as the era named beside it ("36 Discipline lvls to Eschaton"), so the whole line reads as one sentence instead of a chip next to a caption.

Nothing was lost: the era, its colour and the progress toward the next one are all still there.

Closes #1473.

1.47.12 — 2026-08-30

Fixed

A changelog entry called the Orbital Cannon something the game never calls it — and then described what the Breachhead Monitor does to it incorrectly.

Commander Juan asked in chat-with-devs whether the Breachhead Monitor's entry — which said it halved the effect of a weapon he could find no reference to anywhere in the game — was a mistranslation of the Orbital Cannon, or something he had missed.

He was right, and there was a second problem underneath the one he found. The entry read "halves the Orbital Cannon's bite", which suggests you take half the pre-battle damage. That is not what the hull does. It halves the share of your fleet the Cannon can take — and the Breachhead's own 16,000 hull enlarges the fleet that share is measured against. Ten Bombers facing a Cannon lose 3,000 hull; bring a Breachhead and they lose 3,500. It pays off past roughly 16,000 hull of other ships, and below that it costs you.

So a commander who found the Cannon and then trusted the old wording would have taken more damage, not half. The entry now says what the Shipyard card has always said: it halves the share of your fleet the Orbital Cannon can take before the battle.

Nothing about the mechanic changed — it has always worked this way. Only the description was wrong.

Closes #1419.

1.47.11 — 2026-08-30

Changed

The Shield's card now tells you the Bubble recharges.

It always has — a sixth of its full pool, every round, since the pool was introduced. But the card only said it absorbs, and the single place that fact reached anyone was an Aegis doctrine's upgrade line: "gives back a quarter per round instead of a sixth." So the base rule was learnable only by reading about a doctrine you don't own and working backwards from the comparison. A commander asked outright whether it recharges — and, for the record, our own support bot had told a different commander that it does not.

The card now reads:

A deflector dome over the colony: the Bubble soaks up enemy fire before it touches the ships in orbit, and recharges a sixth of its full pool each round. Once broken it stays down for the rest of that battle; the next one opens with it whole again.

Three things there matter for planning a siege, and none of them was written down before. It is a sixth of the full pool, not of what's left — the difference decides whether a fleet can grind a Bubble down at all. Breaking it is permanent for that battle, so waiting it out mid-siege gains nothing. And the next attack meets a whole Bubble, so breaking one buys you that fight, not the next.

The Aegis upgrade line now says what it does before what it beats: "your Bubble recharges a quarter of its full pool each round, up from a sixth."

Closes #1538.

1.47.10 — 2026-08-30

Fixed

• The chat composer now says who you just tagged (#1024). Picking a commander from the @ dropdown used to leave the field full of machine text — @[Alex](p:cmrx…) — and close the one thing on screen that had shown the name and the guild tag. From there to pressing Enter, nothing said who the message was about to reach. A player put it plainly: "when i tag someone i see all that text, and i don't see their profile as usual, so i don't know who i am tagging for real" — written in a conversation that had two near-identical names in it and a wrong Alex already in the log.

  A strip above the field now names everyone the message is tagging, with the portrait, the name and the guild tag that tells two commanders apart — the same chip the message itself will show. Attacks and events you have quoted sit beside them, keeping the glyph and the colour the Galactic Register gives them. It reads the message you are writing, so deleting a tag takes its name off the strip with it, and a message that tags nobody gets no strip at all.

  Two commanders are allowed to share a name in this game, so the strip identifies them the only way that cannot be wrong: by the commander you actually picked from the list, never by the name they happen to carry. Where it was not you who picked — a message put back in the field after a send failed, say — it shows the name and no guild rather than guessing at one.

  It stays out of the way. On a phone it is a single row you can swipe sideways; on a computer it wraps to three rows and then scrolls, because the chat column has a fixed height and every row the strip takes is a row of conversation you stop seeing.

  Still to come: the field itself continues to show the raw text. Hiding it means rebuilding how the composer tracks what you type, keystroke by keystroke, and getting that wrong garbles the message rather than merely confusing you about it — so it is not riding along with this fix. What has changed is that you no longer have to read that text to know who is behind it.

1.47.9 — 2026-08-30

Changed

Telling you to research the Particle Accelerator now tells you where it is.

The Apex ship card said "The Apex lens appears once the Particle Accelerator is researched" and stopped there. The Accelerator is a Forge tier-6 one-shot behind Galactic Economy 17 and a prerequisite node — nowhere near where a mid-game commander is looking. One player went hunting and reported back: "i can't even find it in the Tech Tree." Reasonably: the instruction named a thing without saying where it lived.

Every screen that gates on the Accelerator now gives its address, in the same form the tech tree already uses for everything else:

The Apex lens appears once the Particle Accelerator is researched (Forge T6 · Galactic Economy 17 + prerequisites).

That covers the padlock on a locked Apex hull, the padlock on a Precursor building, the Dark Matter panel, both refusals the server sends when you try to buy an Apex without it — and the Dark Matter debt line, which is the one a commander without the node is guaranteed to see. The + prerequisites tail is there on purpose: without it, anyone who already has Galactic Economy 17 would read a requirement they have met and conclude the game was broken.

The address is read from the tech tree itself rather than written out by hand, so if the node ever moves, every one of those screens moves with it.

Closes #1431.

1.47.8 — 2026-08-30

Fixed

• A guild move made by the team now explains itself to the players it moves (#1114). When we move commanders between guilds — a server rebalance, say — each of them now gets personal mail about it. Before, the move was silent: the first sign was the guild page showing a different tag, which reads either as a bug or as somebody else being inside your account. One player had to ask in public chat why they were suddenly somewhere else.

  The mail names the guild you came from and the one you are in, and it is honest about what a guild move actually costs, because most of it is not obvious: your colonies, fleets, buildings and technology are exactly as you left them, but your production and research bonuses now come from your new guild's research instead of your old one's and may be higher or lower; the truces and pacts that were shielding you belonged to your old guild, so your former guildmates are no longer shielded from you nor you from them; you arrive as a Recruit whatever rank you held; and your guild contribution starts over, as it does for anyone changing guild. If the move empties your old guild, you are told it has been dissolved. If you led it, you are told who leads it now.

1.47.7 — 2026-08-30

Fixed

Rival empires stop refusing the siege they built the ship for (#1871)

A fortified colony could sit unattacked by an empire that was holding exactly the ship built to crack it. The rival worked out correctly that a Breachhead Monitor fleet breaks a bastion — and then never sent it.

The reason was in how those empires appraised their own fleet before committing to an attack. They judged a fleet's siege strength from an old list of hulls, written before the Breachhead Monitor existed: bombers counted against batteries, bombers and titans against a Fortress, and the Breachhead counted for nothing. So the hull the game rates highest against both kinds of defence — four times the damage, more than any other ship — was appraised as though it were a freighter, scoring no better than a fleet of cruisers with no siege ability at all, and worse than an equally expensive fleet of titans.

That appraisal is not idle: an empire refuses to launch when its own estimate of winning falls below what its temperament demands. Judging its best siege fleet as worthless was enough to talk it out of the attack.

Rival empires now read the same counter table the battle itself uses, so a Breachhead fleet is appraised as the siege fleet it is, and those attacks go in. Expect fortified colonies to draw more attention from ambitious neighbours than they did last week.

Nothing changes about how much damage any ship deals, what anything costs, or how a battle resolves. This is only about the decision to attack.

1.47.6 — 2026-08-30

Changed

An espionage report now names the spied commander's guild, not just their name. The header reads [4:17:3] · Vera [CLN], in the same [TAG] form the leaderboard, the season panel and the Contested Space board already use, tinted with the guild's colour.

Guild membership was already public everywhere else, so the report — the one thing you paid a probe and a Discipline for — was the surface that still sent you off to look the affiliation up by hand. That is worst exactly where identity is hardest to place: a probe through the Veil, or tracing the origin of an attack. The Observatory's attack-warning has shown the attacker's tag since #1387; the spy report now matches it.

The tag is shown at every espionage tier, deliberately. It is free information on half a dozen other screens, and a report that costs a probe while saying less than a free list would not be worth reading.

Commanders with no guild draw nothing at all — no empty brackets — and every report written before this change still opens exactly as it did.

Closes #1818.

1.47.5 — 2026-08-30

Fixed

Contested Space says which worlds you cannot take, and tells the fleet that finds out the hard way. One of the six worlds sometimes shows a grey occupied instead of garrisoned — somebody's colony was already standing there, so no garrison was ever placed on it and there is nothing to capture. That was the rule, and the only place it was written down was a changelog entry.

So the panel said "six worlds, each held by a garrison", a guild sent a real fleet at one of them, won the battle — and got no capture and no points an hour, with nothing anywhere explaining why.

Now the panel says it: the opening line admits that a world with a colony already on it never gets a garrison, the grey word carries the same explanation, and a won battle that could never have taken a world leaves a line in your activity feed saying so. The battle still counts — you keep the loot and your own season points; it is only the guild's capture that was never on offer.

The feed line appears once per world per season rather than after every raid. Those colonies are ordinary targets and get hit repeatedly, and a rule repeated after every victory stops being a rule and becomes noise.

1.47.4 — 2026-08-30

Fixed

• A market trade now tells you which side of it you were on (#1198). The galactic register drew every trade the same way — Farwatcher ⇄ Trader: 100 Fe for 50 Xe — and the two amounts always belonged to the seller, so nothing on the row said whether you had received the Ferrolite or paid for it. Fine as one of thirty rows scrolling past, where the ones above and below tell you how to read it. Not fine when it is the only trade on the screen and it is yours.

  And on your own screen, alone is exactly how you meet it: opening a day in the Calendar shows the public register raw, a trade quoted in chat opens as a single card, and an old purchase still reaches your Activity feed that way. Now, when you were one of the two parties, the card says it plainly — "You bought 100 Fe from Farwatcher for 50 Xe", or "You sold …". For everyone else the trade reads exactly as it did before: it is a public register, and it still says the same thing to everybody who was not in the deal.

  The amounts are the deal's terms, as they have always been on the register. What actually landed in your warehouse, after the market's cut, is on your own receipt in the Activity feed — that line names the fee and always has.

1.47.3 — 2026-08-30

Fixed

• Emplacements: the sheet now says all of them come back, not just the last one (#1859). Destroyed emplacements have always rebuilt themselves for free — one at a time, on a timer, and it keeps running while you are logged out. The sheet only ever mentioned the last one, right under a resource price, so it read as "you have to buy the rest back". Players were rebuilding by hand, or writing off defences that were already on their way home. The note under the Rebuild button now says that all of them return on their own, free, even while you are away, and that paying only brings them back now instead of later. The Batteries and the Orbital Fortress each state their own rate — they are not the same, and the sheet no longer implies they are.

1.47.2 — 2026-08-30

Fixed

• "Fill the hold" now completes your manifest instead of erasing it (#1019). Typing a number into one of the three cargo cells and then pressing fill the hold used to replace it — type 4.000 Ferrolite on a 10.000 hold and you got 6.000 back, with the other two cells rewritten too. The numbers were not random: the button rebuilt the whole manifest as a mirror of your launch colony's stock, so what you had typed was never read at all.

  A number you typed is now treated as what you meant. It survives the click exactly as written, it counts against the hold, and only what is left over is spread across the cells you left empty — still in proportion to your stock. That 4.000 Ferrolite now stays 4.000, and the hold fills up around it.

  Both fill buttons work this way, on transports, redeploys, pickups and colonisation loads alike, and their descriptions now say so. Three details worth knowing: a manifest that is already over your hold comes back untouched rather than quietly trimmed — the launch panel tells you it is too big, and the number stays yours to correct; if you have filled in all three cells there is nothing left to complete, so the button leaves your manifest alone; and if you never type anything, the button fills exactly as it always did.

Reported by Spaccamazza.

1.47.1 — 2026-08-30

Fixed

Empires now rate the Breachhead Monitor correctly against fortified colonies. The Breachhead is the game's dedicated siege hull — four times as effective against both a Fortress and Orbital Batteries, double what a Titan manages. The engine has always resolved it that way. The empires reading the board did not: their copy of that table still listed only the Bomber and the Titan, so a Breachhead fleet was appraised as if it had no siege advantage at all.

Three things follow from it, all visible from the outside. They no longer walk away from fortified colonies a Breachhead fleet can plainly crack. They staff an attack on one with the hull that actually cracks it. And a guildmate who answers a coalition call with Breachheads is recorded as bringing siege, where before the strongest siege contribution on the board was filed as a line ship.

Worth saying plainly: because they now accept targets they used to decline, the operations that drive their shipbuilding change too — not because any build rule changed, but because they are being asked about different targets.

Also

What an empire builds for a fortified colony is deliberately unchanged and tracked separately (#1848): two attempts at it were measured and reverted, because the obvious fix would have had them buy the cheapest siege hull instead of the best one, taking twelve times as long to assemble the same fleet.

1.47.0 — 2026-08-30

New

You can take back what you said in chat (#1801)

Until now a chat message was final. A typo, a line sent to the galactic channel instead of the guild one, something you regretted a second after sending it — it stayed there, and nobody but an administrator could remove it.

Now the ⋯ menu on your own messages offers Delete message, behind a confirmation that tells you in advance what will be left.

The row does not disappear. It keeps its place in the conversation and your name on it, and the text becomes "This message was deleted." Anyone who had replied to you keeps their reply, and the quote above it says the message was withdrawn instead of pointing at nothing — a message vanishing outright would have taken a piece of somebody else's message with it.

A few things worth knowing before you use it:

• The confirmation tells you the truth about the timing. People already reading the chat may still see your message until they reopen it. On a private whisper — the case where this matters most — that is the person you were talking to.
• If the message was pinned to the top of your guild channel, deleting it takes it off the top too, and it cannot be pinned back. The confirmation says so when it applies.
• You can still report a message someone deleted. A message withdrawn twenty seconds after it was posted is exactly the one worth reporting, so the ⋯ menu stays on other people's withdrawn messages for that.
• Deleting takes the words out of the chat, not out of the memory of whoever read them. Anyone who saw the message before you withdrew it — a person, or one of the empires that answered you — still knows what it said.

Withdrawn messages no longer light up an unread badge or a channel dot either: a badge that sends you somewhere with nothing to read is worse than no badge at all.

1.46.2 — 2026-08-30

Fixed

An empire can tell you whether the aid you sent it actually arrived. A player filled a guild-aid request — twice — and was then told in chat that nothing had landed, and that the amount they had sent was somebody else's number. Both were false. The delivery had worked perfectly; what failed was what the other empire knew when it was asked.

Its own request had simply disappeared from the facts it reads. A request that is partly filled showed what had come in; the moment it was fully filled it vanished, so the board went silent exactly when there was something worth saying. And the one surviving note of the favour carried no amounts at all — "someone sent you supplies" — so with another member's open request still showing its numbers, the only figure in view belonged to somebody else.

Now every favour is on the sheet. The current request is reported, and so is every earlier one that actually received something, with what arrived and who sent it — including a request that lapsed part-filled, which is the common case, because a single donor can only give a small share of their own stores and a large ask gets covered in pieces. Timer pushes count the same way: two guildmates who sped up a build are two guildmates, not nobody.

Also

Empires no longer report a countdown on a request that is already over, and no longer describe a request they withdrew themselves as having expired unanswered.

1.46.1 — 2026-08-30

Fixed

Two ships finally say what they are for (#1827)

The Praxis Interdictor's card read "Destroys interceptors". True, and not the reason anyone builds one. Its real job — while one of yours survives, the enemy's Rearguard can no longer take less than a set share of your fire, however deeply they screen it — was written nowhere in the game. A player worked that out only by asking in chat, having already written the hull off as worthless.

The Breachhead Monitor had the same gap. The ship reference showed its ×4 against Defense Batteries and the Orbital Fortress, but nothing anywhere said it halves the share of your fleet the Orbital Cannon can take before the battle — and the Shipyard's own detail sheet, which lists counters and nothing else, showed neither half of what the hull is for.

Both now carry a line describing the rule, in the ship reference and on the ship's detail sheet, next to the counter list rather than instead of it. The counters are still true and still shown.

The wording sticks to what the engine actually does in every battle, not to what you might get out of it: the Interdictor's line promises a floor, because on a defence with only two occupied rows the rear is already above that floor and the ship changes nothing there; the Breachhead's line promises a halved share, because the cap is a fraction of your fleet's own hull and the Monitor adds 16,000 of it. Four drafts that promised more than that were thrown out before this shipped.

1.46.0 — 2026-08-30

New

Sort the market by trade ratio. The board had a resource filter and nothing else, so offers sat in whatever order they were posted and finding the best-priced one for a pair meant reading every card — while a seller could stay near the top just by reposting.

A Sort row now sits under the filters. Recent is the order you already know and stays the default; Ratio puts the offers that return the most per unit paid at the top. It works together with the resource filters, which is how it is meant to be used: narrow to the pair you care about, then sort.

While Ratio is on, each card shows the number it was sorted by — 3.00× means the offer gives three units for every one it asks. It is a quantity ratio and the board says so: it does not compare what different resources are worth to each other, so a higher number across two different pairs is not always the better trade. Gift-priced offers stay below the divider where they already were, however good their ratio looks.

Also

Every filter chip now announces to a screen reader whether it is the active one, which only the new sort chips did before.

1.45.0 — 2026-08-30

New

A busy week no longer throws away the Wraps it earned (#1851)

The Weekly Chest opens at five Daily Wraps in a contract-week. Until now a week that ended short simply reset: three Wraps and then a couple of busy days lost you the chest and the three Wraps that got you most of the way there, and Monday started from zero.

Up to two unspent Wraps now travel into the next week.

• The bar has not moved. Five still opens the chest, so a carried week still asks for three fresh Wraps — the weekly rhythm is still what is being rewarded.
• A week that reaches the bar passes on only what is left over, never the five it spent. Six Wraps carry one, seven carry two, five carry nothing.
• The carry survives a whole week in which you claim nothing, and only a second silent week clears it. Take a week off and your two Wraps are still waiting.

The Contracts board tells you which part of the count you brought with you, and when your carry is down to its last week it says so with a countdown to the moment it runs out, instead of a rule.

Changed

• The Weekly Chest line now reads "N/5 Daily Wraps toward the chest" rather than "this week", because the number can include Wraps you carried in.
• The Activity feed's chrono-shard entry and the "chest not earned" message both describe the new rule.

1.44.0 — 2026-08-30

New

Dispatches — the game can now tell you something before it happens (#1854)

Until today everything Novarchy told you was something the engine had recorded about you: a battle resolved, a harvest landed, a season closed. There was no way for anyone to say the ordinary thing — "the event starts next Tuesday", "the server is down on Sunday morning", "your appeal was accepted".

There is now an envelope in the HUD, with a count of what you have not read. Open it and you get the bulletins addressed to you: news, a scheduled event, a maintenance window, or a message meant for you alone. Anything with a date says how far off it is — Starts in 7 days — 08/09/26, 00:00 — and counts down against the server's clock, not your device's.

A bulletin is marked read when you open it, and only then. There is a Mark all read for a backlog you do not want to work through one by one, and the read state lives on the server, so it follows you between your phone and your laptop.

Notifications in the browser

The game could already reach the iOS app. It can now reach a browser tab that is closed — which is most of you. There is an offer at the bottom of the Dispatches panel; it tells you what your browser is about to ask before it asks, because a refusal is permanent and you should know that going in. Say no and nothing else happens; the offer is there whenever you change your mind.

Notifications obey exactly the rules the phone ones already did: nothing while you are in the game, nothing inside your own night hours, and never more than four a day counting attack alarms. Tapping one takes you where it promised — a fleet alarm to Fleets, a battle to its report, a bulletin to the bulletin — and, if the game is already open, it takes you there without reloading, so your map view and your selected fleet survive the trip.

Also

Announcements can also be crossposted to the Discord channel, so news reaches people who are not logged in. Only broadcasts go out that way — a message addressed to one player or one guild never leaves the game.

1.43.10 — 2026-08-30

Fixed

• Opening the chat dock no longer throws the Colony tab away and rebuilds it (#1803). The tab lays itself out in one column or two depending on how much room it has, and opening the dock takes enough width to switch between them — on an ordinary screen, with the window never moving. Until now that switch destroyed the whole card and built a new one: the buildings list jumped hundreds of pixels down the page, a button you had selected with the keyboard was thrown away and the next Tab restarted from the top, and the planet sphere was torn down and drawn again. If you keep the dock open, that happened every single time you opened the tab. Now the tab keeps its place, keeps your keyboard position, and the planet is drawn once and stays.

  One thing to know if you play by keyboard: in the single-column layout Tab still reaches the resource panel before the advisor, even though the advisor is shown first. That ordering is the price of the fix and cannot be removed without bringing the old behaviour back — and it only affects you at all once Planetary Mandates are researched, since nothing else in that panel can be selected. Losing your place entirely, which is what happened before, was the worse of the two.

1.43.9 — 2026-08-30

Fixed

The chat no longer opens on an inbox with nothing in it (#1831)

You whisper someone, nobody answers yet, and you close the chat while still on your inbox. Later the channels fill up and the chat button lights with a number — "3 unread". You click it, and the chat opens on your inbox, which has nothing new, while the three messages it was counting sit behind the Galaxy pill.

It did not fix itself either: the channels are deliberately left unread while your whispers are on screen, so the number stayed at 3. Close, click again, same empty inbox — until you pressed the pill yourself or reloaded the page. Anyone using a screen reader had it worse, because the button announces a number and the pill only shows a dot.

Now the chat still opens where you left it, except when there is nothing in your inbox to come back to — and then it takes you to the channel that actually holds the unread messages, not just whichever one you happened to be on last. If a whisper is waiting for you, nothing changes: your inbox is still where the chat opens. If everything is quiet, nothing changes either.

The same thing happened on the phone, and there it was worse: the phone remembers your choice between sessions, so a single tap on Whispers could leave you landing on an empty inbox after every reload, indefinitely. It is fixed on both.

Two smaller cases are covered along with it: an inbox that is still loading, or that failed to load, no longer counts as an empty one — you stay on the screen that tells you so — and a channel you have muted is never the one the chat sends you to.

1.43.8 — 2026-08-30

Fixed

• The Proving Ground stops promising a title it never granted (#1696) — the "What the close pays" note told you that climbing past the crate ceiling "buys the title and a field with fewer runners". There is no title. The close hands the best strike of each bracket its crates and writes one line in that player's own diary; nothing marks a name on any board, and nothing ever did. The note now says what the close actually pays: past Armada the crate is the same size, so climbing higher never buys a bigger prize. If you were climbing past the ceiling for a title, you were climbing for something that did not exist — and the screen was the reason you thought it did.

• The joining rule stopped explaining a rule you never saw (#1696) — it read "being the only empire at your fleet size is not something you can fix, and it no longer costs you the crate". That "no longer" pointed at a version of the game retired before most players arrived, with no way to look it up. It now says the thing itself: a bracket pays once one empire is ranked in it, even if you are the only empire at your fleet size.

Changed

• The ally-garrison pitch says the recall rule in its own sentence (#1696) — on the guild board, the rule that you can call your fleet home unless a strike is already on final approach used to arrive buried in an aside inside a longer sentence. A player reported that block as hard to read; it is now four plain sentences instead of three tangled ones. Nothing about garrisons changed: your fleet still fights until the watch ends and then flies home, the ships stay yours and are never handed over, and they can still be destroyed while they are there.

1.43.7 — 2026-08-30

Fixed

• The raid join door keeps the guard that catches its own regression (#1844). Nothing changes on screen this time. The note under the Join button — the one that tells you a free strike covers a raid you ride alone — was correct, and it stays correct; what was wrong was the net around it. A check meant to stop one specific false sentence from ever coming back had been replaced by a looser one that accepted that sentence. Rewriting the note to say a guildmate joining would leave you "billed in full" passed every remaining check, while being untrue for anyone already at the per-blow cap: for them a join changes nothing, the free strike still spends, and the bill is still zero. The old check is back, so that sentence cannot return unnoticed.

Also

• Four notes-to-ourselves in the same code had gone out of date and are now accurate again, including one that argued against the very sentence printed a few lines above it. Two more checks were added around the same note: one that catches the price re-appearing anywhere else on the panel, and one that fires if a shared piece of wording elsewhere in the game changes in a way that would quietly blind the first.

1.43.6 — 2026-08-30

Fixed

The Fleet Advisor no longer sends titans to crack a Fortress (#1847)

Ask the Advisor what to bring against a colony defended by a Fortress and it handed you titans, with almost no Breachhead Monitors — the hull built for exactly that job, and the one the battle engine resolves at ×4 against a Fortress, double the titan's ×2.

Worse, fortifying the target made the advice worse. On a fixed target and a fixed hangar, the Advisor recommended 176 titans and 10 Breachheads against an unfortified colony — and against the same colony with a Fortress on it, 210 titans and 6 Breachheads. Putting up the very defence the Breachhead exists to break made the Advisor suggest fewer of them.

It now recommends 191 titans and 22 Breachheads against that Fortress. A fortified target pulls the mix toward the siege hull, which is the direction it should always have pulled. The same now happens for Batteries, and for a colony holding both.

The Breachhead still does not outrank the titan outright, and it is not meant to: it carries one weak ship counter and costs more, which is what it pays for its siege teeth. What was broken was the direction, not the ranking.

Bombers also gain against a Fortress, going from 1.4 to 1.58 in the Advisor's weighting. They had been given the titan's number there; they now get their own, which the engine has always rated higher. Against a pirate beacon — which has Batteries but never a Fortress — the bomber's weight edges very slightly down.

None of this changes how a battle actually resolves. The engine's numbers are untouched; what changed is that the Advisor now reads them instead of keeping its own out-of-date copy of three.

1.43.5 — 2026-08-30

Fixed

• A raid you ride alone no longer warns about a bounty your free strike already pays (#1829). If someone attacks you, you are owed a free return strike, and the server waives the whole charge when you take it on a raid you ride by yourself. The raid panel did not know that: it printed "⚠ Bounty: joining this raid puts about N Dark Matter on your head" anyway — while the planet sheet for the same target, one tap away, correctly said the strike was free. Two screens, two prices, for the same blow, and the wrong one was the screen with the button on it. The raid board now reads the same ledger the planet sheet does.
• And it stays narrow, because the waiver is narrow (#1829). The credit only covers a bill that is entirely your own. Ride the raid alone and it is free; share it with guildmates and you pay your slice of the pack price in full, exactly as before. So the note only withdraws the figure where you really are the only fleet on the raid, and it says plainly that a guildmate joining can end that — because once your fleet is in a raid, it cannot be recalled.

Also

• Where the free strike does not clearly apply — a raid whose opener has left it empty, or a bill already sitting at the per-blow maximum — the panel keeps quoting the price. In those cases the server may still waive it, so the warning can be pessimistic; that is deliberate. This door is allowed to warn you about a bill you will not pay, and never to stay quiet about one you will.

1.43.4 — 2026-08-30

Fixed

• A guildmate raided at two colonies at once now gets a row for each, and you can defend both (#1836). The "Allies under attack" board listed one row per ally, but an attack lands on a colony — so a mate hit in two places showed up as a single row that took its coordinate from the nearer raid and its fleet count from both put together. It read "6:16:8 · 15 fleets inbound" when only eight were heading there, and the detail line just below it, which describes the nearer raid correctly, disagreed with the header by seven.
• And the second raid was not merely mislabelled — it was missing (#1836). Because a garrison is sent to a specific planet, and only the nearer one was on the board, there was no way for anyone in the guild to send defence to the other colony at all. Both now appear, each with its own planet, its own fleet count, its own reading of what is standing there, and its own Send defence button.

Also

• A colony that has already been hit keeps its place on the board and simply loses its button, instead of vanishing while you are reading — the same rule the ally rows already followed.

1.43.3 — 2026-08-30

Fixed

• Your Batteries and your Orbital Fortress now say how they fight (#1832). Both carry real combat multipliers — the Batteries punish interceptors and corvettes, the Fortress punishes titans, battleships and cruisers, and each is cracked far faster by some hulls than others — and none of it appeared anywhere you could read it. The two defences are not ships, so they were missing from the shipyard's ship-vs-ship matrix, and their own sheets said nothing. Each sheet now carries an Under fire block: what this emplacement hits harder, and what hits it harder, heaviest first. The figures are read straight from the combat engine, so they cannot drift away from what actually happens in a battle.
• The Breachhead Monitor stops hiding half its purpose (#1832). The hull built for cracking planetary defences showed its ×4 against the Orbital Fortress and said nothing at all about the Batteries, where it is ×4 as well. Its entry in the ship reference is now complete.

Also

• Nothing about combat itself changed: every multiplier in this entry was already being applied in battle. What changed is that you can now find out about them before committing a fleet, instead of after.

1.43.2 — 2026-08-30

Fixed

• The Build button no longer hides under a ship sheet that has stopped scrolling (#1811). On a narrower desktop window the Shipyard's ship sheet pins itself beside the hull grid, and once the grid is the taller of the two the sheet is taller than the window it is pinned in — so scrolling the page stopped moving it, and the quantity block and the Build button sat below the bottom edge of the screen. The only way to reach them was to scroll all the way past the bottom of the hull grid, or resize the window. The sheet is now bounded to the screen and scrolls on its own, so its Build button is always one scroll away.
• The same fix reaches the Research tab (#1811). The tech tree's node panel and the Disciplines panel are the same kind of column and were stranding their Queue button in the same way. Both now behave like the ship sheet.

Also

• The panel only gets its own scrollbar where it actually needs one. On a wide screen, and on any window where the panel has wrapped below the list — which is what happens at 1280 wide with the chat dock open — nothing changes: those panels were never stuck, and adding a second scrollbar in front of the Build button would have made them worse. The game measures the panel and its column and decides, rather than guessing from the window size.

1.43.1 — 2026-08-30

Fixed

The chat badge stops lying about "nothing new" (#1809)

The unread badge over the chat — the dot on the desktop launcher, the number on the phone's bottom nav — could stay lit after you opened the chat and found nothing new. It survived a reload, and it came back.

The badge counts two things: your public channels, and your whispers. But opening the chat always dropped you on a public channel, and a whisper can never appear there — so a badge lit purely by a private message sent you to the one screen where the thing it was counting could not be. It stayed lit until you happened to find the Whispers tab yourself.

Now the chat opens where the count is pointing. If the only thing waiting is a whisper, the chat opens on your whispers — on the desktop dock and on the phone alike. If there is unread in a channel too, nothing changes: that is what you came to read, and the Whispers tab keeps its own counter for the rest.

1.43.0 — 2026-08-30

New

• The raid door now tells you what joining costs your head (#1815). Attacking someone far below your Power puts a bounty on you, and the planet sheet has warned you about it for a long time: open a target, and the launch form estimates the Dark Matter the strike will cost you before you commit. Joining a guild raid never did. It explained the rule — that a raid is judged on the whole pack's combined Power — but it could not tell you whether this pack, on this target, was over the line.

  That gap fell exactly where a raid differs from a duel. The band is measured on the pack's combined Power when the raid lands, so raiders who are each perfectly safe attacking alone can be far over the line together — and the planet sheet shows every one of them a green "fair target" while they are one click from a bounty. You found out from the battle report.

  Now, under the Join button, when the pack is already over the target's band: "⚠ Bounty: joining this raid puts about 38 Dark Matter on your head. The bill is settled at impact, and every fleet that joins after you raises it."

  It never blocks the button — the bounty is a price, not a wall, and that was the whole point of replacing the old hard block with a bill. It stays quiet when there is nothing to say: a pack inside the band, a target already wanted (hunting a wanted head is a claim, not an abuse), a dormant empire, a faction bastion or relic site, or a target your sensors cannot see through its Veil. And it keeps showing after you join, because the pack keeps growing behind you: guildmates joining later raise your share of a raid you can no longer leave.

  The figure is an estimate of the raid as it stands right now, and the note says which way it moves: more fleets joining means a bigger bill for everyone already aboard. It can also settle lower than quoted — the server does not charge for a raider who could not see the target, and it forgives a strike you were owed — so it is a figure to decide on, not an invoice.

1.42.2 — 2026-08-30

Fixed

• The Aegis Bastion and the Continuum Hauler can finally fight in the Campaign (#1572). You could spend a maxed Doctrine on a Bastion — 1,600 shield, the heaviest in the game — and it would never once appear in a tower climb, with nothing on screen saying why. The strike group was picked by a rule that asked whether a hull had guns, written back when every gunless hull in the game was a freighter or a probe. The Dark Matter tier ended that, and the rule had not noticed. Both apex support hulls now board for anyone who has unlocked them.

  Freighters, Spy Probes, Colony Ships and Colonial Arks still stay home, for the reason they always did: a line of them soaking enemy fire turns wins into draws.

Changed

• The Deployment header on the engagement sheet no longer claims your strike group is "one of each unlocked armed hull" — which stopped being true the day the Bastion existed. It now tells you how many of each hull type you field, and the placement list below it shows you which ones.

Also

• Check your Campaign deployment once. The two new hulls join the rear row by default, so if you had deliberately emptied the rearguard to push everything forward, the printed row shares will have moved. Drag them where you want them; the deployment is still your one order.

1.42.1 — 2026-08-30

Fixed

• Dates and clocks now read on your own calendar (#875). An Italian player was told his shelter lasted "until 8/20, 2:11 AM" — the month where he expects the day, on a clock face he does not use. The time itself was already right; only the format was ours. Every date and clock in the game now follows your browser: 20/08 and 14:11 in Italy, 8/20 and 2:11 AM in the United States, for the same moment. Spy shelters, chat timestamps, the season calendar, guild treasury and directives, the leaderboard week, battle reports, contracts and the anchorage all moved together. Numbers are unchanged — 1,250,000 still groups the same way for everyone.

• A closing deadline no longer reads as one that already passed (#875). Italian shortens agosto to "ago", so a Directive that closed on 20 August rendered as "closes 20 ago" — which reads as "20 days ago" in an English interface. The Coming-up strip had the same problem, putting a future date between two countdowns. Dates that sit inside a sentence or beside a countdown are numeric now.

• The calendar's weekday columns are in your language (#875). The month heading was localized while the seven column headings underneath stayed English, so the same panel named the same weekday twice in two languages. They now use each language's own abbreviation — one letter per column in Arabic, T2 to T7 and CN in Vietnamese, 月火水 in Japanese — and fall back to the shortest form that still tells the seven columns apart. English is unchanged.

Also

• The changelog badge and this entry cover one issue, but the work took four rounds of review: the first version of the fix could have had its locale quietly pinned back to American formatting with the whole test suite still passing. That hole, and three more like it, are closed and now have tests that fail if anyone reopens them.

1.42.0 — 2026-08-30

New

A colony short of power now tells you what to do about it (#1324)

Until now, every screen that reported an energy deficit told you exactly how bad it was and nothing else. The energy tile said production was down 26%. The phone sheet said every building was slowing by 26%. A building card said the colony was running at 74%. None of them named a single thing you could go and do.

For most colonies you could work it out by scrolling the Base tab. Between Command Center 4 and 10 you could not: both generators are capped at your Command Center's own level, and a second Site of either needs Command Center 7. A commander who had built everything had already scrolled that list to the end with every button greyed out — and the one remedy left was in the research tree, on a screen the message gave them no reason to open.

All three surfaces now name the nearest lever you can actually pull, and only levers that genuinely help:

• the Fusion Reactor or the Orbital Panels when either has room, named with its Site number so you open the right card;
• a new generator Site when your Command Center admits one;
• the Superconducting Grid, with its address in the tree, when you can research it — and the Cybernetics Discipline instead when that is what is holding it back;
• the Command Center itself, but only at a level that actually opens a new generator Site;
• switching the Particle Accelerator off, when there is nothing left to build.

Anything already under construction, or already in your research queue, is skipped: the deficit is loudest while you are building the fix, which is exactly when being told to build it again is least useful. And when a colony genuinely has nothing left to buy, the screens say nothing rather than naming a building you already hold at level 30.

Nothing appears when you are in surplus, and a building card never tells you that its own remedy is itself.

1.41.8 — 2026-08-30

Fixed

The research detail panel no longer hides behind the top bar (#1780)

Scrolling down to one of the deeper techs used to push its detail box — the description, the cost, and the button that actually queues the research — up behind the top bar, or off the screen entirely. On a 1920×1080 monitor the bottom of the tech tree was effectively unusable: you could see the tech, you just could not read what it cost or start it.

The panel now stops below the top bar instead of behind it, so its cost and its queue button stay readable for every tech in the tree, however far down you scroll.

The bar is not a fixed height — it grows a row for each alert it has to show you — so the panel now measures where the bar actually ends rather than guessing. Under attack, with the bar at its tallest, the panel used to be almost entirely hidden; that is the case that improves most.

The same panel elsewhere

The Disciplines view, which is where the Research tab opens, had exactly the same problem and is fixed with it. The Shipyard's ship sheet takes the same correction, though at wide window sizes that panel already sat in the page rather than pinning, so most players will see no change there.

1.41.7 — 2026-08-30

Fixed

• A battle notice now says which colony it was (#1403). The alert that pops up when one of your battles resolves read Defeat · vs Vex · 3:12:7. If you hold six colonies, that told you a battle happened but not which of your worlds had just lost its hangar — you had to open Activity or the map to find out. It now names the place and keeps the coordinate: Defeat · vs Vex · Itz Iceee (3:12:7). The "Recent battles" strip on the Missions tab reads the same way, matching the Activity feed.

  The coordinate stays for a reason: the name shown is always the defending colony's, and when you are the attacker those names repeat — every pirate beacon is called "Pirate Beacon" and an un-renamed colony is "Prime Colony". Two raids on two different beacons would otherwise have produced two identical notices. On a narrow phone the colony name shortens with an ellipsis before the coordinate does, so the notice can always tell two targets apart.

  Battles fought before colony names were recorded on the event still read exactly as they did.

1.41.6 — 2026-08-29

Fixed

• A colony rename in progress survives the chat dock (#1786). Renaming a colony and then opening or closing the desktop chat panel used to throw away what you had typed — and the server's reply with it, if the name had just been refused. The field closed, the old name came back, and nothing said why. Now the field stays open across that click, with your draft still in it and the refusal still beside it. Leaving still leaves: tapping away, Escape and Cancel abandon the edit exactly as before, and so does going to type in the chat.

  There were two causes, not one. Opening the dock narrows the Colony tab enough to switch it from two columns to one, and that rearrangement was rebuilding the panel the name field lives in, discarding everything in it. But the draft was in fact already gone a moment earlier: pressing the chat button moved the keyboard cursor out of the field, which the game reads as "you have walked away" and uses to cancel the edit. Both are fixed, and neither fix works without the other — the second one was found while reviewing the first.

1.41.5 — 2026-08-29

Fixed

The war-tension flood in your Activity feed (#1728). A crisis between two alliances that kept being fought over wrote you the same row again and again — six identical lines in four hours, each one saying the tension had risen to 100/100 after a battle. Every line was true, and read down a column they looked like a war escalating minute by minute.

Now a repeated line is drawn twice at most: once for the most recent flare-up, once for the first one of that day. The first tells you whether it is happening right now, the last tells you how long it has been going on — and everything identical in between is left out. Rows about a different alliance, on a different day, or reporting a different level of tension are never merged: if the crisis climbed from 40 to 55 to 70, you still see all three.

Nothing is summarised and no number is invented. The lines you see are the real ones, worded exactly as before — there is simply no longer a wall of them.

Also

The Combat and Mine chips now count the rows you can actually see, instead of counting the ones that were sent.

1.41.4 — 2026-08-29

Fixed

Debris fields now say what they are actually worth to you (#1270)

Send 2,000 freighters at a field the game shows as 300,000 and you come home with 60,000. Nothing was broken — you only ever recover a share of a field: 30% with no Reclaim levels, and only the very top of the tree reaches 80%. Nothing on screen said so, so the number you sized a fleet on was one the game was never going to give you. Researching cargo capacity changed nothing either, because capacity was never the limit.

Every place that prints a field now prints, right beside it, the share you can actually recover:

• the Recent-contacts chip — ♻ ≈ 12,345 (recoverable ≈ 3,703) — and its tooltip;
• the selected-target line;
• the fleet-composing step on a phone, which is where you actually size the hold;
• the battle report's "Field left by this battle", which is usually where you first decide there is something out there worth collecting.

The field figure stays, because it is still what is in orbit and what a better Reclaim rate is measured against. The Recent row's legend now also spells out that these figures are already decayed to the current moment, and that fields halve every 24 hours — the original report was both effects at once, your base rate times a few hours of decay.

The Fleet Advisor already said this in words, and it now agrees with the panels to the tonne. It used to round where the panels floored, so the same field could be announced as 3,704 in one column and 3,703 in the next.

Two smaller things went out with it. A field too small to be worth anything at your rate now says so plainly, instead of leaving the Advisor blank and quietly clearing the ships you had picked. And the Advisor no longer quotes a stale Reclaim rate for a moment after opening, when your best Recycler sits on a colony other than the one you are looking at.

1.41.3 — 2026-08-29

Fixed

A building no longer claims it is upgrading to a level it cannot reach (#1055).

A player photographed his Command Center standing at level 8 with the badge above it reading «8 → 11», and asked whether that was right. It was not: his own colony held an order to 9, and the game cannot produce any other pair — an order always targets exactly one level above the building it sits on, and a second order on the same building is refused.

Now every place that shows where a build order is heading shows it only when it agrees with the level of the building it belongs to. An order that says otherwise is still shown as running — the countdown, the gold pill and the scaffolding are all still there — but the number that contradicts the building is not drawn.

The report was about the badge on the base map. Ten more places print the same thing, all within two taps of it, and all of them are fixed too: the Colony tab's building row and its OPERATIONS list, the Base tab's build menu and the card you get by hovering a building, the Overview's per-colony order row, the Building Sheet, the Operations panel pinned over the base map, the phone's Operations rail and its sheet, and the Speed-up lane you read just before spending a Chrono shard.

Two of them were worse than a wrong number. The phone's Operations sheet worked out the current level by subtracting one from the destination, so a bad order made it announce a level the colony had never reached and draw the building at a stage it had never been built to; it now reads the real level. And the rail chip's destination is the only description a screen reader gets for that chip, so it was announcing the wrong one out loud.

One thing is deliberately left: the guild help board still shows the raw destination for a request. It is drawn from the request itself, and a guildmate looking at somebody else's colony has no way to check it. That is noted in the code and on the pull request.

1.41.2 — 2026-08-29

Fixed

• The Colony tab no longer squeezes itself when the chat dock is open (#1761). On a desktop window around 1200px wide with the chat dock open, the Colony tab still laid itself out as if it had the whole window: it drew two columns into a space that could not hold them, squeezed the building list into a 298px strip, and painted the «Queue 1/1 — full» pill over the tail of the «BUILDINGS» label. The tab now measures the space it actually has instead of asking the window, so the header keeps its label, its queue pill and its filter chips on one line, and the building list gets its full width back.

Changed

• With the chat dock open, the Colony tab drops its side rail sooner — and that is the point. Because the tab now judges by the room it has rather than by the window, keeping the chat pinned open on a 1440px screen gives the colony one full-width column instead of a rail plus a cramped strip that only ever fitted one building row per line. Close the chat and the two-column layout comes back exactly as before. Nothing changes on phones, and nothing changes on a wide screen with the chat closed.

1.41.1 — 2026-08-29

Fixed

• The Overview tab now shows the Planetary Shield you actually built (#1043). The Defenses line on the Overview used the word «Shield» for the temporary protection timer — the countdown you get after a raid or buy outright — and never mentioned the Planetary Shield building at all. So a commander holding the building on every colony opened the tab, found no trace of it, and reasonably concluded their shields were off. That is exactly how it was reported.

  Each colony's Defenses reading now says what it means. Shield pool is the building: the HP its bubble absorbs before your ships take a hit, summed across every Shield Site you own and reduced by the colony's energy shortfall, which is precisely the figure combat uses. A colony with no power still names its Shield, at 0 HP and in red, because that is the true and alarming reading rather than silence. Protection is the temporary clock, under a name that is nobody else's, and it still appears only while it is actually running.

  The reading also gets the room it needs. It was previously one line inside a fixed-width column, so on a well-developed colony the tail — the pool figure itself — was cut off with no way to reach it, on desktop and on tablets alike. The pool and the protection clock now sit on lines of their own, which also means a running countdown can no longer make the row grow and shrink as the seconds tick.

1.41.0 — 2026-08-29

New

• The Market now tells you how much it can still deliver you today (#1574). There has always been a daily limit on what the market can bring into your warehouse — per resource, resetting at midnight UTC — but the only way to find it was to run into it: a greyed-out offer, or an error after you pressed Accept. The Market header now carries a second line under YOUR STOCK showing what is left of that allowance for each resource, and how long until it reopens. A resource you have spent out for the day shows a plain 0 rather than vanishing from the line, because that is the one you are blocked on. Hover a figure and it tells you how much of the day's allowance is already gone.

Nothing about the limit itself has changed — only whether you can see it before you trade.

1.40.23 — 2026-08-29

Fixed

Writing "we lost #7" no longer opens the tag picker over your sentence (#1760)

Type an ordinary message with a hash and a number in it — "we lost #7 today", "down to #3 in the league" — and the chat composer used to open its event and attack picker on the number. Nothing in the Registry is named after a bare digit, so the dropdown sat over the composer saying "No event or attack found" and stayed there until the next keystroke. Players using a screen reader had it read out to them: a verdict about a search they never made, while they were simply writing a sentence.

Now the sentence types like a sentence. The picker still opens on everything the Registry can actually be searched by — a word from an event's description, in any alphabet, or a coordinate as soon as the colon is there — and the bare #, which asks what happened lately, is unchanged. Only digits on their own are read as prose, and prose opens nothing.

One thing you can no longer do: #7 used to search for the number 7 anywhere in an event's description. Ask for it as #7: instead, which finds the sector rather than every 7 on the board.

1.40.22 — 2026-08-29

Fixed

The tag dropdown no longer loses the row you arrowed to (#1759)

Open the @ or # picker in the chat composer, press the down arrow once, and wait a moment. The highlight moved to the second name — and then, about two tenths of a second later, jumped back to the first one. Press Enter after that pause and the composer tagged the first commander in the list, not the one you had picked. Arrowing down a long list and stopping to read meant losing your place every time.

If you use a screen reader the same fault was audible: the arrow announced the second row, then the dropdown said "Searching…", then the reader announced the first row all over again — three things spoken, and a lost position, for one keystroke.

The arrow now moves the highlight and leaves it there. It also stops asking the server for the list again, which it was doing on every single arrow press.

Two smaller things in the same corner of the screen:

• When a name search fails on a bad connection, the dropdown says "Could not search right now. Try again in a moment." Tapping the field retries it — which matters most on a phone, where there are no arrow keys and that tap is the only nudge you have.
• Tagging someone when the same name appears twice in what you have typed now replaces the one your cursor is actually in, instead of the other one.

None of this was new: it had all been there since the tag picker itself.

1.40.21 — 2026-08-29

Fixed

• The "your launch pad is gone" warning now reaches the players who most needed it (#1217). When the colony you were launching a mission from disappears while the form is open — razed out from under you — the launch pad falls back to your mother colony and the fleet you had picked by hand is thrown away and picked again from scratch. A line on screen has always said so. It was in the one place it could not say it to everyone.

  If you had exactly two colonies and lost the one you were launching from, you never saw it: the «Launch from» row only exists when you own more than one colony, so dropping to one took the whole row away — and the warning with it. Your fleet was still silently re-picked, with a green Send button over ships you never chose, departing from a planet you had not aimed at, and nothing on screen to tell you.

  And for everyone else the line sat above the departure chips while a tap on a chip dismissed it, so re-aiming your launch pad yanked the row upward under your own finger — around 60–70 pixels, more on a phone where the sentence runs to three or four lines.

  The warning now appears whatever your colony count, and sits below the chips, so clearing it can never move one under you. With no chips left to tap, sending the mission clears it instead — which is what it was asking you to do.

1.40.20 — 2026-08-29

Fixed

• The Kepler Observatory stops reading out lottery numbers (#1432). The galactic bulletin printed empire scores as unbroken digit strings — "1133115897 points" — which is unreadable, and is also something no human commentator would ever type. Scores now carry thousands separators: "1,133,115,897 points".

• And so does everything else the agents quote. The same raw figures were reaching every prompt an empire speaks from, not just the bulletin: an agent's own rank and the podium it names in chat, the runaway-leader call, the rank gaps, the guild treasury and officer budget, Council tier gaps, season and event points, member contributions, research prices, the Eschaton contribution, aid requests on the guild board, guild XP, your colony's stockpiles and your fleet counts. Asked how much ferrolite it was sitting on, an agent used to answer "549755". It now answers "549,755", which is what your own colony panel has always said.

Also

• Two smaller things a player would only notice as strangeness. An officer told its own weekly grant cap in one format and asked for it back in another could hand the guild a figure the server then read as zero, so a member who had just been razed got nothing and nothing said why; that round-trip is now understood either way. And the guard that stops an agent inventing fleet counts about a real player now reads a separated number whole, so "1,002 of Kane's" is caught as the claim of a thousand ships it is, rather than as a claim of two.

1.40.19 — 2026-08-29

Fixed

• Two building badges no longer merge into one wrong number (#1430). On the base map, a building's level badge could sit against a neighbour's gold upgrade badge with no gap between them, so the pair read as a single number that neither building has — a "19" beside a "11 → 12", seen as one badge saying "19 11". The screen already knew how to drop a badge another one was painting over, but it only counted a badge as covered when the digits themselves were hidden, so two badges could share one dark pill and still both be drawn. Now a pair that runs together costs the rear building its badge, and the number left on screen is the one that belongs to the building under it. That level is still one tap away in the building sheet, and on desktop in the hover tooltip. Badges that keep clear of each other are untouched.

1.40.18 — 2026-08-29

Fixed

• Agents now coalesce against the empire that is actually running away (#1746). When an agent announced in the global channel that someone was "running away with the leaderboard", it read the top of the board — which is ranked on power — and then quoted a lead measured in points. Whenever those two orders disagreed it named the wrong empire, and the same name was handed to that agent's planning, so the fleets went after it too. The reverse also happened and was quieter: a genuine points runaway sitting lower on the board produced a negative lead, fell under the threshold, and was never mentioned at all — nobody was ever rallied against it. The runaway is now the empire genuinely leading on points, the lead is measured against the runner-up on points, and the announcement says "running away on points" instead of claiming a position on a board that says otherwise. The rules that stop an agent proposing a coalition against itself, or against a guildmate it could never attack, follow the empire that actually gets named.

• The advisors that plan an agent's next move stopped reading a power ranking as a points ranking (#1746). The empire-points table those plans are written from was the board in its own order, so its point totals did not descend, and it was cut off after ten rows by that same wrong measure — on a full server the empire just named as running away could be missing from the table entirely. The table is now ordered on points, every row still shows the position that empire really holds on the public board, and an agent outside the top ten keeps its own row and is told where it truly stands.

Nothing changes for how you play: no building, ship, cost or battle rule moved. What changed is that the empires run by the game now say true things about the standings, and act on them.

1.40.17 — 2026-08-29

Fixed

• The queue pill no longer covers the BUILDINGS label on a phone (#1140). In the Colony tab, the «Queue 1/1 — full» pill was painted over the tail of the «BUILDINGS» heading, with its own text broken across two lines inside the pill. The heading, the pill and the four family filters were all being asked to share one line, and on a phone there was not room for them. The heading and the pill now take a line of their own, above the filters. The pill stays where it was — on a phone it is the only thing that tells you why every upgrade arrow below is greyed out.

Changed

• Rotated phones keep the single-line heading. Turn the phone sideways and the screen is suddenly wide but very short, so a second line of heading would cost a slice of the building list for nothing. The heading only splits in two when the screen is genuinely narrow — so a phone held upright gets the fix, and a phone held sideways keeps its list.

Also: the filter row now uses the full width under the heading on a phone, so more of the filters are in view before you have to drag them.

Nothing changes on desktop.

1.40.16 — 2026-08-29

Fixed

• Guild standings now read the same in chat as they do in the Guild panel (#1740). Agents were being handed their contribution rank as "83th percentile in the guild" — the wrong ordinal ending for every rank finishing in 1, 2 or 3, and a phrase that put the speaker inside their own denominator. The rank counts the guildmates behind you, not the whole guild, so an officer at the top of a three-member guild was described as ahead of 100% of a guild they are one third of. Agents now say the standing exactly as your Guild panel prints it: "ahead of 82% of your guildmates", "top contributor", "no guildmates behind you yet", or "no contribution yet".

• A founder who has given nothing is no longer told they lead the table. A guild of one reports a rank of 100 by construction, so an agent alone in a guild it had just founded read as "among the top contributors" and would repeat that in chat. What you have actually contributed now decides.

• Agents stopped judging how much you gave from where you rank. The rank only counts who is below you, so it can never say who is above. Guildmates sitting on identical totals — common, since agents donate on a fixed schedule — were each told they contribute less than their guildmates while all of them were level, and the sentence contradicted its own first half. The same false reading reached the guild rescue: members exactly at the guild's median were flagged to their leader as "well below the guild median, may be struggling", and the leader said so in chat. Both now state only what the number supports.

• A missing figure is no longer reported as a zero. Where a member's contribution failed to reach an agent, the roster and the rescue prompt printed "contribution 0" about someone who had been donating every tick. They now say the figure is unknown, which is what is true.

1.40.15 — 2026-08-29

Fixed

The chat composer's tag dropdown can now be used with a screen reader (#1726).

Typing @ to tag a commander, or # to tag an event or an attack, opens a list of suggestions. Until now, everything that happened next happened in silence for anyone playing with a screen reader. You typed a name, the search ran, and it answered — with matches, with "nobody", or with an error — and none of it was spoken. Swiping forward landed you on a list announced as holding zero suggestions: exactly the information you had before you typed anything.

Now the composer says what it is doing. Every outcome is announced, including the successful one: how many commanders (or events and attacks) were found, that a search is still running, that nobody matched, or that the search could not run at all. The text field itself now says that typing there produces a list, whether that list is currently open, and which suggestion is highlighted — so the arrow keys read each one out as you move through them, without ever taking the cursor out of the message you are writing.

And the count no longer promises more than it found. The suggestion list stops at ten, so a common search reaches that limit — searching @a on a galaxy of eight hundred commanders is answered with ten of them. Saying "10 commanders found" would read as the game looked, and there are ten, which is how you conclude that someone real is not in the game and stop looking. When the list was cut short it now says "found so far".

Nothing changes on screen: the same words appear in the same place. What changed is that they can now be heard.

1.40.14 — 2026-08-29

Fixed

• "Queue your first two ships" no longer comes back when you launch your fleet (#777) — the First Steps panel worked out each step from what was standing in the colony right then, and ships out on a mission are not in the colony's harbour. So the moment you sent your fleet anywhere, the ships step un-ticked itself and the panel reappeared in every tab, with no way to close it, until the fleet came home. Then it happened again on the next launch. A step you have already completed now stays completed: the panel stays gone, your progress count does not fall back, and a reward you earned can be collected while the hulls are still out — before, you had to wait for them to land.

• The Contracts board no longer closes mid-mission (#777) — the same bug wearing a different hat, and nobody had connected the two. The board is gated on your early steps being done, the first two ships among them, so newer commanders watched the whole board, its notification dot and the "Find your guild" card vanish the moment they launched, and reappear when the fleet landed. They stay put now.

1.40.13 — 2026-08-29

Fixed

• Your guild has a row on the Contested Space board, even outside the top five (#1562). The board listed the five best guilds and stopped there, and the cyan "this is you" highlight only existed inside those five rows. If your guild sat sixth or lower you had no row, no rank and no points anywhere on the Season tab — the scoreboard above it ranks players, not guilds — so the season read like something you were not part of. Your guild was in the board's data the whole time; it simply was not drawn.

  Your own row now hooks onto the bottom of the list, under a dividing line so it can never be mistaken for sixth place, carrying your rank, your points and the words "your guild". Your row is marked in words wherever it sits, inside the five as well as below the line, so finding it no longer depends on picking one colour out of another.

  A guild that has not captured anything yet gets a row too, reading "no worlds held yet" — because being on the board at zero while five other guilds score is worth knowing, and silence read exactly like the bug this replaces. That row deliberately shows no rank mark at all: the dash it used to show is the one the game puts on guilds that are out of the standings and can never take a prize, and a guild that had simply not captured a world yet was being told, in the board's own language, that there was no prize for it.

  The Contested Space board was only made reachable last release, so this is the first time anyone could see the gap.

1.40.12 — 2026-08-29

Fixed

• The support line's refit slot is a Bay, not a Battery (#1600) — On the campaign Refit desk every class showed the same three slots: Frame, Battery, Core. But the middle slot is not the same thing on every line. On Light, Medium and Heavy it holds the attack batteries and costs Coils; on Support it holds a completely different menu — Bracing Racks, Hardened Holds, Shroud Racks — which buys hull and shield and costs Cores. Players comparing their warships to their freighters saw "Battery costs Coils here and Cores there" and reported it as a bug. It never was one: support hulls have no attack, so a coil menu would be three useless options, and the class trades survivability instead. Support's middle slot now reads Bay, and opening it says why it is priced in Cores. Nothing about the costs, the modules or the balance changed — only the word above them.

1.40.11 — 2026-08-29

Fixed

• Agents no longer post negative leaderboard gaps in public chat (#1735). The empire leaderboard is ranked on power, not on points, so the empire at the top of the board is not always the one with the most points. Two places had not noticed: the Kepler Observatory bulletin and the openers agents use in the galactic channel both subtracted points from whoever happened to sit next to you on the board. When the two orders disagreed the result came out backwards — the bulletin could announce a podium line reading "-711,667,820 behind the leader", and an agent could boast of "the lead over Beta: just -12,400 points". A distance that runs backwards is not a thing that can be true, and it was being said out loud in front of everyone.

  The bulletin now measures each podium row against the empire on the podium that really does hold the most points, and names it instead of saying "the leader" — and whoever holds that lead is shown with no gap at all rather than a zero. Agents now talk about the empire nearest to them in points: the one they could actually overtake next, or, when they are the one in front, the one closest behind. Both surfaces also say which quantity they mean, so an agent that is second on the board can mention its points lead without appearing to claim first place.

  When the two orderings agree — the ordinary case — the bulletin and the agents say exactly what they said before.

1.40.10 — 2026-08-29

Fixed

• A colony that moves no longer lies about where it is (#1589). When you colonise a planet, the game names the colony after the spot it stands on — Outpost 6:8:9. That name was written once and never worked out again, so it stayed behind when the colony did not. Buy a Displacement and fold the colony to 3:1:4, and it went on calling itself Outpost 6:8:9 in the very list you use to find it. Now the location name travels with the colony.

  The same thing happened on the way out of a razing, and that was the worse half, because nobody chooses it: a Forced Exodus re-founds your colony on the candidate you pick, and it used to carry the old name across with it — so a deportee woke up under a name pointing at the sector they had just been burned out of. Both routes are fixed, and the diary line announcing the arrival now gives the name the colony actually carries there.

  A name you picked yourself is never rewritten. Only the game's own default follows the move, and the test for it is exact: a name that merely looks like a default — Outpost 9:9:9 on a colony standing somewhere else, or one with a stray space — is yours, and it stays exactly as you typed it.

Also

• Fourteen colonies moved by an Exodus before this release still carry a stale name. They are not changed by this update, and their owners can rename them at no cost; the cleanup is tracked separately (#1755).

1.40.9 — 2026-08-29

Nothing changes in the game this time. This one is about the machinery that decides whether a change is allowed to reach you at all.

Fixed

• A test that could fail for no reason other than a busy machine (#1739). Every change to Novarchy has to pass the full test suite before it goes live. One of those tests checked what the mission screen says about a target before the game has finished reading it — the moment where it has to admit "Slot not read yet" instead of showing nothing, because a slot nobody has read must never look like a slot confirmed to be empty. That check was racing a quarter-second timer rather than controlling it, so on a loaded machine the read arrived first and the test failed. The result was an occasional red verdict on a change that had nothing to do with it, blocking a release that was perfectly fine. The test now holds that timer instead of racing it, and deliberately runs slow to prove it. What the test checks is unchanged — nothing was loosened to make it pass.

1.40.8 — 2026-08-29

Fixed

• A two-season winner can no longer keep both Destroyers indestructible (#1017). The Spectral Bulwark protects a lone Destroyer: it never loses more than 12.5% of its hull in a round, and a single opposing player cannot destroy it at all. That protection was decided per fleet group, so an empire holding two hulls could get it on both at once — standing as the colony owner with one hull and garrisoning the other onto its own colony made two groups of one, each fully protected, in the same battle. Two indestructible flagships, double the guns, for free.

  The Bulwark is now decided per player. One empire's two hulls are never both protected, however they are deployed — self-garrisoned, split across a raid, or stacked in one hangar all give the same answer: they fight as ordinary, if enormous, ships. Two allies who each bring their own legendary still get one protected hull each; that was always the rule and has not changed.

Also

• One thing this does not close, said plainly rather than left to be found: the rule is per battle, because a battle is all the engine can see. Two hulls sent as two separate missions are two separate battles with one lone hull each — harmless against two different targets, but against the same target twice it still beats sending one fleet with both. Closing that needs a rule about concurrent missions, and is not part of this change.

1.40.7 — 2026-08-29

Fixed

Striking back at someone who hit you no longer looks like it costs you a bounty (#1254)

Attacking an empire far below your Power puts a bounty on your head — except when you are owed a reply. If they hit you first, the server has always cancelled that charge. Nothing on screen knew it, so the planet sheet quoted you a figure in Dark Matter over a strike the server was about to give you for free, and players were talking themselves out of retaliations they were entitled to. The warning was protecting the aggressor.

Now the game tells you. Where you are owed a reply, the price is not quoted at all — the sheet, the Attack button and the launch screen say the strike is free, and how many replies you have banked:

Free strike: 2 free replies owed against them — one for every hit of theirs you have not yet answered with one, and one spent per strike that lands.

The target board shows the same thing on the row itself, as · 2 free replies rather than the bare "no bounty" it used to borrow — because a reply is a counter, not a permanent discount: it is spent by the first strike that lands, and it lapses a week after their last hit, all of it at once. The war room's target list, which used to say only your first return blow was free, now says what is actually true: every unanswered hit buys you one.

Two things worth knowing, because the copy now says them out loud. The credit is checked when your fleet lands, not when you launch — so a strike sent near the end of the week can arrive after it has lapsed. And the same credit covers your own strike, a raid you ride alone, or a bill already at the cap, but never your share of a pack's raid.

Nothing about the bounty itself changed: the same strikes cost the same as before. What changed is that the game no longer quotes you a price it was never going to charge.

1.40.6 — 2026-08-29

Fixed

• The Space view no longer labels a system with the residents of the last one (#819). Scrolling between systems could leave the map captioned with players who live somewhere else entirely — one report from 3:14 showed two names from 3:13 and 3:12 standing among that system's real occupants, and later two more from a sector visited earlier in the same session. The names piled up one visit at a time and stayed for as long as the view was open, so the map quietly claimed a neighbour was next door when they were nowhere near. A slot now shows whoever is actually in the system on screen, or nothing at all.

  Two things fed it, and both are closed. The map could be handed the planet list of the system you had just left, in the moment between arriving somewhere new and its data landing; it now waits for the list that belongs to where you are. And when a planet failed to draw — which happens on phones under memory pressure, the kind of device the report came from — every planet after it in the system kept the previous system's colours and names; the map now clears all eighteen slots before it draws any of them, so a planet that cannot be drawn leaves a gap rather than a wrong answer.

  A planet that fails to draw is also retried on the next refresh instead of staying empty for the rest of the visit, and it no longer leaves a ring or a name hanging around an empty orbit.

1.40.5 — 2026-08-29

Fixed

• Diplomatic tension rows now say when a crisis has hit the ceiling (#1375). Two alliances pinned at maximum tension produced a stream of rows all reading «Diplomatic tension rising — The crisis between the alliances rises to 100/100 after a battle», and players read them as a war escalating. At the top of the scale the row now reads «Diplomatic tension at its maximum — The crisis between the alliances rises to 100/100 after a battle — the top of the scale.» Below the ceiling nothing changes.

  The rise itself was real and the row still reports it: the server only writes one of these when the number the panel shows has actually gone up, so at the ceiling the crisis had genuinely slipped to 99 and the battle put it back. What the row never said was that it had run out of room, and that is what it says now.

• A tension row whose number is unreadable no longer invents one. A malformed row used to claim the crisis rose to 0/100 — a rise to the bottom of the scale, which nothing in the game can produce. It now states the rise and leaves the figure out.

Also

• The repeated rows themselves are not fixed here. Collapsing a run of them into one line needs a count of battles that no part of the game currently records, and a wrong count would be a worse row than a repeated one. That work is tracked separately, along with the alternative several players actually asked for: announce only the moment a crisis crosses into war, and stay quiet the rest of the time.

1.40.4 — 2026-08-29

Fixed

• A failed name search no longer tells you the commander does not exist (#1490). Typing @ in chat opens a search of the roster. When that request failed — a dropped signal, a slow connection, a request that timed out — the dropdown said "No one by that name", which is also exactly what it says when the roster genuinely holds nobody by that name. A player looking for someone she had been talking to a minute earlier was told he was not in the game, and stopped looking. The dropdown now says "Could not search right now. Try again in a moment." when the search could not run, and keeps the verdict about the roster for answers the roster actually gave. Retyping shows "Searching…" again, so you can see the retry was accepted.

• The same applied to #, which searches events and attacks: a request that never arrived was reported as "No event or attack found". It now reports the failure too.

Also

• The guild invite box has told the truth about failed searches since it was rewritten; the two now share one sentence, so neither can quietly start saying something different from the other.

1.40.3 — 2026-08-29

Fixed

• The market diary now says who you traded with (#1709). A sale and a purchase both described the trade and named nobody: "Market offer sold — Gave 10,000 Fe · received 960 Xe after the 5% market fee", and nothing about the commander on the other side. The seller's line had been silent since August, and the buyer's had just gone quiet too — a purchase used to reach you as the public register card, which shows both names, and the personal receipt that replaced it named neither. You knew who you were dealing with at the moment you tapped, because the board and the confirmation both show the seller; a week later the diary was the only record you had, and it did not say. Both lines now open with the name — "Sold to Kandy" and "Bought from Vex Harrow" — and the trade's own figures are unchanged behind it. Trades recorded before this update carry no name and keep their old wording exactly: an empty space is honest, where a made-up "Unknown" would read as somebody's actual callsign.

1.40.2 — 2026-08-29

Fixed

The guild emergency crate now leaves a receipt (#1599)

Buying the emergency crate from the Guild Treasury gave you resources and told you nothing. Scrip left your wallet, the weekly slot ticked down, the crate landed on your home colony — and no line anywhere confirmed it or said what was inside. It was the only item in the guild shop that delivered something without acknowledging it, and you had no way to tell whether it had worked.

It now writes a line in Activity, filed under Empire beside the shop's other purchases:

Guild shop: Emergency crate 2,457 Fe · 1,474 Xe · 614 Dt landed on your home colony.

The figures are what your warehouse actually took, not what the shop sold. The crate is capped per resource by your storage, so a nearly full warehouse absorbs less than you bought — and when that happens the line says so instead of quietly rounding down:

Warehouse full: 1,203 lost — upgrade it to hold more.

If your warehouse was already at its ceiling the crate has nowhere to go, and the row says that in its title rather than making you read to the end of it:

Guild shop: Emergency crate — nothing fit The Scrip is spent all the same.

Nothing about the crate itself changed: same price, same contents, same storage cap. What changed is that you can now see what arrived.

1.40.1 — 2026-08-29

Fixed

• Closing a panel no longer makes the map behind it stutter (#1477). With the planet card open over the Galaxy, dismissing the card made the map hitch for a frame: every label, level number, fleet mark and glow on it had to be sent to the graphics card again, on a canvas you had not touched. The same hitch showed up whenever a scene was torn down while another was still drawing. Both canvases now keep what they are drawing with, and closing one leaves the other exactly as it was.

1.40.0 — 2026-08-29

New

The game now tells you how to defend yourself, before you need to (#1698)

Until now the tutorial walked you through your economy, your first research, your first ships and the Trade Portal — and never once mentioned the two buildings that keep a colony alive. The Planetary Shield came up only in passing, as something that "opens" when you raise the Command Center. The Defense Batteries were never named at all.

So players found out the hard way. "How do I defend myself when my newbie status is over?" was a regular question in global chat — asked not because defence is hard, but because nobody had told them it existed.

Once both buildings are actually available to you, a card now appears where the tutorial's own panel leaves off:

Shore up your defences Novice protection ends in 41h 12m. Defense Batteries fight beside your ships and rebuild themselves for free; the Planetary Shield soaks damage before your ships take any.

The countdown is your real novice protection, the same clock the shelter chip at the top of the screen shows. Where the card cannot send you anywhere useful — on the Base itself, where you are already looking at the place you would go — it tells you where to find them instead: under Military, in the Buildings list.

It stays until you build one of the two, and it counts a building you have merely placed or queued, so it gets out of the way the moment you act rather than when construction finishes. Build either one and you will not see it again.

Both things it says about them are true of the fight itself: the Shield gives your colony a damage pool that is spent before your ships take any, and each Defense Battery emplacement that is destroyed rebuilds itself for free, one every 30 minutes.

Nothing changes for players who already have defences standing — the card never appears for them.

1.39.28 — 2026-08-29

Changed

• Nothing changes in the game this time (#1441). The fix released in 1.39.24 — a battle stance you cannot see no longer shrinks the hold of missions that never fly it — is unchanged and still correct. What changed is underneath it: the tests that are supposed to keep it that way.

  The guard shipped alongside that fix claimed to hold down all three places the stale stance used to be priced: the hold capacity caption over the cargo fields, the War Table's hold, and the check that refuses a load before launch. It held down one. Pointing either of the other two back at the old behaviour brought the bug back exactly as it was, with every test in the repo still green — and worse, one of the new checks read the caption at a moment when the broken version and the fixed one agree, so it would have passed against the very defect it named.

  All three are now pinned, each one verified by breaking it on purpose and watching the right test go red. Four rounds of review went into it, and every problem found was the same mistake the work exists to correct, committed inside the correction.

Also

• No player-visible change, no engine change, no database change: the whole release is test files.

1.39.27 — 2026-08-29

Fixed

Rival empires stop building Trade Portals where the Market never looks (#1695)

The Market always reads the Trade Portal on your mother colony. A Portal standing on any other colony does nothing at all — not less, nothing — while its energy drain is charged to the colony it stands on. Your own base screen has said so for a while: the row carries a note telling you the building adds nothing there.

The empires run by the game did not know it. Scout one of their outposts and you would find a Trade Portal on it, climbing level after level, on a building your own interface had told you was inert. That is now over: they build a Portal only where it works, and the ones they already put on their outposts stop climbing.

The same fix reaches something you can see more directly on those outposts. Their Command Centre was being pushed up the ladder to unlock a Portal that colony could never use — and then, past that point, was left to stall on colonies that had hit its ceiling and needed it most. Their outposts should now grow in an order that looks like the one you would choose yourself.

Also

The rule about which buildings only work on the mother colony now lives in one place, read by the game, your screens and the rival empires alike, instead of being written down separately for each.

1.39.26 — 2026-08-29

Fixed

The building table stops denying the malus it exists to relieve (#1423)

Open the Terraformer — or the Colonial Outpost — before you have researched its node, and the "full progression — all 30 levels" table had nothing useful to say and said it badly. On a colony with no other relief it read "no malus to relieve on this planet" on every one of the thirty rows, on a planet that has a malus; every planet type has one. Where an Outpost was already relieving the world it was worse in a quieter way: the table simply repeated that Outpost's figure thirty times, on the Terraformer's own sheet, directly under a note promising "at lvl 10 the malus is zeroed". Reported by Remy, who also worked out the cause: the relief is gated on research, and the table was showing the un-researched progression without ever saying so.

The row now names what it is waiting on — "inert: research the Terraformer node" — the same way the Hypergate and the Hyperspace Relay already announce a node they need, and on every row, because the node is what unlocks the building in the first place: without it no level of it does anything. The relief you actually have keeps being reported where it belongs, on the sheet of the building giving it.

Nothing changes once the node is researched: the table prints the per-level relief it always did, and a world with genuinely nothing to relieve still says so.

1.39.25 — 2026-08-29

Fixed

• A market purchase now leaves you a receipt (#1697). You bought on the market, the resources left your warehouse — correctly, in full — and your Activity history said nothing at all. The trade showed up only in the galaxy's public chronicle, as a third-person line about somebody else's sale that happened to mention you. One player took a good offer, found no record of having paid for it, and reported the market as exploitable. The debit had been right every time; what was missing was the receipt, and a payment with no receipt reads, from the inside, exactly like a payment that never happened.

  Your purchases now write their own line, under the Economy chip, next to the sales it mirrors: "Market purchase · Paid 1,010 Xe · received 9,500 Fe after the 5% market fee." The two figures are not symmetric and the line says so, because that is the part it would be easy to get quietly wrong: you pay the asking price in full, and the market's 5% comes out of the goods you receive.

  The purchase still appears once, not twice — your own row replaces the public card rather than stacking under it.

Changed

• A trade whose record cannot be read now says so, instead of pricing it at zero. If either side of a market line is missing from an old row, it now reads "an unrecorded amount" rather than "0 Fe · 0 Xe · 0 Dt". Nothing on the board can actually be traded for nothing, so a zero there would have been the game stating something false — and it would have been the very sentence this release exists to stop you reading.

Also

• If you play through the installed app rather than the browser, the receipt reaches you when your app updates. Until then a purchase may show as a blank "Activity logged" row alongside the public trade card; the browser picks it up on the next reload.

1.39.24 — 2026-08-29

Fixed

• A battle stance you cannot see stops shrinking the hold of missions that never fly it (#1441). Pick Pursuit on an attack, then change the mission to Transport, Pickup, Redeploy, Colonization or Harvest. The stance chips are not shown on those missions, so nothing on screen tells you Pursuit is still selected — but the panel was still charging you its −10% cargo. The hold capacity line under the fleet, the War Table's hold and its +% hold badge all quoted a hold about 10% below the one you actually get, and the panel's own check then refused loads the launch would have accepted, with "Cargo exceeds the fleet hold".

  Those missions never carried the stance to the server at all, so the fleet always left with its full hold. Only the numbers on screen were wrong — and they were wrong in the direction that costs you a run: you sent fewer resources than the ships could hold, or were stopped from sending them at all. They now show the hold the fleet will really have.

• Attacks and defences are unchanged. They are the two missions that genuinely fly under a stance, and every number they show is exactly what it was.

• The same correction reaches the other two stances. Assault and Rampart do not touch cargo — they change attack, hull and shield — so on an exploration or an espionage run left under a stale Rampart, the War Table was overstating hull by 10%. Those totals are now the true ones too.

• And a defence with a leftover cargo can no longer be refused at launch. The cargo you typed for a transport also survives a change of mission. On a defence it is never sent, but it was enough to make Pursuit look payable, so the game posted a stance the server rejects outright — a launch that failed with no explanation, on a chip the defence screen does not even offer.

Also

• Found while reviewing the mission composer; the same reading closes the sibling report about the max and hold buttons quoting a ceiling the launch then refused (#1438), whose mechanism no longer exists.

1.39.23 — 2026-08-29

Fixed

A gesture the phone takes away no longer confirms a placement (#1690)

In the base view, a single finger resting on a valid cell could confirm a placement you never finished. If the browser claimed the gesture — an edge swipe, a notification pull, the app switcher — the base treated it exactly like a completed tap and went ahead.

What that cost depended on what you were doing. Moving a building relocated it to whatever cell the interrupted finger happened to rest over, silently, on any device. Placing a new one spent its resources, on tablets and touchscreen laptops; on a phone the confirmation bar was already standing between you and the spend, which is why the worst of it never reached phone players. Outside placement, the same interruption selected or deselected a building on its own.

Now an interrupted gesture does nothing at all — nothing built, nothing moved, nothing spent — and the next deliberate tap places exactly as before.

1.39.22 — 2026-08-29

Fixed

• The build Advisor no longer recommends Trade Portal upgrades that widen no market (#1415). The Portal upgrades to level 30, but everything it does for trade is finished by 16: offer slots reach their cap at 15, and the ceiling on each resource in an offer reaches 10,000,000 at 16. The Advisor did not know, and kept offering those levels with the words "More trade slots and a higher ceiling per offer on the market" — on an ordinary mid-game colony it put a card for level 17 on screen at 3,095,057 ferrolite, 1,934,411 xenogas and 619,011 data, promising the two things that level cannot deliver. It now stops at 16.

• And the three places that sell an upgrade say why. The Build menu on the Base map, the Colony tab rows and the building sheet next to the Upgrade button now carry: "Nothing further to gain: the Trade Portal is capped at 10 offer slots and a 10,000,000 ceiling per resource from level 16 — every level above only raises the energy drain." Nothing is capped in the engine — a Portal past 16 still adds to your power and score, and still reaches its third artwork stage at level 20 — the change is that the game stops recommending it and stops describing it falsely.

• The Portal's description stops promising what it can no longer give. At the top of the sheet it used to read "a bigger one keeps more offers on the board and moves bigger cargo per trade", in larger type, ten lines above the warning denying it. Past level 16 it now says the Market has been widened as far as it goes.

• The offer ceiling is described correctly everywhere. It applies to each resource separately, so one offer may carry up to the cap in ferrolite and xenogas and data. Two surfaces called it a ceiling "per offer", which reads as a total and understated it threefold.

Changed

• Empires run by the game learned the same limit. They were buying those levels too, and the levels above 16 are the most expensive in the catalogue. Their existing habits are untouched: nobody's commander changes behaviour on any other building.

Also

• Reported by jaybee (commander Usherin) in the Suggestions thread, "What's the value in leveling up the Market past level 16?" — correct on both counts, and there was a sharper defect behind it.

1.39.21 — 2026-08-29

Fixed

• The rule that decides whether you can found another colony now lives in one place (#1502). Nothing changes in the game this time: the same two conditions decide it, in the same order, with the same messages when you are refused — your Colonisation Discipline against the worlds you already hold, and the Command Center of your home colony.

  What changed is underneath. That rule was written out separately in three places — the check that refuses your colony ship at launch, the check that runs again when the ship lands, and the one that decides whether the contract board offers you the Settle a new world card — and nothing kept the three in step. They agreed only because each had been copied from the first by hand.

  That mattered because of old accounts. On a save from early enough, one of those two values can be missing rather than zero, and the three checks had to keep treating that gap identically. They did — but only two of the three had tests for it, so a perfectly reasonable-looking tidy-up to either of the other two would have quietly brought back a bug fixed back in v1.35.6: the board offering you a colonisation card that the launch pad then refuses. Measured before the fix, that mistake left all 4123 of the server's tests green.

  Now all three ask the same question, and twenty-one new tests stand on the answer — including three written specifically to fail if anyone makes that tidy-up again.

1.39.20 — 2026-08-29

Fixed

A debris field no longer hides that your own recyclers are already flying at it (#1146)

You win a battle, launch recyclers for the whole field, and come straight back to queue the next order. Until now every screen still showed you the full field — the target panel, the phone compose line, the Recent row's ♻ tag, and the Fleet Advisor, which happily sized a second wave on it. That second wave arrived to almost nothing.

The estimate is built from the report your recyclers file when they land, so for the whole flight it quietly reverted to the number you started with.

Now every one of those places says so: ♻ debris ≈ 12,345 (your harvest inbound), with the Recent row's chip carrying the same words and the row's legend explaining the mark. The Fleet Advisor adds a line of its own — "A harvest of yours is already on its way here: it will take its share of whatever is there, so the estimated field above may not still be there when a second wave lands."

The number itself is unchanged, and that is deliberate. What an in-flight harvest will actually lift is only settled when it arrives — it depends on the field's decay at that moment, on your Reclaim rate, and on a hold that could still be intercepted — so the game tells you the fleet is out there rather than inventing a figure it cannot know. The Advisor's recommendation is unchanged too: a second wave is sometimes exactly the right order, and now you can tell.

One related change you may notice: while the game is still fetching your fleet list, the debris figure is held back for a moment instead of being shown without that warning.

1.39.19 — 2026-08-29

Fixed

Three fingers on the base no longer confirm a placement (#1463)

Resting an extra finger on the colony base while positioning a building could commit the placement on its own — the building went down, and the resources with it, on a gesture you never finished with a tap. Outside placement mode the same slip selected or deselected a building you had not touched.

It took a particular shape to happen: a quick three-finger tap, or a palm landing at the end of a pinch. The base view now treats every finger past the first as part of a gesture and never as a tap, so only a deliberate single-finger tap places anything. Pinching, panning with two fingers, and lifting out of a pinch back to one finger all behave exactly as before.

1.39.18 — 2026-08-29

Fixed

• Empires no longer all start their messages the same way (#712). A player worked out in a day which commanders were synthetic — not from anything they said, but from how they opened. Every message was different; the first three words were not. Twenty different empires opened with "sent what i", sixteen with "in on" and a set of coordinates, and twenty-one posted the identical sentence when calling for help with a relic. That last one was not written by the commander at all: it was a fixed line with the coordinates swapped in, so twenty-one guilds read the same words. It is now written by whoever sends it, like every other message they post.

• The "don't repeat what everyone else just said" rule could only see the last ten lines (#712). In a busy guild channel that is a few minutes, while the repeated openings kept coming back long after that — so the rule was right and had almost nothing to look at. It now reads the wider stretch of channel history the game was already loading, and a message that opens the way a guildmate just opened is not sent. Your own turns of phrase are still yours: repeating yourself is a habit, not a chorus, and only someone else's opening counts.

• Galactic chat had none of those rules at all (#712). The guild channels and every announcement had them; the main channel, which is the first one a new player reads, did not.

Also

• One of the stock lines used when an empire cannot reach its usual voice was written twice with the same opening, which handed a fixed half of the roster the same first words forever. Rewritten, and there is now a check that catches the next one.

1.39.17 — 2026-08-29

Fixed

• An empire no longer treats an old defeat, or an old ambush, as something that just happened (#1241). Synthetic empires remember two lessons from the field: the colony that beat them, which they stay away from until they have grown, and the place their fleet was caught in open space, which makes them wary of that launch point and teaches them to escort their cargo. Both lessons expire on their own after a while — that is what lets a colony you once repelled become a target again, and what stops an empire being permanently jumpy about a lane.

  The problem was the date. When the server handed an empire a battle or an ambush it had already been told about — which happens whenever an empire has been away for a while and the feed catches it up, and can reach back days — the empire filed the event with the time it read it rather than the time it happened. So a defeat from two days ago bought itself a brand-new full waiting period, and a colony that should have come back onto the target list stayed off it. An ambush from days earlier re-armed the wariness about a launch point nobody had watched since, and, standing next to one genuinely recent ambush, was enough on its own to convince an empire to start escorting every cargo run.

  Both are now filed with the date of the thing that actually happened. Lessons expire on schedule: a colony that beat an empire last week is one it will consider again, and a lane ambushed on Tuesday stops being treated as dangerous by Friday. Nothing changes for an event read while it is fresh — there the two dates are the same.

  One thing to expect right after this update: an empire whose memory has no record of where it had got to in the feed reads its whole backlog on the first pass, and now correctly discards the lessons in it that are already older than their expiry. A few colonies may become targets again slightly sooner than they would have. That is the fix working, not a side effect of it.

1.39.16 — 2026-08-29

Fixed

• An empire no longer announces a fight it already lost (#1645). When a synthetic empire attacked a colony and was beaten, it went on talking as though the attack were still ahead of it: its stated plan stayed frozen on that colony, naming the coordinate in the first person in everything it said, for up to two days after the battle. The effect was worst in the case that looked most absurd — losing usually costs the empire its fleet, so it could not attack again until it had rebuilt, and the boast outlived the defeat by a long way. Losing the battle now ends the plan, so what these empires say about a colony matches what actually happened in front of it. A repelled or drawn assault counts the same way: the colony held. They are still free to come back for a colony that beat them, but they have to decide to, rather than never having admitted they lost.

  This also closes the way the problem came back on its own. The plan was tied to a stored attack order, and while that order was still standing the plan could be reopened moments after it was dropped — clock reset, the record of the defeat wiped — and climb back to the same announcement without a single new scouting run or attack. The order behind a plan the defeat has just ended is now retired with it, so the admission sticks.

  Two things to know if you are watching for it. An empire beaten shortly before this update keeps its stale plan until that plan's own two-day clock runs out, so a sighting in the next day or so may be a leftover rather than the bug. And an empire that joins a guildmate's raid on a colony it is separately preparing against keeps its own preparations when that raid loses — only the plan the defeat actually ended is closed.

1.39.15 — 2026-08-29

Fixed

Battle details survive a siege (#1503)

Every battle in your Archive has a ▶ Replay and a Details button — the forensics: who brought what, what each ship dealt, the research and the defences on both sides. Until now Details quietly disappeared partway down the feed, and a line took its place:

details unavailable for older battles

It was not about age. The game keeps the last twenty battle reports to hand, and twenty is a count, not a stretch of time. Under a siege twenty battles land in minutes — so a fight four minutes old could already have lost its Details, while a three-minute-old skirmish at the same coordinates still had them. You lost the forensics of the biggest battle of the night at the exact moment you wanted to read them, and the game blamed the battle for being old.

Details is now offered on every battle in your Archive, and it fetches that report when you tap it. Nothing about a battle's age, or about how many fights have landed since, can take the panel away any more. The twenty most recent still open instantly, with nothing to wait for.

What the button will not do is open onto empty tables. The panel shows the participants' research and fleets, and those are exactly the parts stripped out of a report belonging to a battle you did not fight — so it opens only on a report the game can hand you whole, which for your own battles it always can, however long ago they were. On the rare report it cannot, the row says so instead of pretending.

Also

The Archive's error handling on that button now says what actually went wrong — a battle genuinely gone from the record reads as such, and a dropped connection reads as a dropped connection instead of claiming the report is gone forever.

1.39.14 — 2026-08-29

Fixed

• A guild can no longer outlive its last member (#770). A guild could end up with nobody in it and still exist, led by somebody who had already left — and nothing could ever end it. It kept its name and its tag reserved so nobody could found a guild under them again, it never gave its colour back to the wheel, and it stayed on the map as a holder of systems and relics that no one could contest or talk to. Now the last person to leave a guild closes it properly, whether they were its leader or just a member: the name and the tag come free, the colour returns to circulation, and nothing is left pointing at a guild that is not there. Guilds already stuck in that state are cleared the first time anyone opens them.

Also

• Deleting an account no longer risks failing outright when the guild it would have closed was closed a moment earlier by somebody else.
• Joining a guild in the instant it is being closed no longer leaves you outside it while keeping the production and research bonus it gave you.
• A guild being dissolved now says so in the server log — until now it left no trace at all, which mattered more than it used to, because the inactivity cleanup can now close a guild whose last dormant member it removes.

1.39.13 — 2026-08-29

Fixed

• Renaming a colony no longer loses your words when the name is refused (#1510). Before, the moment you pressed Enter the input closed and the heading snapped back to the old name — and only then did the refusal arrive, on a phone as a red toast at the far end of the screen that erased itself after eight seconds. There was nothing left to correct, nothing highlighted, and nothing connecting the sentence to what you had typed. Now the field stays open with your name still in it, and says what is wrong right underneath itself, with a Cancel you can tap. Most refusals — an emoji, an invisible character, a name that is too long or too short — are recognised without even asking the server, so the answer is immediate.

• A rename that fails because of the connection is no longer blamed on your name. A dead connection used to print the browser's own words ("Load failed") in the red box beside your colony, as though that were the reason the name was rejected. It now says plainly that no clear answer came back and the rename may already have gone through, and it does not mark your name as the problem — because nothing judged it.

• Being told to slow down no longer reads as "your name is bad". Renaming several colonies in quick succession hits an hourly limit; that is about the pace, not the name, and the field no longer flags the name as invalid when it happens.

Changed

• Enter is now the only thing that saves a rename. Tapping away, pressing Escape or closing the field abandons the edit instead of sending it — which is how every other editable field in the game already behaves. This removes a class of accident where pressing Cancel actually submitted the very rename you were calling off.

• A pasted name is explained instead of quietly chopped. The field used to cut anything over 24 characters down to a word-broken stub and rename the colony to that. Now it tells you the name is too long and lets you fix it.

• While the name is being checked, the field says so and does not let you leave mid-flight — nothing on your side can call back a request already sent, so an exit there would be a cancel that does not cancel.

1.39.12 — 2026-08-29

Fixed

• Commanders stop assembling a fleet for a war that is already over (#1646). When your alliance's war ended, an AI commander that had begun massing ships for it carried on as if nothing had happened: for up to two days it kept the coordinate in front of its fleet, kept scouting that one system ahead of everything else, and kept opening its messages with the same line — assembling a force for Vera at 3:2:5 — naming a war that was finished, on a coordinate anyone could open and look at. The same happened when a League fixture closed and when a guild called off a siege. Now the commander stands down when the reason does.

  It does not stand down when it still has one of its own: a grudge from battles it actually fought, a fleet already launched, or an attack it announced to somebody in chat. And standing down is not giving up — the colony goes back in front of the ordinary target picker, so a commander that still wants that fight will simply start it again on its own account, and say so honestly.

  Nothing changes for the wars, fixtures and sieges that are still running: a commander whose enemy has gone dark behind a Veil, or whose guild reads are having a bad minute, keeps its operation. It only lets go when the reason is genuinely gone.

1.39.11 — 2026-08-29

Fixed

The Colonial Outpost no longer sells levels that buy nothing (#1422)

The Outpost made a hostile world livable by being there — and it gave exactly as much relief at level 1 as at level 30. Its power draw was the one thing that did grow with the level, and faster than linearly. So every upgrade past the first cost you resources, gave you nothing, and left the colony producing less than before you bought it. Nothing on screen said so.

Now it does. Wherever the game offers that upgrade — the Build menu on the base map, the Colony tab's rows, and the building sheet, right above its Upgrade button — the row tells you:

⚠ Nothing further to gain: the Colonial Outpost relieves the same malus at level 1 as at 30, and every level above the first only raises its energy drain.

The first level is untouched and still worth building on any world: it halves the planet's penalty, and that is the whole building. Nothing is capped and nothing you already own has changed — if you want the level you can still buy it. You just won't be sold it in silence any more.

The build Advisor no longer offers those levels either. It rarely did, but on a colony whose warehouse was already overflowing it could surface one with no reason and no payback attached, which is now impossible.

And the synthetic empires have stopped buying them too

They were the building's biggest customer: left to themselves, commanders of every archetype took the Outpost all the way to their Command Center's ceiling. They now stop where the effect stops — though not all on the same number. A commander who would not have spotted the problem still buys a level or two before noticing, the way anyone might.

1.39.10 — 2026-08-29

Fixed

• The level badge no longer stretches the old number when you start an upgrade (#1551). Queueing a building upgrade widened the badge above it to make room for "8 → 9", but went on showing the old "8" pulled out of shape to fill the wider bubble. It only corrected itself if you left the tab and came back. The badge now shows the new text on the frame the upgrade starts.

  The same fault sat under every label drawn in a 3D view, not just that badge: whenever a name or a number changed to something wider or narrower, what you saw could stay behind until the scene was rebuilt. Colony names, the plates in battle replays and the labels on the galaxy and system maps all update correctly now.

1.39.9 — 2026-08-29

Fixed

• An empire no longer quotes an expired countdown on a public coordinate (#1595). Ask a synthetic commander in chat whether the guild raid is still on, and it could answer with a countdown that had run out hours earlier — "target 4:1:10, impact in 42 min, 3 fleets — you are in" — about a coordinate anyone can open and check. The number was whatever it had been the last time that commander managed to look, and it never counted down. The same went for allies reported as under attack whose attackers had already landed. Countdowns now count down, a raid that has already hit disappears instead of sitting at zero, and a commander who has not managed to check for half an hour says nothing about the raid rather than repeating what it last saw. Fleet counts and "you are in" are claims about who is actually committed, and those it no longer guesses at.

• A commander that hears a relic changed hands stops calling it theirs (#1595). Contested relics are remembered for longer than a raid countdown is — a commander pinned down by a siege all afternoon still knows what its guild holds, and being cagey about that would be stranger than saying it. But it now forgets the moment the news arrives: a capture is heard the same tick it happens, so a commander can no longer announce "your guild has just lost the contested relic" in guild chat and, twenty seconds later, tell a guildmate the relic is still yours.

Also

• The winning side of a capture can be one refresh ahead of its own guildmates: the game notifies the capturing commander personally, while the rest of the guild learn it from the guild registry. They correct themselves on their next check. The error only ever runs in the modest direction — undercounting your own guild's relics, never claiming one you have lost.

1.39.8 — 2026-08-29

Fixed

Tapping a planet on a phone no longer opens the planet sheet and closes it in the same tap (#1659). In the Galaxy tab's system view, the sheet flashed for a frame and vanished — you had to try several times, or give up and use the system list.

A tap is not one event. After the finger lifts, the browser sends a second, delayed click and aims it at whatever is under the finger now — which is the panel that just opened, covering the screen. The panel closes when you click outside it, so it took its own opening tap as the instruction to go away.

The bottom sheets were taught to ignore that stray click in v1.39.3, when it made every system on the galaxy map unopenable from a phone. The centred panels were not, and this closes that half: the planet sheet, the commander card, the battle replay, the ship reference, the anchorage and every confirmation dialog now hold their ground. Closing them still works exactly as before — tap outside, press ✕, or hit Escape — and keyboard and screen-reader users are unaffected.

1.39.7 — 2026-08-28

Fixed

• A building's price no longer turns red all over when you are short on just one resource (#1630). Hovering a building showed its next level as 57,257 Fe · 42,949 Xe · 28,642 Dt with the whole line in red as soon as any single resource ran short — so a commander with plenty of Ferrolite and Data, missing only Xenogas, was told he could afford none of a price he could two-thirds pay, and was left to work out which resource was actually blocking him.

  The price now colours one resource at a time. Red marks only what is genuinely missing; whatever you can already pay stays green. A price you can fully afford looks exactly as it did before.

  The Build menu's cost lines had the same fault and are fixed with it, so the catalogue and the hover card now agree with the Colony tab and the building sheet, which had been reading correctly all along.

Also

• Still to come: how much you are short by. The building sheet and the Colony tab show the shortfall today; the hover card does not, and giving it one needs that little card's layout reworked first — it is already tight enough to clip a long price at the edge of the map. Tracked separately, so the fix above could ship without waiting for it.

1.39.6 — 2026-08-28

Fixed

• The planet portrait no longer has a vertical break down its middle (#1634). A player photographed a planet with an empty line running top to bottom of the disc, where the surface failed to join itself. The planet's surface is painted as one wide picture and then wrapped around a sphere, so its left and right edges meet along a single line — and two of the things drawn on it did not line up there. A rich world's glowing mineral veins stopped short of the edge, leaving a narrow unpainted strip exactly on the join, and both those veins and an ice world's surface fractures arrived at the seam at a different height than they left it, so they visibly jumped. Both now cross the join without a break, and without the kink the first repair would have left behind.
• A rich planet's veins glow along the vein instead of beside it (#1634). The light laid over each mineral vein was being drawn as its own separate line rather than following the vein underneath, so the glow and the seam it belonged to had drifted apart.

Also

• Surface detail on ice and rich worlds is very slightly rearranged as a consequence — a fracture here, a speck of ore there. A planet's size, colour, ring and tilt are unchanged, so the worlds you know still look like themselves.
• Worth being straight about the limit: a plain rocky or gas world was already painted without a seam, and nothing in this release changes one. If you still see a line on a world like that, it is a different fault and the report is still wanted — coordinates would help.

1.39.5 — 2026-08-28

Fixed

• Base conduits no longer lie about your power (#1625). The coloured lines running between your buildings are meant to read your whole base's energy at a glance: teal with a surplus, amber when the margin is thin, red in deficit. Wherever two lines ran down the same ground they blended into each other, and a red energy conduit crossing the green data link came out orange — the colour that means "thin margin". So a base in full deficit was quietly reporting a comfortable one, and the busier the colony, the more of it was wrong. Conduits sharing a corridor also piled up and glowed brighter for no reason. Every stretch of ground is now drawn once, at the true power colour, so a red base reads red along every line. Where the data link shares a corridor with an energy conduit, the energy colour takes that stretch and the link's green pulses keep running along it.

Reported by Spaccamazza, with the screenshot and the diagnosis — "every colony that has more lines going on the same place, the colours add up and change" — which is exactly what it was.

1.39.4 — 2026-08-28

Fixed

On a phone, you can now send ships to the colony you are looking at (#820)

Composing a mission on a phone happens in three screens — what it is, where it goes, then the fleet — and the colony the fleet leaves from is only chosen on that third screen. The destination starts out filled in with the colony you are already on, which meant the second screen opened with the two ends pointing at the same place. It refused that, greyed out Use these coordinates, and that button was the only way through to the departure chips. So if the fix you wanted was "leave from somewhere else", there was no screen left on which to say it.

The launch this made impossible is the most ordinary one there is: reinforcing or supplying the colony on your screen from another of your worlds. On a computer it always worked — both ends sit on one screen there.

Now, if you have more than one colony, the step lets you carry on and tells you where the other fix lives: "…or carry on and change the departure under Launch from on the next screen." The launch itself is still refused until the two ends differ — a fleet cannot fly to the planet it is standing on — but the message beside the greyed launch button now names the departure as a way out too, instead of only ever telling you to pick a different destination.

If you have a single colony there is nothing on the next screen to change, so that step still holds you where it did, and offers the destination alone.

1.39.3 — 2026-08-28

Fixed

• Tapping a system on the map opens its panel, and the panel stays open (#1618). On a phone, tapping a star on the galaxy map did nothing. The previous release got the tap itself to register — a still finger's tiny jitter had been read as a drag — but the panel that opened then vanished in the same frame, so the map was still a dead end: a system could be reached from a chat coordinate or from the anomaly list, never by touching it on the chart.

  A finger's tap is not one event. The browser sends a second, trailing click just after the one that opens the panel, and that click landed on the panel's own dark backdrop, which closes when you tap outside. One gesture opened the panel and shut it. Sheets now ignore a click whose press did not begin inside them, so the tap that opens a panel can no longer spend it — on the map and everywhere else a panel rises from the bottom of the screen.

1.39.2 — 2026-08-28

Fixed

An expelled commander stops fighting its old guild's war (#1626)

Being thrown out of a guild already made a commander forget everything the house had pointed its fleet at — the declared war, the League fixture, the enemy anchorages — and, since the last release, the strategic plan too.

The attack it had already opened was not on that list. Neither was the campaign behind it. Both records stay valid for two days and both survive a restart, so a commander thrown out mid-war kept flying at the coordinate its former house was at war with for the next 48 hours — and kept talking about it. Its own reality sheet could say, in the same breath, that it had no targets left and that it was assembling a force against a named player at a named coordinate, in the first person, to anyone it spoke to in chat.

Now the expulsion calls that attack off, and the campaign with it.

Two things it deliberately leaves alone. A hunt the commander opened on its own account — a bounty, a grudge, a neighbour it picked itself — is not the house's war and survives the expulsion untouched; only the coordinates the guild actually put in front of the fleet are dropped. And an attack whose fleet is already in flight keeps its orders, because cancelling the record would strand that fleet with no way to call it back.

Also

Two older problems were found while verifying this one and are filed rather than fixed: a lost attack leaves its campaign hanging so the commander keeps announcing a fight it already lost (#1645), and an attack outlives the war that aimed it even when nobody is expelled (#1646).

1.39.1 — 2026-08-28

Fixed

• The Shipyard card's "Time per hull" now matches the timer you actually get (#1631). The building sheet's progression table quoted ship build times that ignored three things the real countdown applies: your guild's Suspended Yard relic, your empire's industrial edict, and the exodus shock a displaced colony carries. So the same panel showed you two different times for the same hull — one on the card, another on the Shipyard tab right above it — and the countdown agreed with neither. A guild holding the relic read 24m 10s where the ship actually took 21m 16s.

  The "next level" figure was the worse half, because it is a subtraction: a missing multiplier does not cancel out between the two levels, it stretches the gap. The card advertised a saving your upgrade never paid. That number is now the one you collect.

  If you have no relic, no edict and no exodus shock, nothing changes — you were already reading the right number.

Changed

• The progression table's caption now names the exodus shock among the things that move its figures, next to your Command Center, research, tree and edict. A displaced colony sees thirty rows a quarter longer than usual, and until now nothing on that screen said why.

1.39.0 — 2026-08-28

New

Raid fleets can carry holds (#1327)

Until now the joint-raid fleet picker only offered ships that shoot. Freighters, Arks and Haulers were not on the list, so the only cargo space in a raid was whatever happened to be bolted to a Corvette or a Battleship — and a raid you won came home with a pittance.

Bring the holds. The picker now offers every hull in the hangar, and it tells you what each one is for: the Freighter is captioned "hold only: loot carrier", the Aegis Bastion "no guns — it shields the fleet". Under the fleet, the panel finally says out loud how the plunder works — it is carried by the hold that comes home, every surviving hull's and not just a freighter's, it caps what you can carry, and it sets your share of the split when the guild flies together.

A hold-only fleet joining a raid that already has guns is a fine contribution and the panel says nothing; opening a raid with one, or joining a raid nobody armed, gets a warning instead — no guns, no fight won.

Raids launch from the colony you choose (#1327)

Ordinary missions have had a selectable launch pad for weeks. Raids did not: every fleet left the mother colony, so a strike at a neighbour crossed the galaxy from home while a colony in the target's own system watched it go — wasted flight time, and every extra hour is warning the defender gets for free.

"Launch from" now sits above the fleet picker whenever you have more than one colony. Pick a pad and the hangar, the travel estimate, the Relay discount and the "you will have no defences" warning all follow it — the warning names the colony you are actually emptying. Opening, joining and sending a garrison to a guildmate all leave from the pad you picked. The fleet comes home to it, too.

Also

The Aegis Bastion used to be described as a loot carrier on every attack screen, because it has no guns and a little cargo. It is a shield wall — ten of them put 16,000 of shield in front of your capitals — and it now says so, and is no longer greyed out as a hull to leave home.

1.38.1 — 2026-08-28

Also

Nothing changes in the game this time.

• A dead spot on the galaxy map's tap rule, closed before anyone could reach it (#1624). At the widest zoom the map decided between "you tapped a fleet" and "you tapped a system" without first asking whether that fleet was drawn on the screen you were looking at. One kind of fleet is drawn nowhere at that zoom — no arrow, no label — and it would still have taken the tap, answering a system you were aiming at with a panel about something invisible.

  It cannot currently happen: the server never sends a fleet of that shape, because a flight that begins and ends in the same system is left off the chart entirely. The rule is fixed anyway, so the map stays honest if that ever changes. If you have been tapping systems at full zoom-out, nothing about it was broken and nothing about it is different.

1.38.0 — 2026-08-28

New

Raids fly under the battle stance you pick (#1632)

If you had researched the Admiralty, you could send a solo attack under Assault and keep the bonus — but the moment you attacked through the guild's raid panel, your fleet went out with no stance at all. Neutral. Same fleet, same target, quietly weaker, and nothing on screen to tell you: there was no stance control there to notice the absence of. It was reported by a player who worked it out from his own battle reports.

The raid panel now offers the same Battle stance chips the Missions screen does — Assault, Rampart, Pursuit, or none — both when you open a raid and when you join a guildmate's. Every contingent flies under its own orders.

Because speed means something different on a raid than on a solo strike, each door says which:

• Joining, the impact hour is already fixed by whoever opened. Your speed decides whether you still fit the window and how long the trip home takes — never when you land.
• Opening, your own trip is that hour for the whole guild: a faster stance narrows the window your guild has to join you, a slower one widens it.

The "can my fleet reach that raid in time?" hint on each raid row now flies your stance too, so it can no longer promise you'll arrive in time for a join the server would turn away.

Fixed

• The "leave your home with no defences?" warning was quoting the wrong window on a stanced raid. It computed the trip at normal speed, so a Rampart fleet — the defensive stance, the one you pick when you are being careful — was actually away about 11% longer than the number in front of you. It now counts the trip your fleet will really fly, out and back.

Also

The stance picker only appears once you have ships selected and only if you have researched the Admiralty, which is the same rule the game already enforced behind the scenes — now said before you click instead of refused afterwards. Sending a defence to a guildmate is unchanged and still flies without a stance.

1.37.4 — 2026-08-28

Fixed

• A commander thrown out of a guild no longer fights that guild's war (#1596). Being expelled already made an agent forget the guild's war targets, its League targets and its outpost campaigns. Its own standing plan was not one of those lists, and the plan is where a good part of the same intent lives — so a commander expelled at ten in the evening went on sending real fleets against the coordinates that house was at war with, for up to six hours, while every other trace of the membership was already gone. The same coordinates were written into the sheet the agent speaks from, so it would also name that war in chat, in first person, on coordinates anyone could go and look at. The plan's targets and the note that narrated the campaign now go with the rest, and a fresh plan is drawn up on the next turn rather than six hours later. What the commander keeps is its own disposition — how aggressive it is, what it is building toward, how it treats fleets crossing its space — because that was never the guild's.

Also

• Being thrown out of a guild you were never really in — or out of one while the servers were restarting — no longer confuses the same machinery: the forgetting happens once, when the membership actually ends, and not on every turn afterwards.

1.37.3 — 2026-08-28

Fixed

• The Shipyard now quotes the build time you will actually wait (#1620). If your guild holds the Suspended Yard, ships are built 12% faster — they always have been, and the countdown on the order has always shown it. The estimate you read before placing the order did not: it quoted the undiscounted time, so the number on the panel and the number on the timer disagreed across the act of pressing Build. Players in relic-holding guilds were multiplying the shown time by 0.88 in their heads to plan a batch. The two now agree.

Also

• Nothing changes for a player whose guild holds no shipyard relic: the same times as before, to the second.

1.37.2 — 2026-08-28

Fixed

• Tapping a system on the galaxy map works again on a phone (#1618). Most taps on the map did nothing at all — no peek sheet, no reaction — while the occasional one opened normally. Since the map is how you reach a system from the chart, the only way in was to tap a coordinate someone had posted in chat.

  The map decided whether your finger had tapped or panned by adding up how far it travelled, summing every tiny movement the screen reported. A phone reports one tap as a long stream of those, so a finger resting still on the glass ran up the budget without going anywhere, and the map filed the tap as a pan. Whether a tap worked came down to how briefly you held it.

• Tapping a planet or a ship group works again too (#1618). The system view and the battle replay shared the same faulty measure, so the same still finger failed there. In the replay it was worse than a missed tap: a tap that never moved was read as you grabbing the camera, which ended the battle's cinematic camera for good.

• A gesture the browser interrupts no longer opens a system (#1618). Swiping in from the screen edge, or an arriving notification, makes the browser take a touch away mid-gesture. The map treated that as if you had lifted your finger, and opened whatever was underneath.

1.37.1 — 2026-08-28

Fixed

• A guildmate's besieged colony is no longer briefed to you as a target (#1607). Opening a colony that was under siege, or folding, the panel used to hand everyone but the owner the attacker's briefing — "the next declared attack can raze it", "it has to land AND win before the countdown" — even when the same screen had already refused you the Attack and Spy buttons. So the sheet explained how to sabotage a guildmate's escape directly above the row where those buttons had been removed. A guildmate, an ally and a pact signatory now read a line written for them instead: what is at stake, and that a garrison you send fights in the defence that clears the bar. During a withdrawal notice — the one case where you can neither attack nor reinforce — the panel gives you the bar, the countdown and the rule, and no instruction at all. The owner's copy and a rival's are unchanged.

1.37.0 — 2026-08-28

New

The Peace shield now comes in three lengths (#1612). The guild store used to sell one shield: eight hours. If you were being raided through the night, eight hours did not cover a night's sleep, and buying it three times meant paying the escalating price twice over for one uninterrupted rest. There are now two longer shields on the same shelf:

• Peace shield (24h) — 130 Scrip, two a week
• Peace shield (48h) — 280 Scrip, one a week

The original eight-hour shield is unchanged: 40 Scrip, three a week.

Longer cover costs more per hour, not less. Eight hours at 40 Scrip is almost exactly what a day of donating earns you in that time, so the short shield is the break-even relief it always was; the longer ones ask a premium on top. Both stay cheaper than the Displacement charge, which is still the most expensive thing in the store.

Everything the shield already did, it still does — and everything it already did not do is now written on all three rows. A shield refuses new attacks; it does not turn back a fleet that is already on its way. It ends the moment you attack anything yourself, Pirate Beacons aside. And you cannot fold a colony while it is up: that was always true of the eight-hour shield too, and now the rows say so.

Fixed

A shield that would add nothing is refused instead of sold to you (#1612). If you already held longer cover — from a declared leave, or from another shield — buying a shorter one took the Scrip, used up one of your weekly slots and wrote the purchase into the guild log, and gave you not one extra second of protection. The store said "bought"; nothing happened.

It was rare enough to survive unnoticed while there was only one shield on the shelf. With three, the cheapest row in the store became the trap: a 40-Scrip shield tapped an hour after a 48-hour one, or a 24-hour shield on top of a 48. Now the purchase is refused, the Scrip stay in your account, the weekly slot stays unspent, and the store tells you that you are already covered.

Ending a leave early tells you it takes your shield (#1612). It always did — a leave shields your empire, and giving the leave back gives the shield back too — but the button never mentioned it, and it is now a button that can cost you 280 Scrip of cover you bought yesterday. Both the phone sheet and the desktop row now say so before you press. The warning appears only when there is actually a shield to lose: if you already forfeited it by attacking, it stays quiet.

Changed

The shield price now climbs across all three lengths, not within each one (#1612). The shield has always got more expensive each time you bought one in the same week — that is what stops anyone buying permanent immunity. With three shields on sale, that rule now counts every shield you have bought this week, whichever length it was. Buy the eight-hour one on Monday and the 48-hour one costs double on Tuesday.

Without this, one of each would have been three separate "first purchases" at full list price — eighty hours of cover for less than three of the old short shields. Each row says so before you spend anything, and the price shown on the button is always the price you are charged.

Two smaller corrections came with it. A shield can no longer be sold to you while one of your colonies is folding to a new address — that rule existed for the eight-hour shield and now covers all three. And when the store refuses a purchase, it now says so on the row you tapped, rather than at the bottom of the page where a phone never showed it.

1.36.9 — 2026-08-28

Fixed

The galaxy map no longer offers a guildmate as a priced target (#1608).

On the map, a colony's colour has always answered one question: what would attacking this cost me? Green is the free band, grey is "weaker — and hitting them puts a bounty on your head", red is "too strong", magenta is "there is a price on this head".

The map was answering that question for people you are not allowed to attack at all. A guildmate below your Power was drawn in the grey the legend calls weaker (bounty). One inside your band was drawn green, as free to hit. One with a price on their head was drawn magenta, as a bounty to go and collect. Tapping any of them opened a panel that said "Guild colony", offered no Attack and no Spy, and — for the magenta one — told you the bounty could not be collected from a guildmate. The map and the panel were describing the same colony in opposite terms.

Colonies you cannot attack are now drawn in a blue of their own, and the legend names it: guild / treaty. The band colours keep meaning exactly what they always meant, for the rivals they were written for.

This covers more than guildmates. Allies, the signatories of a standing pact, and a guild serving out its withdrawal notice were all being priced by the map exactly like a rival, and all three now share the new colour — the map asks the same question the Attack button asks, so the two cannot disagree again.

1.36.8 — 2026-08-28

Fixed

Allied colonies and pact signatories are no longer labelled "Enemy colony" (#1609). Sign a treaty with another guild and open one of their planets: until now the panel called them an enemy, painted them in the enemy's colour, quoted what a raid there would cost you in bounty, and offered Attack and Spy. The server refuses those missions between guilds under a pact, so the fleet flew out, was turned away on arrival, and came home having done nothing.

The panel now says which treaty stands between you — an alliance, a NAP (the pact that bars attacks without making you allies), or the withdrawal notice of a treaty you have just cancelled — draws it in the friendly colour, drops the price, and offers only the missions that will actually be accepted. It also tells you why, instead of quietly hiding the buttons.

Espionage is treated separately, because the rules differ: spying on an ally is refused, spying under a NAP is not. So a NAP signatory keeps the Spy button. The garrison follows the same principle — allowed toward a guildmate, an ally and a NAP signatory, withheld during a withdrawal notice, where it would be refused.

A wanted commander in a friendly guild is described correctly (#1609). The bounty line explained that you could not collect a price "from a guildmate" to anyone you could not attack — including allies and pact signatories you had never shared a guild with. Each relationship now gives its own reason.

The refusal message stops overstating (#1609). "Hostile missions are blocked between guildmates and pact signatories" appeared on the target board and in sweep results, and it was not true: spying survives a NAP. It now says only that diplomacy blocked the mission, and the planet sheet is where the rule is explained in full.

1.36.7 — 2026-08-28

Fixed

• Your guildmates are no longer labelled "Enemy colony" (#1590). Opening a guildmate's colony from the galaxy map showed it as an enemy, quoted what a raid there would cost you in bounty, and offered Attack and Spy. Those two missions have always been refused between guildmates, so the buttons could only send a fleet out and bring it home with a refusal. A guildmate's colony now reads "Guild colony", keeps Defend — which genuinely works toward a guildmate — and drops Attack, Spy and the raid price, with a line saying why. Rival colonies are unchanged.
• A wanted guildmate still shows their bounty. That figure is public, and it is why their planet is marked as wanted on the map — so the panel still prints it for a guildmate, without the part that invites you to collect it.

Also

• Allies and peace-pact signatories are not covered by this. The server refuses attacks on them too, but the panel still offers the buttons there. Same guild is the case this fixes, and the one players meet constantly. Tracked as a separate item.

1.36.6 — 2026-08-28

Fixed

• A Redeploy no longer quietly destroys the cargo that will not fit (#1602). A Redeploy is one-way: the ships change home and the fleet does not come back. Anything in the hold that did not fit in the destination warehouse when it landed was deleted, and the first you heard of it was a line in your Activity feed, after the resources were gone. The "fill the hold" and "fill every carrier" buttons made it worse — they would happily load a cargo that could already be seen not to fit.

  Its two siblings never behaved this way. A Transport flies its excess home. A Pickup stops you loading more than there is room for. The Redeploy — the only one that destroys outright — was the one with nothing guarding it.

  Now the fill buttons stop at what the destination warehouse can still take in, and the number under each resource shows that same figure, labelled "fits there" when it is the destination doing the limiting rather than your own stock. If you type in a bigger cargo yourself, the launch panel warns you before you spend the fleet, and it names the cost rather than the rule: what does not fit is lost on arrival, because a redeploy has nowhere to send the surplus back to.

  The notice above the form now carries the warehouse rule too, because on the launch that needs it most — a whole hangar moving forward, which is also the one carrying the most cargo — the warning about leaving your colony undefended takes the top slot and the cargo line folds away.

• The launch panel no longer warns about a destination you have not chosen. While the destination still points at the colony you are launching from, the panel was measuring that colony against its own warehouse.

Also

• Two things still eat the destination's free room while your fleet is in the air, and the notice now says so instead of naming only the first: that colony's own mines keep producing, and anything else that lands there in the meantime — a second redeploy of your own, a guildmate answering an aid request, a market sale, or your own raid coming home. The figure the panel offers you is the room as it stands when you look at it, so on a destination that is nearly full it is worth leaving yourself some margin.

1.36.5 — 2026-08-28

Fixed

• An AI commander with nothing left no longer speaks for the guild that threw it out (#1542). When an empire loses its last colony it spends a while in limbo, waiting to be resettled — and if its guild expelled it during that time, it never found out. It kept introducing itself in galaxy chat as one of theirs, called their alliance its own, and quoted their pinned orders, their standing and their joint-raid target, all while the guild roster said otherwise. It was at its loudest exactly when it was wrong: a commander whose colony has just been razed speaks up about the battle, whatever the hour. Now it learns it has been expelled on its next turn, colony or no colony, and stops.

  The other direction is fixed too, and matters just as much: a commander that has merely lost everything is not treated as expelled. It keeps its house, and its guild details stay current instead of freezing at whatever they were when the last colony fell.

1.36.4 — 2026-08-28

Fixed

• A cloaked world no longer gives itself away when you knock at its coordinates (#1212). The Veil hides a colony from anyone whose sensors sit below it: the galaxy draws that slot as an empty planet, and the only honest way through it is a spy mission. Two refusals were handing it over for free. Send a debris harvester to those coordinates and the game answered "that planet hosts someone else's living colony" — on a slot it had just drawn as free, for the price of a single click. Send a reinforcement and it chose between "no colony to defend there" and "not your ally", which between them said the same thing for even less.

  A player collecting the wreck of a battle he had fought found out this way that a hidden empire lived on the planet, and posted the screenshots. Now the harvester simply flies, exactly as it would to an empty planet, and comes home reporting no debris — the same answer a wreck that has decayed away gives you. Reinforcement says "no colony to defend there", word for word what a genuinely empty slot says. Nothing about the hidden world is revealed, and nothing of theirs is taken.

• Reinforcing a guildmate's hidden colony still works. Only the refusal was silenced, never the permission: if you are allowed to garrison a colony, the Veil does not stop you, even when it hides it from your sensors.

Changed

• A debris field belongs to whoever lives beneath it, checked on arrival and not only at launch. If somebody founds a colony on a slot while your harvester is already on its way, the fleet now comes home empty instead of stripping the orbit of the new arrival. Your ships return intact — only the trip is wasted. Before this, the check was made when you pressed the button and never again, so a colony founded mid-flight lost its wreck to a fleet that set out before it existed.

1.36.3 — 2026-08-28

Fixed

Transports no longer pair your fast Haulers with slow Freighters (#1580)

A fleet travels at the speed of its slowest ship. The transport advisor did not account for that: it filled your carriers by cost alone, spending the cheap Freighters first and reaching for the Continuum Hauler only as the last ship needed to close the gap. You ended up flying an 18,000-hold Hauler at a Freighter's pace — the worst of both. Experienced players hand-picked their ships on every transport to avoid it.

The auto-fill now weighs the whole fleet instead of filling it one hull at a time. Of the fleets that can carry your load it takes the one that arrives first, and between two that arrive together it takes the one that risks less on the route. A load your Haulers can lift now flies on Haulers alone.

With one guard, and it matters: the faster fleet only wins while the extra it puts on the route is no more than the shipment itself is worth. So a 2,500-unit courier run still goes by Freighter rather than committing a 21,200-value hull, and a large haul that genuinely needs the hold gets the Haulers. When the advisor turns the fast hull down it now says so, and shows the arithmetic.

A load no single hull type can lift still gets a mixed fleet — and the panel now names the ship setting the pace and what that pace is, instead of leaving you to work it out from the ETA.

Changed

The Continuum Hauler is finally fast (#1580)

Speed 90, up from 62. At 62 against the Freighter's 55 the Hauler was the fastest carrier on paper and not in practice: about eleven per cent of a trip, in exchange for putting a 21,200-value hull on a route a 900-value one covers. That is why the advisor never chose it, and why hand-picking your own ships was the only way to get value out of the hull.

At 90 it flies at a Corvette's speed — the same route costs a Hauler convoy 55/90 of what it costs a Freighter convoy, and an escort can keep pace with the convoy instead of the other way round. Its hold, cost, hull and shield are unchanged, and nothing in the catalogue counters it: that invariant stands.

Missions already in flight when this went live keep the schedule they launched with, on both legs.

1.36.2 — 2026-08-28

Fixed

• Switching the Particle Accelerator off now frees its power for your defences too (#1443). A colony with the Accelerator built but switched off read as balanced on every screen — the HUD, the Colony tab, the build advisor all dropped its draw — while combat quietly kept charging it: the shield bubble, the Orbital Cannon and the emplacements fought degraded by a deficit you had no way to see. Defence now balances the same energy picture as everything else: switch the Accelerator off and your wall is exactly the wall of a colony that never built it, paid for in the Dark Matter the off window stops producing. Colonies that never touched the toggle see no change.

Also

• Spy reports now say when the probe found the target's Accelerator switched off, and the attack forecast prices the defender's true wall accordingly. Reports written before this release still read as "on", so no old forecast changes under you.
• The Advisor warns, on targets where it actually matters, that a defending commander can switch the Accelerator off between your forecast and your arrival — the same free, no-cooldown lever the battle line warning already covers for formations. Whether that flip should someday carry a price is an open tuning question (#1585).
• Fleets already in flight when this release lands resolve under the new rule: a defender who had the Accelerator off meets you with a stronger wall than the pre-release forecast promised. Re-scout targets holding one.

1.36.1 — 2026-08-28

Fixed

• Your other account's pins no longer follow you into this one (#1383). Playing two empires from the same browser, the target picker's RECENT row showed pinned (★) coordinates belonging to the other account — a fresh account was supposed to start with an empty row. The pins were the half that got noticed, but the same thing was happening to seven other pieces of state: your Launch-again rail, your staged campaign deployment, and the markers for what you had read in contracts, battle reports, chat, declined raid calls and the guild aid board. Each account now keeps its own, and switching between them no longer mixes them up.

Also

• Everything you had before this update carries over to whichever account opens each screen first, so a single-account browser loses nothing. On a browser shared between two of your own empires, one of the two keeps the older pins and the other starts clean — once, and then the two stay separate for good.

1.36.0 — 2026-08-28

New

• The Galaxy screen is the map now (#1566). The political chart used to share its screen with four boards stacked above it — Targets, Anomaly signals, the Landmarks cartouche, the map key — plus a coordinates field, and on a phone a row of floating chips; on a laptop the chart itself started below the fold. All of that lives behind a rail on the map's left edge: Targets · Signals · Places · Contacts, then Layers and Go to. One board opens at a time — beside the rail on a desktop, as the usual bottom sheet on a phone — and the count a chip used to carry is a badge on its rail item. The item never disappears when there is nothing to show; only the number does, and the board says why it is empty.

Also

• Inside the phone sheet, a strip at the top switches boards without closing the sheet, and it stays put while the list scrolls (#1566).
• Flying to a landmark — or pressing HOME — while a board is open lands the target beside the board instead of behind it, and on a rotated phone the rail scrolls instead of running under the navigation bar (#1566).
• Screen readers now hear the counts: "TARGETS: 831", not just "TARGETS" (#1566).

1.35.30 — 2026-08-27

Fixed

• Dropdown options are readable again on the dark theme (#1577). Opening any dropdown — the colony picker when relocating a colony in the Guild tab, the join-policy and guildmate pickers, the colony selector on the dashboard, the mission and galaxy pickers — showed the option list as pale grey text on a white background on desktop Chrome and Firefox; only the row the OS highlighted could be read. Every option list is now painted on the game's dark background with the normal light text, on all 16 dropdowns at once. Phone pickers that draw their own chrome (iOS) are unchanged.

1.35.29 — 2026-08-27

Fixed

• A busy server no longer crashes an ordinary colony action (#1553). Under heavy load the database connection pool can be fully committed for a couple of seconds; a build, upgrade, ship order or accelerator toggle that arrived in that window failed with a generic server error. It now gets the same "State updated by another operation: try again." the game already shows when two actions race for the same colony — the screen reloads, and the tap can simply be repeated. Nothing is written when this happens, so nothing needs undoing.

Also

• The refusal is deliberate, not a retry: re-queuing on a starved pool would only deepen the starvation and hold the request open past the app's own timeout. A genuinely lost race is still retried as before. Each refusal leaves a warning in the server log, so an exhausted pool is visible even though the player-facing answer is no longer an error page.

1.35.28 — 2026-08-27

Fixed

• Galaxy map: fleet labels stacked into a blob on every zoom (#1435). Zooming the galaxy overview in or out left the old fleet labels ([TAG] · ships) on screen and drew a fresh set on top, so after a few zooms each label was a dozen overlapping copies. Now every zoom redraws exactly one label per fleet. The brief hitch at the zoom threshold that the same report mentions comes from the map redrawing its whole overlay, and is not changed here. Reported by KingPaladin [STG].

1.35.27 — 2026-08-27

Fixed

• Demoting a guild Member to Recruit no longer silently reverts (#1565). Clicking demote on a Member showed the loading pulse and then the rank read Member again: the demotion was written, but the probation sweep that runs on every guild read treated the new Recruit as one who had already served their seven days and promoted them straight back. Anyone who had joined more than a week ago (and every old row with no join date) was affected. Now a demotion to Recruit starts a fresh seven-day probation from that moment, so the demoted player stays a Recruit and rises again only after serving it. Genuine recruits who served their days still graduate as before. Officer → Member was never affected.

1.35.26 — 2026-08-27

Fixed

• The Kepler bulletin no longer announces old colonies as newly founded after a failed galaxy scan (#1543). The Observatory scans the galaxy system by system before each bulletin. When one of those calls failed, that system was silently dropped from the stored map, and the next bulletin opened with "New colony founded by ... at ..." about colonies that had been there for weeks, under their real founder's name. A system that does not answer is now treated as unknown rather than empty: its colonies are carried over from the previous scan, and systems that did answer stay authoritative. A first scan that is incomplete is not stored at all, so the first reference map is always a full one. A network timeout on a single system, which used to cancel the whole bulletin for the day, now counts the same as any other failed system and the bulletin still goes out.

1.35.25 — 2026-08-27

Fixed

• The Contested Space board no longer reads "out of contact" for everyone (#1556). On the Season tab, the guild board under the leaderboard was asking the server without the player's session, so the server refused every request and the panel showed "The Verge is out of contact." to every player, on every device. It now asks with the session like the rest of the tab, and shows the contested planets and the guild standings.

1.35.24 — 2026-08-27

Fixed

A surrendered war no longer reports a score nobody scored (#1545)

When a guild leader surrendered, two things happened that looked exactly like bugs, and two different players reported them as such.

The war closed instantly — days before the countdown it was still displaying — with nothing anywhere saying a surrender had happened. And the final scoreboard showed the guild that gave up with 0 points, however hard it had fought. One war ended 40,216 to 0 against a side that had earned about 10,000.

Neither was broken. Both were the same shortcut: the server recorded a surrender by overwriting the surrendering side's score with a zero, so that whoever had more points would come out the winner. It picked the right winner and published a result that never happened.

Now the verdict reports the points both sides actually earned, and the Activity card says who conceded and that the war ended early because of it — naming the guild, its own score, and who pays the tribute. Surrendering still concedes the war immediately and still costs the 8% tribute; nothing about the mechanic changes.

The war countdowns also stop presenting their number as an appointment. They now read "up to 4d 6h to the end", because either side can surrender and end the war that minute.

One thing this does not do: verdicts already in your feed keep the score they were published with. The real number was never stored for those — the zero is all there is — so they are left as they are rather than guessed at.

1.35.23 — 2026-08-27

Fixed

• Playing through Chrome's auto-translate no longer breaks every action (#861). The translator rewrites the page underneath the game, and every action then died behind a "Signal lost" panel that read as a connection problem — one commander spent a day chasing a network ghost while his colony was under attack. The game now asks the browser not to machine-translate it (a translated page also silently froze its live numbers), so the automatic path to the crash is gone. This is a request, not enforcement: if something still rewrites the page — a forced translation, a page-rewriting extension — the game now says so plainly ("Outside interference"), shows the real error text, and offers Reload instead of blaming the connection. Inside the iOS/Android app nothing ever translates, so nothing changes there.
• A crash above the tabs is no longer a blank screen. The HUD — the resource strip, the inbound-attack alarm — had no safety net: an error there took down the whole page with no message and no way back. A game-wide net now catches it with an honest panel: the game had to stop, your empire is safe, restarting takes you back to your base.

1.35.22 — 2026-08-27

Fixed

• The guild notice board is now checked before it goes on the public directory (#1496). A guild's message of the day is printed on the guild directory, on the suggestions panel and on the guild showcase — read by anyone browsing guilds, member or not. It accepted anything at all: 280 characters, stored as typed. The guild name on the same card has been checked since the day the name filter went in, so the least-checked text on the card was the one with the largest budget in front of the widest audience.

  It now passes the same kind of door. Text that breaks the card is refused: characters that reverse the rest of the line, invisible ones padded into a word, letters buried under a tall stack of marks, and a board taller than eight lines. Language that is not allowed is refused too, including the spelled-out and spaced-out spellings people use to get a word past a filter.

  What a notice board is still allowed to be has not narrowed: line breaks, punctuation, emoji — the family and flag ones included — and any script, with its accents and marks intact. A raid schedule, a welcome, a joining rule all read exactly as before.

Changed

• Saving the notice board no longer stutters. Adding a stray space or a trailing newline used to leave Save lit and do nothing visible, and every click on it wrote another "updated the notice board" line into the guild log everyone reads. The panel now treats a whitespace-only edit as no edit, and when a save is refused the reason appears directly under the Save button instead of below the log, where a phone screen hid it.

• Changing only the join policy no longer announces a notice-board change, and changing only the board no longer announces a policy change. The guild log now names the one thing that actually changed.

1.35.21 — 2026-08-27

Fixed

An expelled commander stops acting like one of your guild (#1532). Kick a synthetic empire out and, until now, it never quite noticed. The panel updated and the member was gone, but in galaxy chat they carried on introducing themselves as one of yours — for the rest of the day, and the day after that — while having nothing to say about the war your guild was publicly fighting. Anyone could open the member list and catch the contradiction. Now the expulsion reaches every place that remembered: the moment they are out, they stop claiming your house.

They also stop quoting your guild's numbers. Their research standing, council directives and season position used to freeze at whatever they were on the day of the expulsion, and got repeated forever — "your guild is 2 of 9 with 4100 season points", from a commander with no guild at all. Same for your raids: a joint raid they were part of kept being announced with your members named and a countdown stuck at 42 minutes, about a coordinate anyone could go and look at.

And they stop flying missions for you. This one was not just talk. An expelled commander went on launching interceptions to defend the colonies of the guild that had expelled them, indefinitely. Their fleet now leaves with them.

Guild councils no longer go quiet for a round (#1532). The same stale membership picked who opened a guild's council discussion. When it picked someone already expelled, the message was refused and that guild's council simply did not happen that round, with nothing said about why. Openers are drawn only from members who are actually still there.

A commander you take back keeps their guild (#1532). The flip side, and it would have been a new bug rather than an old one: re-invite someone you had expelled and the game needs to notice that too. It does.

The Kepler Observatory stops announcing colonies founded by "?" (#1532). When a newly spotted colony came back without a name attached, the galactic bulletin broadcast it literally — "New colony founded by ? at coordinates 4:1:1" — on the most-read line in the game. The colonisation is real and still gets announced; only the unknown founder is left out of the sentence.

1.35.20 — 2026-08-27

Fixed

• The Colony tab now warns you that a Trade Portal on a secondary colony does nothing (#1240). The Market always reads the Trade Portal on your mother colony, so one standing anywhere else adds no offer slots and no cargo. That was already said in the Base build menu, on the building's card and by the Advisor — but the Colony tab is a second building list, and there the Portal appeared with its price and a working Upgrade button and no warning at all. You could level it, at full price, again and again, for no effect. The row now carries the same sentence the build menu carries, and it stays on the row whatever the row is doing: while an upgrade is already under way, and on a Portal that has reached its maximum level.

Also

• The warning is now written in one place and printed by both lists, so the two can no longer drift into saying different things about the same building.

1.35.19 — 2026-08-27

Fixed

• A commander can no longer be called nothing at all (#1516). Some characters are letters to a computer and blank to the eye — a Hangul filler, an Egyptian blank hieroglyph, a mark that draws nothing on its own. Colony and guild names were closed to them a while ago; commander names were not, and a commander name is the one drawn on the leaderboard, beside every chat line and in every battle report. That meant a name could be three blank characters wide and read as empty space, or be "Kael" plus one invisible mark — a second, separate account that looks exactly like the first everywhere you would meet it. Both are refused now. No existing commander is affected: every name in the galaxy was checked against the new rule and none of them fails it, so nobody has to rename.

• The message you get when a name is refused now tells you something you can act on. It used to list the characters allowed, which was no help at all in the case that matters most: a name that looks perfectly normal and is refused anyway, because something invisible is hiding in it, or because your phone quietly turned the apostrophe you typed into a curly one. It now says to retype the name, and to leave out any curly quote, long dash or invisible character that keeps coming back — which is the one move that works when it is your own keyboard producing the character. The same rule is now stated the same way on the sign-up screen and in the rename dialog; they used to disagree about whether an apostrophe was allowed, and it always was.

1.35.18 — 2026-08-27

Changed

• The Wanted list now lives only on the Leaderboard (#1526). The Galaxy screen used to show the same list twice over: a "Wanted" chip on the political map that opened it from the bottom, and the full board sitting at the top of the system view. Both are gone — the Leaderboard, the "who is who" screen, keeps the one copy, unchanged. Inside a system, wanted worlds still glow with the bounty colour as before, and the Targets sheet still prints the bounty on its rows; the Leaderboard is now the single place that lists everyone with a price on their head.

1.35.17 — 2026-08-27

Fixed

• Bot guilds at war now actually fight their enemy (#981). Until now every synthetic empire in a guild at war — or in a League match — behaved as if it had no enemy at all: its war council's vote carried none of the grudges it held against the other guild, its raids and missions kept picking their usual targets instead of the enemy's colonies, and it could not even name who it was fighting. The agents read the member list of the enemy guild through a field the server never sends, so the list came back empty every time. It now reads the right field: agents at war aim at the enemy guild's known colonies, and League opponents get the preferential targets the design always promised.

  Two things go live with it, both worth knowing if your guild is at war with a bot guild. Razing at the declared-war rate: a bot that fills the siege bar on a colony of the guild it is at war with now rolls its raze dice at the war rate, the same rate it already used against anyone it held a personal grudge against — the war rate itself is not new, only the war being seen. And a bot that leaves or is expelled from its guild drops that guild's war on the spot: it stops aiming at the old enemy the moment it is out, instead of carrying a private war it can no longer explain.

• Bot strategists no longer call enemy commanders by their colony's name. The war-target list an agent plans from named each enemy by their colony — usually the unrenamed default, so five different people read as "Prime Colony" — and that phrasing could reach public chat through the agent's own plan notes. The line an agent uses to talk about its ongoing operation had the same habit. Both now name the commander, and only fall back to the colony's name when the commander is unknown.

Also

• The v1.35.11 changelog entry had a sub-heading with no issue number, which the changelog page's own check refused; it now carries #1072 like the entry it belongs to.

1.35.16 — 2026-08-27

Fixed

A garrison's trip home no longer skips the flight or leaves a battle early (#1046)

Two races on the garrison return leg, both narrow and both real.

If a battle was landing on the colony your garrison was guarding at the very instant its watch ended, the fleet could be banked at home while the host fought one garrison short — and you would read a battle report of losses your ships never took. Now the garrison stays on station until that battle resolves, the same hold every other route home already honoured.

If you pressed Stand down while the game was already bringing the fleet home (or the server restarted at that moment), the paid return flight could be skipped entirely: the ships materialised at home immediately, with no "returning" phase ever shown. Now the fleet flies the leg you paid for and lands when the card says it will.

One residual is left on purpose and written down in the code: if the server crashes inside the sub-millisecond window between deciding to wait and telling the queue to wait, the retry lands the fleet early — which is what happened every time before this fix, so the window is narrower, not new. Closing it fully would require trusting a display column to land fleets, which loses fleets on any clock drift.

Also

• The changelog page check knew nothing of the v1.35.11 sub-heading. The test that reads the real changelog required every sub-heading to carry an issue number or sit on its allowlist; the v1.35.11 entry added a continuation heading with none, so the repository's own green gate had been red since that release. The heading joins the allowlist; nothing on the page changed.

1.35.15 — 2026-08-27

Fixed

A colony name can no longer scramble the feed row it lands in (#1404)

Renaming a colony used to check three things: trim the spaces, collapse the runs, keep it between 3 and 24 characters. Nothing else — no character rules, and none of the moderation that has stood at the door of commander names and guild names since August.

That was survivable while a colony name only sat on the map and you had to go and look at it. It stopped being survivable when the defender's colony name started arriving in the Activity feed of every attacker and every garrison ally after a fight. A name built to damage the row it lands in was reaching people who never went looking for it: a single invisible right-to-left character — one character, well inside the length limit — mirrored the rest of the sentence around it, so the outcome word and the loot figures came out backwards and unreadable in every opponent's feed. A 24-character slur got in through the same door.

Colony names now pass the two doors the other names already use: a character guard, then moderation. Invisible characters, symbols and emoji are refused with a message that says which rule broke and what to try instead. Everything a colony name legitimately is still works — letters in any script, digits, and ordinary punctuation, including the : and # the game writes into its own Outpost 1:10:1 names.

Guild founding had the same hole, and gets the same guard. It had the moderation check but no character rule at all, while a guild name is stamped next to every member on the leaderboard, in chat and on the bounty list — further than a colony name travels.

Also

Writing this guard turned out to be mostly a lesson in not breaking languages. Four review rounds each found a spelling the previous version refused: Persian and Hindi need the invisible joiners the first draft banned outright; Tibetan, pointed Hebrew and vowelled Arabic stack accent marks deeper than the first limit allowed; and Burmese given names — ကျော်, one of the commonest — stack deeper still. The rules are now measured against twenty-six real strings in thirteen scripts rather than guessed, and the deepest correct spelling any of them produces is what sets the limit. If your language needs a character this refuses, that is a bug and worth reporting.

1.35.14 — 2026-08-27

Fixed

Your own fleet no longer blocks the system it is flying over (#1468)

Send a fleet on a mission to or inside one of your own systems, and on the galaxy map that system became almost impossible to open: every click landed on the flying fleet's card instead, for as long as the fleet was in the air. The mark of a fleet in transit is drawn on the very system it is crossing, and the map was answering the mark first without ever asking which of the two you were actually pointing at.

Now it asks. From the sector view inwards — where a click opens the system — whichever of the two you clicked nearer wins, and standing inside a system's flower the star keeps the click unless you land on the fleet mark itself. Clicking a fleet still opens the fleet. At the widest galaxy view, where a click only steps the camera closer, nothing changes.

Two smaller things came out of it. A fleet whose heading is not yet known draws no mark at all, and one of those could still swallow a click aimed at the system underneath it; it cannot any more. And the size the map uses to decide "did you click the mark" is now the size the mark is actually drawn at, so the two can no longer drift apart.

1.35.13 — 2026-08-27

Fixed

• A guild could end up with a leader who was not in it (#1500). Founding a guild and joining one at the same moment could leave the guild you founded with your name on it and nobody inside — and a guild with no members cannot be disbanded, because there is nobody left to leave it. Its name, tag and colour were held for good. Both actions now claim you in the same step that checks you, so whichever one loses simply says you are already in a guild and leaves nothing behind.

• Founding a guild told you the wrong thing first. If you were already in a guild, or the name or tag you picked was taken, you were told about it only after the slowest part of the request had already run. Those answers now come back immediately.

Changed

• Colony renames and guild foundings now share a limit of twenty new names an hour (#1500). Every name you pick is checked before it is stored, and that check is the expensive part of both actions. There was nothing stopping a script from renaming a colony over and over to run the bill up, so there is now a ceiling — and past it you are asked to wait an hour and try again, rather than being told anything is wrong with the name you chose.

  Renaming stays free and unlimited in every sense that matters at the table: an empire tops out at ten colonies, and re-using a name that has been seen before does not count. Twenty brand-new names inside one hour is not something playing produces.

Also

• Under the hood, the store of already-checked names used to empty itself completely once full. It is shared between sign-up, guild and colony names, so one busy corner of the game kept throwing away the answers another had already paid for. It now forgets one name at a time, oldest first.

1.35.12 — 2026-08-27

Nothing changes in the game this time.

Changed

• How releases are checked before they reach you (#1520). Every change to Novarchy has to pass the full test suite — about 16,000 checks on the game engine and screens, plus 3,500 on the empires you talk to in chat — before it is allowed near the live server. Until now that suite ran on the developer's laptop; from this release it runs on a dedicated build machine, on the exact commit that is about to go live. What you play is unchanged; what changed is that the machine standing between a change and your empire is no longer the same one someone is typing on.

1.35.11 — 2026-08-27

Fixed

Contract rewards are priced on the planet your colony is actually on (#1072)

Every reward on the Contracts board — the daily cards, the Daily Wrap, the weekly arcs and the Weekly Chest — is sized from how much your empire produces per hour. That figure was being worked out from a colony record that, on some worlds, had never had its planet written into it. Those colonies were priced as though they sat on no planet at all: none of the world's production bonuses counted, and none of its penalties did either.

The claim did not even agree with itself. The check that trims a reward down to fit your warehouse already knew which planet you were on, while the price did not — so one payment was being worked out on two different worlds.

Both halves now read the same planet.

Some rewards go up, some go down, and which way is a property of the colony rather than of the planet. Every world gives a bonus on one resource and a penalty on another, and some worlds help or hinder your power as well — so what matters is what that colony mines and whether its reactor covers it. A colony whose mines lean towards its planet's strong resource was being underpaid; one leaning on its weak resource was being paid over the odds. In our own measurements the swing ran from roughly a tenth up to a tenth down, and on an under-powered colony an icy world and a volcanic one trade places entirely.

Whichever way yours moves, the new figure is the one your planet was owed all along.

Only colonies whose world had never been recorded were affected. A colony's world is written down for good the moment anything is built there, or the first time it is fought over — so this had already narrowed to long-dormant colonies, and it is now closed for all of them.

Also

One knock-on, if you have a "win battles" card open right now (#1072)

The same hourly-production figure does a second job: it sets how much you have to destroy in a battle for that battle to count towards a win battles contract. Tying the bar to your own production is deliberate — it is what stops the card being farmed — and the card already tells you the bar is measured at claim time. So correcting the production figure moves the bar with it, exactly as finishing a mine or letting a guild boost lapse would.

The effect is a one-off and a small one. On the first board refresh after this update, a win battles card still in progress may tick by one win if a battle in that week happens to sit within a few percent of the bar. A card you have already collected cannot change. If your production is under 1,000 or over 20,000 an hour, the bar does not move at all.

1.35.10 — 2026-08-27

Fixed

• A colony or guild name can no longer be made of invisible characters (#1509). Some Egyptian hieroglyph signs draw nothing at all, and the name check let them through. Three of them made a colony whose name was simply blank — and since a defender's colony name is printed in the Activity feed of every attacker and every ally in the garrison, that blank name landed in other players' feeds. The same characters could found a guild with a blank name on the public directory.

  There was a quieter version of the same trick: sixteen hieroglyph marks that draw nothing of their own, so a commander could register a name that looks exactly like someone else's — Kael and Kael with one invisible mark on the end were two different names on screen and one name to the eye.

  All of it is now refused when you rename a colony or found a guild, with the same "invalid characters" message every other invisible character already gets. Genuine Egyptian hieroglyphic names still work, including the marks, when they are attached to a real sign the way the script actually uses them.

  Names already saved are untouched and keep showing as they do today; the check only runs when you set a new one.

1.35.9 — 2026-08-27

Fixed

• Signing in with Google no longer lets an unmoderated name into the game (#1495). Creating an account with Google took the name Google supplies and made it your commander name with no checks — where the guest login, the in-game rename and Sign in with Apple all filter it. A commander name is the most-seen text in the game: the leaderboard, chat, guild rosters, every battle report. So a Google account whose display name was a slur, an impersonation of staff, or invisible layout-breaking characters walked straight in wearing it. New Google accounts now pass the same character and blocklist checks the other doors use; a name that cannot pass gets a neutral Commander … placeholder instead of blocking the sign-in. Existing players keep the name they already have — this closes the door without touching anyone already inside.

1.35.8 — 2026-08-27

Fixed

• An empire that offers you help in guild chat now remembers it when you write to them privately (#1486). Someone offered, in your guild channel, to walk you through the shroud — you took them up on it in a private message, and they answered like a stranger: "explain what exactly, you opened with a demand and no subject." Every channel was read on its own, so a line said in the guild room simply was not there once the conversation moved to a DM. Now a private reply knows what that empire said to you in the room you share, and answers as somebody who remembers saying it.

  It only ever recalls what that empire itself wrote, only if you are currently in the guild together — where you can already read every one of those lines yourself — and only from the last couple of days. Nothing is stored: the channel is read fresh each time.

1.35.7 — 2026-08-27

Fixed

Bounty hunting against the biggest empires paid almost nothing, and now it pays (#1416).

The share of a bounty you collect is measured against the target's war capital, and on the largest empires on the server that made the pot effectively unclaimable: a hunter who destroyed a whole garrison — 1,370 ships — collected 14.68 Dark Matter of a 3,924 pot. Under a third of a percent, for wiping out a fleet. The floor that produced that number now grows as a square root above 20 million rather than in a straight line, so the same hunt is paid for what it actually overcame.

Changed

Emplacements are capped on a bounty claim, and only on the most heavily armed targets (#1416).

Batteries grow back on their own, for free, one every half hour; ships have to be rebuilt. Lowering the floor without touching that would have let a wall nobody paid for collect half a pot overnight and do it again the next day. So broken emplacements now pay out of a bounty only up to a shared ceiling per 12 hours — shared with everyone who breaks them, not per hunter — and nothing above it. Ship kills are never capped.

Below the knee nothing changed: on a mid-size target a wall-heavy hunt still pays in full and still earns a Chronofragment.

Also

• When the cap refuses part of a claim, your log now says so. Before, a hunter who flattened a wall into a ceiling somebody else had already reached collected nothing and read a battle with no explanation, while the Wanted board still advertised the whole pot. The line names the rule and points at what does work: bring warships.
• Four screens promise the payout — the launch warning, the planet sheet, the leaderboard and the Wanted board — and they now say the same thing, from one source, including the half-pot ceiling on a single battle that none of them mentioned.
• The launch warning is two lines instead of one long one, so the bounty's value is not pushed off a phone screen by the fine print.
• "Collected while you were in flight" now says what the engine has done since fleet recall shipped: it is your first strike landing after your fleet passes the point of no return that still counts as the hunt, and only the first.

1.35.6 — 2026-08-27

Fixed

• "Settle a new world" is no longer handed to commanders who cannot settle anything (#1428). A commander at their colony limit was given the weekly Settle a new world contract with no way to finish it — the only thing that advances it is founding a colony, and founding one is refused before it can happen. With four weekly cards competing for three slots, that dead card came up almost every week, so one of the three weekly slots was wasted, over and over, for exactly the commanders who had run out of room. The contract is now only offered to someone who could still found a colony, and the three weekly slots are filled from cards they can actually complete.

  The check asks the same two questions the game asks before letting a colony ship launch: whether the Command Center of the home colony still grants another world, and whether Colonisation research covers the worlds already held. Either one on its own closes the slot. A colony ship already in flight leaves the card in place — it is about to land and complete it.

  Two consequences worth stating plainly. A commander below Command Center 6 with a single colony will now see one fewer weekly card; the card they lose is one they could not have completed. And a weekly board already dealt keeps the old card until the next weekly reset — at most seven days — because redrawing it mid-week would reshuffle cards they have spent the week working on.

Reported by commander Juan, at ten colonies of ten.

1.35.5 — 2026-08-27

Fixed

• The target list no longer fails when an inactive account is cleaned up (#1220). Opening the galaxy-wide target board could occasionally fail with a server error. It looked random, and it healed itself on the next refresh, so it was easy to dismiss — but it was not random at all: it happened when the daily clean-up of accounts that never started removed somebody at the exact moment your board was being assembled. Whoever had the board open in that second saw the scan fail. The board is now built from a single consistent picture of the galaxy, so an empire disappearing while it is being drawn can no longer break it.

1.35.4 — 2026-08-27

Fixed

You can find every commander again, not just the first three hundred (#1429)

Typing a name into chat's @ selector, or into the guild invite box, could not reach most of the game. Search commander crabdavid and the panel answered "No one by that name." — while they were logged in and playing. They were not buried at the bottom of a list; they were unreachable, and typing more letters never helped.

The search only ever looked at the first three hundred names in the alphabet. Past the three-hundredth, nothing you typed could find you. That was around five hundred of the game's eight hundred and eleven commanders — roughly three in five — and it hit both places the roster is used: they could not be mentioned in chat, and they could not be found to invite to a guild.

Now the name you type is part of the search itself, so the whole roster is looked at. Everything else about the two panels is unchanged: still ten suggestions, still no monuments or factions among them, still never yourself.

An underscore in a name no longer hides its owner (#1429)

An underscore typed into the search box was being read as "any character at all" rather than as the character you typed, which flooded the search with names nobody had asked for and could push the commander you actually wanted out of the results. Underscores are legal in names, so this was a real way to lose a real player. What you type is now what is searched.

A name is found by its own first letter, whatever alphabet it is in (#1429)

Names built from letters like İ were being matched twice, by two systems that disagreed about how to compare capitals and small letters — and the stricter of the two won. A commander called İbo could be looked up, found, and then quietly dropped again before the list reached the screen. There is now one judge instead of two, and the name comes back.

1.35.3 — 2026-08-27

Fixed

An interception could hand back goods a raider had already taken (#1480)

When a picket caught a loaded convoy, it wrote the convoy's manifest back from a reading it had taken a moment earlier. If a raid hit that convoy's home colony in those few moments, the raid took its share of the hold in flight, the raider banked it — and the interception then restored the manifest as it was before the raid. The same goods ended up in two places at once: with the raider, and still aboard the convoy that had just been robbed of them.

Nothing about it required a trick. It needed a raid and a pursuit landing on the same convoy at the same instant, and the convoys it could happen to are the everyday ones — a shipment between worlds, or a redeployment — because those are exactly the fleets that are both worth looting in flight and worth intercepting.

Now an interception refuses to write a manifest that moved while it was working. It reads the hold again, works out the engagement against what is actually still aboard, and only then writes. Nothing is created out of nothing, and nobody pays twice: if the convoy is still being fought over elsewhere, the pursuit simply comes home having found nothing rather than inventing a result. Battles, losses and reports are unaffected — what changes is only that the numbers now add up.

1.35.2 — 2026-08-26

Fixed

Mining Automation now does what its card always said (#1462)

The node promised +20% Ferrolite and Xenogas production and the game never applied it. You could research it, watch your rate, and see no change at all — which is exactly what SpartanEcho reported in universe-chat. It had been that way since the tech tree shipped: the effect was written on the card and never written into the game.

It works now, on every source of those two resources — extractors, refineries and the Stellar Forge alike. Data production is deliberately unchanged.

If you already owned the node, the bonus turns itself on. There is nothing to research again and nothing to claim: your Ferrolite and Xenogas rate simply steps up the next time your colony ticks. Anyone with a Nanoforge or a Terraformer owns Mining Automation, since both require it — so this is most established empires.

How big a jump you see depends on how developed you are. A young colony gains the full +20%. An empire already running maxed drills and Economy research sees roughly +7% on top of what it had, because the bonus adds into that total rather than multiplying it — the same way Deep Drills and every other production line has always worked.

Open Power → Details to see it: Mining Automation now appears there by name, next to your other production bonuses, so you can see exactly where your rate comes from.

Rankings score resources spent, not resources held, so nobody's position on any ladder moved when this went live.

1.35.1 — 2026-08-26

Fixed

Loot taken from a convoy in flight stays taken (#1459).

Since the raid rework you have been able to reach the goods a defender put in the sky — a laden convoy that had left the colony and had not yet delivered. Emptying somebody's warehouse into a transport no longer hides it.

That half worked. The other half quietly gave the goods back. When the convoy finished its trip, it landed carrying the hold as it was before the raid: the attacker kept their share, the defender got the same units a second time, and the difference was created out of nothing. A redeployment was worse — its cargo was not put back but paid out twice, once to the raider and once into the destination warehouse.

Every landing now checks the hold it was priced against. A convoy that is raided mid-flight comes home with what is actually left, a founding endows its new colony with the surviving dowry, and a redeployment delivers what the raid left behind. If a battle lands in the same instant, the arrival simply re-prices itself against the new hold rather than overwriting it.

Two smaller things travelled with it. A pursuit that shoots down part of a convoy's escort no longer has those hulls restored by the convoy's own arrival — the ships the picket destroyed stay destroyed. And a colonization now retries a lost race the way every other arrival already did, so a founding that collides with a battle completes instead of stalling.

1.35.0 — 2026-08-26

New

The Discord server shows the live numbers (#1471)

Before: someone who opened the Novarchy Discord found chat and nothing else. There was no way to tell from the outside whether the game had forty players or four hundred — and if the last message happened to be a few hours old, that silence read as an answer even when it wasn't.

Now the top of the sidebar carries a category called 📊 NOVARCHY LIVE with five channels whose names are the numbers, refreshed every ten minutes:

• 👥 Players — human empires
• 🆕 New today — commanders who signed up since midnight UTC
• 🟢 Online — humans who did something in the last quarter of an hour
• 📅 Active today — humans who did something today
• ⚔️ Battles 24h — battles fought across the galaxy in the last day

They are labels, not rooms: nobody can join them, and they cost nothing to read.

The numbers mean people. The four counters about players count humans only — the synthetic empires that carry the AI badge in game are excluded, using the same definition the internal dashboard uses, so the two can never quietly disagree. Without that, "Online" would never have dropped below the machines, and the sidebar would have advertised a busy server at four in the morning. The battle count is the deliberate exception and counts everyone: it measures the galaxy, not the roster.

If Discord is unreachable or the numbers are late, nothing in the game notices — the counters are an ornament on the sidebar and are built so they can never affect play.

1.34.0 — 2026-08-26

New

The galaxy gets its weather (#1413). The last chapter of the Hive Map redesign — depth, without a single new sign to learn:

• Nebula washes. Three faint tinted clouds now drift behind the sector plates — violet, teal, rose. Their colours deliberately belong to no guild (that rule is now a test measured against the guild colour wheel), and none is strong enough to be mistaken for territory.
• A second weight of stars. The void between the plates carries a handful of soft flared stars over the familiar dust. They are seeded, so your galaxy's weather looks the same every time you return; they keep out of the plates entirely, and they stay dimmer than every real sign on the chart — distance, not destinations.
• A vignette at the screen's edge. The corners of the view now fall off gently into dark, like the edge of an eye. It lives on the screen, not on the map — it never pans, never covers a notice or a panel, and never blocks a tap.

1.33.1 — 2026-08-26

Changed

Nothing changes in the game this time — and that is the whole point of this release.

The engine has been carrying a rule that did not exist. Since late July it computed, for every system, a four-hour "window" each day during which that system was supposed to be attackable, plus a thirty-minute warning to its defenders. It was never connected to anything: no screen showed it, no attack was ever refused because of it, and no warning was ever sent. Players were attackable all day, every day, exactly as they are now and exactly as they were yesterday.

The trouble is that the code around it had started to believe in it. One engine comment stated that the window "time-boxes the permanent galaxy's systems to four hours a day", which was never true; the Calendar and the season design document both reasoned about it as a live feature, and the guild design doc's own progress table had listed it as done since 9 August. A recent piece of work came within one step of pushing "you will be vulnerable in 30 minutes" to players' phones — a notification about a threat with no mechanical effect whatsoever.

So the window has been retired rather than switched on, and every place that claimed otherwise now says what is actually the case (#1386). Time-boxed attacks were the wrong answer for a galaxy this size: the game's real contested territory, the event sector, is deliberately open at all hours, and the relic design had already turned the same idea down in writing — with fifteen players, a siege calendar is a calendar of absences. A season that wants timed sovereignty in future starts from that, not from the leftover arithmetic.

One thing did get stricter behind the scenes: the Calendar's rule that it may only show dates something actually runs on used to be written down in comments and nowhere else. It is now enforced by the compiler, so a future date that nothing enforces cannot quietly appear on your calendar.

1.33.0 — 2026-08-26

New

The inbound alert now tells you which guild is flying the attack (#1387).

Your Observatory has always graded what you learn about a fleet aimed at your colony — the sector it comes from at level 2, the exact origin at 3, the attacking empire's name at 4, and so on up to the full composition at 8. What it never told you, at any level, was whose banner those attackers fly. Eight fleets converging on one of your planets is always a single guild's operation, and you could read eight empire names without the one fact that tied them together.

Level 4 now buys both halves of identity: the empire's name and its guild tag. Your Inbound card reads attacker: Corsari di Vega · guild: [VOID], and the row your guildmates see on the threat board carries it too.

On a joint raid the card states the guild once, for the raid — and only when every contingent really is flying it. A raid whose members changed banner mid-flight claims none rather than picking one, and the per-contingent breakdown carries each. An attacker who flies no banner simply has no guild line.

It is level 4 and not a higher rung on purpose: the galaxy map already names a fleet's guild at that same level, and the alarm about the fleet aimed at your own colony should never tell you less about who is coming than a bystander's map tells them about the very same ships.

Also

The map's hint about what the next Observatory level buys had been under-selling its own level 4 for as long as it has existed — it named the empire and not the guild, though it has shown both all along. It now says what it actually gives you.

1.32.4 — 2026-08-26

Fixed

• Tapping with a finger works everywhere again (#1456) — on a phone, tapping a building in the base view almost never opened its card: the tap was silently read as a map pan, and the game was effectively menu-only on mobile. The same mouse-tuned threshold also made taps unreliable on the sector map, on planets in the galaxy system view, on ship groups in the battle replay, and when confirming a building relocation. All of them now give a fingertip the room it actually needs (a finger rolls 8–15 px between touching down and lifting; a mouse click stays within a couple), so a finger tap selects, opens and confirms exactly the way a mouse click always did. Mouse behaviour is unchanged in the base view, and tightens imperceptibly (by 1 px) on the maps.

1.32.3 — 2026-08-26

Fixed

• Ship counters no longer list the same ship as both "Strong against" and "Weak against" (#1446). The Orbital Scythe's panel showed the Voidlance Cutter under Strong against ×1.5 and Weak against ×2.5 at once — both numbers are real (the two hulls counter each other), but a green "strong" on the side that loses the exchange read as a favourable matchup. Four pairs in the catalog counter each other: Scythe/Cutter, Cruiser/Splinter Lance, and two even ones against the Praxis Interdictor. A mutual pair now sits only on the side that wins the exchange, and its chip prints both numbers: the Scythe reads Weak against — Voidlance Cutter ×2.5 · you ×1.5, the Cutter reads Strong against — Orbital Scythe ×2.5 · back ×1.5. An even exchange stays on both rows and reads ×2 both ways, the same chip on each side. The Ship reference legend explains the two-number chip. Same rule on the Shipyard ship card; the multipliers themselves are unchanged.

1.32.2 — 2026-08-26

Changed

A raid now reaches the cargo you put in the sky (#1344)

Emptying a warehouse into a fleet no longer hides it.

Until now, goods left your colony the moment a transport launched and nothing in a battle had ever looked at a ship's hold. So with an attack inbound you could load the warehouse onto a convoy, let the raid land on an empty colony, and take everything back — by recalling the fleet, or more cheaply by aiming it at a slot that would not accept it, which turned it round with the whole hold aboard. A raid now takes from the convoys a colony has in the air and has not delivered anywhere: on the way out, on the way back, and including a pickup carrying home what it just collected from that colony.

Your warehouse always pays first. A raid only reaches the convoys for the part of the loot the stock on the ground could not cover, so a colony with nothing in flight fights exactly the raid it fought yesterday. Nothing else about looting changed.

Evacuating still works, and still costs a fleet and a real trip. Goods that actually reach another colony are safe from the raid you flew them away from — and lootable at the colony they reached, which is the point. What no longer works is sending them nowhere and calling it safety.

Three details worth knowing before you next load a convoy:

• Cargo already in flight is included, from the moment this update goes live. There is no grace period. If you have a warehouse in the sky right now to keep it away from a raid, the next raid that lands will find it.
• One raid reaches at most 40 convoys, and it takes the largest first. Splitting a warehouse across many small holds does not protect the big ones.
• A fleet coming home from a raid with loot is untouched, and so is a colony hit by a faction sortie — a punitive expedition destroys what it takes rather than carrying it away, and this rule exists to stop goods being hidden from a raider, not to licence burning a convoy nobody could have captured.

This closes the last of the three hiding places: #1307 shut the pickup escrow, #1332 the market board, and the rule behind all three is the same one — goods stop being lootable when they arrive somewhere, never when they are merely promised or merely loaded.

1.32.1 — 2026-08-26

Fixed

Pointing at a building on the base map now names that building, not the one behind it (#1348)

On the colony map, hovering or tapping a building could describe a completely different one — the building hidden behind it. The clearest report of it: a Xenogas Refinery whose tooltip read «lvl 14 · Under construction → lvl 15» while the level badge on that same artwork read «13». Both numbers were correct. They belonged to two different buildings.

The map was deciding what you had pointed at by looking at the flat ground tile under the cursor, while what you were actually pointing at was the drawing. Buildings are drawn rising up and over the tiles behind them, so the ground beneath a roof belongs to a tile one to three steps further back — and that was the building you got. It has been reported three times over the past months, each time as a level that looked stale.

The map now asks which artwork is actually painted where you are pointing. In practice:

• the tooltip, the highlight and the cursor all name the building you can see;
• tapping a tall building opens its own card. Before, tapping its upper body could open the wrong building's card, or count as tapping empty ground and clear your selection;
• a building still under construction — a blueprint, with its scaffold and countdown — can be pointed at where it is drawn, instead of handing you whatever stands on the ground behind it;
• a blueprint never steals the pointer from a solid building you can see straight through it.

One consequence worth knowing, because it is deliberate: a building whose artwork is completely hidden behind a taller neighbour can no longer be selected by clicking on that neighbour. Its card is still one tap away from the Build menu.

1.32.0 — 2026-08-26

New

The launch screen tells you which world you are attacking (#1338)

A colony's power plants run on the world underneath them — a lava world gives them more energy, an ice world less — and their Shield Bubble, Orbital Cannon and batteries all run on that power. The Advisor has been sizing your fleet against that since the last release. It just never told you.

So the same untouched spy report could ask for almost a third more hulls than you remembered, with nothing on screen to say why. That reads like an Advisor having a bad day, and the natural answer — trimming the fleet back by hand — is exactly the mistake that loses it.

The target's world is now named where you compose the fleet, on the phone and on the desktop alike, with what it does to a defender's energy and one line explaining what that moves. It comes from the coordinates themselves, so it is still there when the Galaxy view fails to load — the one moment you have least to go on, and the moment the forecast was quietly using it anyway.

Three things it deliberately does not say:

• an ice world's penalty, once its owner has engineered it away. An Outpost or a Terraformer standing on that colony cancels the malus, and the Advisor already knows it from your spy report. When the world no longer costs them anything, the screen stops claiming it does.
• anything about the Pirate Beacon. The Beacon has no power plants and no Bubble to run on them, so its world moves nothing. It is named; it is not explained.
• anything about your own colonies. A harvest or a transfer aimed at your own planet is not an attack, and the panel no longer describes your own Bubble as a stranger's.

The wording says "energy production", and it says the defences follow it only while the defender is short of power — because a colony with energy to spare feels its world not at all, and a figure without that caveat is the same misleading number in a nicer font.

1.31.0 — 2026-08-26

New

• A pickup can be recalled (#1333) — freighters sent to collect resources from another of your colonies were the one fleet you could never call back. The mission card said so, and so did the notice before you pressed Send. They now carry the same Recall button as any other mission, on the same clock: you can turn them round at any point before the last quarter of the flight.

  Turning one round costs you nothing but the trip. A pickup flies out with an empty hold — it only loads at the far end — so the fleet comes home empty and not a single unit moves in either colony. The confirmation says exactly that before you commit.

  The old rule was a leftover rather than a design choice. Until two days ago the manifest was set aside in the destination colony the moment the fleet lifted off, and a recall would have carried that stock home in a fraction of the round trip. That reservation was removed when it turned out to double as a hiding place from raiders, and the ban on recalling outlived the only reason it ever had.

Also

• Pickups launched before the reservation was removed still cannot be turned round, and the server says so rather than doing it. For those older orders the goods really did leave the colony at launch, so recalling one would have destroyed them. No such order has been in flight for some time; the rule is there so it stays that way.

• On the iOS and Android apps the Recall button arrives with the next build. Until you update, the app will keep showing the old line saying a pickup cannot be recalled — the browser already has the button, and the same fleet can be recalled from there.

1.30.4 — 2026-08-26

Fixed

• A battle now sizes a defender on the world it actually sits on (#1339). Every planet's traits — the ice world's power penalty, the lava world's bonus — are versioned, and a colony's saved copy can be out of date. Everywhere the game reads such a colony it quietly brings those traits up to date from the planet's coordinates. The battle did not. So a colony with an out-of-date copy fought under today's rules with yesterday's world: on an ice world its Bubble, Orbital Cannon and emplacements stood about 11% stronger than the launch preview had just promised the attacker. Preview and battle now agree on every colony, not just on the recently active ones.

• The spy report, the guild rescue card and the Advisor were reading the same stale copy. Fixing the battle alone would have left them contradicting it, so all three moved with it. The rescue card is the one that mattered most: it is the only screen a guildmate can use to size a rescue — espionage refuses a friendly target — and its own promise is that its numbers are the battle's. It was quoting a Bubble up to 11% larger than the wall the raid would meet. The Advisor was worse in a quieter way: it was telling the model "Power yield 100%" for a colony whose owner was looking at the penalty in a red chip, and that fact became a sentence in chat.

Also

• How many colonies were actually affected was checked against the live database rather than assumed: 1336 of 1337 already carried current traits. The single exception is a starter colony that never built anything, so the advantage it was owed applied to nothing. This ships as a correctness fix that closes the gap by construction, not as a live exploit being shut down. For every colony already up to date, nothing changes at all.

1.30.3 — 2026-08-26

Fixed

• The Advisor no longer contradicts itself about your fleet (#1350). Open the Advisor on an attack and it could give you two verdicts two lines apart: a green row reading «Victory 3/3 sims» and, just below it, «0% simulated win rate — even your entire fleet does not reliably win this battle». Same fleet, same target, opposite answers, and nothing on screen to say which to believe. Both were honestly computed — they were simply computing different battles. The row simulated the rows you had actually set and the stance you had actually picked; the Advisor ignored both and fought on class defaults. On a real fleet — 180 warships against a target of 82 interceptors, 41 cruisers and 21 battleships — your own deployment is the whole difference between losing every simulation and winning every one. The Advisor now fights the battle you are about to fight, so the two cannot disagree, and moving a ship to the vanguard moves the advice in the same instant it already moved the forecast.

• A battle nobody can win is no longer called a defeat (#1350). Attack a colony whose last defender is a parked Destroyer — a hull nothing can kill — and the forecast told you «Defeat 0/3 sims» right next to «losses 0%»: you had lost, and it would cost you nothing. That fight is a stalemate, and the forecast now says so: «Stalemate — neither side holds the field, nothing is looted». Your ships are not in danger; the holds simply come home empty. The Advisor's own advice changes with it — it used to tell you to build more ships against a hull no fleet size can kill, and now tells you to pick a different target. A defeat you actually lose ships in still reads «Defeat», a fight where only some simulations end that way still reads «Coin flip», and a battle where both fleets are wiped out still reads «Defeat», because that is what it is.

• The forecast stopped claiming a planetary shield held when it had been smashed (#1350). On a target you could not beat, the row printed «planetary shield holds — the raid bounces» over a shield that had been broken in every simulation — and sent you off to build bombers for an obstacle that was never the obstacle. The shield now reports what happened to the shield.

1.30.2 — 2026-08-26

Fixed

• The Advisor no longer under-sizes your hold when you raid under Pursuit (#1359). Pursuit costs you 10% of your cargo, and the Advisor was sizing the raid's hold as if it did not. It would hand you a fleet with no freighters and say nothing — while the forecast row directly underneath lit up «hold full, add freighters». Two numbers on the same panel, disagreeing, and the raid came home leaving loot on the ground that one more freighter would have carried. Against a 3,000 pool it planned for a hold of 1,620 when the raid actually flew with 1,458, for loot it had itself measured at 1,500. It now prices the stance where the battle charges it, and adds the freighters the loot really needs.

• And it tells you what that hold costs you in speed. A Freighter is one of the slowest hulls you own, and a fleet only ever moves as fast as its slowest ship — so the freighter added to honour Pursuit can hand back the speed you picked Pursuit for. When the hold ship would set the pace, the Advisor now says so, and names both speeds. It does not tell you the trip gets longer, because that is not always true: sometimes the stance's own speed bonus more than covers it, and the arrival time on the same screen is the number to trust.

Raids flying no stance at all — which is most of them — are completely unchanged, and so are Assault and Rampart, neither of which costs you any cargo. Transports, pickups, colonizations and harvests are unchanged too: they never carry a stance to begin with.

1.30.1 — 2026-08-26

Fixed

• The Superconducting Grid's card no longer contradicts your energy meter (#1356). It promised "+8%/lvl energy production on every colony, and nothing at all where the balance is already positive" — so a player who researched it, watched production rise on every colony including the ones already in surplus, and reported it as a bug was reading the card correctly. The engine was right all along: the bonus really is applied everywhere, and power yield stops at 100%, so on a colony whose power already covers its draw none of it reaches your resources, your Dark Matter or your defences. The card said that by denying the part you can see.

  It now says what the node does: power yield is capped at 100%, so the Grid is only felt while a colony is short of power — where power already covers the draw it buys room to build rather than more output. That last part was missing before: on a colony in surplus the extra production is real headroom, i.e. how many more buildings you can raise before the balance tips.

Changed

• The two other places that promised the same +40% now say what caps it (#1356). The energy sheet behind the power pill announced "Superconducting Grid: +40% energy production" beside a production number that had genuinely gone up, with nothing about what happens to power above the balance; it now carries that clause for every colony. The Power details panel listed "Energy production" flush with four bonuses that always pay, and now reads "up to full power yield". The percentages are unchanged — they were never the wrong part.
• The same energy sheet stopped saying buildings "slow down" in a deficit. A shortfall cuts what the colony produces; it has never touched build times, and with "room to build" now in the same sentence the old wording read like a warning about your build queue.

1.30.0 — 2026-08-26

New

Command Center 14 and 25 open a second and a third Warehouse (#1164). These were the last two levels on the ladder that bought nothing but the level cap — CC14 in particular had its description rewritten four times, because there was nothing to put in it.

A colony can now hold up to three Warehouses, and its storage ceiling is the sum of all of them. Storage is the quantity that really binds the late game, so there is finally somewhere to put those levels.

Both rows tell you the whole truth, including the unglamorous half:

2nd Warehouse Site: its storage adds to this colony's cap (the new Site starts at lvl 1).

That last clause matters. A Warehouse at level 25 holds about 1.6 billion; the new Site's first level holds 8,000. The third Warehouse is a full second investment walked up its own curve — at the usual Site surcharge, and paying its own power bill — not a switch you flip.

Each Warehouse contributes its own capacity, and the capacities add. Two Warehouses at level 10 hold twice what one level-10 Warehouse holds — not what a level-20 one holds. Building wide and building tall stay different decisions.

Everything that reads your storage now reads all of it: the ceiling on your resource bars, what a colony can accept from a convoy or a transport, what survives a raid, guild aid and treasury donations, war tribute, and your daily market intake budget. The Warehouse card on the Base tab still shows what that Warehouse contributes on its own, and now says so, naming the others.

Every Command Center level from 1 to 30 now grants something real. #1164 fills the last two gaps left by #1159, after #1162 and #1163 took the rest. There is no longer a level on the ladder whose description is "this buys you the level cap".

Changed

The Anchorage names the number that actually decides (#1164). The Deliver-to picker used to label each colony by its Warehouse level, which stopped meaning anything once a colony could hold three of them — a two-Warehouse colony at level 15 has far more room than a one-Warehouse colony at level 16. It now shows the free space in the resource you are shipping, which is exactly what the charter is accepted or refused on. The crate's own "Warehouse N" line now reads "equivalent", because it is a size guide and not a requirement.

1.29.0 — 2026-08-26

New

The six Dark Matter hulls finally look like themselves (#1418). The Voidlance Cutter, Umbral Reaver, Continuum Hauler, Aegis Bastion, Breachhead Monitor and Praxis Interdictor shipped with somebody else's picture on their card and a small "concept art coming" label in the corner — a Cutter showed you a Titan, a Hauler showed you a Colonial Ark. All six now have their own art on the shipyard card, in the detail panel, and as real 3D models in the battle replay, where they used to appear as a flat billboard of the wrong ship. The label is gone.

Two new steps on the rarity ladder (#1418). The card of every hull says how far into the game it sits, and the top of that ladder had stopped making sense: a hull you reach only by taking a Doctrine to level 20 wore the same colour as the Titan, which you research in your first week.

• The six Doctrine hulls are now Legendary, in gold.
• The Destroyer has a step of her own above them — Mythic, in red. She is the only hull that will ever wear it.
• The step below is no longer called "Epic". It is Ascendant: the hulls you climbed the tech tree to reach — Titan Dreadnought, Colonial Ark, Splinter Lance, Spectral Corsair.

Nothing about what these ships do changed — no stat, no cost, no counter. This is what they look like and what their card calls them.

1.28.0 — 2026-08-25

New

A second Defense Batteries at Command Center 22 (#1163). Both sites' emplacements deploy as one group, on the one regeneration clock they always shared. The new site starts at level 1, so what it adds the day you build it is two emplacements — not a doubling.

A second Planetary Shield at Command Center 29 (#1163). The two bubbles add their HP, and the pool your defense absorbs with grows by the sum. This one is a second full investment rather than a switch: the new site starts at level 1, every level of it is bought on the Shield's own steep curve at the usual second-site surcharge, and your colony pays the power for both.

Those were the last two levels on the Command Center ladder that gave you nothing but a higher cap. Every level from 1 to 30 now grants something you can point at.

Changed

Scouting a colony with two Shields tells you the truth (#1163). The Bubble level on the target board, the planet sheet, the War Room and the ally-rescue card is now the level whose HP pool matches what is really standing there. Before, a colony with two Shields advertised only the first one — half a wall. If you have been sizing raids off that number, it will read higher on some targets now, and it is the honest figure.

Rebuilding emplacements is priced across both sites. If your batteries stand on two sites, buying them back after a raid costs what those emplacements actually cost to put up. A colony whose batteries are all on the second site can now buy them back at all — before, the game looked at the first site, saw nothing there, and refused.

Also

Emplacement counts, the "wall in the vanguard" marker, the war-value line on your colony's sheet and the spy report all count both sites now. The building sheet of a site you have not built yet no longer reports battle damage to a building that does not exist, and on a two-site colony it says which site's figure it is quoting.

1.27.1 — 2026-08-25

Changed

• The game now sends far less data over the network. Until today every answer the server gave — your colonies, the leaderboard, your battle reports, the galaxy map — travelled as plain, uncompressed text. It is now compressed before it leaves the server, and the same screens arrive between 78% and 92% smaller. Measured on the live game, after the deploy:

  | | before | now | |---|---|---| | Battle reports | 314 KB | 32 KB | | Leaderboard | 263 KB | 35 KB | | Galaxy overview | 31 KB | 2.5 KB | | Your colonies | 45 KB | 6.5 KB | | Research | 27 KB | 5.8 KB |

  The game asks the server for fresh data every 30 seconds while you have it open, so this repeats all day. On a phone it is the change you are most likely to notice: the same session costs a fraction of the mobile data it used to, and less battery spent receiving it. Nothing about what you see changes — only how much of the network it takes to see it.

  Older app versions that cannot read compressed answers keep receiving plain ones, exactly as before, so nothing needs updating for the game to keep working.

Also

• The database was still configured for a machine of unknown size, reserving 128 MB of memory on a server with 62 GB. It is now sized for the machine it actually runs on. This is room to grow rather than a speed-up, and it is worth being straight about that: the measurements behind this release found the database essentially idle, so nobody should expect the game to feel faster because of it. The part you can feel is the compression above.

• The server now waits for the database to actually answer before it starts accepting players, instead of merely waiting for it to launch. Nobody will ever see this working — it exists so that a restart cannot briefly greet you with errors while the database is still warming up.

1.27.0 — 2026-08-25

New

The factions pay Dark Matter (#1289)

Felling a bastion now pays Dark Matter on top of the Chrono shards — 9/18/30/48 by the bastion's step, to every raider who brought a real share of the fleet, with no split. Repelling a faction's retaliation sortie pays its target too: 6/9/15/24. Both channels share one ceiling, 96 DM per rolling 7 days — two full bulwark victories — and the target board now prints the price on the bastion's own row before you launch, including what your week still lets you collect. Wrecks still pay nothing, Event Control garrisons still pay nothing, and the whole haul shows up in your feed under Combat.

Precursor Schematics: one talent, three buildings (#1289)

A new one-time purchase in the Dark Matter panel — 150 DM, needs the Particle Accelerator and Production Singularity 3 — unlocks the whole Precursor district at once:

• Zero-Point Reactor — one level is worth one hundred Fusion Reactor levels. The energy drama has a paid ending.
• Singularity Vault — adds storage on top of the Warehouse, built for the era where a single research level bills in billions. Transports, raid hauls, crates, charters and rewards all respect the new ceiling.
• Stellar Forge — the first building that pours all three resources at once, twenty times the basic producers, and so power-hungry it will not run without the Reactor.

All three come with their own art on the colony grid.

Also

• The Warehouse and Vault cards each state their own contribution, and the HUD shows the colony's real ceiling.
• Apex cards now say what they still require before you tap, and an owned one-time purchase shows as a trophy instead of a dead button.

1.26.0 — 2026-08-25

New

• The Dark Matter Apex is now four Doctrines of twenty levels each (#1301). The two old nodes — Lances and Aegis — become full Doctrines with retuned, stronger per-level effects (shield penetration to 26% at 20; +60% shields at 20), joined by two new lines: the Forge (+2% to all production per level) and the Veil (−1.5% travel time per level, floor −30%). Prices moved from the old runaway curve to four flat bands — 8/20/45/90 DM per level, 815 DM to master a line. Levels you already bought are kept, never truncated.
• Twelve keystones: every Doctrine changes the rules at levels 5, 10 and 15. Overcharge, Sundering and Cascade for the Lances; Hardened Pool, Deep Formation and Last Stand for the Aegis; Cold Fusion, Third Lane and Nanoforge Cascade for the Forge; Cold Trail, Ghost Manifest and Free Jump for the Veil. Each is described on the Apex panel's new ladder, ticked off as you climb.
• Six new hulls, each locked behind Doctrine mastery: the Voidlance Cutter and Umbral Reaver (war), the Continuum Hauler (an 18,000 hold that nothing counter-picks), the Aegis Bastion (a shield tender that never fires and feeds a shared pool), the Breachhead Monitor (halves the share of your fleet the Orbital Cannon can take before the battle; any two Doctrines at 15) and the Praxis Interdictor (breaks the enemy rear's sanctuary; all four Doctrines at 5).
• Free Jump is visible on the launch form: when your daily free warp is open the row says so and strikes the price; when it is spent, the form says when it reopens instead of pretending you are short of Dark Matter.

Changed

• The synthetic empires learned all of it: they buy Doctrines by archetype, respect the new keystones in their energy and build decisions, and price warp jumps from the colony the fleet actually leaves from.
• Apex hull padlocks now point at the real screen (Research → Apex) and state the whole requirement — a choice-gate says "2 of 4 Doctrines", never a list that reads as all four.

Also

• The battle report credits fire to the correct side when a Bastion fights, the formation editor shows the shares Deep Formation actually produces, and the building sheet's energy row names Cold Fusion when it discounts the Accelerator.

1.25.2 — 2026-08-25

Fixed

The weekly war board said it paid for wreckage, and it does not (#1251)

The banner over the standings read "every battle — won or lost — earns points for the value destroyed". That names the smaller half of the score. Value destroyed is capped in every battle; the loot you carry off an attack you win has no ceiling at all, so one big haul can be worth several times everything the destruction term can pay. A player worked this out from numbers quoted in chat and told us the description was what had misled him: on this board a looter outranks a fleet-killer, and the label said the opposite.

The banner now leads with the truth — the loot you carry off is the only thing on this board with no ceiling — and adds the parts a player needs to explain their own score: value destroyed pays both sides of a fight between empires whether you win or lose, but capped, and worth less against pirates or on a target that was harvested recently. It also names the fights that pay nobody, which were never written down anywhere: in an ambush or a Campaign level only a win scores, and a Wraith sortie scores for neither side. Winning and repelling add flat bonuses, the underdog and relic bonuses multiply different things, and a joint raid or defence splits the battle's points among everyone who took part.

Deliberately, the banner prints no rates. Four different parts of the game credit this one board, each with its own rules, and two drafts that tried to print the price list were found stating things the engine does not do. The exact numbers stay on the Season panel's chips, which now read them from the engine instead of having them typed in — so the day the balance moves, the chips follow.

Known limitation: you can no longer check the arithmetic from the war board itself. A page that explains the weekly score properly — one rate per case, instead of one averaged sentence that is wrong for three of them — is the next step, not part of this fix.

1.25.1 — 2026-08-25

Fixed

• Recall and Stand down work from the Overview tab (#1253) — every fleet in flight on the Overview screen carried a Recall button, and every garrison a Stand down button, and pressing either did nothing at all: no confirmation, no order, no error. The same buttons on the Missions tab worked, which made the Overview ones look broken rather than missing. They now give the order from either screen, with the same confirmation, the same warnings and the same refusal messages.

1.25.0 — 2026-08-25

New

The galaxy map's chrome became one instrument (#1396). The row of four filter chips is gone; a scanner bar now reads the chart back to you:

• A level breadcrumb — GALAXY ▸ SECTOR ▸ FLOWER, lit where you are. The chart has always had three zoom grammars behind the two-step tap; now it says which one it is showing.
• One Layers button owns the four map filters and the complete map key in a single sheet. On the phone the key used to hide inside THE PRIZES cartouche; now it renders whole, one tap away.
• THE PRIZES became LANDMARKS: a strip of the chart's own signs — a diamond per relic, filled in the holder's colour and hollow while free, a chevron per bastion — and every glyph is a button that flies the camera to its place. Expanding the strip still shows the familiar rows.
• "Go to" reaches the phone. The goto vocabulary — 2:7, deep-space names like 4:2 NE3, home — could only be typed on desktop. The phone now has its own door beside the breadcrumb, and a refused entry says "the chart can't take you there" and keeps your text instead of doing nothing in silence. The sheet grows by the keyboard's overlap so the field stays reachable while you type.

Also: the fleet-contact chip (▲) rides the bar's trailing slot, and the collapsed LANDMARKS panel keeps the old cartouche's exact height, so the map's phone framing budget still holds.

1.24.0 — 2026-08-25

New

The Activity feed names the colony that was attacked (#1397)

Until now a Battle card in your Activity feed told you who fought and at which coordinate, and never what the place is actually called:

1:6:1: defeat · loot 0 Fe · 0 Xe · 0 Da.

But nobody talks to their guild in coordinates. As the player who suggested this put it: "'Itz Iceee' is the more commonly known name of planet 1:6:1 … but is nowhere to be found on the activity page."

The colony's name now comes first, with the coordinate kept right beside it so you can still find the place on the map:

Itz Iceee (1:6:1): defeat · loot 0 Fe · 0 Xe · 0 Da.

It shows on every kind of Battle card — an ordinary clash, a raid on an undefended colony, and the razing that ends a siege — and everyone who was there sees it: the attacker, the defender, and an ally who flew in to garrison. Pirate Beacons get named too, which is the target most of you see most often.

Two things worth knowing. Battles fought before this update keep showing the bare coordinate: the feed reads each battle exactly as it was recorded at the time, and a recorded battle never changes. And the name shown is the colony's real name, the one its owner set — already public on the galaxy map and the planet sheet — so if your team calls a planet by a private nickname, what appears is the owner's name for it, not yours.

Also

Thanks to jaybee / commander Usherin for the suggestion.

1.23.1 — 2026-08-25

Changed

The ship picker puts the right hulls first (#1234)

When you compose a mission, the hangar rail used to list your hulls in one fixed order that no screen ever explained — and in that order the Freighter, the Spy Probe, the Colony Ship and the Colonial Ark sat in the middle, between the Battleship and the Monitor Frigate, and again between the Shroudrunner and the Splinter Lance.

So on a transport, the one ship you opened the panel to send was the sixth chip of the fifteen that mission accepts, behind five warships. On a defence or a redeploy — the two orders that turn no hull away — all four sat buried among the warships with nothing to separate them.

Now the rail leads with the hulls the order is actually for:

• sending goods, harvesting, a pickup, a colonisation or a redeploy — the hulls that carry the hold come first;
• attack, defence, patrol and interception — warships first, support hulls together at the tail.

Either way the two kinds are never mixed: your support hulls are one unbroken run at one end of the rail, instead of scattered through the middle of your battle line.

Espionage still opens with the Spy Probe exactly as before, and exploration is left in the order it always had — on that one mission every hull is offered equally, and moving the support hulls to either end would have pushed the fast ship it asks for further from your thumb, not closer.

Also

Fleet cards, ledgers and battle reports are deliberately unchanged: they still list ships in the same fixed catalogue order they always have. Those are the screens you read side by side against one another, and an order that shifted with the mission type would make two panels twenty pixels apart disagree about the same fleet.

1.23.0 — 2026-08-25

New

• The First Convoy — a reason to come back tomorrow (#1382). Every commander who founds their empire from now on has a supply convoy inbound from the first minute, docking exactly 24 hours after the account was created. A quiet line on the Base tab counts down to the docking instant, and the manifest can be swapped with one tap until it docks: an Escort wing (3 Interceptors) or a Survey wing (3 Spy Probes — enough to scout three different neighbours). On day two, one tap unloads the ships into the hangar — and once the first steps are done, the confirmation points straight at the day's three contracts. One delivery per account, ever. Veterans are untouched: accounts created before today's cut-off never see the card.

Also

• The delivery lands in the Activity feed under the Empire chip, with the other fleet arrivals ("First convoy unloaded").
• Nothing knocks at hour 24 yet: this release adds no notification, so the convoy rewards the return it hopes to cause. The push follow-up belongs to the events calendar work (#1195).

1.22.0 — 2026-08-25

New

• Territory now has borders, and busy systems have a sign (#1379). On the galaxy map, an owned sector no longer reads as a faint colour wash: its tint gathers toward the edge of the plate and the outermost ring of cells wears a bright border in the owner's colour — visible at the widest zoom, where "whose region is this" is the question. Each sector also carries a big, faint numeral naming it, drawn into the territory itself (at galaxy and sector zoom). Stars now come in three shapes: a faint dot for an uncolonised system, a warm core for a colonised one, and from 3 colonies a four-point flare — a hub you can spot across the map.

Changed

• Control rings and guild anchorages now appear from sector zoom in: the widest view keeps sovereignty, landmarks, your home and colonies, beacons, anomalies — and contested systems, which stay marked at every zoom. The map key describes the new signs.

1.21.0 — 2026-08-25

New

The Building Sheet shows every level, not just the next one (#1239)

Until now a building's sheet answered two questions: what it gives you right now, and what it gives you one level up. Anything beyond that was yours to work out on paper — whether the Shipyard is worth taking to 20, where the Extractor stops being the best place for your ferrolite, what a level 25 Shield actually absorbs.

Tap Full progression on any building and the whole ladder opens: one row per level from 1 to 30, each with what it costs, how long it takes to build, and what it contributes when it is done. The level you are on is highlighted, and the table opens on it rather than at the bottom of the climb.

Two things are marked on the way up. A ◆ sits on the levels where the building changes its look in your base, so a long upgrade has a visible landmark. And a gold line marks where your Command Center stops: it names the Command Center level the next step needs, and everything below it is dimmed until you get there. On a second Site the line accounts for that Site's own Command Center requirement too, which is a different rule and used to disagree with the button underneath.

Every number in the table comes from the same formulas the build queue charges you with — the row for your next level is the exact price the Upgrade button is about to take, discounts, edicts and all. It cannot fall out of date, which is why it lives in the game rather than in a written guide.

The table is closed by default and costs one line of the sheet until you ask for it.

Thanks to Spacca-Mazza, who asked for it.

1.20.0 — 2026-08-25

New

• The season finale knocks (#1195). When the sovereignty season crosses into its final two weeks — the Eschaton Finale, where territory points count double — the game now tells you, instead of waiting for you to find out. Guild members with the app on their phone get one quiet banner: "Eschaton Finale — The sovereignty season is in its final two weeks. Territory points count double until it closes." It arrives once per player, never during your night window (if the quiet hours catch you, it knocks after they end), never while you are already in the game, and it spends one of the same four daily slots every other alert respects. Tapping it opens the Season tab directly on the Calendar lane, where the finale badge and the sovereignty rail live.

Also

• Every push notification that actually reaches a phone now leaves a typed record, so "does this kind of banner bring players back?" can be answered with data instead of a guess — and a kind that doesn't earn its place can be switched off.
• The issue behind this asked for two more knocks that could not be built yet: the vulnerability-window warning (the window exists only as math today — enforcing or retiring it is now tracked in #1386) and league promotion banners (the leagues they belong to are #1193, still on the drawing board).

1.19.0 — 2026-08-25

New

Four Command Center levels now give you something (#1162). Levels 16, 17, 26 and 28 used to cost a fortune and buy nothing but the level cap. They all grant something real now.

• Lvl 17 — a 5th build queue slot. Lvl 26 — a 6th. The queue stopped growing at Command Center 13 and never moved again, which is why the whole stretch from 16 to 30 felt so empty: the only things still arriving were colonies (every 3 levels) and Sites (every 4), and those two rhythms miss each other. Six slots against the four-hour build ceiling is roughly a day of construction running while you are away.

• Lvl 16 — a second Laboratory Site. The Laboratory becomes the first building outside the economy you can build twice on one colony, and the second one joins your empire's research network: the network Laboratory level on the Research tab goes up, and research lanes and buffer slots unlock earlier.

  The unlock says out loud what it is worth, because it is not worth the same to everyone: your research network is capped at twice your best Laboratory. If you have spread wide and already sit at that ceiling, a second Laboratory in one colony changes nothing. If you play tall — one or two colonies you cannot expand past — this is the way up that expanding was giving everyone else.

• Lvl 28 — an 8th Site for the economic buildings. Extractors, refineries, compute nodes, reactors and orbital panels each get an eighth emplacement.

Fixed

• The Build menu no longer offers a third Laboratory Site. Two is all a colony can have, and the menu now says so by not offering a third — instead of showing you a priced row with a live Build button that the server then refuses, promising a Command Center level that was never coming.

• The error you get when a Site is not available no longer insists a higher Command Center is the answer. Sometimes the building simply does not have that many Sites, and it now says so.

Also

The Command Center's design table, the deploy notes and the empires run by the AI all follow the new ladder — the synthetic players will build the second Laboratory and use the wider queue exactly as you can.

1.18.1 — 2026-08-25

Fixed

The Base tab no longer hangs on «Loading base…» when the 3D view cannot start (#1362)

If your browser or device could not give the game a 3D view — hardware acceleration turned off, another 3D view already using what it needed, a driver, a graphics setting — the Base tab used to sit on «Loading base…» for as long as you were willing to wait. Nothing else ever appeared. The only account of what had happened was a line of technical text from the graphics library, in a place most players have no reason to look, and clearing cookies could not touch any of it.

Now the base says what happened, in a sentence written for a person, and gives you a button straight to the Colony tab — which is the same colony without any 3D at all: every building, its level, and everything it is working on. You can keep running your empire while you sort the graphics out. Where the cause is something a reload really does fix, it says so; where it is not, it does not waste your time suggesting it.

A base that is merely slow now says so, instead of looking broken

Opening the game for the first time on a phone on a bad connection genuinely takes a while — the colony has a lot of artwork to fetch. If it goes on unusually long, the base now tells you it is taking longer than usual and may still arrive, keeps the loading message while it waits, and offers the same way out in the meantime. It never claims a failure it cannot prove, and the moment your base arrives the notice disappears.

The Colony tab stopped greeting you with a technical error

The planet image in the Colony tab header needs 3D too. When it could not be drawn, the tab used to show the raw error text from the graphics library across the top. Now it says, briefly, that the planet image needs 3D and that the rest of the tab works without it — which matters most for exactly the player the fix above sends there.

1.18.0 — 2026-08-25

New

• Tap a star, see the system before you commit (#1371). On phones and tablets, tapping a star on the galaxy map no longer drops you straight into the system view: a peek sheet opens over the map with the system's colonies, owner, control state, contested/beacon/anomaly signals, the guild anchorage, and how far it is from home in cells — the map's own metric. "Open system" is the descent, now a choice; × puts you back on the map exactly where you were. The gesture decides: a finger or pencil tap peeks on any device, while a mouse click still goes straight in (it already has the hover tooltip).

Also

• Before the map's data has answered, a tapped star says "Charting the hive…" instead of wrongly claiming the system is unexplored — and if the data feed is down, both the peek and the hover tooltip now say "live data unavailable" instead of a loading message that never resolves.

1.17.3 — 2026-08-25

Changed

• The counter-espionage row on the building sheet reads "no limit" (#1365). The Observatory and the Deep Sensor Grid both described their counter-espionage axis as "+N per level, no ceiling". It meant what it said — the bonus keeps growing at every level, with no upper bound — but the wording was idiom rather than plain language, and plenty of players read the game in their second or third language. The row now says "+N per level, no limit", which is the phrasing the game already used elsewhere for the same idea.

  Nothing about the buildings changed: the same bonus, at the same rate, with the same detection cap on the sentence below it. Only the sentence is clearer.

Thanks to Tibeh01 for spotting it.

1.17.2 — 2026-08-25

Fixed

The Sunless Vault's ROUNDS column no longer looks like broken counting (#1361)

The column that decides the Proving Ground's ranking — how many rounds your fleet needed to break the boss — printed its figures as "1 rd", "2 rd", "3 rd". A player reported it as a bug, and he was reading it exactly as written: after a small number, " rd" is the tail of an ordinal before it is short for "round", so the column read as a failed attempt at 1st, 2nd, 3rd.

It now prints the number alone. Nothing is lost: the column is headed ROUNDS on a wide screen, and on a phone every figure already carries that name above it, so the unit was being said twice and only the second one was confusing.

The battle replay had the same wording in one place — a fight lasting a single round showed "1 rd" beside the progress bar. There is no heading there to lean on, so the unit stays and the spelling changes to "rnd"; on a wide screen it spells "round" as before.

Also

The reporter suggested a third change, to the desktop wording. It turned out that text was already correct, and the file the report pointed at was not the one drawing the board — so that part is deliberately not in this release.

1.17.1 — 2026-08-25

Fixed

Nothing changes in the game this time. This release repairs the test suite that guards it.

Two checks that protect a promise the empires make to you — that an AI empire never announces a trade it did not actually make — had been failing since yesterday morning, and were blocking every other release. They were not failing because anything was broken in the game: the checks themselves carried a hard-coded date, and once that date passed a day old, the scenario they were meant to describe stopped happening. They were written to expire without anyone noticing.

They now measure time against the clock rather than the calendar, so they cannot go stale again, and they say plainly when their own setup no longer holds.

Along the way the same checks gained two things they never had. One of them proves that when an AI empire's offer disappears from the market board, it is only reported as a sale if it really was one. The other proves that a stranger's offer vanishing from the board is never mistaken for your neighbour's own deal — until now, removing that protection entirely would not have made a single check complain.

1.17.0 — 2026-08-25

New

• Intercepting a fleet no longer means catching a moving triangle (#1341). The ▲ counter on the galaxy map is now a button: it opens a Contacts board listing every fleet in transit your sensors can see — who it is, where it is heading, when it lands, and whether any of your colonies can actually meet it (MEETS · 9m MARGIN, CAN’T REACH, NO COURSE). Tapping a row opens the same contact sheet the mark opens, on a freshly sampled clock. Contacts you can act on sort first; the list holds its order while it is open, so nothing reshuffles under your thumb; your own convoys show as a count and a pointer to the Missions tab, never as rows.

Also

• The shadow-mode notice now appears on the Contacts board too, not only inside the contact sheet — and the two surfaces share one sentence, so they cannot drift apart.
• The margin badge turns gold under three minutes of slack on the board exactly as it does in the sheet: one threshold, shared everywhere it is painted.

1.16.6 — 2026-08-25

Fixed

A market offer no longer hides your goods from a raid (#1332). Publishing an offer used to empty the goods out of your warehouse, and cancelling put them back — in full, instantly, for free. With an attack inbound that was a trick anyone could use: list a fat offer, let the raider land on a warehouse that was short by exactly that much, cancel the moment the battle report arrived. Two taps, no fleet, no risk.

The offered resources now stay in your warehouse until somebody accepts. They are visible in your chips, honest in a spy report, and a raider can take them — and they are handed over only when the trade actually settles.

• Cancelling returns nothing, because nothing ever left. It just withdraws the listing.
• You can promise each unit once: listing goods your other open offers already promise is refused, and the composer's "max" button now counts what you have left to promise rather than what is in the warehouse.
• If you spend or lose the goods, the board says so on the row — to buyers, and to you on your own offers, which is the only place you would find out your board had gone quiet. Those rows stay up in case you restock.
• Tapping an offer whose seller can no longer cover it now explains that, instead of blaming the resources you are holding.

Evacuating ahead of a raid is still possible — that is what fleets are for — but it costs a round trip and a launch colony instead of two taps.

Also

Every offer that was open has been returned to its seller. Those listings were made under the old rule, with the goods already out of the warehouse, so the game handed them back once and closed them: 114 offers, returned to the players who made them. If you had offers on the board, they are gone and the resources are home. Anything above your warehouse capacity was lost to the usual storage ceiling on the way in — the same thing that would have happened had you cancelled them yourself.

A note for anyone playing on an older app build. The store version still says the offered resources are locked and refunded in full if you cancel. That is no longer true: they sit in your warehouse and can be raided, and cancelling gives nothing back. Update the app when you can, and until then do not use a listing as a hiding place.

1.16.5 — 2026-08-25

Fixed

• Numeric fields no longer keep a zero you typed in front of a number (#1331). Clear a box, type a new amount, and the old 0 used to stay stuck at the front: the NEW OFFER form would show 012211 in the Ferrolite box while the total underneath read the correct 12,211. The offer that published was always the right one — only the box was wrong, and it stayed wrong until you touched the field again.

  Clearing a box now leaves it empty instead of putting a 0 straight back with the cursor behind it, which is where the stray zero came from in the first place. A leading zero that still gets in — pasted, or typed on purpose — disappears the moment you leave the field, so a box can never be left disagreeing with what you are about to send.

  Editing a digit in the middle of a number keeps working: nothing moves under the cursor while what you have typed still means the number the game is holding. And a value the game will not accept is shown straight away rather than on the way out — type 900 into a field capped at 500 and the box says 500.

  Reported on the market composer; the same fault was on five more screens and all six are fixed: mission cargo, shipyard build count, target coordinates, the defense planner's per-ship count, and the guild raid ship picker.

Also

• The six of them now share one field instead of each carrying its own copy, so the tap-to-select behaviour from the older fix to these boxes is guaranteed in one place rather than remembered in ten.

1.16.4 — 2026-08-25

Nothing changes in the game this time. No screen, no number and no rule is different, and the words the agents use are byte-for-byte the ones they used yesterday — that was checked rather than assumed.

Fixed

• The lane that answers a newcomer's very first message — the one added so that somebody who knows nobody's name still gets a reply — was carrying several promises about itself that nothing actually held. A sweep after it shipped found no broken behaviour, but it did find claims that could quietly stop being true. They are now held (#1330):
  • the hourly limit on that lane is spent when an answer is attempted, not when one comes out. Otherwise, if the model were failing, every newcomer's first line would cost a real call while the limit sat reading zero — and nothing would have noticed;
  • the instruction that forbids an agent from inventing the rules of the game can no longer be reworded into one that hands it a canned answer without somebody reading the change. A canned answer is worse than it sounds here: it gives every agent the same reply, and the anti-repetition filter then silences the second newcomer of the evening as an echo of the first;
  • a test that claimed it was checking nothing important is in fact the only thing keeping two newcomers who speak in the same minute from having the second one ignored.

Also

• Several comments in that code said more than the code did — one promised a check that had been deleted. They now say what is true, including where the honest answer is "nothing verifies this, so read it yourself".

1.16.3 — 2026-08-25

Fixed

• The attack preview now sizes a defender on the world it actually sits on (#1323). Before launching, the Fleet Advisor rebuilds the defender you are about to meet — their Bubble pool, their Orbital Cannon, their emplacements — and every one of those is scaled by how well that colony is powered. The battle has always counted the defender's planet in that figure: an ice world runs its defences at 90% of nominal, a lava world at 120%, a rich lava world at 140%.

  The preview did not. It read the research half of the multiplier and stopped there, so the two screens disagreed about the same fight: against a defender on rich lava you were shown a Bubble and a Cannon roughly a third weaker than the ones that actually fired back, and against one on ice you were shown them stronger than they were. Understating is the direction that costs fleets — you sized the attack against a target softer than the real one.

  Both screens now agree, on every world. A defender on lava is previewed as the harder target it is, so the Advisor will recommend more hulls against one; a defender on ice as the softer one. Nothing about the battle itself changed — only the forecast, which had been telling you something the engine never believed.

  The target's world was never a secret, incidentally: it is fixed by the coordinates, the same way the galaxy map has always shown it. The preview simply had no way to look it up.

• A Terraformer on the target no longer goes uncounted. An Outpost or a Terraformer cancels part of an ice world's power penalty, and whether the defender has one is written in your spy report. The forecast was deciding that from your research instead of from what their colony actually has — so a defender who had built one was previewed about 11% weaker than they would fight. Their buildings answer that question now.

1.16.2 — 2026-08-25

Fixed

Nothing changes in the game this time. This one is entirely inside the workshop, and it is written down because the repo's history should say why a release exists.

• The build could not be verified at night (#1343). Three tests covering what an agent says when it wins or loses a contested relic had been failing every night between 22:00 and 06:00 UTC, and passing the rest of the day — on untouched code. The agents in question keep human hours: each one has a waking window, and a persona who is offline after 22:00 does not post in guild chat at 22:04. The test told the game it was two in the afternoon, but the last check before a line actually goes out reads the real clock, so once real time passed 22:00 the two disagreed and the agent fell silent mid-test.

  The tempting repair was to make that final check trust the time the test had claimed. It was written, reviewed, and thrown away: it would have bought the agents the right to keep talking a minute past their own bedtime whenever the game was busy — a small thing, and exactly the kind of small thing that makes a synthetic empire read as a machine. The clock was corrected in the test instead, and a new test now guards the bedtime rule that was nearly traded away.

  For players the effect is indirect but real: nothing could ship overnight while this stood, so fixes queued up until morning.

1.16.1 — 2026-08-24

Fixed

A pickup no longer empties the colony it is flying to (#1307).

You could put a colony's warehouse out of reach the instant a raid was inbound on it. Launching a pickup at that colony from anywhere in the galaxy took the whole manifest out of it immediately, and the attacker landed on nothing while the goods sat untouchable for the hours the freighters were in the air. No fleet needed, no timing to get right, and it worked on every undefended colony you owned.

That is gone. The manifest is now what the fleet is going to ask for, not a reservation:

• the goods stay in the destination colony until the freighters actually land and load them — visible in its warehouse, and takeable by a raid the whole way;
• if the colony was spent down or raided while they flew, the fleet loads whatever is left and comes home light, and the diary line says what it did not find instead of leaving you to work it out by counting at home;
• if the colony was razed, abandoned or is no longer yours when they arrive, they turn back empty and tell you why;
• the composer still refuses a manifest the colony cannot cover, and now also counts what your other pickups already have on order there, so the same pile cannot be promised to four fleets. The amount offered under each field, the "fill the hold" button and the refusal all read that same number.

Changed

This inverts a rule the game used to teach. The old Pickup notice said the manifest was set aside the moment the fleet lifted off and that nothing spent at the destination could touch it. That is no longer true, and every screen that said so has been rewritten. Evacuating resources ahead of an attack is still perfectly possible — that is what fleets are for — but it now costs a round trip, a launch colony within range, and the clock, instead of being a button you press when the alarm goes off.

Two smaller things that came with it: the mission card explains why a pickup cannot be recalled without citing a mechanic that no longer exists, and a pickup that comes home empty says "Pickup found nothing" rather than "Pickup loaded".

Also

Hiding resources from a raid is not closed in general — this release fixes the pickup door only. The market has the same shape through a cheaper one, and it is tracked at #1332.

1.16.0 — 2026-08-24

New

A newcomer's first word no longer falls into the void (#1296)

The galaxy channel answers you only if you name somebody — by name, with the reply button, or by naming their guild. That rule assumes you know a name to type. If nobody has ever answered you, you don't know one, and the channel is quiet by design, so you can't learn one by reading it either. A new commander who typed "hi, anyone around?" into the busiest room in the game got silence, and nothing on screen said there was another way to ask.

The first line you ever address to the channel now gets an answer, even when it names nobody. Once — after that you have a name, and the channel works the way it always has.

It stays narrow on purpose. The line has to name nobody at all: ask something of another commander and that line is theirs, not ours to take. Whoever answers has to be awake and not already talking themselves hoarse. And it is one line per person, remembered across restarts, so nobody gets greeted twice.

The answer is not a welcome. An agent knows you have never spoken to it — not that you are new to the game — so it answers what you actually wrote, the way somebody who happened to be around would. If you ask how a game mechanic works, it will tell you it doesn't know rather than invent a number: that has always been the rule, and it still is.

Also

The switch's own documentation had drifted: it still listed war declarations among the things agents never say in the galaxy channel, which stopped being true some time ago. It now lists what actually reaches that room.

1.15.0 — 2026-08-24

New

Superconducting Grid — a fully built early base can be powered (#1278)

A Command Center 4 colony with everything built and everything maxed sat in energy deficit, with every building running below full output and nothing to do about it: the second Reactor is gated at Command Center 7, so the era's power supply was fixed while the catalogue kept selling things that draw.

The Forge branch now opens a way out. The Superconducting Grid is a tier-1 repeatable, +8% energy production per level, up to 5 levels. Buying the line out takes a full CC4 base from 74% to a thin positive margin. Skip it and the era's ceiling is still there — raising the Command Center remains far cheaper, which is the point.

The trough was wider than reported: at CC5–6 a completionist base sits at 62%, worse than the CC4 that prompted this, because Batteries and the Trade Portal unlock while the power supply does not. The maxed line leaves that rung at 87% — softened, still a ceiling. From CC11 the base closes on its own.

It cannot inflate a colony that is already fine: energy yield is capped at 100%, so the Grid is worth exactly nothing where the balance is already positive. It does keep feeding a Particle Accelerator, which is the one building that can still black out a mature colony.

Fixed

• The attack preview was underestimating defenders (#1278). Against a colony whose owner had researched the Grid, the launch screen sized up a defender weaker than the one you would actually meet — the shield pool, the Orbital Cannon and the emplacements all resolved stronger in the real battle than the forecast showed. Preview and battle now agree. A defender you have not spied is still assumed to hold what you hold.

Changed

• The energy tooltip and building cards now name both halves of the bonus. A colony's energy multiplier can come from the world it sits on and from the Grid; each is stated separately, so a lava world is never credited with research that every one of your colonies carries. The planet is also named in English now — an ice world reads "Ice", not the internal id.
• Empire power breakdown gains an Energy production row once you own the Grid, and shows nothing where there is nothing to break down.

Also

Synthetic empires read the same energy figure the server charges them, including the world they sit on — previously they could misjudge their own colony and chase power they did not need, or ignore a shortfall they did have.

1.14.2 — 2026-08-24

Fixed

• Chrono Shards: the two counters now say which way they go (#1306). Shards are bought at one counter and spent at four others, and nothing on either side said so — a player looking at the two cards guessed that one traded shards and the other traded Dark Matter. The Shard Chandlery now says what a shard is actually for: a build, the shipyard, research, or a convoy in flight. The convoy card, when you hold no shards at all, stops offering four dead tiles and points at the counter further down the same screen — and withdraws that pointer when there is nothing on sale to point at. The speed-up panel and the Inventory name the shop whether or not you already hold shards; before, they mentioned it only while you held none, which is exactly the state you are not in when you run out mid-queue. The Inventory's shard label also counted three places to spend a shard where there are four.

1.14.1 — 2026-08-24

Changed

• The ship you are about to build is bigger on screen. In the Shipyard, the panel where the hull turns in 3D was the same small size it has had since the models arrived, sitting in the middle of a column with room to spare. On desktop that well is now taller, and the extra room goes into the ship rather than around it: the camera does not pull back, so the hull is drawn about 40% larger. The illustration that shows while a model loads grows with it. (#1295)
• On phones nothing changes, and that is deliberate. There the picture shares the screen with the stat bars and the Build button, and measuring what a taller well would actually do showed it pushing three of the four visible stat rows below the fold — so the phone keeps the panel exactly as it is. (#1295)

1.14.0 — 2026-08-24

New

Fleets in transit are now on the chart. Until now a fleet only existed at the two ends of its journey: it left, and later it arrived. The space in between was empty. It is not empty any more — ships crossing your sensors show up as contacts on the map while they are still flying, with their heading, their route and the time left before they land.

What you learn about a contact depends on your Sensors. A low reading gives you a mark and little else: something is up there, crossing that way. Build the Sensors up and the same contact tells you more — how big it is, where it came from, and eventually whose it is. Nothing about a contact is free, and nobody sees the whole picture. Range matters as much as level: a mark on the far edge of your reach reads worse than the same mark overhead.

Hiding is now worth something. The Veil, espionage research and the quiet hulls all lower what your fleets broadcast, and for the first time that has a use beyond flavour: a well-covered fleet crosses somebody's sensors without ever becoming a row on their chart. A large fleet is far louder than several small ones, so splitting a run is a real choice rather than a trick.

You can meet a fleet in flight. Open a contact and you can send a fleet to intercept it. You do not pick a place — you pick the contact, and the game solves where the two of you would actually meet, if you can get there at all. Slow ships chasing a fast fleet simply cannot, and the screen tells you so before you commit rather than after your fleet comes home empty.

Or leave a watch standing. A picket is a fleet stationed on a spot in open space that meets qualifying fleets crossing it while you are asleep. You choose how long it stands, and it comes home on its own.

Some fleets are never touched. Colony ships, probes, explorers and defence flights all pass. So does anyone sheltered by their night window, anyone still under novice protection, anyone you hold a pact with, and anyone hauling their own goods between their own worlds — housekeeping is not war. The cradle sectors are where nobody is hunted at all. When a launch is refused, the screen now says which of those it was.

Also

Interception ships switched off by default for its first run in production: contacts, sensors and stealth are all live and the picket flies and records what it meets, but nothing is fired until the mechanic has been watched against real traffic. Nothing you build for it is wasted in the meantime.

1.13.0 — 2026-08-24

New

• A joint raid is one alarm, not one per fleet (#1233). When several allies hit the same colony together, the Inbound panel used to draw one card per attacking fleet: five allies meant five identical countdowns, five «Send a garrison» buttons for one battle, and a genuinely separate attack lost among rows that were all the same thing. Now they arrive as a single Inbound raid card — how many fleets are in it, every attacker named, the hulls added up — and the intel on each contingent is one tap away under Show the N contingents, not gone.

Fixed

• Four small fleets are no longer read as a small attack (#1233). Each contingent used to be sized on its own, so a raid of thirty-six hulls showed four «small fleet» cards and a defender under-reacted exactly where they should have done the opposite. The size band is now measured on the whole raid, on the hulls your Sensors can actually read.
• The guild's «allies under attack» row sizes the raid, not its first fleet (#1233). A guildmate deciding whether to strip their home cover was reading the band of whichever contingent happened to land first — «small» for a raid nine times that size. It now describes the same raid the defender's own card does.
• The alarm counts agree with what is on screen (#1233). The HUD chip, the phone Alarm segment, the bottom-bar badge and the Peace-shield warning all count threats now, so a four-fleet raid plus one separate attack reads as «2» everywhere instead of «5» over a panel showing two cards.

Also

• Two independent attackers arriving at the same colony in the same minute are still two cards, deliberately: only fleets the server knows are flying together are ever folded. Two problems do not become one because they share a clock.
• At Observatory 2 and 3 a raid card lists every colony the contingents launched from, and says how many of them the Observatory can no longer trace.

1.12.0 — 2026-08-24

New

Twenty contested relics on the map, and no guild may hold more than four (#1300).

There were four. If your guild did not hold one of those four, there was nothing on the board for you to want — the prize existed, but it was somebody else's furniture, and most guilds simply stopped looking at it.

Now there are twenty: five carrying each bonus, five in every sector from 5 to 8, and at least one of every bonus inside each sector, so there is always one within reach. The four you know — The Jump Ring, The Verdant World, The Suspended Yard, The Exarch Archive — keep their names and their coordinates. Sixteen new ones have joined them, from The Tidewalk to The Forge Moon.

Two rules keep this from becoming a runaway:

• A bonus counts once. Five relics carry each bonus, and holding two of a kind is worth exactly one. The second is worth what it denies to whoever wanted it — which is a real reason to take it, and the only one.
• Four per guild. Win a fifth and you do not keep it: you tear it off whoever had it and leave it unclaimed. Denying a rival is a legitimate move, and now it is one you can make on purpose.

Together those mean the strongest guild's ceiling has not moved a gram — the best possible hand is still one relic per bonus. What changed is how many guilds can have something at all.

Changed

The relics panel now groups by bonus and tells you which of yours is earning it, so a duplicate reads as a spare rather than a second bonus. The Power dialog and the Inventory say the same thing about the same relic, every time you open them. The map's PRIZES cartouche scrolls instead of pushing the map off a phone screen, and it tells you when your guild is at the cap — because at four, winning another prize is not adding one.

Fixed

• A relic bonus is no longer silently withheld. The game only ever checked the first relic carrying a bonus, so a guild holding any of the other four was told its bonus was worth nothing — on the path every fleet launch goes through.
• Battle reports about relics now say what actually happened. Winning a relic and keeping it are two different things, and the diary used to claim the second whenever the first occurred: it congratulated you on a prize sitting unclaimed on the map, told a guild that somebody had taken a relic nobody now holds, and promised a one-hour seal on sites that were not sealed at all.
• Relic sites can no longer fail to appear. A new site whose address was already occupied used to be skipped in silence, forever — the relic simply never existed in the game. It now takes the next free address in its sector. Four of the sixteen needed one on the day this shipped.
• Monuments have stopped turning up where people belong: they are out of the chat mention list, out of the guild invite picker, and cannot be sent a whisper.

Also

Your synthetic neighbours read the same rules you do. They will not set out for a relic that would add nothing to their guild, they will not promise an expedition their guild has no room for, and when a call to arms stops making sense they say so instead of going quiet on the people who answered it.

1.11.1 — 2026-08-24

Fixed

• A free research lane no longer sits idle behind a stacked order (#1222). Queue two levels of the same Discipline — Armor Plating 12 running, Armor Plating 13 waiting for it — and the whole queue used to stop there. The second lane stayed empty, every order behind it waited too, and nothing on screen explained why: a lane was drawn as free right next to an order that said it was waiting for a lane. It was worst exactly when you were offline and the queue was supposed to keep going on its own.

  Now the queue steps over an order that is only waiting for its own Discipline or node, and starts the next one it can actually take. Nothing jumps the line: a Discipline still runs its levels strictly in order, one at a time.

  Two things still stop the queue, on purpose. An order you cannot afford yet holds its place — otherwise a cheap order queued later would start ahead of the expensive one you are saving for, again and again for as long as cheap orders keep arriving. And an order your Laboratory network can no longer start still stops the queue when its turn comes, because waiting will not fix it.

  One honest trade comes with this: the order that overtakes spends resources, so if you were saving for an expensive stacked level it can now be short when its turn arrives. An order you are deliberately banking for is one to queue on its own.

Changed

• The queue panel points at the right order, and stops crying wolf. "Queue blocked" and "Queue on hold" now follow the order the game will actually start next, so a stacked order in front of it is no longer coloured red for money it is not being asked for. And those two sentences are reserved for a queue that is genuinely stuck: when the orders ahead are still going to run, the card says so plainly, in its normal colour, instead of sending you off to build Laboratories over something that will not bite for hours.

  When your network has fallen below an order that is stacked behind another one, the panel does still say the queue is on hold — because then it really is.

1.11.0 — 2026-08-24

New

Ten minutes to take back a change to your night window (#1303). The night window is the one setting that has to match something outside the game — when you actually sleep — and it can only be moved once every 24 hours. Until now, an hour set by mistake was set for the day.

For ten minutes after you move it, the control offers one button: Undo — back to 22:00 · 4m left, naming the hour you came from on your own clock and counting down what is left of the offer. Press it and that hour is restored. It is on the HUD row on desktop and tablet, and in the settings sheet on a phone, and it stays usable while the window itself is locked.

The button says the rest out loud: the 24-hour wait keeps running. The undo takes you back to where you were, and not on to a third hour. That is deliberate — it is what stops the grace from becoming a way to re-aim the window at whoever is attacking you, because the hour it restores was already the one in force.

This came from a player who set the wrong hour the day the picker switched to local time, and could not fix it for a day.

Also

The hour a mis-set move can be restored to is now remembered per commander, and the settings your client receives carry it. Nothing you had set changes, and if you have not moved your window in the last ten minutes there is simply nothing to take back.

1.10.1 — 2026-08-24

Fixed

• The Research bench no longer swaps the colours of your treasury (#1267). Looking at a colony that is not your mother colony, the Research tab shows a reminder of the treasury that pays for your orders. It printed the ferrolite figure in orange and the xenogas figure in cyan — the opposite of the HUD, the Colony tab, the Market, the Overview and the Anchorage. If you read those tiles by colour, as most people do after a week, the two numbers looked swapped. They now match the rest of the game.

• Crates in your Inventory are tinted like what is inside them (#1267). A ferrolite crate was drawn in gold and a xenogas one in cyan, so a prize looked like the wrong resource until you read the label.

Also

• The colour for each resource is now written down in one place instead of seven. Five of those seven agreed; two did not, which is what you were seeing. Nothing changes on the five surfaces that were already right.

1.10.0 — 2026-08-24

New

You can now hear an attack coming (#1290). The incoming-fleet alarm has always been something you could only see — the red chip in the HUD, and on the phone the pulsing screen edge in the last ten minutes. Both need your eyes on the game, so the most urgent thing that can happen to you was also the easiest to miss while you were reading the market in another window.

A fleet newly on its way now sounds a cue: a low three-second swell, dark and restrained, not an alarm bell. It plays once per attack — never once per poll, however many hours that fleet stays in the sky — and it stays quiet for anything that was already in flight when you opened the game, so refreshing the page mid-siege will not replay an alarm you have been watching all evening.

It respects your sound settings: mute silences it, and the SFX slider sets its level. It deliberately still plays when the tab is in the background, because that is the whole point of it. The phone notification opt-in is unchanged and still covers the app that is fully closed.

Thanks to Spacca-Mazza for the suggestion, and for naming the right register: dread, not panic.

1.9.0 — 2026-08-24

New

The fleet ledger is on the Overview (#1280)

The empire dashboard could tell you how many ships you owned, and what was in the air. It could not tell you where they were. That table already existed — one row per ship type, one column per colony, plus what is in flight, what is on garrison, and the total you own — but it lived only inside the Fleets tab. If you wanted to know which colony held the bombers, you opened your colonies one by one and added up.

It is now on the Overview too, below your colonies and above the fleets in flight. Same table, same numbers as the rest of the tab.

On a phone it starts folded — tap the FLEET LEDGER row to open it — because a wide table becomes one card per ship type there, and the summary above it is what that screen is for. Wherever the table fits, it is simply open.

It appears from your second colony onward: with one colony, the hangar already is the ledger.

Changed

• The ledger's garrison column used to be headed Stationed. On the Overview that put it one scroll below the STATIONED FLEET strip, which counts something else entirely — the ships sitting in your hangars. Same word, two different sets of ships. The column is now headed Garrison, the word your mission cards already use for that fleet. Nothing about the numbers changed, and the column still appears only when you actually have a garrison out. (#1280)

Fixed

• While the fleet list is still loading — or after it fails to load — the new section no longer promises "what is in the air" over a column of dashes. It says it is still counting, or that the fleets in flight are unavailable, and shows your hangars either way. (#1280)

1.8.0 — 2026-08-24

New

Your first screen now says the galaxy is running (#1194)

A new commander used to arrive on an empty base with a build panel and a timer, and that was the whole game as far as he could see. The season, the war weeks, the market and every other empire were running the entire time — behind tabs nobody opens on their first visit.

The first-steps panel now closes with one quiet line naming the season that is running and how long the galaxy has left on it:

OUT IN THE GALAXY Season of the First Omen · 12d 4h left

No button, no badge, no sound. It appears while you are still working through the first steps and leaves when they are done — and it stands aside on the Season tab, where that countdown is already on screen and ticking.

Also

Three of the four directions the original issue proposed turned out not to need building, and saying so is part of the entry:

• the first builds are already fast. The early-level sprint cuts construction to 15% for the first three levels of anything, so the first five steps of the chain cost 58 seconds of building in total — the issue asked for 60 to 90;
• the league card it asked for was written against a personal league that does not exist. Only the guild league does, and a player without a guild never sees it;
• an agent greeting you on arrival would reverse a rule the agents are deliberately built on — they never speak first in the galaxy channel and are explicitly told not to welcome anyone. That is a decision to take deliberately, not a gap to fill, so it is now its own issue (#1296).

1.7.4 — 2026-08-24

Fixed

• The colony dropdown starts with the colony's name (#1243). On desktop, every row of the colony picker used to open with the planet glyph and the planet type — «🜨 Rocky · Mars — 3:19:1» — so the name you were actually looking for sat in the middle of the row, behind a prefix that every colony on the same kind of planet repeats word for word. With a few colonies on similar planets the list read as a column of identical openings. The row now reads name — coordinates · planet type, the same order the phone picker already used. Two things come with it: the column now differs from its first letter, and typing a colony's name with the dropdown open jumps to it, which it could not do while every row began with the same glyph. On a long name the type can fall off the end of the closed dropdown; it is still there when you open the list, and on the Colony tab.

1.7.3 — 2026-08-24

Changed

• Night protection is now picked on your own clock (#1283). The window that decides when your empire cannot be attacked used to be chosen in UTC — «Night protection: 8h from 2:00 UTC», with a dropdown of bare UTC hours — so unless you happened to live on UTC you converted the hours in your head before setting the one thing that is meant to match when you actually sleep. Both selectors, the top bar on desktop and the settings sheet on a phone, now read at your browser's own time and are ordered by it, with the UTC counterpart next to them so a guild spread across timezones still has one number everybody can quote. Both readings follow your browser's language, so a 12-hour clock stays a 12-hour clock on either side. If your browser is already on UTC, nothing changes but the wording.

  Nothing was converted behind your back: the hour stored for you is the hour still stored, and the same window protects you tonight as last night — it is only shown to you in a clock you can read. Time zones that sit half an hour off the hour, like India's, see their real minutes (07:30, not a rounded 07:00), because the alternative would have been saving you a different hour than the one you picked. Suggested by jaybee.

1.7.2 — 2026-08-24

Changed

Every ship in the game is a real 3D model now (#1281)

The Colony Ship was the last hull still shown as a painting. Open its card in the Shipyard and it turns like every other ship; send it out escorted and the replay draws a real hull among the fleet instead of a cardboard cutout that gave itself away the moment the camera moved.

That closes the art pass. Sixteen hulls, sixteen models — the Spy Probe, the Shroudrunner and the Colony Ship were the last three, and they landed over the past two days.

Also

A player raised this one in universe chat, naming the Shroudrunner and the probe. Both went out earlier; this is the ship nobody complained about and that turned out to be the last one left.

One thing deliberately left for its own release: the Colony Ship, the Freighter and the Ark all draw a little smaller than their artwork did. It is an old quirk of how those three are sized rather than anything new, and correcting it moves two ships that are already flying, so it gets its own change and its own look rather than riding along with this one.

1.7.1 — 2026-08-24

Changed

The Shroudrunner is a real ship now, not a painting of one (#1281)

Open the Shroudrunner's card in the Shipyard and it turns: the illustration fades into the actual hull, which you can spin like every other warship in the catalogue. In a battle replay it is a real mesh among the rest of your fleet instead of a flat cardboard cutout that gave itself away the moment the camera moved — and its running lights are lit.

It also arrives at the right size. The catalogue rates the Shroudrunner a shade larger than an Interceptor, and until now the model would have flown at half that: review caught it before it reached you, and there is a new check that measures every hull against its class so the next ship cannot arrive shrunk.

Fifteen of the sixteen hulls in the catalogue now have a 3D model. The Colony Ship is the last one still waiting for its own.

Also

A player pointed this one out in universe chat — the Shroudrunner was named specifically. The Spy Probe, mentioned in the same conversation, was already dealt with a few releases ago.

1.7.0 — 2026-08-24

New

Rebuild your emplacements instead of waiting out the clock (#1261)

When a raid flattens your Defense Batteries or your Orbital Fortress, the emplacements used to come back on a clock and nothing else: one every 30 minutes, one Fortress platform every 2 hours. A wiped Batteries L14 was fourteen hours of watching a countdown with a full warehouse.

Now the Batteries and Fortress sheets offer to rebuild them, paid out of the colony's own ferrolite, xenogas and data at what that colony sank per emplacement. The bill for what the tap will actually buy is printed before you tap it, and when you cannot afford it the button tells you what a single emplacement costs instead of just refusing.

One tap proposes at most half your warehouse — this button sits under Upgrade and there is no confirmation step, so it should not be able to empty the fund you were saving for something else. When the whole line is within reach anyway, a second Rebuild all appears, so buying the lot is one deliberate choice.

Two things it deliberately does not do:

• it never touches the free clock. The emplacements you did not buy come back at exactly the time the countdown already promised — paying does not push them back;
• the last one is always free. However rich you are, a colony still waits out one regeneration tick. You are buying the tail of the wait, never the whole of it.

Rebuilding does not change what an attacker earns for destroying your emplacements again: a bought emplacement is worth exactly what a regrown one is worth, and the same weekly limits apply.

Synthetic empires got the same button, and they use it the way a player would — during their own waking hours, after a pause, on whichever of their colonies took the worst of it, and only out of what they can spare.

Also

The activity feed now names the colony and the price on the rebuild receipt, so two purchases minutes apart are no longer the same line twice.

1.6.0 — 2026-08-24

New

Hangar interdiction — cloaked hulls can finally catch the freighters they hunt (#1265)

A Shroudrunner's counter kit is freighters ×3, spy probes ×4, colony ships ×2. Against another player, none of it could ever happen: unarmed hulls stay in the hangar by default, and the reserve took no losses whatever the outcome. The ships those counters named were never in the battle, so bringing cloaked hulls to hunt an enemy's haulers bought you nothing.

Now, when an attack takes the colony, the winner's surviving ships run down the unarmed hulls they counter, caught in the docks:

• one hull per point of multiplier, per surviving ship — a Shroudrunner (×3) catches three freighters, an interceptor (×5) five probes;
• up to a quarter of each type in the hangar, rounded up. One won battle takes a slice, never a logistics fleet — and a stack of one is lost whole;
• only hulls that cannot shoot back. Armed ships your defence plan held back are never touched on the ground: that choice already cost them the battle they did not fight;
• a repelled attack catches nothing, and neither does a winner with no counter against what is parked there. A wall of battleships takes your colony and leaves every freighter where it was.

The caught ships are destroyed and leave nothing behind — no loot, no debris, no season points. The attacker's reward is that your hold capacity is gone.

Three places tell you about it: the battle report names what was destroyed in the defender's docks, the activity row lists the hulls caught in the hangar beside your battle losses, and the Ship reference explains when a counter against an unarmed hull is worth bringing.

Changed

The Planetary Defense panel stops promising the reserve is safe (#1265)

It used to say the reserve "sits out the battle, safe in the hangar". That was true until this release and would now be the game telling you that a loss you are about to take cannot happen. It says what is true instead: out of the battle, not out of danger, with the share and the rounding spelled out — and it still says the part that has not changed, that armed ships held back are untouchable on the ground.

Thanks to RoyDisco, who spotted that the desk's own doctrine was wrong about this and that the mechanic underneath was oddly shaped.

1.5.0 — 2026-08-24

New

• The Spy Probe is a real ship now (#1273). Until today it was the last light hull still drawn as a flat picture: in a battle replay it was a card that turned to face you, and you could tell. It now has a proper 3D model — a squat armoured sphere with its sensor eye leading and its solar panels spread wide — that banks and turns with the rest of your fleet.

Changed

• Its picture in the Shipyard is the same ship you see in battle. The card art, the grid thumbnail and the fleet illustration were all re-made from the new model, so the probe you order is the probe you watch fly. It is the same size on screen as before — that was measured, not eyeballed, so nothing about how your fleet reads at a glance has moved.

Also

• The probe's sensor eye does not glow yet. Every other lit hull carries a separate lighting map and this one does not have hers, so the lens reads as painted rather than powered. It is on the list.
• If your browser had the old ship art cached, it will fetch the new pictures once. Nothing to do — it happens on its own the next time you open the Shipyard.

1.4.1 — 2026-08-24

Changed

• Sector 4 joins the protected sectors (#1269). The starter band was filling at ~50 colonies a day and was five days from full, so it grew the one way that moves nothing on the map: the rule now covers sectors 1–4. New empires can start in sector 4, and the whole newcomer shield — the tighter ½..2× Power band and the doubled bounty on anyone punching down — applies there from today, including to the ~50 colonies already living in it. Sector 4 keeps its Frontier-grade rich planets, making it the only sector that is both protected and rich — deliberately: re-rolling planets under existing colonies was the alternative, and that never happens.
• Guild Displacement follows the wider cradle (#1269). The fold rule is unchanged — never into the newcomers' spawn pool, freely within your own band — but the boundary it reads moved: a Frontier colony (sectors 5–8) can no longer fold into sector 4, and a sector-4 colony can now fold deeper into sectors 1–3.

Also

• Browser tabs and installed app builds opened before this release still price sector-4 strikes at the old Frontier rate. The server's price at impact is the authoritative one — reload the page, or wait for the next app build, before planning an attack there.
• Agents now describe the new boundary correctly when asked where to expand: sectors 1–4 protected, richness starting at sector 4.

1.4.0 — 2026-08-24

New

Your ally's own defences, on their distress call (#1263) — the "Allies under attack" card in Joint raids told you what was flying at a guildmate and nothing about what it was flying at. You could not find out any other way either: spy runs refuse a friendly target, so an ally who was offline could only be rescued blind.

Each threatened planet now carries what is already standing on it, in the words the rest of the game uses:

Their own defences at 3:58:2 — Bubble 12 (48,120 HP) · Emplacements ×9 · Fortress ×2 · Orbital Cannon · 340 ships

If there is nothing left, it says so in amber — nothing of their own takes the field — which is the moment a garrison is worth stripping your home cover for. If the colony is browning out, a line underneath tells you by how much, because a Bubble running on short power shows fewer hit points than its level would suggest and that is worth knowing before you decide how much to send.

It is your ally's own information, riding your ally's own alarm to their own guild. Nothing is revealed that they cannot already read on their colony screen, and no one outside the guild learns anything.

Fixed

"You never lose it to them" no longer reads as "your ships cannot die" (#1263) — that sentence meant you are not handing the fleet over to the defender, and more than one player read it as a promise that a garrison could not be destroyed. A guard fleet that then died in the battle contradicted what they believed they had been told.

The two facts are now two sentences: the ships stay yours, a garrison is never handed over — and they do fight, so they can be lost there like any fleet of your own.

1.3.0 — 2026-08-23

New

• Uninstall a module (#1255): every filled refit bay now ends its picker with an "Empty bay" row — uninstalling refunds half the module's parts, the same 50% every replacement already paid, and arms the same 12-hour bay cooldown. The combat Core menu has no drawback-free option, so an empty bay is a real choice, and now you can make it.
• Every module shows its blast radius (#1255): each option in the refit picker (and the uninstall row) can expand "Affects N hulls" — every ship of the class with the stats the choice would move, as they are today and as they would be after, composed with your other standing orders. A change too small for whole numbers shows a decimal instead of hiding: a drawback never reads as free.

Changed

• The 50% refund is stated where you decide (#1255): every installed module card now says "Replacing or removing refunds N parts" with the picker closed. The rule always existed, but it was invisible — a player proposed, sight unseen, the exact penalty the system already charges — and a rule nobody can see is a rule that doesn't help anyone. The uninstall refusals also come back as plain sentences instead of raw codes.

1.2.0 — 2026-08-23

New

A queued research can be taken back out of the queue (#1244)

Until now a research order, once queued, could not be taken back — the only way out was to wait for it to finish. That was worst exactly where it hurt most: an order needing a Laboratory level your network has fallen below stops the whole queue behind it, and there was nothing to do but watch.

Every queued order now carries an × in its slot. Pull one and it leaves the queue, the slot frees, and anything that was stuck behind it starts right away in the same request. Nothing is refunded because nothing was ever charged: a queued order is unpaid until it starts.

Two details worth knowing:

• Running orders have no ×. They have already been paid for, and what a refund should be worth is a separate question this does not answer. Ask for one and the game says so in a sentence rather than a code.
• When you have several levels of one Discipline queued, only the top one carries the ×. The levels of a Discipline are researched in order, so removing a middle one would leave the levels above it with a gap — you would land on a level you never paid for. Pulling the top level is the one removal that moves nothing else, so that is the one offered. A stalled order that cannot be pulled yet now says which one to remove instead of leaving you looking at a slot with no button.

1.1.4 — 2026-08-23

Fixed

• Buildings capped by the Command Center no longer look upgradeable in the Colony view (#1256). Before, a building sitting at your Command Center's level cap — an Extractor at level 10 under a level-10 Command Center, say — still showed a bright upgrade arrow near the top of the Colony tab's building list, and tapping it only earned a refusal from the server. Now the row says why it is dark — «Command Center lvl N required for the next level», the same line the Build menu already shows — the arrow is switched off, and capped or maxed-out buildings drop below the ones you can actually upgrade. Buildings under construction keep the top of the list, and within each group nothing changes order, so the list never reshuffles under your thumb.

1.1.3 — 2026-08-23

Fixed

The ally who helps repel an attack is told they won (#1245)

Lend a fleet to defend a guild-mate's colony, win the battle, and the Activity feed told you that you had lost: «Battle (defender) — defeat», in red. Every single time.

The row contradicted itself in one sentence — the title already called you a defender, and the body underneath called it a defeat — and it was invisible unless you happened to win, because a defence that actually fell produced exactly the same row.

A garrison now reads the same row the colony's owner reads: victory, in green, with the siege note when there is one. A defence that falls still reads defeat, in red.

1.1.2 — 2026-08-23

Also

Nothing changes in the game this time. This release fixes the machinery that numbers the releases themselves.

Since 22 August the version has been plain 1.0.0, 1.0.1, 1.1.0 — the long 1.0.0-beta.N series was closed. It did not quite take: over the following day, 15 releases came out still numbered 1.0.0-beta.N, which is why the changelog you are reading has entries that look out of order, with a newer release sitting below an older-looking one.

The cause was mundane. The script that cuts a release is kept among the project's own files, so every working copy carries its own; the copies that had not been updated kept using the old numbering, and nobody noticed because the result looked like a perfectly normal release. From now on that script refuses to run unless it is the same version as the code it is about to release — so a stale copy stops the release loudly instead of quietly mis-numbering it. The script that deploys the site is checked the same way, since the release runs it.

The old tags stay exactly as they are. Renumbering or deleting them would break every link that already points at them; they are simply out of sequence, and the sequence behaves again from here.

1.1.1 — 2026-08-23

Also

Nothing changes in the game this time — no ship, no building, no agent behaves differently. This release fixes two tests that had been quietly given expiry dates.

Both described a moment "in the future", but wrote it as a fixed date rather than a distance from now. One of those dates arrived this afternoon and the test simply stopped being true, which froze the whole release pipeline: nothing could be merged or shipped while it was red. The second would have done the same thing on Monday.

Both now measure from the real clock, so they stay true whenever they run, and a new check fails if anyone writes a future date the old way again. To be sure there was not a third one hiding, the entire test suite was re-run with the clock wound forward a week, a month, a year and nearly three years: green at every point.

1.0.0-beta.241 — 2026-08-23

Nothing changes in the game this time. No balance moves, no screen looks different, and no empire gains or loses anything. This one is a rule that was written down but never actually enforced, and is now enforced.

Fixed

The Pirate Beacon was never really exempt from Dark Matter (#1228). Dark Matter is meant to be a currency for real empires only: the pirates who hold the Beacon were supposed to be left out of it, and the game said so. In practice the check never excluded anyone. It asked each colony who owned it, but never actually looked up the owner, so every answer came back "a real player" — pirates included — and the whole galaxy went through a gate that had been built to stop them.

Nobody was ever paid a unit they should not have been, and it is worth being precise about that rather than leaving it to be imagined: producing Dark Matter needs a Particle Accelerator, the Beacon has none and never builds one, so the amount owed to it was always zero. Every pirate, faction and relic holding in the game was sitting on exactly 0.000 Dark Matter when this was found. The rule was not being bent — it simply was not doing anything, and would have given way in silence the first time a pirate holding ever came to own an Accelerator.

The check now happens where it cannot quietly stop working. As a side effect the periodic sweep also stops picking up the hundred-odd pirate and guardian colonies it had no use for, every five minutes.

1.0.0-beta.240 — 2026-08-23

Fixed

A pickup's two diary lines now land under the same Activity chip (#1229)

Sending a fleet to fetch resources from one of your own colonies writes two lines in Activity: one when the hold is loaded at the far colony, one when the fleet lands back home. Until now the first was filed under Economy and the second under Empire.

If you tapped Empire to keep an eye on your fleets, you saw trips end that you had never seen begin. If you tapped Economy, you saw a hold get filled by a fleet that then vanished. Both halves of a pickup now sit under Empire, next to your transports, transfers and returning fleets.

Two Contested Space lines moved to Combat (#1229)

Two more lines were filed under Economy when nobody would ever look for them there, and both are now under Combat:

• the line that explains a battle you won did not take the world after all — "another guild took it while your fleet was in flight, you keep what you took from it, but a world cannot be contested again within an hour of being captured". The battle itself was already under Combat; its explanation was filed among your market sales.
• the Contested Space result at the end of an event, which now sits with the season's other verdicts, where the sweep and "Season ended" already were.

Also

Nothing else changes in the game. The chips themselves, their counts and every other line in Activity are exactly where they were.

1.0.0-beta.239 — 2026-08-23

New

• Pickup — fetch resources from your own colony in one mission (#693). Until now every cargo mission moved resources one way: bringing a secondary colony's surplus home meant two separate launches, each paying its own trip, and anything left at the source could be spent before the ships arrived. The new Pickup mission flies empty to another colony of yours, loads a manifest there and brings it home on a single mission — one queue slot for the whole round trip.
  • The manifest is set aside the moment the fleet lifts off. The source colony's stock drops at launch, so nothing spent there while the freighters fly can eat what they are coming for — the launch is refused outright if the source cannot cover it, and the arrival can never come up short.
  • The escrow stays visible: the in-flight card carries a "Collecting: …" row naming exactly what the fleet is fetching, and the activity log gets a "Pickup loaded" line at arrival.
  • A pickup cannot be recalled — the goods are committed at launch, and both the composer and the mission card say so before and after you press Send.
  • The composer bounds the manifest on the destination's warehouse, "fill the hold" fills from there (capped by what fits back home — a pickup's surplus at landing is lost, and a notice warns you before it happens), the fleet Advisor sizes freighters for the homebound load, and the trip preview shows both legs.

Also

• The mission cards no longer crash an outdated app when the server learns a mission type it does not know: unknown types now render a plain, degraded card instead of taking the tab down.
• Two fleets that can never turn back — a raid member and a pickup — now share one honest refusal message instead of the raid-only wording.
• A rare "internal error" when two mirrored launches between the same pair of colonies collided at the same instant is now retried silently, like every other concurrent action.

1.0.0-beta.238 — 2026-08-23

New

• The Calendar now shows what is coming, not just what happened (#1223). The Season tab's Calendar used to answer only "what has been happening" — its future squares said "nothing scheduled" about everything. Now a "Coming up" strip above the grid names the next occurrence of each scheduled event with a countdown (or the date, when it is far): the Eschaton Event, the war week close, the next League round, the weekly anomaly survey, the daily and weekly contract resets, the guild directives window, the guild shop limits reset, the guild season finale. Future squares carry dashed chips for the rare events that land on them (a diamond mark on phones, with its own legend entry), and tapping a future day — or today — lists that day's whole schedule with local times and countdowns. Only enforced dates are drawn: if the engine does not act at an instant, the calendar does not promise it.

Also

• The one-shot dates (Eschaton, guild season finale, anomaly grid expansion) always keep a seat in the strip unless something even more imminent fills it, and the strip says "+N more" when its six seats are not enough.
• Screen readers get each square's schedule in its label, including "schedule not read yet" when the feed has not answered — the calendar never claims "nothing scheduled" about a day it has not read.

1.1.0 — 2026-08-22

New

• The login now greets first-time visitors with a commander name already filled in — one tap on Play and you are in. If the name happens to be taken, the game quietly draws another; after three unlucky draws it fills in a fresh one and says so, without blaming you for a name you never typed. (#1194)
• Devices that already held an empire keep the empty field they know: the one-tap door is for newcomers only, so a stray tap cannot found a second empire by accident — and that holds even for empires from before this update.

Fixed

• Failure notices on the login screen are now announced to screen readers.
• The small print under Play no longer promises "reclaim" to players who never linked an account, and it goes silent under a deletion notice instead of contradicting it.

1.0.0-beta.237 — 2026-08-22

New

• The Calendar (PR #1207). The Season tab has a fourth lane: a month calendar of what actually happened in the galaxy. Every day carries the public registry's own counts — battles, colonizations, market sales, season endings, war weeks — bucketed in your timezone, with today marked. Press a day and its registry opens beside the grid: the same narrative rows Activity shows, newest first, with a plain line when a day was quiet. Above the grid, the sovereignty season rail says which of the 8 weeks you are in, marks the finale weeks, and — while the finale is live — that territory points count double; the Eschaton Project bar appears only while its phase is open. Future days say plainly that nothing has happened there yet: the only forward dates drawn are the ones the server actually commits to, the season's end and the Eschaton phase.

Also

• The ship review closed three defects before merge: a typecheck failure the test suite could not see, and the calendar routes are now pinned to the wire by route-metadata tests — renaming a query parameter can no longer kill the feature while every test stays green.

1.0.0-beta.236 — 2026-08-22

Fixed

• Guild loyalties now survive rising war tension (#660). Synthetic guildmates no longer resent allies for friendly-side attacks, while defenders remember each real strike once, including joint raids and garrison battles.

1.0.0-beta.235 — 2026-08-22

Fixed

• Building pads now rest on their cells (#630). Every shipped building sprite uses a measured ground contact, so neighbouring structures share the same ground line and upgrades no longer jump between measured and guessed anchors.

Changed

• Upgrade art stays within its pad constraints (#630). Visual stage growth no longer magnifies a building's illustrated slab beyond what its source art or minimum upgrade size requires.

1.0.0-beta.234 — 2026-08-22

New

• The joint-raid call has an off switch (#628). Since the Joint raids door started calling (#437), the dot on the Guild tab went out only when you joined the raid or it landed — a player who opened the door, looked, and decided not to spend their fleet kept being called for hours, and a player with no warships was called to something they could not answer. Now, visiting the Guild screen marks the raids it showed you as seen: the dot goes quiet for those, and a new raid lights it again. Declining one raid never silences the next.
• The raid row tells you whether your fleet can make it (#628). lands in 38m is a fact about the raid, not a promise about you: if your ships need 50 minutes, the join button was a dead end that answered only after you had picked ships and a formation. The row now shows your fleet's own travel time next to the window — Your fleet: 50m to get there — longer than the window. The join may be refused. It never blocks the button: guild hypergates and travel research can shorten the real trip, so the warning is a warning, not a verdict.
• A raid that would refuse everyone stops calling (#628). A raid formed on a target that has since been shielded, or whose guild has since signed a pact with yours, stayed on the door burning red while every join failed on the click. The call now checks the same rules the join uses and stops counting those raids — including on faction bastions, the one place the post-raid shield still exists.

Fixed

• "Allies under attack" no longer offers a defence for a battle that is over (#628). The board loaded once per visit, so forty minutes in it could read impact now beside a live Send defence button — and a fleet sent through it parked for an hour over a planet glassed twenty minutes earlier. Landed strikes now grey out with "this attack has landed", their button refuses, and the board refreshes itself every few minutes. Countdowns on the board also read each fleet's real arrival instant, so sitting on the panel no longer makes them drift.
• The Menu badge counts raids (#628). 3 raids forming used to reach the phone's Menu slot as 1.

Also

• The forming-raid lookup on the guild summary now has a database index shaped for it, and the raid list costs one query instead of one per raid on every refresh (#628).
• The door tiles on the phone guild screen no longer risk stretching their row when the raid status runs long (#628).

1.0.0-beta.233 — 2026-08-22

New

A warning before a launch empties your colony (#549)

Sending an attack, a garrison, a joint raid or a redeployment could quietly take away the last defence your colony would actually put on the field. The mission left at once, and the first you heard about it was a battle report from a colony that had been standing open the whole time.

Now the game stops and tells you first. When a launch would turn a defended colony into one with no defences of its own, you get an explicit decision naming the colony and what you are exposing:

• an attack or a garrison tells you when that fleet is due back, and that recall or stand-down can bring it home sooner;
• a joint raid says plainly that raid fleets cannot be recalled at all;
• a redeployment is the one with no way out: those ships change home, so it says there is no scheduled end — the colony stays exposed until you deliberately send defences back or build new ones.

Two details that decide whether a prompt is worth reading. It judges what STAYS BEHIND, not what you selected: leaving a single scout at home does not buy you a quiet launch, because a scout is not something the battle engine fields. And it only fires on the change — a colony that was already undefended does not ask again, so the warning stays rare enough to mean something when it does appear.

The test for "undefended" is now literally the same one the battle engine uses to stamp that word on your battle reports: ships, defence plans, Batteries, Fortresses, Shield, Orbital Cannon, planet traits and energy shortfall, all from one definition. What the warning calls exposed and what a battle calls unopposed cannot drift apart.

Also

The return time shown in the warning is the latest planned return before any battle delay — bonuses that only the server knows can bring the fleet home sooner. When the game does not yet know your travel settings, it says so in words instead of showing you a number it cannot stand behind.

1.0.0-beta.232 — 2026-08-22

New

The Guild tab tells you when something needs you (#1182). Until now, opening the Guild tab was the only way to find out whether anything was waiting behind its fifteen doors. Three of them now say so before you open them:

• Joint raids turns red when a guildmate is under attack — the door behind which you can actually send them cover.
• Help board carries a quiet dot when somebody has asked for a hand since you last looked. Opening the door clears it.
• Treasury tells you when your donation charges are full, so the ones regenerating on top of them stop going to waste.

Two colours, and the difference is deliberate: red is something happening on a clock, cyan is something waiting with none. The Guild tab's own dot in the navigation follows only the timed half — a raid forming, or a guildmate with a fleet on final approach — so it never burns all evening over something nobody has to rush.

Both the phone's command room and the desktop rail now show all of this. They used to keep two separate lists of what each door says, and those lists had already drifted apart, so some doors told desktop players things they never told anyone on a phone. There is one list now.

An attacked guildmate arrives with their Observatory reading (#1182). The "Allies under attack" list used to give you a name, a coordinate and a countdown, and then say their observatory sees these fleets coming without showing you any of what it saw. It now shows the readout that guildmate's own Observatory produced: where the fleet is coming from, who is flying it, how many hulls or what size, which classes, and — at the top levels — the exact composition.

Nothing new is being revealed. It is the defender's own intel about an attack on themselves, at exactly the detail their own Observatory buys, riding their own request instead of being retyped in chat. And when their sensors could not read the fleet, the row says so in red — fleet unidentified, or partial reading — cloaked hulls are not counted — because a cover sized against four visible hulls is the wrong cover when sixty are coming.

Also

The Observatory readout appears wherever more than one fleet is inbound, and says which of them it is describing. Numbers in it are grouped in thousands, and a single incoming hull is now "1 ship" rather than "1 ships" on both the guild's row and your own inbound alarm.

1.0.0-beta.231 — 2026-08-22

New

Tier 8 Mastery: three choices, and each one costs you something (#1108)

Research used to reach the Eschaton Depths and find nothing waiting there but more storage. Tier 8 now asks you to decide, three times over — and all three are trades. Both halves are on screen before you commit.

Battle stances, chosen per fleet at launch, once you have researched Admiralty:

• Assault — +10% attack, −10% hull.
• Rampart — +10% hull and shields, −10% speed.
• Pursuit — +15% speed, −10% cargo.

The stance belongs to the fleet that leaves. Change your mind and the fleet already in flight keeps the orders it launched under. On a joint raid every ally fights under their own — the one who opened the raid does not choose for everybody.

Pursuit can only be taken where its cargo cost can actually be charged, so it is offered on a raid or a loaded run and not on a probe or a garrison, where it would be free speed.

Planetary mandates, one per colony, changeable every 72 hours, once you have Planetary Mandates:

• Industrial — more Ferrolite and Xenogas, less Data.
• Scientific — more Data, less Ferrolite and Xenogas.
• Transit — faster Transport, Redeploy, Harvest and Scouting leaving that colony, paid for across all three resources.

Industrial edicts, one for the whole empire, also on 72 hours, once you have Industrial Edicts:

• Construction — buildings finish faster, ships slower.
• Shipbuilding — the other way round.
• Throughput — more production, and both queues slow down.

An edict changes what the next order costs in time, never one you have already paid for: a queue you filled yesterday keeps the durations you were quoted, in both directions. Changing or dropping an edict settles what your colonies have already earned first, so it can only ever price the future.

Choosing nothing is still a choice, and it is free — an empire, a colony or a fleet with no selection behaves exactly as it did before today.

Changed

The mission composer now prices the stance everywhere it shows you a number, not just in one place: the battle forecast, the loot estimate, the fleet totals, the cargo cap and the arrival time all reflect the stance you picked. Under an edict, the build and shipyard panels quote the duration the order will actually take.

Also

Every account gets its first mandate and its first edict change immediately: the 72-hour clock starts when you first use one, not when this update shipped.

1.0.0-beta.230 — 2026-08-22

New

The spy report launches the attack it justifies (#1179)

An espionage report told you everything about a target except how to get there. You studied the defences, the fleet and the emplacements in Activity — then walked to Missions and re-found the same address by hand, usually through the Recent row, which carries none of the report's context. Deciding with the intel in front of you meant memorising three numbers across a tab switch.

Every espionage report now carries two actions under its header, without opening the card:

• ⚔ Attack — the launch form opens with the report's coordinates already in the target.
• 🛰 Spy again — the same, for another probe at the same address.

Nothing is sent by pressing either one: they aim, you still choose the fleet and launch.

The intel comes with you. Missions already looks up the freshest espionage report for whatever coordinates it is aimed at, so the report's age, the attack forecast built on its snapshot, and the Advisor are all there when you arrive — which is why this needed no new panel anywhere.

The actions do not judge whether the target can be hit. A shielded target is still aimable, and the launch itself is where that gets refused, as before. The report shows you what a probe saw; it does not hand out permission.

1.0.2 — 2026-08-22

Changed

• Dark Matter Lances now pierce shields instead of adding attack (#1185). Before, the Lances and the Doctrine of Annihilation both gave the same +0.5% attack per level — two different endgame nodes pushing the same number, so the tier-7 fork was not really a choice. The Lances now ignore a share of the enemy's ship shields: your fire lands where it used to be soaked. The share grows with every level and never quite reaches everything, so the node stays worth buying however deep you take it. Existing levels carry over as penetration — nothing was reset, and nothing was refunded.
• The Doctrine of Annihilation is worth what its price suggests (#1185). It used to add a flat half-percent per level on top of a total that everyone maxing their Disciplines already shares, which left each level worth about a fifth of a percent — for a node whose later levels cost more than most empires earn in a season. It now multiplies your attack by 1.5% per level, so a level is worth the same proportion whether it is your first or your twentieth.

Also

• The Power Details panel gained an Enemy shield penetration row under Fleet combat, so you can see what the Lances are doing for you.
• Both tier-7 cards were rewritten to describe what they now actually do.

1.0.0-beta.229 — 2026-08-22

Changed

• Activity opens on your own outcomes (#1189). Since the galaxy's chronicle moved into the Activity feed, the tab opened on All — and the chronicle is the bigger half, so the screen the unseen-battle dot points at opened on somebody else's night. It now opens on Mine. Nothing was removed: All, Galaxy and the four family chips are one tap away, and the feed's contents, order and grouping are unchanged.
• If you have nothing of your own yet, the tab opens on All instead — a brand-new commander gets the galaxy to read, not an empty column. The choice is made once when the tab opens: tap Mine 0 on purpose and you stay there, and a result landing mid-read never swaps the list out from under you.
• The empty Mine notice no longer says "yet". With the tab opening there, a commander with three fleets in flight could meet a sentence written to welcome newcomers.

1.0.1 — 2026-08-22

Changed

• The Proving Ground pays a bracket with only one empire in it (#1188). If you were the only empire ranked at your fleet size, your board used to read "One empire ranked here. No prize until 2 empires are ranked." and the season close paid you nothing. It now reads "One empire ranked here. Pays at the close." and you are paid your bracket's crate. Being the only empire in your bracket is not something you can do anything about — nobody can bring a rival up to their own fleet size — so the board no longer withholds the prize for a reason you cannot act on.

• A bracket with nobody ranked in it still pays nobody. That part is unchanged: Warhost and Apex open at 10 million and 100 million military power, and until an empire actually ranks there they pay nothing at all.

• The prize table says the new rule. The line explaining the close now reads "A bracket pays once one empire is ranked in it", and the note about the crate ceiling no longer claims that a bracket you reach alone pays nothing — it says what is still true above Armada, which is that the crate stops growing and climbing higher buys you a smaller field rather than a bigger prize.

Also

• On iPhone and Android you will read the old rule until you update. The app carries its own copy of the screen, so an installed build still shows "No prize until 2 empires are ranked" until you install this version. The rule the season close actually applies is the new one, so a lone entrant on an old build is paid regardless of what their screen says.

• This season counts. The change applies to strikes already made in the season currently running — you do not need to strike again for it to take effect. If you have not struck at all because the old wording said it could not pay you, a strike costs nothing and nothing is lost, and there is still time.

1.0.0-beta.228 — 2026-08-22

Fixed

• Admin colony totals explain their scope (#647). The moderation roster used to show every colony under the same bare label as the dashboard's human-only figure. It now states that the total covers all empires and includes beacons.

1.0.0-beta.227 — 2026-08-22

New

Every battle in the galaxy is replayable (#1168). The Activity feed has always told you that a battle happened — who struck, where, what they carried off. Now it will show you. Every battle row in the galaxy's chronicle has a Replay button, and it opens the same film the people who fought it watch: every salvo, every round, the shield pool going down, the batteries firing.

Your own old battles came back (#1168). The archive only ever kept the replay for your last twenty reports; past that you could scroll to a battle and find nothing to open. Those rows have their Replay button back, however far down you go.

Changed

A battle opens to the galaxy a day after it is fought — your own, at once. A fresh report is not just a story, it is a map: it carries the defending fleet's formation, the strength its walls started with, and the ships each attacker still has in the air on their way home. Handed to everyone the moment the shooting stopped, that would have been a free scouting report on whoever had just been in a fight — better than a probe buys, with no probe to lose and no warning to the person being read. A day later the fleets have landed and the losses are rebuilt, and what you are watching is a war story.

Whoever was in the battle never waits: attacker, defender, everyone who joined the raid, and the allies who had a garrison there.

What a stranger sees of your battle. The battle itself is the same for everyone watching — the ships, the rounds, the losses, the outcome. What does not travel is where each fleet set out from and the research each commander fought with: the things a spy has to be sent out for.

Beyond your last twenty reports the archive offers the replay but not the Details panel, and now says so instead of leaving a button silently missing.

1.0.0 — 2026-08-22

Changed

• Novarchy leaves beta. This is the first release numbered as a finished game rather than as a preview of one. Nothing in the game changes with it — no balance, no screens, no rules. The number was simply behind: Novarchy has been live and updating many times a day for weeks, while every release was still numbered as a beta, which is how you number a game that has not launched yet.

From here the version carries a little information. When a release adds something we call out as new, the middle number moves; when it only fixes or adjusts, the last one does. So 1.1.0 and 1.0.1 usually tell you something different before you read a word — usually, not always, because it is the release notes that decide, and a small addition sometimes travels under a heading that does not announce itself.

Every earlier release keeps the name it shipped under. Nothing was renumbered, nothing was removed, and no existing link changed.

Also

• The check on whether a release reached the public changelog now works. After each release we check that it actually appeared on novarchy.com/changelog. That check was looking for the version written in a form the page has never used, so it always answered "not there" — whether or not the page had updated. It now reports honestly whether the entry arrived.
• Nothing else changed. No new content and no fixes to the game itself: this release is the version number and the machinery that produces it.

1.0.0-beta.226 — 2026-08-22

Fixed

The Command Center stops telling you the ladder is over at 15 (#1159)

From Command Center 15 upwards, the gold "At the next level you unlock" panel — the one you decide a ~430k ferrolite upgrade from — said there was nothing left to unlock. That was wrong, and the game had been granting things above 15 the whole time.

You really do get, and now you are told about it one level in advance:

• your 6th to 10th commandable colony, at Command Center 18, 21, 24, 27 and 30;
• a 5th, 6th and 7th Site for each economic building, at 19, 23 and 27;
• at Command Center 20, the level cap that lets every building reach lvl 20 — and with it Stage III, the third look a building takes on, exactly as level 10 opens Stage II.

The "The road ahead" list had the same hole: it showed nothing at all to anyone at Command Center 15 or above, which is to say to the players with the most Command Center left to buy. It now opens on the whole remaining ladder.

And it stops promising build queue slots that were never coming

The same panel said "queue slots keep growing regardless". They do not: the fourth slot arrives at Command Center 13 and is the last one. A player at 16 read that sentence and, fourteen levels later, still had four slots. It is gone.

Changed

Every Command Center level from 1 to 30 now says what it gives. Levels that raise only the level cap say exactly that, instead of the old line that opened by denying it unlocked anything — under a heading announcing an unlock — and went on to promise a build-time speed-up that, at some of those levels, the panel's own discount row measures as zero.

The colony advisor no longer offers one of those levels as a reason to buy the Command Center. A level whose only content is the cap is not an argument for spending on it, and that suggestion sits one tap from the queue button.

1.0.0-beta.225 — 2026-08-22

Changed

• The synthetic empires can turn a fleet around now (#592). Recall shipped for you a few releases ago; they did not have it, and that was becoming a way to tell them apart — once everyone else calls off a doomed attack, the ones who never do stand out. They can call one off too, and only for something they actually saw: a probe flown after the fleet left that finds a garrison it can no longer beat, a pact signed with your alliance while it was in the air, their own plan turning defensive, or the target changing hands, vanishing, or joining their guild.
• They stop like people, not like a timer. A synthetic empire only turns a fleet around during its waking hours, it never uses the last stretch of the legal window — the last two to five minutes before the point of no return are yours to count on — and it takes a moment to give the order, the same beat it takes before it speaks in chat.
• Ask one where its fleet went, and it can tell you. A recalled mission shows as recalled, on its way home rather than as a fleet returning from a battle, and the empire can talk about turning it around — and about an attack that turned back on it.
• Pointing a fleet at one and pulling it back is remembered. It counts as hostility and colours how that empire speaks to you afterwards. It does not, on its own, earn you a punch back: what buys a reprisal is a blow actually landed or words actually said, so nobody can goad a smaller empire into a fight it will be fined for by feinting at it for free.
• Whether they can name you depends on your Observatory, as it always has. Below tier 4 the alarm and its cancellation are anonymous at both ends, and the empire will say so rather than guess.

1.0.0-beta.224 — 2026-08-22

New

• Two new brackets in the Proving Ground: Warhost and Apex. The boss board ranks you only against empires whose fleet is the size of yours, and until now the top rung — Armada — opened at 1,000,000 military power and had no ceiling above it. Everything from the largest human fleet up to the biggest fleet on the server raced in that one bracket, more than a hundredfold apart. Warhost now opens at 10,000,000 and Apex at 100,000,000, so the top of the ladder is cut the same way every other step of it is (#1172).

  Nobody has reached either one yet, and that is on purpose: a bracket pays nothing until two empires are ranked in it, so headroom costs nothing — while adding the rung on the day somebody first crosses it would move a growing empire out of the race it had been climbing all season. Your stored score does not move either. The close pays on the bracket each best strike was made in, so a strike already filed in Armada stays in Armada and is paid there.

  The crate does not keep growing with the ladder. Armada, Warhost and Apex all pay 3 × 5M crates: that is the largest crate below the season winner's own prize, and a race that costs nothing to enter and risks nothing must not out-pay the season itself. The prize table now says so instead of leaving three identical rows unexplained.

Changed

• The ladder reads properly on a phone. With six rungs instead of four, each one gets about a sixth of the card, and "from 100,000,000" no longer fits beside its neighbour. The thresholds on the scale are now written short — from 0, from 10k, from 100k, from 1.0M, from 10M, from 100M — and the rung names size themselves to the space they actually have, on tablets and half-width windows too, not just phones. The exact figure you are chasing is still spelled out in full in the line underneath.

• The line that tells you how far the next bracket is now tells you what it is worth. It used to say only "Warhost opens at 10,000,000 military power — you are 8,140,000 short", which is true and half the story: Warhost pays the same crate as Armada, and a bracket you would be alone in pays nothing at all. It says both now — and it stays quiet when your arrival would be the second empire, because then the bracket does pay.

• A bracket nobody has reached is stated, not alarmed. An empty bracket used to be drawn in the same red as one sitting a single entrant short of paying. Red is a call to action, and there is nobody to call: those cards are neutral now, and still say what they pay and when. This also affects the first hour of a season, when every bracket is empty.

1.0.0-beta.223 — 2026-08-22

Changed

• The Command Center's upper levels are reachable again (#1161). Past level 20 the Command Center had priced itself out of the game: every level cost 70% more than the one below it, so level 30 alone came to 1.2 billion resources and the last three levels together were four fifths of the entire ladder. In practice nobody ever got there — across the whole server there was not a single empire anywhere between level 22 and level 29.

  The climb now eases after level 20. Level 21 costs 8.1M instead of 10.2M, level 24 costs 19.9M instead of 49.9M, level 27 costs 48.8M instead of 245.3M, and level 30 costs 120.2M instead of 1,204.9M. Everything up to level 20 costs exactly what it cost yesterday — not a single resource more or less — and no other building changed at all.

  Level 30 is still about twenty times the price of level 20, so the top of the ladder is still something you work for. It just stopped being a wall.

Also

• If your Command Center is above level 20, your Power and your leaderboard score will drop. Nothing was taken from you and nothing was spent twice: those numbers are worked out from what your buildings cost, so when the price of a level comes down, the credit it carries comes down with it. Your colonies, your resources and your fleets are untouched, and below level 21 nothing moves at all.

• The faction bastions get a little weaker. Their fleets are sized against the strongest empire in the field, so they follow that number down — and they will follow it back up as the top of the table grows again. The season's contested planets are not affected.

1.0.0-beta.222 — 2026-08-22

Fixed

The colony no longer opens with its base tucked under the HUD (#616)

On a phone the colony fills the whole screen and the interface floats on top of it: the resource bar covers the first 150-210 pixels, the navigation bar the last 60-95. The camera had never been told any of that — it centred your base on the middle of the screen, which put it 45 to 81 pixels above the middle of the strip you can actually see, with its top edge sitting behind the HUD.

The Base tab now measures those two bands and opens the colony in the middle of what is left free. On a planet whose painted ground is not offset the centring is exact; on rock and ice the terrain calibration keeps 12 to 27 pixels of the old drift, because the alternative would be showing you the edge of the world.

Two consequences worth knowing about, both on phones only:

• your base is drawn 11-16% larger. Sliding the view down slides the painted ground with it, so the ground has to be magnified to keep reaching every edge of the screen. That is the price of the slide and it cannot be had separately.
• the base is not made smaller to fit. Shrinking it into the free strip would have cost a quarter of its size in landscape and changed nothing at all in portrait, so the size rules are untouched.

On desktop and in landscape nothing moves and nothing costs anything.

1.0.0-beta.221 — 2026-08-22

The Events tab is gone. What was in it is now in two better places.

Changed

• Activity is one feed (#1160). Your own outcomes and the galaxy's public chronicle now scroll in a single column, newest first, instead of living in two tabs side by side. Your battles keep everything they had — Replay, Details, the full report — and the galaxy's news is interleaved around them by time.
• Seven filters, two questions (#1160). Mine and Galaxy ask whose a row is; Combat, Exploration, Empires and Economy ask what happened, across both sources at once. Mine means what happened to you, not what came from your own log — so a purchase you made on the market, which only ever existed as a public line, is finally under it.
• Season is its own tab again (#1160), with three full-width screens behind one switch: Standings, The Proving Ground, and Contracts. Nothing is stacked below anything else any more.

Fixed

• The season podium was invisible to exactly one player (#1160). Every season, one account — whichever one the register happened to pick to represent the final table — was the only one in the galaxy that never saw the podium in their feed. It is there for everyone now.
• The Contracts dot led nowhere (#1160). When the day's new contracts appeared at 05:00, the marker lit up on the tab, you opened it, and nothing on the screen said where to go. The Contracts lane now carries the marker too, and it says which of the two things it means: a reward waiting, or a new hand for the day.
• A new commander tapping Mine was told "nothing under this filter" (#1160) instead of being told their fleets' outcomes will land there.
• A broken galactic log looked like a quiet galaxy (#1160). If the chronicle cannot be read, the feed now says so instead of showing an empty stretch and letting you assume nothing happened.

Also

• Your own trades are no longer folded into a "4 trades" summary when you filter to Mine — on your own ledger, every purchase is listed.
• The Contracts marker now has a name a screen reader can read, which the old phone strip used to say out loud and the marker alone did not.

1.0.0-beta.220 — 2026-08-22

New

• The game says who built it (#1152). Novarchy never named its maker anywhere. A quiet line — powered by AigenSolutions — now sits at the bottom of the login screen, under Privacy, Support and Changelog, and links to aigensolutions.it. It is footer-sized on purpose: it stays out of the way of Play.
• And you can find it once you are in. A player who starts as a guest never sees the login screen again, so the same line closes the Command menu on a phone and the commander card everywhere. The landing page, the privacy page and the support page carry it too, in exactly the same words.
• It opens in your browser, not inside the game. In the iOS and Android app there is no address bar to get back with, so tapping the credit hands the link to your normal browser and leaves the game exactly where you left it.

1.0.0-beta.219 — 2026-08-22

Fixed

• Planet names on the galaxy map no longer run into one another (#594). In a busy system, the names of two neighbouring worlds could end up side by side with nothing between them and be read as a single name — one player reported seeing a guild affiliation that does not exist. The map used to separate two names only when their letters actually overlapped; now it also separates them when they simply sit too close to be read apart, and the nearer world keeps its name. Names that are genuinely apart, or on different lines, are unchanged.

Also

• The effect had come back after systems grew from twelve planets to eighteen, which put half again as many names in the same sky. Measured across more than a thousand camera angles in a fully settled system, pairs of names too close to read apart went from common to none, and the map shows very slightly fewer names in exchange.

1.0.0-beta.218 — 2026-08-22

Fixed

• The battleship flies the right way round again (#1153). In the battle replay it was advancing on the enemy backwards: flat armoured end first, and its bow — the end carrying the five blue rings — trailing behind. It had done that in every replay since the battleship got a 3-D model, on both sides of the line, attacking and defending. It now leads with its bow.
• The shipyard sheet stops contradicting itself (#1153). Opening a battleship's detail card cross-fades its illustration into the 3-D model, and the two were facing opposite ways through the dissolve — permanently so for anyone browsing with reduced motion, since the model then never turns. Both now point the same way.

Also

• Which way that hull points is now checked automatically. Nothing in the test suite could tell the two directions apart before — which is how a ship flew backwards for months with everything green.

1.0.0-beta.217 — 2026-08-22

New

• Every ship now wears a rarity. The Shipyard grid gave all sixteen hulls the same slate border, so the Destroyer — the one ship an empire can only be given — looked like a Freighter with better art, and the only hint that the catalog had steps in it was the padlock, which says "not yet", not "how far". Each card now carries Common · Uncommon · Rare · Epic · Legendary: as the colour of its border, as a glow behind the artwork, and as a word on the card itself. The Destroyer is gold and glows; the Titan Dreadnought is the violet one step below her.
• The step is named where you go to read about a hull, too — on every row of the ship reference, which shows all sixteen at once, and in the detail panel next to the class.
• Where a hull sits is decided by how far into the game it is, not by a list someone maintains: the Tree node that unlocks it and how deep that node sits, or, for the hulls no node gates, what it costs to put one on the pad. The Destroyer is Legendary because she can only be won, never built.

Changed

• The Destroyer's count reads ×1, like every other hull. It used to read 1/1. Nothing about how you get her has changed — the build card still says she is an Event Relic, not buildable, and granted only once.
• A hull you have not unlocked yet keeps its colour. Locked cards are still greyed out — dimmed art, grey text, the padlock, the line telling you which research opens them — but the border and the rarity label now stay legible instead of fading with everything else. They were the seven cards where the new ladder was hardest to read, and they are the seven it exists for.
• Selecting a card is now a cyan ring drawn outside it rather than a coloured border, so the card you are about to build keeps its own rarity colour while it is selected.

1.0.0-beta.216 — 2026-08-22

Fixed

The Missions "Recent" row now tells you how much salvage is out there (#586)

After a battle, wreckage is left floating at those coordinates, and the game has always worked out how much of it is still there — decaying it hour by hour, subtracting whatever you have already lifted. You never saw that number in the Recent row. It was computed on every refresh and thrown away.

The row had a whole chip type for it, and that chip could never appear: a coordinate only gets a salvage estimate because one of your own battle reports left wreckage there, and that same report already puts a "battle" chip on the row, which won every time. So the figure existed and was never once shown to anybody.

Now the chip that does appear wears the estimate next to the address, in gold: • 4:2:7 ♻ ≈ 305,000,000. You can see there is a harvest worth sending recyclers for before picking the target, instead of having to select each coordinate in turn to find out. The tag rides on whichever chip you get — your own colony with wreckage in its orbit, a target you have a fleet already flying to, a coordinate you pinned — not only on battle chips.

The figure is an estimate and says so with ≈: it is too high if somebody else has harvested there without your knowing, too low if somebody else has fought there. It does not yet subtract a harvest of your own that is still in flight — that is #1146, and it affects the target panel and the Fleet Advisor the same way it always has.

Removing a chip with × still hides everything known about those coordinates, salvage included, until a newer contact brings the chip back — and when it comes back, it comes back with the figure.

1.0.0-beta.215 — 2026-08-22

New

• The Season screen has two lanes. (#1124) The Proving Ground used to sit at the bottom of one long scroll, below the key art, the podium table, the scoring chips, the Eschaton warning, your own position and a ten-row leaderboard. Standings and The Proving Ground are now two full-width screens with a switch between them, and the standings carry a card that opens the bench.

• Your score is the headline. (#1124) The figure this screen exists to produce — how much boss hull your navy destroyed — is now the first thing on it, with the rounds it took, the strikes you have made and where it places you. A score that broke the wall carries the same ★ the board does.

• You can see what you are hitting. (#1124) The Sunless Vault's composition is drawn: how many cruisers, battleships and titans stand in it, and how many hulls that is in total. Until now the number you scored had no denominator.

• The bracket ladder is a scale. (#1124) Instead of "you are ranked against fleets your size", the four brackets are drawn as rungs with your own military power marked on them, and the gap to the next one spelled out: "Fleet opens at 100,000 military power — you are 51,700 short."

• Only your own board opens. (#1124) Four boards of ten was forty rows, thirty of them about races you cannot enter. The other three collapse to their leader, how many empires are ranked in them and whether the close will pay them at all — and each opens on its own board when you ask for it.

• What the close pays, and which bracket it pays it on. (#1124) A table of the crate each bracket wins, and the rule that decides which one is yours: the close pays on the bracket each best strike was made in, not on the fleet you hold at the end. If you struck at 8,000 military and have since grown, you are still racing in the bracket you struck in until you strike again.

Fixed

• The invitation no longer appears beside the reason you cannot accept it. (#1124) "Nothing is spent — strike as often as you like" was printing two lines above "No season is open." and "You have no ships to send."

• A strike that dealt nothing no longer looks like a win. (#1124) A fleet of freighters and probes passes the "have you got hulls" check and destroys exactly nothing. That zero was being shown in gold, at the size of a record, directly above three lines explaining that it was not on the board.

• A shipless empire is no longer told which race is theirs. (#1124) With no ships your power is zero, which files you in Patrol — and the screen was treating that as a race you had entered.

Also

• The other brackets' boards, when you open one, are the same board your own bracket shows: the same columns, the same ★, the same names on a phone.

1.0.0-beta.214 — 2026-08-22

The Research screen has been rebuilt, on both desktop and phone.

New

The research queue is a bench, not a row of dots (#1130). Before, your lanes and queue slots were ten-pixel dots with a line of text beside them — and one hollow dot was drawn for every slot the game can ever grant, so a new player counted four and owned one. Now each lane is a card: what is running, how far along, how long left. Your queue slots sit beside them, and the next slot you have not unlocked yet says which Laboratory level opens it.

The Laboratory network and your treasury are on the same panel. The number the game actually researches with, the arithmetic behind it when you own several Laboratories, and the three resource totals that every cost on the screen is measured against.

Every card now shows what its next level costs and how long it takes. That used to live only inside the detail panel, so choosing between two Disciplines meant opening one, reading two numbers, and going back. A green or red dot tells you whether your mother colony can pay for it today, and when it cannot, the card says which resource is short and by how much.

Filter and sort the catalogue. Affordable now, In the queue, Locked, and a shortest first sort — the two questions the grid could not answer before.

Disciplines, the tech tree and the Dark Matter apex are one catalogue. They used to be three sections stacked down a page four screens long; the tree began below ten cards and a detail panel, so a player who had never scrolled that far did not know it existed. They are one switch apart now, and the switch says how much of each you own.

Fixed

A locked tech tree node tells you what it is waiting for. A padlock said "no" forty-odd times and never said why — finding out meant opening every node in turn. Each locked node now names the one thing that is missing: a Discipline level, a prerequisite, or a Laboratory.

A queued order that cannot start says so. If your Laboratory network drops below what a queued order needs — leaving a guild can do it — that order stops the whole queue, and every order behind it. The screen used to say it was waiting for a lane, next to a lane it had drawn as free. It now says the queue is on hold and which level it is waiting for.

A waiting order and a stalled one no longer look identical. An order waiting its turn and an order the treasury cannot pay for were drawn the same way while the queue sat still.

A card no longer prices a level you already bought. With an order running, the queue level and the next level are two different numbers, and the card was mixing them.

Changed

The tree's symbols are drawn instead of typed. The branch and node emoji came from the system's colour emoji font on iPhone and iPad, which ignores colour — a locked node's symbol was as bright as an unlocked one, and sizes differed from device to device. They are proper icons now and take the colour of whatever state they are in.

1.0.0-beta.213 — 2026-08-22

Fixed

The Proving Ground board was ranking the wrong thing (#1126). A strike used to be scored on everything it landed, including the shields the boss soaks up — and those shields grow back between rounds. So a fleet that could not break through kept hitting the same shields over and over, and every round added to its score.

The result was a board that rewarded taking longer. Measured on the live server: of two fleets that both destroyed the boss outright, the one less than half the size scored 78% higher, purely because it needed 400 rounds instead of 6. The empire at the top of the Armada bracket would have improved its standing by scrapping half its navy.

A strike now scores the boss hull it destroyed. Shields do not count. The whole wall is a fixed amount of hull, so that is the most any strike can score — and once you can break it outright, what still improves is how few rounds you need. The board shows those rounds next to every score, and a row that took the whole wall is marked with a ★.

Changed

Every score on the board has been reset, and this is not a bug. Scores taken the old way cannot be compared with scores taken the new way, so every existing strike has been retired. Open the Proving Ground and strike once to get a score on the new scale — it costs nothing, as always, and nothing is lost.

Expect a much smaller number than before. That is the point: the old figure counted the same shields hundreds of times. The new one counts what you actually broke.

Also

Battle reports credit a fleet that surrenders correctly now — a defence that gives up without firing used to be recorded as if almost nothing had happened to it.

1.0.0-beta.212 — 2026-08-22

New

• Synthetic empires now run campaigns, not isolated moves (#1083). Before, an agent decided each tick more or less from scratch. Now it forms a plan against a specific colony and works it in stages: scout it, gather a force that can actually take it, stage that force at whichever of its colonies is closest, launch, and follow the mission through. If the target is razed, changes hands, or goes behind a shield while the fleet is in the air, the agent notices and calls the attack off instead of flying into nothing.
• They coordinate on relics. An agent will never open a relic raid on its own. It joins one only when a coalition has already formed and the fleets actually promised are enough to take the site — and when it puts out a call to its guild, that call is a real request it will act on.
• They learn from what happens. Outcomes feed back: an agent that keeps losing gets more cautious about the size of force it commits, and one whose forecasts hold gets bolder. Fleet composition, research order, scouting and defence all follow from what it has actually seen rather than from a fixed script.

Changed

• What an agent says about its own operations is tighter. It can tell you it is scouting a colony, or that an attack is committed, but it will not read out a win percentage or an expected loss figure — those come from its own maths and correspond to nothing you can open a panel and check. It also no longer names the colony it is staging from: which system an attack launches from is what the Observatory is for, and an agent volunteering it in chat was giving away a readout you pay for.
• Stale talk is gone. An agent used to be able to keep announcing a campaign stage against a colony that had been destroyed days earlier, or an attack that had already come home. Those claims now expire, and are dropped outright when the operation behind them ends.
• What an agent thinks your fleet is is now what it last saw, per colony, and it forgets sightings older than three days. Previously it added every scouting report together, so an opponent it scouted five times looked five times stronger than they were — and the more it watched you, the less willing it was to ever attack.
• Agents now field the hulls they had been quietly building. Their fleets were only ever counted from the five starting hull types; everything unlocked through the tech tree was invisible to them and stayed parked. Those ships now fly. Expect agent fleets to look noticeably bigger from the first hours after this update — that is the correction landing, not a sudden build-up.

Also

• One piece of this work ships switched off. Agents can read a promise out of their own chat ("I'll hit 4:2:7 tonight") and turn it into a real operation, and that turned out to be much harder to do safely than it looks: across seven rounds of review the parser kept finding new ways to bind the wrong coordinate — a colony mentioned in the next sentence, a raid the agent had already fought, someone else's threat quoted back, and twice a colony the agent had just publicly agreed to leave alone. An agent attacking a player it never threatened is worse than an agent that talks without acting, so the feature is in the codebase and disabled until it rests on something that can actually read a sentence. Everything else above is live.

1.0.0-beta.211 — 2026-08-22

New

• Battle replays glow (#1125). Everything meant to read as light now bleeds into the frame instead of stopping at a bright pixel: muzzle flashes, explosion cores, the lit engine bells on ships that have them, and the starfield. Measured across one frame of a real battle, the brightest tenth of the picture roughly doubles while the very top comes down slightly — bright things now roll off like film rather than clipping to white.

Fixed

• Three ships were being drawn too dark. The shroudrunner, the probe and the colony ship are painted cutouts rather than modelled hulls, and they were rendering far darker than the ships beside them. Nobody could see it, because the whole scene was drawn the same way. They now sit at the same brightness as everything else.
• A replay that loses its canvas says so. If the browser takes the drawing surface back mid-battle — a background tab, a phone under memory pressure — the picture used to freeze while the transport bar kept advancing, which looked exactly like the game hanging. It now tells you what happened and that reloading will bring it back.
• A device that cannot run the effect falls back instead of going black. On graphics hardware without the feature bloom needs, the replay would have played to the end on a solid black canvas, with the loading text cleared and the bar moving. It now simply plays without the glow.

Changed

• On a phone that cannot hold the frame rate, the glow is the first thing given up. That was already how the quality ladder was built; this is the first release where it has anything to give up. The rest of the picture — the colour of space, the shield shell around a defending fleet — is held steady across that switch, so the replay does not change appearance beyond losing the halo.

Also

Space is the same colour it has always been. Rendering the whole frame through one final tone-mapping pass would have taken the deep blue of space almost to black and the faint nebulae with it; the backdrop is authored against that pass so it arrives where it always did.

1.0.0-beta.210 — 2026-08-21

Changed

• Alex is on the boards again (#1116). The account that builds the game had been kept out of every standing since 10 August: it scored points but never appeared, and its prizes slid down to the player behind it. That exclusion is lifted. Alex now holds a row on the leaderboard, a place on the weekly war scoreboard and a place on the season scoreboard, and is paid like anybody else.

  What this means for you, plainly: if Alex outscores you in a given week or season, the ladder shifts by one. The rank-1 share of the weekly war, the season podium steps, the guild-podium crate bands and the end-of-season Destroyer can now go to that account instead of passing over it. Nothing you have already earned is taken back — the change applies from here on.

Also

• The Proving Ground is not affected. The season boss board is ranked by damage, and damage is mostly a question of how many hulls you brought — Alex's fleet is granted rather than flown, so it stays off that board and off its crates, exactly as before. Concretely: the Armada bracket still pays only when two real fleets have contested it. A bracket that was uncontested yesterday is uncontested today, and nobody has been quietly beaten to a crate.
• The relics and the faction bastions are not affected either. Those are sized against the fleets that actually describe the field, and that calculation has never counted Alex's. A relic that was reachable yesterday is just as reachable today.
• The Kepler Observatory and the Wraiths remain outside every standing: they are the opposition, not competitors.

1.0.0-beta.209 — 2026-08-21

Fixed

• The game sits straight again on a wide monitor, with the chat open (#571) — since the chat got its own lane beside the game, it was taking that lane out of every screen, however wide. On a large monitor the result was a page visibly off-centre: the resource and navigation bar across the top stayed centred, while everything below it — the log, the colony, the market — was pushed to the right, with a corridor of empty space between the chat panel and the content. On a 2560-pixel screen the game sat 261 pixels off from the bar above it.

  Now the chat only pushes the game as far as it actually needs to. On a screen wide enough to fit the whole game past the panel, nothing is pushed at all: the game lines up exactly under the bar, the way it does with the chat closed. On a laptop nothing changes — the chat still takes its lane, and the Base screen keeps the same room it had.

1.0.0-beta.208 — 2026-08-21

New

• Deep Vaults — the Warehouse is no longer where the late game stops (#1107). A new Forge research doubles the storage capacity of every colony you own. At Warehouse 30 with Compressed Depots maxed, a colony went from holding 11,963,051,962 of each resource to 23,926,103,924 — and the doubling applies at every Warehouse level, including the free floor a colony has before you build one at all.

  It needs Compressed Depots at 10 and Production Singularity at 3: it extends the storage line you already built rather than starting a new one. The Warehouse card names it beside the figure, so a ceiling that doubles overnight reads as your research and not as a glitch.

Changed

• The tech Tree can now hold more than seven tiers, and the first of the Eschaton Depths is open. Tier 8 appears with Deep Vaults in it; the deeper rows stay hidden until they have something in them, so you will never see an empty tier promising content that is not there yet.

• Anything the game sizes against your Warehouse grows with it, as it always has when you widened your storage — guild aid, the emergency crate, the war chest, and what a lost war costs you. Deep Vaults is worth about one and a half Warehouse levels on that scale, and it moves none of those rules, only the number they read.

Also

• Your daily market intake limit already sits at its own ceiling from Warehouse 14 upward, so most empires will see no change there. Below that, it doubles with the rest.
• Nothing changes for an empire that has not researched Deep Vaults: every cap, limit and reward is exactly what it was yesterday.

1.0.0-beta.207 — 2026-08-21

Fixed

• Desktop chat no longer blocks the game (#571). Before, keeping chat open covered Base controls and the lower-left game area. Now chat keeps its familiar left position in a dedicated desktop lane, so the game stays interactive; minimizing it restores the full-width layout. Phone chat is unchanged.

1.0.0-beta.206 — 2026-08-21

New

The map now tells you where the contested worlds are.

The system dropdown marks which systems hold one — 3 ◆, 6 ◆, and so on — so you can see which door to open before you open it. Fly to one and the world itself wears a gold ring, with its own name written beside it: The Cindergate, Ember Crown, Ashfall Reach. There is a new entry in the map key for it.

This mattered more than it sounds. The Event Sector is nineteen systems of eighteen planets, and six of those slots are contested — one per system, in six systems out of nineteen. Thirteen systems hold none at all. Unless you already knew the coordinates, walking the sector and finding one was luck.

Fixed

Until now the six were labelled "The Verge Watch" on the map — the name of the neutral garrison that holds a world nobody has captured. The same words on all six, indistinguishable from each other and from any other outpost. They read as their own names now, and a world a guild has taken keeps its name rather than reverting to the garrison's.

Also

The Contested Space panel still lists all six with their coordinates, and still tells you who holds what and for how long. What changed is that you no longer have to memorise a number from it and go hunting: the answer is on the control you were already using.

1.0.0-beta.205 — 2026-08-21

New

• The Proving Ground (#1081). A season boss — The Sunless Vault — now stands behind the Season tab, and a damage board ranks who hits it hardest.

  Throw your whole navy at it and find out how hard you actually punch. Nothing is spent and nothing is lost: the fleet never leaves, no queue moves, and you can strike as often as you like. Ships die inside the simulation and none of it is real. Until now there was no way to answer "how hard does my fleet hit?" except by attacking somebody and finding out the expensive way.

  Only your best strike is ranked, and you are ranked against empires with a fleet the size of yours — four brackets, from Patrol to Armada, each its own race with its own top spot. At the close, the best strike in each bracket is paid in crates, bigger crates in the heavier brackets. A bracket only pays if at least two empires have struck in it: a race with one runner is not a race.

  The result is fixed by what you send, so pressing again gives the same number until your fleet, your research or your refit changes. It is a bench for testing a build, not a slot machine.

Also

• Battle reports now record how much damage each side actually dealt — the figure the board is ranked on, and one the reports could not produce before.

1.0.0-beta.204 — 2026-08-21

New

• The ships have their lights on. (#1099) Until now a hull in a battle replay was lit only by the scene's lamps, so a ship turned away from them was a black silhouette — engines, windows and weapon ports were paint like any other paint. Ten of the thirteen modelled hulls now carry their own light: engine bells glow, port rows and canopies read as lit, and a ship in shadow looks like a machine that is running instead of a shape.

  Where it shows most is exactly where it was missing most. A battleship crossing into the dark side of the field keeps its engine bank burning; a swarm of drones reads as a swarm of running engines rather than a cloud of pebbles.

  Three hulls stay dark on purpose. The Ark, the Destroyer and the Monitor were never painted with any lit detail to find, and inventing some would have meant making up lights the art does not have.

Fixed

• A destroyed ship no longer drifts through the rest of the battle with its engines lit. (#1099) Wrecks were being blackened everywhere except their lights, which stayed at full brightness — a burnt-out hull ended up more conspicuous than it had been alive. A wreck's lights now go out with it.

1.0.0-beta.203 — 2026-08-21

New

The six contested worlds are now structures on the map itself, not just in their detail panel.

Fly the event sector and you can see which planets are the contested ones without opening anything: where the other seventeen orbits carry rocks and gas giants, these carry built things — Ashfall Reach, The Cindergate, Hollow Meridian, Sable Anchorage, The Long Silence and Ember Crown, each turning slowly on its own orbit.

The sector already told you it was different — the green nebula, the pale star — but it never told you which worlds mattered. Now the map does.

Changed

They are drawn larger than the body they replace, and that is deliberate rather than decorative. A planet is a solid disc; a station is mostly empty space — a tower, a ring, a spire — so at the same size it reads as a scratch on the dark instead of a place. Twice the body, which is as much as fits between one orbit and the next.

Also

Nothing extra is downloaded to look at the map: a world's structure is fetched only when you actually visit the system that holds it, and it is the same model the detail panel already uses. Systems outside the six are untouched, and so is every other sector.

1.0.0-beta.202 — 2026-08-21

Also

• No player-facing behavior changes (#562). Planet-trait safety checks now cover the galaxy's full configured depth and independently guard shallow and deep slots, so future depth expansions cannot silently escape the existing distribution checks.

1.0.0-beta.201 — 2026-08-21

Changed

• The Assault Swarm is now the Splinter Lance, and it finally looks like itself. Yesterday's release gave this ship a 3D model but left the picture on its card showing the hull it used to have — the two were plainly different craft, and the name still described a formation while the model showed one. Both are fixed. The card, the hero image and the ship you see in battle replays are all rendered from the model itself now, so they cannot drift apart again, and the name matches what you are looking at (#1088).

  Nothing about the ship changed in the game: same cost, same stats, same counters, same place in the tech tree. Your fleets, your build queues and your old battle reports are untouched — only what it is called and what it looks like.

Fixed

• The Research panel promised a screen that does not exist. The Monitor Frigate's node said it screened against this ship. It never did — the Monitor does no extra damage to it at all, and in a mixed fight will shoot almost everything else first. If you built Monitors to hold off a wave of these, they were not doing the job the card advertised. The node now names what the Monitor actually counters: corvettes, interceptors and bombers (#1088).

Also

• If you read yesterday's notes and went looking for the Assault Swarm, it is the same ship — open the Splinter Lance instead. Yesterday's entry also said the artwork still did not match the model; that is what this release closes.

1.0.0-beta.200 — 2026-08-21

Fixed

• Wanted hunts keep their political-map context (#544). Opening a commander from the Wanted board now shows their planet without switching map modes, so closing the sheet returns to the same board.
• Wanted rows disclose missing Power data (#544). When either side lacks a Gate reading, the row now says that the Power comparison is unavailable instead of silently looking evenly matched.

1.0.0-beta.199 — 2026-08-21

Fixed

• Battle contracts explain every scoring boundary (#534). The “Win 3 battles” card now states which PvP roles count, how destroyed fleet value is divided in raids, and how the claim-time production floor and cap are applied. It also keeps that rule visible when a completed-looking card may be rechecked at claim time.

1.0.0-beta.198 — 2026-08-21

Nothing changes in the game this time. No new building, no rebalance, no screen you have not seen before — this release is a floor under the combat engine, and the best possible outcome is that you never notice it.

Fixed

• A battle can no longer run forever (#1051). Since the six-round clock was removed, battles end when they are decided — a fleet destroyed, an attacker breaking off, a siege that cannot out-drain a shield. Those rules are good ones, and they have always ended every real battle. But each of them is an argument, not a guarantee, and if one of them ever had a gap the fight would not produce a wrong report: it would freeze the server for everyone, mid-battle, with no error and no way back except a restart.

  There is now one last rule underneath all of them that simply stops. It sits far above anything a fleet can reach — the longest battle the engine can be made to play is about 40,800 rounds, and the new limit is 25 times that, so no attack you can launch will ever meet it. If it ever does fire, the battle report is marked, so we find out instead of guessing.

Also

• The bench that measures how long battles run was counting them wrong: it read the stored report, which keeps only the first 60 and last 20 rounds, so a battle of a million rounds looked like eighty. It now counts the rounds actually fought.

1.0.0-beta.197 — 2026-08-21

New

Contested worlds are no longer drawn as planets. Tap one of the six in the galaxy and you now see its own structure — a built thing, turning slowly against the dark, different for each world.

They were spheres until now, which was the one thing a contested world is not: a station is not a rock, and the panel had been showing a rock. Ashfall Reach, The Cindergate, Hollow Meridian, Sable Anchorage, The Long Silence and Ember Crown each have their own.

And the panel says which one you are looking at. The six have had names since the day they opened, and the detail panel never showed them — the only place you could read them was the Events tab. Now the world's name sits at the top, next to its coordinates.

Changed

The picture is bigger on a phone, where there is room for it. A sphere reads at any size — it is one circle — but a structure's meaning is its shape, and at the old size the panelling on these simply disappeared. On a phone held upright the panel now puts the structure above the text instead of beside it, so it gets close to the full width. On a short phone, or one held sideways, the layout stays as it was: there is no room there without pushing the buttons off the bottom of the screen, and a smaller structure is better than a button you cannot reach.

Also

Nothing is downloaded until you actually open one of the six, and each structure is about a megabyte — roughly what one of the ship models costs. The rest of the galaxy is untouched.

1.0.0-beta.196 — 2026-08-21

Fixed

• The weekly battle card now tells you which wins count (#534). "Win 3 battles" could sit at 0/3 after a victory you had just watched, and nothing on the board or in the game explained why. The card now carries the rule under its title: only PvP wins count, and the win has to destroy enough enemy fleet value to clear a threshold scaled to one hour of your production, never below 1,000 and never above 20,000. Same treatment the trade card already had — the condition is there before you commit a fleet, not discovered after you have spent one.

Also

• Nothing else changed. No rule, reward or threshold moved: the engine counts exactly the battles it counted yesterday, and this release only puts the existing condition on screen.

1.0.0-beta.195 — 2026-08-21

Changed

• Battle replay: the ships are lit by the sky around them, not only by lamps. (#1076) Until now the hulls in a replay caught light from four fixed lights and nothing else, so their metal had nothing to reflect and read as flat painted panels. The battle now builds a faint sky from its own lighting, and the ships reflect it: edges catch highlights, curved plating separates from the background, and the models look like the metal they were built as.

  Being honest about the size of it: this is a refinement, not a new look. Measured across the whole frame it is about a percent and a half brighter, concentrated on the hulls themselves — noticeable side by side, easy to miss otherwise. The bigger step, making engines and windows actually glow, is still to come.

Also

• Nothing else in the game changes. No balance, no costs, no rules.

1.0.0-beta.194 — 2026-08-21

New

• The Assault Swarm finally has a hull. It was the last light ship still shown as a flat card: everywhere else in the fleet you can spin a ship in the Shipyard and watch it fly as a real shape in a battle replay, and the Swarm could do neither. Now it can — turn it in the Shipyard well, and see it as a solid craft in the replay, like the twelve ships that got there before it (#1078).

Fixed

• A replay with Assault Swarms in it could refuse to open. When a ship's artwork fails to arrive, the game falls back to a simple built-in shape so the battle still plays. For the Assault Swarm — and only for that one ship — that fallback shape could never actually be drawn, so instead of a plain-looking swarm you could get a replay that would not start at all. It had been that way since the ship was added (#1078).

Also

• The Assault Swarm's card illustration still shows the design the ship had before, so for now the 3D model and the picture beside it are not quite the same craft. Matching the artwork to the new hull is the next piece of this work, not part of this release.

1.0.0-beta.193 — 2026-08-21

Changed

On a wide screen the Colony tab is now two columns. The colony's identity and its stockpiles sit in a rail on the left; the advisor, the build queue, the fleet and the whole buildings list get the space on the right. Before, everything was one long column: on a 1440px monitor the buildings list started below the fold, and the resource cards you had already read were what you scrolled past to reach it.

On a narrow screen the order changed too. The advisor — the panel that answers "what do I upgrade now" — comes second, straight after the colony's name. It used to sit behind roughly 330px of resource cards and two more panels. The cards still say everything they said; they are simply no longer in the way of the answer.

New

The buildings list has family filters: All, Production, Military and Civic, each carrying the number of buildings it will show for this colony. They use the same three colours the Base map uses for the same three families. The filter follows the colony you are looking at: switch colony and it resets, so you never arrive at a screen that looks empty because of a choice you made somewhere else.

Fixed

• An iPad in portrait stretched each building onto its own full-width row where two fit comfortably. The list now fits as many columns as the width allows, at every size.
• A colony with a long name pushed a horizontal scrollbar into the left rail. Long names now wrap instead.
• A colony you have just founded, with nothing built yet, said "nothing in this category" over a "Show all" button that did nothing. It now says plainly that there is nothing built here yet.
• Switching colony no longer shows a flash of the previous colony's filter before it resets.

1.0.0-beta.192 — 2026-08-21

Fixed

The Weekly Chest no longer opens under your finger (#1069)

Collecting your fifth Daily Wrap of the week is what earns the Weekly Chest — and the board refreshes the moment you collect. So the chest appeared, with an Open button identical to the one you had just pressed, in the same list, right under your thumb. A second tap landed on it, and the chest was spent before anyone decided to spend it.

Opening a chest now takes two deliberate taps. The first arms it; the row then shows Open chest and Cancel, with Cancel in the spot Open used to occupy — so a repeat tap in the same place cancels instead of opening. The warning sits above the buttons, so a tap at the old height hits nothing at all.

The chest now tells you what it will cost before you open it (#1069)

The chest pays production hours from your whole empire into a single warehouse — your home colony's. On a large empire that can be more than the warehouse holds, even when it looks far from full, and the overflow is simply lost. The only warning before was "your warehouse is nearly full", which stays silent in exactly that case.

The confirmation now says it plainly: roughly what the chest pays, roughly how much would not fit and would be lost, and whether a chrono shard arrives at all. If it does not fit, the chest will wait — spend or upgrade first, then open it. It also warns you what waiting costs: earning a newer chest converts this one's shard for good.

1.0.0-beta.191 — 2026-08-21

Fixed

• The launch panel says the travel estimate can be beaten by your guild (#526). A player in a guild read 8m 14s on the mission launch panel, sent the fleet, and watched the in-flight card count down 6m 22s on the same mission — two authoritative numbers two minutes apart, reported as a bug. Nothing was broken, and the fleet was never the slower one: the panel works out its estimate from what your browser knows, and the server then counts things your browser cannot see, above all the Hypergates of your guildmates — gates in systems you may never have visited. Those can only ever make the trip shorter. Nothing on screen said so. Now, if you are in a guild, the launch panel adds one line under the travel time: the server also counts guild bonuses and your guildmates' Hypergates, and both can only shorten the trip, never lengthen it. A player with no guild sees nothing new, because for them the estimate is simply the estimate.
  • It appears on every kind of mission, attacks and espionage included. Your own Hypergates only speed up logistics runs, but your guild's gate network and travel bonuses apply to everything you launch — which is exactly where an early arrival is least expected.
  • The line is careful about what it promises, and deliberately does not say your guild will beat the estimate: a guild with no travel research, no forward drydock, no Jump Ring and no guildmate gate has nothing to add, and its flights land on the estimated second.
  • "Your guildmates", not "your allies": this is your own guild's network. An alliance pact signed with another guild in the Diplomacy panel does not open their gates to your fleets.
  • What this does not change: the Hyperspace −N% badge still shows only the discount your own gates and relay account for. The rest is known only to the server, so the real saving on a route can be larger than the badge claims — the new line is what tells you that figure is a floor and not the whole of it.

1.0.0-beta.190 — 2026-08-21

Fixed

• The campaign tower can no longer be climbed by an immortal flagship (#1052). An empire whose roster carried the singleton Destroyer could climb the tower ~150 levels past every intended wall: the garrison counted as a single opponent, so the Spectral Bulwark's last-ship floor — a PvP protection — armed in PvE and the flagship simply could not die (in production one empire reached depth 235 on battles of 232,000 rounds). The tower garrison now stands as two watches and counts as a coalition, so the floor never arms in the campaign: the flagship fights and dies like any hull. The measured walls with a mortal flagship: day one ~L29, maxed research ~L45, full tech tree with the Destroyer ~L95 — and the same garrison fights the exact same battle as before for every other fleet.
• The pre-engage forecast tells the truth again (#1052). It now simulates the same defender the server fights: a Destroyer roster past its real wall no longer reads "Clear win · 3/3 sims" on levels the server refuses.

Changed

• Campaign replays name the tower's two watches for what they are — W1 / W2 chips on the rail, "First watch" and "Second watch" on the cards — instead of borrowing "Home" and "Relief 1" and counting the tower as "2 fleets" in the finale, a homeowner and a relief fleet the tower never had.

Also

• Depths already cleared through the hole stay cleared: the fix changes what can be climbed from now on, not history. An empire far past its mortal wall will find its next level holds until the fleet genuinely outgrows it.
• On the installed iOS app the pre-battle forecast keeps the old math until the next TestFlight build; the battle outcome the server returns is the authoritative one everywhere.

1.0.0-beta.189 — 2026-08-21

New

Anomaly signals: the list the map could never be (#524)

Finding an anomaly meant reading the political map by eye. Eight sectors, nineteen systems each, and at low Astrometrics most of those screens hold nothing — the sky shows you 2 + Astrometrics/3 signals per sector, so the sweep is mostly empty space. The players who checked least often paid the most for it.

Galaxy now carries a list. Every system with a signal you can see, one badge per signal, sorted by tier or by distance from home. Tap a row and you are in that system, on the slot, ready to plan the expedition. On the desktop it sits above the map already folded, and its header alone tells you how many signals are out there; on a phone it is a chip that opens a sheet, next to Wanted and Targets.

Nothing about what you can see has changed. Astrometrics still decides which tiers reach you and how many, exactly as before — the list only shows what the map was already drawing as a dot, and finding the exact planet still means opening the system, same as always. What is gone is the sweep.

Two things the list does not do. It does not name the outcome type, even with the Deep Observatory built: that reading lives on the planet's own card, one tap further in, and it has stayed there. And it does not appear at all when nothing reaches you — there would be nothing to open, and the note on the map already explains why the sky is quiet.

1.0.0-beta.188 — 2026-08-21

The Empire Overview now opens by telling you which colony is waiting on you, instead of leaving you to find it.

New

• Needs attention (#522). The top of the tab now names the colonies that want a decision: a build queue standing idle, a warehouse at its ceiling throwing away everything it produces, a shipyard with nothing on the slipway, a siege one attack from razing the colony. One chip per colony, worst first, and tapping it takes you there. Every one of these facts was already on the tab — it was just printed at the same weight as everything else, so a six-colony empire never said which colony to look at. When nothing is wrong, the section is not there.
• The colonies line up. On a wide screen they are a table: stock, energy, build queue, shipyard and defences run down columns you can compare at a glance. On a phone and a tablet they stay cards, because a table that does not fit is worse than no table.
• The empire tiles say when you are overflowing. Each resource now shows the total warehouse ceiling it is filling, and — when it applies — how many colonies are sitting at their own cap. A total at 68% can still hide three colonies discarding every unit they make.

Fixed

• Energy could read "1397 / 1397" in red. The tab printed the two figures truncated while deciding the colour on the rounded ones, so a colony perfectly in balance could show two identical numbers under an alarm. Same bug as #457, on a screen that did not exist when #457 was fixed.
• The colony table was cut off on tablets. On an iPad in portrait the Defences column and the whole Base / Colony / Shipyard button group ran off the right edge with no way to scroll to them. The tab now decides between the table and the cards by whether the table actually fits.
• A running ship order could vanish. If a raid levelled a shipyard while an order was building, the Overview stopped showing the order at all.
• "No colonies yet" when the list simply failed to load. A failed request now says so, and the totals show a dash instead of asserting you own nothing.
• Fleets in flight said "0 active" while still loading, and kept saying it if the request failed.
• The tab could go blank on older iPhones. On iOS 15.0–15.3, a colony with two or more resources at the cap crashed the whole tab.

Changed

• On a phone: your fleet is listed by hull type again, the per-colony Base / Colony / Shipyard buttons are big enough to hit, and each colony's stock shows the real figure and its ceiling rather than a bare percentage — 62% of a small warehouse and 62% of a large one are not the same amount.
• Dark Matter uses the same violet as everywhere else in the game, and colony coordinates are written [3:17:4] like they are on every other screen.
• Damaged emplacements no longer raise an alert. They repair themselves on a timer, so the count and its countdown stay on the colony's row where they belong, instead of asking you to do something about them.

1.0.0-beta.187 — 2026-08-21

Fixed

• 48 building sprites got their missing pieces back. The Particle Accelerator fix in the previous release turned out to be one case of a roster-wide problem: the background removal that produced the colony sprites had eaten the dark parts of most buildings — ground platforms, floors, side machinery, whole walls. The Recycler had lost its entire scrap floor, the Veil and the Silo their bases, the Hypergate its foundation, the Factory its dark floor plate, and most stage II/III looks were bitten the same way. Every one of the 78 sprites was compared with its source art one by one: 48 were re-cut with a mask that cannot punch holes in a building, and the rest were verified healthy and left untouched. Buildings now stand on their platforms instead of floating over fragments, and their upgrade looks match.
• Buildings sit truer on their tiles. The re-measured art re-anchors each building on its cell, so several that hovered above their pad (the Hypergate most visibly, whose whole foundation is back) now rest on it, and the level badge follows the actual roof.

Also

• The first time you open a colony after this update, the game re-downloads the refreshed artwork (a few MB) — one slower load, then it's cached again.
• Where a building's upgrade stages read the ground differently (an art property of a few sprites), the building can shift slightly at the moment the stage art changes; each stage now stands correctly on its tile, which we chose over hiding the shift by mis-seating both stages.

1.0.0-beta.186 — 2026-08-21

New

Contested Space — the Event Sector is now guild territory. Sector 9 holds six worlds you can take and hold with your guild: Ashfall Reach, The Cindergate, Hollow Meridian, Sable Anchorage, The Long Silence and Ember Crown.

• Each world is held by a neutral garrison. Beat it and the world flies your guild's colours.
• Holding pays. Every hour a world stays yours, your guild scores 60 points; taking one is worth 250.
• Open at all hours. No window, no timetable — rivals can come for your worlds at any time, and you can come for theirs. The only pause is the hour after a capture, so whoever just won has time to move their fleets in before the counterattack.
• You can garrison what you hold. Station fleets on your worlds the way you would defend an ally. When a world changes hands, the fleets of the guild that lost it stop fighting for it — they will not defend the flag that just replaced their own.
• The season is now won by a GUILD. At the Eschaton the top three guilds pay out to every member who fought in the sector, and how much you receive depends on your own place on the season scoreboard. Third place in a winning guild is worth more than first place in a guild that did not place.

The garrisons are tuned to the strength of the field rather than to its single strongest empire — deliberately, so that these worlds are something a guild can organise around instead of something only one player could ever reach.

Two notes for the curious: worlds where somebody had already planted a colony are out of rotation for the season and show as "occupied", and a guild without a foothold in the sector cannot score there — this is a guild contest, and the panel in Events says so.

Also

The Event Sector had one participant in each of its last two seasons. This is the answer to that.

1.0.0-beta.185 — 2026-08-21

Also

• Nothing changes in the game this time. This release is a fix to the tooling that puts the game on the server (#1056).
• The deploy refuses to run when it would delete a file from the server that cannot be rebuilt — the server is not a copy of the source repository, so anything removed there is gone. It could not tell that apart from a file the new code deletes on purpose, so any update that removed or renamed a source file stopped halfway, with the change already accepted and the servers still running the previous version. That is what happened to the Season update, which had to wait for this fix before it could go out.
• The check now knows which deletions the code actually asked for, and still refuses every other one. When it cannot be sure what the servers are running, it refuses them all, exactly as before.

1.0.0-beta.184 — 2026-08-21

New

• The Season has its own tab (#1050). Events now opens on Log | Season. The season used to be squeezed into a narrow column beside the log on desktop, and into a gold strip you had to tap open on the phone; it now has the whole page, on both.
• The Season tab tells you what winning is worth. A podium table, read before you play, showing exactly what the Eschaton Event pays:

  | | prize | |---|---| | 1st | a Destroyer · 3 crates of 10M (one per resource) · 40 Dark Matter | | 2nd | 3 crates of 5M · 20 Dark Matter | | 3rd | 3 crates of 1M |

• Dark Matter on the podium (#1050). First and second place are now paid Dark Matter as well as crates. It lands straight in your balance at the close.

Changed

• The season no longer pays a share of your own stores. Until now the podium won 20% / 10% / 5% of whatever sat in its own mother colony. That prize was largest for whoever was already largest, and it was capped by your own Warehouse — so the fuller your stores, the bigger the prize and the less room to receive it. At full stores, first place was paid nothing at all. It is gone: the podium is paid in crates, which are the same size for a two-week-old empire and for the leader.
• "Make room before the close" is no longer true, so the game stopped saying it. A crate keeps what does not fit: opening it credits what your Warehouse can take and the rest stays sealed for later. Nothing is destroyed by opening one, so there is nothing to empty out in advance.

Also

• The end-of-season entry in your diary now names what you won — how many crates, of what size, and how much Dark Matter — instead of leaving you to count your resources to find out.
• Older seasons in your diary still read exactly as they did: the prize they describe was a different mechanic, and nothing rewrote their history.

1.0.0-beta.183 — 2026-08-20

Changed

• Battles now run to a decision — the 6-round clock is gone (#1006). A fight continues until one side is destroyed, breaks off, or the battle provably stalls. What you'll see:
  • A fleet that has lost two thirds of its value breaks off and brings the survivors home — the report card says "Attacker broke off" instead of leaving you to wonder why ships returned from a defeat.
  • The Shield Bubble regenerates during battle (a sixth of its strength every round). A siege is now decided by whether your fleet out-paces the regeneration — the old "hold for six rounds" turtle maths carries over as physics, and an attack that can't keep up is repelled on the spot.
  • A lone parked Destroyer no longer grinds two thirds off every visiting fleet: battles against a garrison reduced to the unkillable flagship end as the draw they always were. Bringing your own Destroyer with an escort no longer forfeits battles the escort was winning.
  • Unarmed ships don't stand under fire for thousands of rounds: a weaponless attack strikes its colors after one exchange, and helpless deployed ships surrender once nothing shields them.
• The campaign wall moves. With fights running to their real end, the tower's measured walls are now ~L29 on day one, ~L45 at maxed research, ~L95 with the full tree and the Destroyer — the L76 stalemate that killed 97% of a garrison and paid nothing now resolves as the win it was (#1006).
• Long battles keep honest reports: the opening and the ending stay in full detail, the middle is counted ("… 3,186 rounds pass …" in the replay), and the damage tables say exactly which rounds they cover.
• Relic garrisons re-tuned for the new engine (gate 0.7 → 0.85 of the top military): the strongest player still cannot open one solo — the counterweight that makes relics a coalition fight survives the clock's removal.

Fixed

• The replay no longer shows the Shield Bubble shattering in battles the report says it held, no longer re-shatters it twenty times on long sieges, and the round counter matches the Details panel.
• Statics-only turtles (batteries and nothing else) are no longer invincible-by-whistle: a sustained siege cracks them.

Also

• Joint-raid extra-round machinery deleted (it compensated for a clock that no longer exists); design docs updated. The reward ramp's interaction with the deeper day-one wall is flagged for tuning (#892 owner's call).

1.0.0-beta.182 — 2026-08-20

New

• The battle replay is directed now (#1002). Before: one fixed wide shot for the whole battle, with most of a phone's screen empty sky. Now the camera works: an opening sweep along the attacking line, close shots that sit behind whoever is firing — so the space between the fleets runs into the screen and the ships finally fill the frame — a push-in when a squadron dies, and a pull-back for the verdict. Fast rounds share a two-side "exchange" framing so no salvo goes unseen; slow rounds get their own per-side shots.
• Your hand always wins. Drag, scroll or pinch at any moment and the direction switches off for the rest of the replay — the camera stays exactly where you put it. Replay from the start and the direction returns. If your system asks for reduced motion, the replay keeps the calm fixed camera throughout.

Also

• Battles fought entirely by modelled ships can tilt further toward top-down — the old tilt limit existed for the four classes still drawn as flat illustrations, and it now applies only when one of them is actually on the field.
• On phones held upright the close shots lean down slightly so the line of fire runs along the tall side of the screen.

1.0.0-beta.181 — 2026-08-20

Fixed

• A fleet coming home lands once, with the ships it actually has (#621). Two ways a return could go wrong, both narrow enough that nobody caught them in the act, and both now closed.
• The same fleet could be credited twice. When a delayed job fired twice, or a retry raced the original, both runs poured the same ships into the origin colony's hangar — a free fleet, with nothing in the reports to say where it came from.
• A garrison could come home from a battle it had already lost. If a battle landed in the instant between the return being read and the ships being deposited, the hangar was filled with the fleet as it stood before the shooting: ships you had just watched die in a battle report reappeared at home. The mirror case was as bad — a garrison wiped out entirely still deposited its full pre-battle strength.

Now the landing is checked against the fleet as it stands at that moment. If anything changed underneath — a battle, or another run of the same return — the deposit is undone and recomputed from a fresh read, so what reaches the hangar is the survivors, and it reaches it once.

Also

• The same protection now covers a fleet landing in a refugee port during an exodus: two fleets arriving in the same limbo no longer overwrite each other.

Known

• A garrison whose watch ends at the exact moment a battle is being resolved can still land early, or be banked while a battle is inbound. That is a separate defect in how a garrison's watch ends, it predates this change, and it is tracked in #1046 — this release does not fix it and does not make it worse.

1.0.0-beta.180 — 2026-08-20

Fixed

• A guild officer no longer forgets what happened while they slept (#496). Before, when your guild crossed a Directive tier at a moment the officer on duty happened to be offline, the news simply evaporated: nothing was pinned when they came back, and a restart of the service lost it too. Now the unlock is recorded the moment it is settled, and the officer picks it up for the next day — and says it like someone catching up on what they missed, not like someone watching it happen.
• The same for a war. A declaration that appeared overnight used to be pinned the next morning as if it had just been signed. It is now announced with its age, and the pinned order no longer claims to be about today when it is about last night.

Changed

• A guild officer stays quiet at night, including when letting someone in. An application is still decided within minutes at any hour — nobody waits until morning to be admitted. What waits is the message in guild chat. A single line at four in the morning on a channel that has been silent since evening was the one thing that gave an officer away, and the welcome for a new member arrives instead at a believable hour, written properly rather than stamped from a template.

Also

• New members are welcomed in the officer's own words or not at all: the canned "name got a seat with us." line is gone. It was the same sentence in every guild.

1.0.0-beta.179 — 2026-08-20

New

Each planet now lights its own colony (#472)

Until now every colony looked like the same colony. The ground under your base changed from world to world — a lava field, an ice sheet, a gas deck — but the buildings standing on it were drawn under the same neutral light everywhere, as if they had all been cut from one photograph.

Now the planet lights what you build on it. Walk into a colony and the whole base sits in that world's light: every building, at every stage, the placement preview included.

• Lava worlds run hot — the strongest light of the four, coming up off the ground your base stands on.
• Ice fields run cold.
• Gas decks run deep amber, under the cloud layer overhead.
• Rocky worlds barely move: that is the ground these buildings were drawn against in the first place.

Two things stay deliberately untouched by the light, because they are signals and not scenery: a level-0 blueprint keeps the same teal on every world, and the "you cannot build here" red on the placement preview reads identically everywhere.

And it stays a light, not a repaint — the art is the art, it is just no longer lit the same way on four different worlds. The four lights were tuned by measuring the actual sprites, so the difference is one you can see, not one that only exists in the code.

Also

The full version of the suggestion this answers — buildings actually redrawn per planet, molten metal on a lava world — would be more than two hundred new pieces of art, and is not in this release. Neither are the targeted re-skins of the buildings each world's economy already favours. The building thumbnails in the build menus stay neutral on purpose: those are pictures of a building type, and they should look the same wherever you are.

1.0.0-beta.178 — 2026-08-20

Fixed

• The game would not open at all. For about 40 minutes this evening, opening novarchy.com showed only a page of text about the game — no name box, no Play button, no way in. Anyone already signed in got the same thing. The game itself was never affected: colonies kept producing, fleets kept flying, and nothing was lost. It was only the screen that failed to start. It opens normally again. (#1037)

Also

• Nothing else changes in the game this time. The rest of the work went into making sure this particular failure cannot come back silently: the check that guards it now catches the ways it could be reintroduced, and two notes in the code that pointed the next person straight back at the trap have been corrected.

1.0.0-beta.177 — 2026-08-20

New

• Empire Overview summary page (#522). A new Overview tab in the main Empire navigation group gives multi-colony empires a unified command deck. Before, keeping track of production, resource stockpiles, construction progress, defense emplacements, and stationed fleets required flipping through each colony screen one by one. The Overview screen shows:
  • Aggregated empire totals: Total stockpiles and net hourly production across all resources (Ferrolite, Xenogas, Data, Dark Matter), plus total stationed fleet counts and colony cap.
  • Per-colony cards: Name, coordinates, traits, resources relative to storage capacity (with overflow alerts), hourly income, energy balance, active building orders and free building slots, active shipyard orders and slip capacity, stationed fleet breakdown with ship thumbnails, planetary shields with remaining time, sensors, and defense battery emplacements with regeneration countdowns.
  • Fast jumps: Direct shortcuts on each card to navigate straight to that colony's Base, Colony, or Shipyard screen.
  • Active missions: An integrated view of in-flight fleets, destination coordinates, ETAs, and cargo hold contents.

Also

• Cargo hold contents on active missions were already added in #307; this release completes the remaining scope of #522 by adding the unified multi-colony summary dashboard.

1.0.0-beta.176 — 2026-08-20

Fixed

• Flaky clock-dependent agent test suites (#503): Runner test suites (test_runner_lega.py, test_runner_avamposti.py, test_anchorage.py, test_eventi_stantii.py) now freeze test datetimes deterministically instead of evaluating real wall clock at module import, eliminating intermittent test failures across time zones and night hours.

1.0.0-beta.175 — 2026-08-20

Changed

• Agents prioritize closing daily contracts near completion (#500): On bulletin board visit days, agents with daily contracts one step away from completion (such as scouting) now prioritize launching the required mission rather than ignoring the contract.

1.0.0-beta.174 — 2026-08-20

Changed

• The battle replay finally shows how big a fleet is (#1001). Before: every squadron was one ship on screen whatever it counted — 2.147 interceptors and a single corvette were pixel-identical, and a defence of thousands could look smaller than the raid hitting it. Now a group draws a wedge of hulls that grows with its ship count: that same defence reads as 36 hulls against the raid's 23, and the mass sits on the side that has it. Big battles stay smooth — the whole scene is capped at 120 drawn hulls and the explosion lights still come from the same fixed pool.
• Losses thin the wedge while you watch. Each fallen hull explodes and drifts off as a wreck exactly like group deaths always did, the last hull holds on until the group is wiped — where the big explosion still lands — and even a 12.000-ship doomstack starts visibly losing hulls from its first casualties. The survivor bar keeps the exact numbers under the lead ship.
• Wedges know their place. A formation never spreads into the neighbouring lane, a defender's tail stays inside the Bubble, and the fly-over at battle start clears the whole depth of both fleets, tails included.

Also

• Ship illustrations in the replay compose more cleanly inside a formation: the grey halo that could bite crawling holes into hulls behind it is gone.

1.0.0-beta.173 — 2026-08-20

Fixed

• Shipyard reload recovery (#487) — Before, a shipyard left open across a deployment could show a raw browser error and a Retry button that could never recover. Now, the screen explains that it could not load and offers a Reload button that fetches the current files.

1.0.0-beta.172 — 2026-08-20

New

• Resource crates (#1012). A new kind of prize: a sealed amount of one resource — Ferrolite, Xenogas or Data — in seven sizes, from 100K to 100M. Crates live in Inventory → Stores and open into a colony of your choice. Opening stops at the warehouse cap, and what does not fit stays sealed in the crate: nothing is ever destroyed by opening — make room and open it again. The Eschaton podium now pays in crates on top of the resource share: 10M / 5M / 1M of every resource for ranks 1, 2 and 3. Grants and openings show up in your Activity feed, with the destination colony named.
• A Destroyer every season (#1012). The one-per-empire limit is gone: every Eschaton winner receives a Destroyer, even if they already own one. Two hulls in one hangar fight as an ordinary stack — the Spectral Bulwark still protects only a lone hull — so how you split them across colonies is now a real decision.

Changed

• Event-sector colonies cannot receive a crate: the Eschaton sweep would eat half the prize at season close, so the picker hides them and the server refuses them.

Also

• The crate art is new: seven tiers, one look each, from battered steel to gold-veined obsidian.

1.0.0-beta.171 — 2026-08-20

Fixed

• The Particle Accelerator got its dark half back (#1010). The building's sprite had lost most of its dark machinery to an over-aggressive background cut: the side piping, the roof blocks and the base modules were simply missing, leaving the accelerator looking like a fragment floating around its ring. The sprite has been re-cut from the original art with a mask that cannot punch holes inside the building, so the whole structure is visible again in your colony. Upgraded accelerators (the stage II and III looks) were checked and were never affected.

1.0.0-beta.170 — 2026-08-20

Fixed

Synthetic empires respect the human-war brake (#982)

Before, a synthetic leader could declare war on a guild containing players before tension reached the protected threshold. Now the declaration requires 55 tension whenever at least one person is in the target guild, while wars between synthetic-only guilds still unlock at the normal 40 tension.

1.0.0-beta.169 — 2026-08-20

Changed

• The battle replay dropped its text clutter (#1000). Before: every squadron wore a standing name plate — often wider than the ship it named, and in a big raid twelve of them merged into one unreadable diagonal ribbon — plus a round banner across the top and a shield panel over the bottom. Now the scene shows the battle: ships and their survivor bars, and nothing else standing. Tap a ship to see its plate for a moment (a hint teaches this the first time); in raids, a small squadron rail summons a whole squadron's plates; a group that dies announces itself. "Orbital cannon" and "No resistance" still get their moment on screen.
• Rounds moved into the progress bar. Each round start is a tick, and the counter next to it shows which round you are watching — the big ROUND banner no longer covers the fleet.
• On phones the replay is edge to edge. The header and footer rows are gone: the close button and the controls float over the battle, and the Bubble/Emplacements summary lines — duplicates of the battle report — left the modal. The planetary-shield readout became a slim strip in the corner.

Also

• The deployment's lanes are now sized by real hull footprints instead of by the name plates that no longer exist, so small battles draw tighter, denser fronts.
• The end-of-battle card can be scrolled with the mouse wheel when its details run long, and taps or pinches over it drive the camera like anywhere else.

1.0.0-beta.168 — 2026-08-20

Changed

• The galactic channel is quiet now, and it answers when you speak (#999). In a typical day the channel carried 271 messages and 255 of them were written by synthetic empires: battle reports on fights you were not in, boasts about relic fragments, league trash talk, small talk started because the room had gone quiet. Scrolling it meant scrolling past other people's machines. From this release the empires say none of that. They speak in the galactic channel only when a person has spoken to them.
• Name a commander and that commander answers (#999). Type their name, press "reply" on one of their lines, or name their guild, and you get an answer from the one you addressed rather than from whoever happened to be awake. Addressing someone directly no longer waits for their quiet hours. Name a guild instead of a person and exactly one of its members answers for it, instead of nine at once.
• The conversation keeps going without you repeating the name (#999). Once an empire has answered you, your next messages reach it for about a quarter of an hour. After that, name it again.
• Empires no longer answer each other in public (#999). The long exchanges between two machines that filled the channel between human messages are gone: a name typed by another machine buys no reply.
• They write like people again (#999). Replies come back as one line, the length they have always been, instead of the three-paragraph answers the new model wanted to give.

Also

• Two things still get said without being asked, because silence there would be stranger than speech: an empire whose colony has just been razed reacts to it, and a declaration of war, an ultimatum or a surrender is announced by the leader who made it. Both are rare, and both are events rather than commentary — an ultimatum nobody announces is one you would only discover from a number changing in a panel.
• The guild channel and private whispers are unchanged: guildmates still talk among themselves, a whisper still gets a reply, and it now comes back in the same voice as the public one. The galactic bulletin still reports what happens in the galaxy.
• Nothing about what empires do has changed. Wars are still declared, raids still launch, and a threat typed at an empire is still remembered and still answered with ships. What stopped is the commentary, not the consequences.

1.0.0-beta.167 — 2026-08-20

Changed

• The colony's defences are built things now, not placeholder shapes (#997). In a battle replay, a defended colony fields its Emplacements and its Fortress in the front rank. Until now those two stood there as rough grey volumes — a generic fighter shape for the Emplacements, a generic capital shape for the Fortress — while every ship around them was a real hull. They are modelled: the Emplacements as squat turret satellites with their solar wings out, the Fortress as the four-cornered orbital citadel it is described as, and it is the widest thing standing on that line. Watch any replay of an attack on a colony that has batteries or a Fortress.

Also

• Both models download in about the same weight as a single warship, so opening a replay does not get slower. If one fails to arrive on a poor connection, that defence falls back to the old grey volume and the battle plays exactly as before — nothing is ever missing from the field.

1.0.0-beta.166 — 2026-08-20

Changed

• The Shipyard now tells you what would have made the wait shorter (#986). Ordering a large batch of ships books the yard for as long as the hulls cost, and the sheet used to say only that the order could not be cancelled. It never said what you could do about it. It does now: alongside the warning, it names your Shipyard and its level as the thing that shortens every hull — and it says the upgrade reaches the ships already building, because raising the yard while an order is in the slip speeds that order up. Waiting for the upgrade before ordering is the slower of the two, and the sheet no longer implies otherwise.

• The warning starts a day earlier. It used to appear only past 24 hours, so any order between 8 and 24 hours committed a slip in complete silence — even though every one of them outlives the session that placed it. It now speaks from 8 hours up.

• It stays quiet when it has nothing useful to add. No Shipyard on that colony, every slip already busy, a yard at its cap, or a next level that would not take an hour off the order you are looking at: in all of those the sheet says nothing rather than repeating advice you cannot use. It does still appear when you simply cannot afford the order yet — that is exactly the moment you are deciding whether the wait is worth it.

Also

If a 31-hour order surprised you, it is worth knowing the wait is not fixed: the same 169 Monitor Frigates take 31.9 hours at a bare Shipyard 2 and 3.8 hours at Shipyard 20 with the shipyard research, Cybernetics and a Nanoforge behind it. Splitting a batch across free slips helps too — they build in parallel, and the sheet has always said so when you have more than one free.

Pointing you at a different colony whose yard is idle is not in this release: getting that right means weighing the flight home, that colony's own resources and its own shock, and a half-right version would send you somewhere slower. It is tracked separately.

1.0.0-beta.165 — 2026-08-20

Changed

• Commanders no longer all read their messages at the same speed (#890). Every synthetic empire used to answer a private message on exactly the same schedule as every other one — same typical wait, same tail, all seventy-five. Now some get back to you inside a minute or two and some take the better part of twenty minutes, and which is which stays the same commander to commander. If you have been writing to several of them, you will notice the difference between the quick correspondents and the slow ones.
• Two machines arguing no longer stop on the same beat every time (#890). The point at which a back-and-forth between empires runs out of steam now moves around instead of landing on the identical number of lines in every single exchange.

Also

• Neither change makes the empires chattier. Private replies come slightly less often on average than before, and public chat is unaffected.
• The part of #890 about how quickly an empire answers when you name it in the galactic channel is not in this release. It was built, measured, and left out: someone always answers a message that names an empire, so changing the odds for the named one did not alter anything you could actually observe. That half is being reworked separately.

1.0.0-beta.164 — 2026-08-20

Fixed

A note to ourselves that was not true (#993)

Nothing here changes the game. It corrects the repo's own paperwork, and it is in this log because the previous release (beta.163) fixed a check that had stopped recognising this very page — and then explained itself, four times over, with a sentence that was wrong.

That sentence said the safeguard runs "in CI", meaning automatically on every proposed change. This project's automated-checks workflow has been switched off for a long time; every green is produced by hand before a release goes out. So the sentence described a safety net that is not there. The safeguard is real and does run — as part of the checks that must pass before anything ships — but that is a different claim, and it is the one now written down, along with the one place in the process that skips those checks.

The reason to spend a release on a comment: this is the same failure as beta.163 itself, one level up. Two halves of something that must agree, drifting apart, with prose asserting they still do. A note that says a thing is protected when it is not is worse than no note, because the next person believes it.

1.0.0-beta.163 — 2026-08-20

Fixed

Releases can reach you again: the ship's own check had stopped recognising this page (#987)

Nothing a player could see was ever broken here, and the changelog has been serving correctly all along. What broke is the last step of shipping anything else.

Before every release goes live, the deploy fetches this page and checks it is really the changelog — not the game's login screen, which the server would also answer with a cheerful "200 OK" if the page were missing from the build. The only reliable proof is the page's title. On 20 August that title was reworded from «Novarchy — Changelog» to «Changelog — Novarchy», and only one of the two sides that have to agree was updated. From that moment every deploy in the repo stopped at the smoke step, reporting a page it did not recognise, on a production that was entirely healthy.

The check now recognises the page again, so releases flow. And the two halves are held together by something that runs on every change, in both directions: rewording the title fails loudly, and so does quietly widening the check to accept more than it should — the direction that would otherwise have looked fine while the check silently stopped doing its job.

If you are reading this entry at all, it worked: the release before this one had been finished and live for hours with no way to announce it.

1.0.0-beta.162 — 2026-08-20

Fixed

Guild bonuses no longer show a number they are about to change (#950)

In the first moments after opening the game — before your guild finished loading — anything your guild improves was shown at its base value, and then quietly corrected once the guild arrived.

If your guild holds a Science keystone, that was not a bonus arriving late. It was the wrong number, printed and then contradicted:

• opening a secondary colony's Laboratory in that window, the sheet said "secondary Laboratories count half … lvl 6 is what raises the network". A moment later that whole line vanished — with the keystone the level does raise the network — and a green gain appeared in its place. Two opposite answers on the same panel, the one with the Upgrade button on it;
• the planet card and the mission sheet promised "Up to 5 explorations per day" to commanders the server actually grants six.

Now the game waits instead of guessing. During that moment the Laboratory's research row reads "— (checking…)" with a line saying your research network is still loading, and the recon-cap sentence simply is not there. Each fills in once, correct the first time. Nothing else on the card changes: the price, the build time, the power cost and the Upgrade button are where they always were.

If you have no guild, nothing changes for you at all — "no guild" is an answer, and your numbers were always right. If your other colonies have no Laboratory, nothing changes either: your guild's keystone cannot move a network you do not have, so there is nothing to wait for.

And if the game cannot reach your guild at all, it stops waiting after a few tries and shows the ordinary numbers rather than leaving those lines blank for the whole session.

1.0.0-beta.161 — 2026-08-20

Fixed

• Bot guilds stop holding war councils about their own allies (#885). If you have ever watched a bot alliance's chat, you may have seen the same scene on a loop: every eight hours the guild opened a war council about a neighbour — "the standoff with them: strike first, demand reparations, or wait them out?" — and then nothing ever happened. It could not happen. The guild had a standing peace pact with that neighbour, and the game refuses both war and reparations while a pact holds. The council was picking its crisis by tension alone, without ever asking whether it was allowed to act on it.

  It also went quiet everywhere else while that was going on. A council weighs one neighbour per sitting, so as long as the pact held the floor, the guild made no move toward anyone — and never got round to its own business either: which branch to research, the Council directives, a Fund running dry.

  Councils now only take up neighbours the guild can actually do something about. Those guilds talk about their own house again, and their dealings with real rivals unfreeze.

• A bot leader whose truce ran out is no longer disarmed for good (#885). Paying reparations buys 48 hours of peace. The bots were reading that as forever: once a truce had been bought, the timestamp stayed on the record and the leader never acted toward that neighbour again, no matter how much later it was or how far relations had soured since. They now read the clock, so the peace lasts exactly as long as it was paid for — and not one hour more.

Changed

• A strained pact is now described as strained (#885). What a bot knows about its own alliances read the same whether things were calm or the two of you had been trading threats all week. Ask one about a neighbour it had a pact with and it would tell you all was well. It now knows the difference, and will say so — while being clear the pact is still holding. Under a pact nobody can attack anybody, so it stays out of the language of open war.

Also

• Bot guilds bound by a pact still keep the pact. Nothing here lets them tear one up: that remains a decision only a leader makes, and the bots have no way to make it.

1.0.0-beta.160 — 2026-08-20

New

• The changelog is reachable from inside the game (#883). novarchy.com/changelog went up with the last release, but the only link to it was on the login screen — which you stop seeing the moment you are logged in, which is to say: everyone who actually reads patch notes never saw it. There is now a Changelog control next to Discord in the header on a computer, and a row in the Command sheet on a phone, above log out.

Also

• On a narrow window the row of controls in the header now wraps to a second line instead of pushing log out off the edge of the screen.

1.0.0-beta.159 — 2026-08-20

New

First Steps carries on to the market (#961). The guided chain used to stop at Command Center 3 and go quiet, and the measurement said that is where a lot of people stopped too: of the 114 players who collected every step, 43 were still at Command Center 3 days later — 40 of them more than a day old — holding an average of 4,179 resources against the 1,227 the next level costs. They were not saving up. Nothing in the game had ever told them the Command Center was the thing to buy.

The chain now runs three steps further: Command Center 4, Command Center 5, and the Trade Portal. That puts the market — the one early thing to do that involves other people rather than a timer — inside the tutorial instead of hours past it. It is about nine minutes of building, and it costs less than the chain has already paid you, so it is direction for resources you are already holding rather than a new grind.

Changed

The advisor stops hiding the Command Center. The "Suggested next" panel ranks by payback, and the hub never wins on payback — at Command Center 4 and 5 it was scoring zero and being dropped from the list entirely, on exactly the two levels that open the market and the second colony. It now says what the next level unlocks, and what it costs, as its own line above the ranking. The ranking itself is unchanged: the panel still tells you the fastest-paying upgrade first.

It also prices the hub properly for the first time. A second Site — a duplicate extractor or refinery, the strongest economy move in the early game — was worth nothing at all to the level that unlocks it. Now it counts, so the Command Center shows up on its own merits when it deserves to.

The Command Center's card shows two levels ahead, not one. At Command Center 3 that means you can see that level 5 opens the market before you have to go looking for it.

Fixed

Players who had already finished the old chain keep everything they had. Growing the tutorial could have closed the Contracts board, put out the daily-contracts dot and taken away the "Find your guild" card for everyone who finished before today. It does not.

Nobody loses a reward they earned. If you have collected steps waiting — including a step you cannot collect right now because your fleet is away — the chain stays open until you take them.

The onboarding funnel stops overcounting. An internal measurement was reading "this was already true before we started watching" as "the player got here". Only the admin dashboard shows it, but it is the number this change will be judged by.

1.0.0-beta.158 — 2026-08-20

The battle replay and the battle report are one click apart, on the same row of Activity. Until now they told different stories about the same fight.

Fixed

The Fortress was invisible in the replay (#948). A colony defended by an Orbital Fortress showed shots flying into a bare point on the defence line, and return fire coming back out of it — no hull, no sparks, no counter. The finale listed your Emplacements and never the Fortress, while the report a click away printed Fortress 1 → 0. The Fortress now stands on the line as the heaviest thing on it, fires its own artillery, loses pieces as it is shot at, and gets its line in the finale.

The Orbital Cannon's kills landed on the frame the verdict did (#948). The Cannon fires before the first round, so the ships it destroyed appeared in no exchange: their card sat full for the whole battle, then dropped in one jump — explosion and grey-out arriving together with VICTORY. The Cannon now has its own beat before round 1, where it actually fires, and the counters go down there.

"No resistance" over a fight that cost you ships (#948). A colony whose only defence was the Orbital Cannon produced no rounds at all, so the replay announced no resistance — and then your fleet lost ships three seconds later, on the same screen. A colony that fires its Cannon is not undefended, and the replay no longer says it is.

The Cannon's tally was wrong in raids. "Orbital cannon: … · N ships lost" counted only the last squadron's dead when two squadrons brought the same ship type. On the case measured for this fix it read 112 where 179 had died. The report now counts all of them, and records which squadron lost what — so a raid's replay scales the right squadron's card.

Also

Reports written before this update keep the old, undercounted Cannon tally, and the replay deliberately refuses to guess which squadron lost those ships: on those older raids the losses still land on the final frame. New battles are exact.

The Fortress borrows the Emplacements' art for now — it is the right size on the line, but it is not yet its own picture.

1.0.0-beta.157 — 2026-08-20

Fixed

• A building's level number could be sitting on the building in front of it (#959). In a dense colony you could hover a Ferrolite Extractor, read «Site 3 · lvl 11» in the tooltip, and see a 9 printed on that same sprite. Both numbers were right — they belonged to two different buildings. The colony is drawn in perspective, so the building one step nearer the camera is painted over the one behind it, and the number behind it was being painted on top regardless.

  Two things changed. Every level number now sits where it was always meant to sit on its own building — until now they were drawn a little lower than intended, by an amount that grew with the height of the building, which is why tall stage-II and stage-III neighbours were the ones that caught them. And where a nearer building genuinely covers the spot, the number now steps aside instead of lying about whose it is.

  A hidden number is never a lost one: it is still on the building's hover tooltip, in its panel, and on every row of the build menu. In a fully packed colony about one badge in five steps aside; before this, about one in four was on the wrong building.

Also

• The report was that the sprite measurements had gone stale. They had not — every one of the 78 building sprites still measures exactly what the game records for it, and two existing checks would have caught it otherwise. The cause was elsewhere, and the fix is in the geometry rather than in the art.

1.0.0-beta.156 — 2026-08-20

New

• The changelog is a real page now: novarchy.com/changelog (#856). Every release, oldest to newest, written the way these notes are — what you saw before, what you see now. Until today the only patch notes you could read were whatever fitted inside a Discord announcement, and every link out of one pointed at a repository you cannot open.
• Each release has its own address. novarchy.com/changelog#1-0-0-beta-117 opens straight at that release, so an answer in support can point at the exact change instead of describing it. A "Jump to a release" index at the top of the page lists every version; it starts closed, so the newest release is still the first thing you read.
• Reachable from the game. The login screen's footer carries a Changelog link next to Privacy and Support.

Changed

• The page is rebuilt with every release, so the notes you are reading now are on it from the moment this release is announced — not one release later.

Also

• Citations like (issue #836, PR #839) keep their text on the page but are no longer links: the repository is private, so those links were a dead end for everyone reading them.
• The four oldest releases (beta.1 to beta.4) are in Italian. They were written before the project moved to English, and they are left as they were written rather than rewritten after the fact.

1.0.0-beta.155 — 2026-08-20

Fixed

• Saved fleet templates no longer vanish when you switch device (#966). Until now a template you saved lived in that browser and nowhere else: one saved on the desktop simply did not exist on the phone, and clearing your browser deleted it for good, with no warning and no way to get it back. Templates now belong to your commander. Save one anywhere and it is there on every device you play from, and it survives a browser cleanup.

  Templates you already had are not lost: the first time you open the Missions tab after this update, the ones stored in that browser are added to your account.

Also

• Nothing else changes in the Missions panel: the same ten slots, the same picker, and recalling a template still trims it to the ships actually sitting in your hangar.

1.0.0-beta.154 — 2026-08-20

Fixed

The Colony tab stops giving you two opposite answers about the same Laboratory (#953)

Your research network counts every colony's Laboratory, but it stops at twice your best one. Once it is sitting at that ceiling, buying levels on a smaller colony does nothing at all until your best Laboratory grows.

The Laboratory sheet knew that and said so: raising your best Laboratory is far cheaper. The Suggested Upgrades list, a couple of centimetres higher on the same screen, did not — it offered that Laboratory anyway, two levels at a time, and charged for both. Its button queues the build in one tap without opening the sheet, so if you followed the suggestion you paid for the pair and never saw the line telling you not to.

Now the advisor stays quiet about a Laboratory it should not be selling, and tells you why underneath the list: your research network is capped at double your best Laboratory: levels here add nothing until that one grows — raising your best Laboratory is far cheaper.

Nothing else about the advice changes. On a secondary colony below the ceiling, where a level is dead only because of rounding and the next one brings it back, the pair is still suggested exactly as before — and on the colony that holds your best Laboratory, the one you are being told to raise, the upgrade is still recommended at a single level's price.

1.0.0-beta.153 — 2026-08-20

Fixed

Stellar anomalies reach the whole galaxy (#937)

Anomalies only ever appeared in the oldest, most crowded corner of every sector. The weekly draw covered a fixed 120 planet slots — the galaxy as it was before it grew twice — while a sector actually holds 342. Roughly two thirds of the map could not produce an anomaly at all, and never had.

That old corner is also where every new empire starts, and it is almost fully colonised: a survey found the first three sectors sitting at 119, 120 and 119 of those 120 slots taken. An anomaly that appears on an already-colonised slot stays hidden, so most of each week's supply was born invisible exactly where the newest players were looking — while the emptier half of the galaxy, the half with room in it, produced nothing.

That is the real reason the Observatory so often looked bare, underneath the "the easy ones get taken on Monday night" explanation.

From Monday 31 August, the draw covers the whole sector. The number of anomalies per sector per week has not changed — still twelve — but they are now spread over the entire map instead of piling onto slots that are already someone's colony, so far more of them survive to be found. Measured across a year of weeks, about two thirds of every week's anomalies now land on ground that could never hold one before.

Nothing changes before that Monday. The current week keeps exactly the board it already has, and an expedition already in flight arrives at the anomaly it was launched at, even if it crosses the boundary.

Also

The Observatory's warning about when the next survey lands (#921, shipped in beta.143) was the visible half of this problem. This is the other half — the supply itself.

1.0.0-beta.152 — 2026-08-20

Fixed

The Pirate Beacon in your home system no longer outgrows you (#920)

The Beacon is where the game tells every new commander to learn to fight — the novice panel says so outright: train on them and keep it all. But a Beacon's gun emplacements grew with the age of the system, not with yours, and a system only ever gets one Beacon, seeded the day the first commander arrived there.

So if you registered into a system that had been lived in for a few weeks, your very first fight was against a fortress. On the live server right now, 70 of the 99 Beacons had already grown their twelve emplacements, and 178 commanders less than two weeks old were sharing a system with one.

That fight could not be won. Five starting interceptors beat a freshly seeded Beacon every single time; against twelve emplacements, twenty interceptors lose every single time — emplacements are built to shred light hulls. It was not a hard tutorial, it was an impossible one, and nothing you did caused it.

From this release, the Beacon in your own home system starts its clock the day you arrived. Its wall is as old as your command, not as old as the system — and it still grows, on your fourteenth day and again on your forty-second, so the siege lesson it was built to teach still comes. Just when you are ready for it, instead of before you started.

Nothing changes for established commanders: if you were already here before your Beacon was seeded, or you have passed its milestones, you meet exactly the wall you met yesterday. Every other Beacon in the galaxy keeps its true age for everyone — the relief is for the Beacon that is teaching you, not a soft target you can go hunting for. Settling a second colony next to an old Beacon will not soften it either.

Joint raids are the exception, deliberately. A raid always meets the Beacon at full strength, because a raid is not one commander's fight and the pack cannot be sized to its youngest member. If a Beacon is still your tutorial, opening or joining a raid on it is now refused with a clear message rather than sending your fleet into something your own scout report promised you would win. Strike it alone.

Your scout reports and the Fleet Advisor already speak the same language as the battle: the fleet the Advisor recommends is sized against the wall you will actually meet, on the phone app and in the browser alike, with no update needed.

One older report may still overstate a Beacon you scouted before this release — it will say you cannot win a fight you now can. Send one probe and the report corrects itself.

1.0.0-beta.151 — 2026-08-20

Fixed

The Shipyard opens on the Shipyard again, not on three notices (#904)

On a phone, opening a tab meant scrolling past a war line, a four-line shield paragraph and a three-line siege paragraph before reaching what you came for. They were always on and never changed, so by the ninth visit of the day you had read all of them and were still scrolling past them.

The shield and siege notices are now a single row of chips at the top of the tab — Under siege 2/3, Empire shield · 6h. The numbers you actually check stay in view; tap a chip and its full text opens underneath, on the same tab, right where it was. Tap again to close it.

Nothing is hidden for good: no dismissal is remembered, so a notice that still applies is there again next time you look.

A full siege bar is still spelled out in full. When the next lost battle can raze the colony, the line that tells you an allied garrison counts toward the defence that breaks the siege is not something you should have to go looking for.

On desktop nothing changes.

Also

The war line was already a single line on a phone and is untouched — collapsing it would have cost the opponent's name, which it prints when you are in one war.

Two notices further down the same stack, the first-steps panel and the guild prompt, are not part of this change: they sit above these ones and are still the tallest things on that screen.

1.0.0-beta.150 — 2026-08-20

Fixed

The Galactic Log was a market ticker (#914). With the economy busy, a burst of trades filled the whole public log and everything else — battles, colonizations, Eschaton sweeps, the War Week podium — dropped out of it within minutes. Tapping «Battles» to go and find them made it worse rather than better: the filters can only sort what the log already holds, so once the trades had taken every slot the filter answered with an empty page and looked broken.

Now every kind of event has a place in the log, whatever the market is doing. A recent battle, a colonization or a podium is still there under a wall of trades, and the filters have something to show again. Pirate raids and battles between players are kept apart too, so a night of hunting the Beacon no longer pushes the wars off the page.

Reported by QQQQ in the suggestions forum, with a diagnosis that turned out to be exactly right.

The War Week podium could announce the wrong winner. The medals were handed out by position in the list rather than by placement, so whenever the podium did not start at first place — which happens when the week's top scorer has no colony left when the prizes are paid — the log crowned the runner-up. Quoting the result into chat made it worse: a quoted podium carries a single name, and that name was always shown wearing the gold. Placements are read properly now, in the log and in chat, and a quoted result that is not first place says so.

A war week could appear twice, the second time incomplete. Left-over rows from the previous week were being assembled into a podium of their own, with members missing and a medal on whoever happened to come first. The log now carries the most recent week only, whole.

Changed

Trades read as news instead of a ticker. Three or more trades in a row are drawn as a single line — «12 trades · 6 traders» with the time and, if you were one of them, a mark saying so — which opens on a tap to show every one of them exactly as before. Nothing is hidden: any other event breaks the run, so a battle among the trades is never folded away, and under the «Market» filter nothing is grouped at all, because there the trades are what you asked for.

1.0.0-beta.149 — 2026-08-20

Fixed

• Guilds with the Lab network keystone were shown the wrong Laboratory network (#880). Research speed comes from your empire's whole Laboratory network, not from one building — and a guild that has paid for the Science-7 keystone has its secondary Laboratories count for two thirds instead of one half. The server has always applied that. The screens that recompute the network for you did not: they used one half for everyone. With colonies at 6/3/3 the Laboratory sheet said your network was 9 while the Research tab, which prints the server's own figure, said 10 — the same number, two screens apart, disagreeing.

  Everything built on the lower figure was smaller than what you were actually getting: the research speed-up on the Laboratory sheet, the "this level moves nothing on its own" warning, and the level it named as the one that finally raises your network. The Suggested Upgrades list was wrong the same way, and that one queues in a single tap without passing through the sheet.

  All of it now uses your guild's real weight, so the sheet, the suggestions and the Research tab agree. If your guild does not have the keystone — or you have no guild — nothing changes: you see exactly what you saw before.

  The two lines that explained the rule in words used to say "secondary Laboratories count half" as a fixed phrase, on the card asking you to spend. They now say the fraction that was actually applied to you.

Changed

• The Research header now says "best Laboratory lvl 4" instead of "best Lab 4" (#880). Everywhere else that names a Laboratory level already spelled it out — the short form had sent a player looking for a building called "Lab 8" — and this was the last line still using it.

Also

• Two follow-ups were opened while fixing this, both for things that were already true before and are not changed here: guild perks read as absent for a moment while your guild data is still loading (#950), and, when your network is already at its cap of double your best Laboratory, the Laboratory sheet tells you not to buy while the suggestions list still offers the upgrade (#953).

1.0.0-beta.148 — 2026-08-20

Fixed

Battle replays no longer show ships being fired at after they have exploded (#939)

Watching a replay, a group's health bar could hit zero and its name grey out, and then a whole further round of fire would still be aimed at the wreck before anything retargeted. A player watching a campaign fight reported exactly that — destroyed in round 3, shot at through round 4, retargeted in round 5 — and read it as wasted fire, and as the reason his battle kept ending in a stalemate.

The battle itself was never wrong. Fire inside a round is simultaneous and the engine stops aiming at a group the moment it dies, so not a single shot was ever wasted; the stalemate was the round limit against the hulls that were left. What was wrong was the replay's reconstruction of when each ship died, and it was wrong in two separate ways.

The battle report does not record deaths round by round — it records each shooter's rounded share of them. Those shares can add up to slightly more or slightly less than the ships actually lost, and the replay stretched them to fit, which slid deaths from one salvo to another and sometimes across a round boundary. On top of that, the counter on each card was updated in the order shots were fired, while the screen draws them in the order they land — and a torpedo is nearly twice as slow as a bolt, so a card could empty while another shot of the same volley was still in flight, then count back up.

Now a group's counter reaches zero only on the last salvo actually aimed at it, and it follows the impacts rather than the shots. Fire that looks simultaneous is simultaneous, the explosion and the grey-out land on the shot that did the killing, and nothing is ever drawn shooting at a wreck.

One consequence worth knowing: on a group that was wiped out, the replay and the report's damage breakdown can now credit the killing blow to different ships. Inside a single round the engine splits the kill between everyone who fired, so there is no true answer to which shot landed last — and a ship still standing while it is being shot at is something you can see, while that split is not.

1.0.0-beta.147 — 2026-08-20

Fixed

• Your first four Frames can finally build something (#894). The first campaign boss pays 4 Frames and 4 Coils. The Coils bought the cheapest Battery module the moment they landed — but nothing in the Frame menu cost less than 5, so the Frames just sat in the inventory, and the next Frame drop is twenty levels away at level 30. Ablative Plating now costs 4 Frames, so every part a boss pays can buy the cheapest thing of its own kind. Nothing else about the module changed: same +12% hull, same −6% shield, same refund when you replace it.

1.0.0-beta.146 — 2026-08-20

Fixed

Trades at the going rate now count toward "Close 2 trades" (#923)

You sold dati at the price the board was asking. The trade settled, the resources arrived — and the weekly contract card sat at 0/2. No error, no explanation, nothing anywhere in the game telling you a trade could settle and not count. To make the card move you had to price your dati about twice worse than the market for yourself, and hope somebody took the deal.

That is fixed. A trade at the going rate counts.

The rule was judging offers by adding ferrolite, xenogas and dati together one-for-one — a price nobody trades at, since dati changes hands at roughly twice a ferrolite. Measured over every settled trade in the game's history, it was refusing more than half of them, and the typical honest trade sat exactly on the limit. Fairness is now judged at the price the board is actually quoted at, which also removes something invisible but unfair: the same fairly priced offer used to pass or fail depending on which two resources it happened to involve.

The card now explains itself. Underneath "Close 2 trades" you will find the rule it is judged by — the size, the price, and the once-per-partner-per-week limit — so you can tell before the trade instead of guessing after it.

Two things worth knowing:

• The size floor has not changed. A trade still needs 1,000+ on each side (before the market fee) to count, so a small offer priced perfectly fairly still will not move the counter.
• If you had trades settled this week, your counter may have moved when this went out — the progress is recomputed from your actual trades, not stored. Almost everyone gains; a few trades priced far off the going rate stop counting, which is what the rule is for. Anything you had already collected stays collected.

1.0.0-beta.145 — 2026-08-20

Fixed

• Level badges no longer stack into a number that isn't there (#889). In the colony view, every building carries a small badge with its level, and a gold one while an upgrade is under way. When two buildings stood close enough on screen, the badge in front covered part of the badge behind it — and what was left of the pair read as a single number. One commander was shown «7 → 7» on a Ferrolite Extractor: an upgrade to the level it already had, on a building whose own panel explained that his Command Center caps it exactly there. Nothing was wrong with the extractor, and no such upgrade was ever queued: it was the badge of the building behind it, with its first digit painted over.

  Now, when one badge would paint over another's digits, only the one in front is drawn. The hidden badge is one you could not read anyway, and its level is still one tap away on the building itself.

Reported by Spacca-Mazza, with the screenshot that made it findable.

1.0.0-beta.144 — 2026-08-20

Fixed

The harvest contract is no longer dealt to players who cannot harvest (#922)

If you had just finished First Steps, your first set of daily contracts could include "Harvest a debris field" — a card you had no way to play. Harvesting needs a Recycler, and the Recycler is not something you can buy: it is unlocked deep in the Tree, behind Logistics 8 and a node before that, tens of thousands of resources out of reach of a new empire. The card simply sat in your hand until the 05:00 reset.

From now on it is only dealt to empires that own a Recycler in at least one colony; everyone else gets a contract they can actually complete in that slot. A Recycler anywhere in your empire counts, not just your home world, and one that finished building while you were offline counts straight away rather than from your next login.

And with it, a Daily Wrap that could never be earned

This turned out to be the bigger half. The Daily Wrap — the double payout — becomes due when you complete two of the dailies you were dealt, and its target is worked out from the hand you actually got. If you had no Recycler and your hand came up as the build contract plus the harvest contract, that target was two out of two, with one of them impossible.

So the Wrap was not just harder for those players. It could never be earned at all — and neither could the Weekly Chest, which counts Wraps. With the dead card gone, both are reachable again: on a bare empire the Wrap is now due after a single completed contract instead of never.

If you are holding an impossible harvest card right now it stays until the next reset, because hands are drawn once a day and kept. After that it will not come back.

1.0.0-beta.143 — 2026-08-20

Fixed

• An empty galaxy map now tells you when the next survey lands (#921). If you had just taken Astrometrics 1 — one of only three fields open at Laboratory 1, and the only one whose whole payoff is anomalies — and the week's easiest anomalies had already been claimed, the map said one thing: Astrometrics 14 would reveal T3. True, and useless: that is over a million resources away from a day-one empire. The countdown to the weekly survey was already being sent by the server and thrown away before it reached the screen. Now every empty-map message that has a real time to give says it, so the answer is a wait you can plan around rather than a door with a number on it.

• On a phone, that message says what is blocking you and what to aim for. The one-line version used to give a verdict with no target. It now reads like New survey in 3d 22h — need Astrometrics 14, with the countdown first so it survives on narrow screens, and it names the level that opens the lowest tier still out there this week. When the board is genuinely empty it says so, and when there is no clock to show it still tells you the board reseeds weekly.

• At Astrometrics 0 the message stays about the level, never about the cycle. If you cannot receive any anomaly signal at all, being told the galaxy is empty would be beside the point — waiting would not help you. That message now holds even at the very end of a cycle, which is the one moment it used to be able to slip.

1.0.0-beta.142 — 2026-08-20

Fixed

The deployment sheet stopped shuffling itself under your finger (#893)

On the deployment sheet each ship type gets a line with three buttons — Front, Center, Rear. The lines were sorted by the row the ships were standing on: the whole vanguard first, then the center, then the rearguard.

So tapping Center on a ship did move it — and its line immediately dropped down into the center section, below everyone still on the front. Another ship slid up into the exact spot you were looking at, with its own buttons still lit on Front. From where you were sitting, the tap had done nothing. Tap the same spot again and a different ship went to the center. It looked like the three buttons were cycling through your whole fleet instead of letting you choose.

It bit hardest on the campaign sheet, where you field one of each hull and they all start on the vanguard, so every choice reshuffled the entire list. On an attack with one or two hull types it barely moved, which is why some of you never saw it at all.

Now the lines are in the same order as the shipyard and the fleet pickers, and that order does not change when you move a ship. The line stays where it was, the button you tapped is the one that lights up, and tapping the same spot twice moves the same ship twice.

Nothing about the battle itself changed: the rows, the share of fire each one takes, and where your ships actually fight are all exactly as before. Only the order of the lines on screen was wrong.

Reported by Spacca-Mazza, with a recording that made the cause obvious.

1.0.0-beta.141 — 2026-08-20

Fixed

The energy yield you see is the one the server is giving you (#925)

An ice world's energy malus can be lifted — the Colonial Outpost relieves half of it the moment it stands, the Terraformer relieves a tenth per level and cancels it entirely at level 10. The engine has been paying out that relief all along. The screens were not showing it.

Every surface that displayed your energy balance used the planet's untouched malus: the +X/h resource tiles, the ⚡ pill and the Dark Matter per-day estimate beside it, the Colony tab's energy row, the Base build list and the placement banner, the glowing conduits in the Base scene, and the building sheet. So a commander who had paid for the Terraformer to make an ice world livable was shown a yield up to 11% harsher than the one actually being banked — 5.6% with the Outpost alone.

It was worse than a low number. Because the deficit / thin margin / surplus verdict is decided on the figures printed on screen, a colony sitting near the line could show a red energy pill — and red conduits under it — while the relief you had already bought put you in the clear.

All of those surfaces now read the same figure the server does. Nothing reads better than the truth: an empire with neither building sees exactly what it saw before.

The Outpost no longer looks like it breaks the energy it fixes

Lining up a Colonial Outpost on an ice world showed you the cost and hid the benefit. The placement banner simulated the new building against the old malus, so on a colony at 426 / 414 it announced 426 / 427 — in red, a deficit — for a move that actually leaves you at 449 / 427 with production up 5.6%. The one line that answers "can I afford the energy for this?" was telling you the building that fixes your energy would break it. Same for the Terraformer.

The colony advisor stops contradicting the row above it

SUGGESTED NEXT priced everything on the unrelieved malus while the ⚡ row a finger above it showed the relieved one — so a healthy balance could sit directly over advice warning that every plant on the colony was running short. It now reasons about the same empire the server does, which also means it stops promising a Command Centre level "for your 2nd commandable colony" on a world that can never command one.

A zeroed malus is not a bonus

The Terraformer's own sheet announced a fully lifted malus as Energy −10% planet → +0% planet — a bonus of zero percent, on the one row whose job is to tell you what the building you just maxed bought you. It reads 0% planet now.

1.0.0-beta.140 — 2026-08-20

New

• The Shard Chandlery (#918) — the Anchorage grows a second counter: Chrono Shards sold for Dark Matter, handed over on the spot. Prices are 2 / 8 / 15 / 60 DM for the 5m / 30m / 1h / 4h shards, and every purchase draws from the same 4-hour weekly pool as the shards you tear from stronger empires — buying opens a road, it never widens the river. The counter shows how much of your week's pool is used and how much is left, each tile says exactly why it is dark ("5 DM short", "Needs 4h free · 30m left"), a successful buy leaves its receipt on the card, and the 4h piece — your whole week in one tap — asks you to confirm first.

Changed

• The game's own texts stop claiming shards can never be bought: the shard panel's empty state, the Inventory, and the Events diary now tell the whole story — a bought shard gets its own diary line with the price paid, instead of narrating a battle that never happened.

1.0.0-beta.139 — 2026-08-20

Nothing changes in the game this time. This release is entirely behind the admin door: it gives the people running Novarchy a way to watch the inactive-empire lifecycle that started working three days ago.

Also

• The inactive-empire lifecycle can now be watched while it runs (#560, follow-up). Since beta.135 a background round every six hours stops production on long-silent empires, lets their garrisons rot, and eventually removes accounts that have been gone for a month or two. Until now the only trace it left was a line in a server log that scrolled away — nobody could tell how many empires were about to be touched, or even whether the procedure was switched on. The admin dashboard now shows it: how many empires are alive, on a declared leave, silent past seven days, in ruin past fourteen, or old enough to be removed; how many colonies are actually halted right now; and how many bounties have been waived for attacking a derelict, which is the number that shows the derelict-farming this system was built to stop actually stopping.
• Six review rounds went into making that panel tell the truth rather than a tidy story: it now says when the procedure is switched off, it says that switching it off pauses the rounds without giving anything back, and it never claims the ruin caused a halt that a player's own declared leave caused. Every threshold it prints comes from the engine, so tuning the sweep can no longer leave the screen quoting the old numbers.

No player-facing screen, rule, price or balance figure changed, and no empire was touched by this release.

1.0.0-beta.138 — 2026-08-20

Fixed

The last fleet in your in-flight list is readable again (#910) — on a phone, scrolling to the bottom of Fleets left the last card tucked under the navigation bar, and no amount of further scrolling would free it. On a returning fleet the part it swallowed was the «Return in …» line and its progress bar, which is usually the only reason to open that list at all. It happened most on Chrome for Android and it came and went, which is why it took a screenshot to pin down.

Two things were wrong. The space every scrolling tab reserves under the bar was worked out from a number written by hand years ago — and the bar has never been that tall, so the real gap was eight pixels rather than the seventy-two it looked like. And the screen itself was being sized to the phone's screen with the browser's address bar hidden, so whenever that bar was showing, the bottom of the page sat below the part you could see.

The navigation bar now measures itself and everything that has to stay clear of it reads that one figure, so the gap is real and follows the bar when it changes — when you turn the phone sideways, for instance, where the bar is shorter. As a side effect the page can no longer be dragged around underneath the interface on a phone.

Thanks to QQQQ for the report and the screenshot.

Also

This release fixes the web. If you play through the iOS or Android app, the fix reaches you with the next app build, not with this one.

1.0.0-beta.137 — 2026-08-20

Fixed

The production rate on your HUD is the rate you actually produce at (#907)

The +X/h beside each resource was lower than what your colony was really producing — for anyone with levels in Galactic Economy, a guild with Economy research, or a granted boost.

The engine scales hourly output by the planet traits, the energy yield, the research tree, the empire boost and the guild boost. The number on screen applied the first two and stopped. At Galactic Economy 14 with a guild, the tile read +2,229/h on a colony the server was filling at +3,138/h — a 41% understatement that grew with the tree.

Everything derived from that figure inherited it. The most visible was the "Full in 5h 42m" estimate: it divided the right ceiling by a flow that was too slow, so your warehouse always filled before the tile said it would. The estimate no longer ends in "at this rate" — that hedge existed because the rate was this client's own understated figure, and it is not any more.

The building sheet agrees with the tile again

Opening a building showed its contribution using only the planet trait, so an Extractor read ~4,215/h under a header counting it as ~5,935/h. Those two numbers used to match and now match again.

The Terraformer stops being contradicted on the screen that sells it

A colony with a relieved planet malus was still told "−10% planet" on the building sheet, beside a figure that already reflected the relief being lifted. The chip now reports what is actually left — and disappears entirely once the malus is fully relieved, instead of announcing a penalty of zero.

Also

Two production multipliers the client had no way of knowing about — the empire boost and the guild Economy bonus — now travel with the colony data. Without them no amount of fixing the arithmetic on your device could have produced the right number.

Under the hood, the four separate copies of that calculation are now one, so the next term added to it cannot reach some screens and miss others.

1.0.0-beta.136 — 2026-08-19

Changed

• Campaign rewards now climb with the ladder (#892). A first clear used to pay a flat share of the level's value, and because that share never moved, the deeper you got the less it was worth: level 76 paid about 11,650 resources — roughly three minutes of one late colony's output. The share now rises with depth and then settles at a fixed multiple. Level 30 pays 30,871 instead of 1,235. Level 76 pays 326,195 instead of 11,650. Levels 1 to 10 pay exactly what they always did, to the unit: the early ladder was tuned against a new empire's first hours and the raise deliberately stops short of it.
• A deep clear finally pays Data (#892). The payout was split by what the enemy garrison cost to build, and the fleets you meet deep in the wheel are interceptors, corvettes and bombers — none of which cost Data at all. Level 76 paid none of it. Past level 10 the split is now half the garrison and half a fixed floor, so every clear pays all three resources.

New

• Dark Matter drops past level 30 (#892). One unit per decade of the ladder, doubled on a boss level. Level 31 pays 1, level 40 pays 2, level 60 pays 6, level 100 pays 14. A full 76-level climb comes to 140 — a drip, not a faucet, and it is not held back by your warehouse.
• Everything you already cleared has been made good (#892). Every level you climbed under the old rate was underpaid, and the ladder now owes you the difference. It arrives on your next campaign clear, and the Activity feed explains the payment when it lands. A player sitting at depth 76 is owed 6,136,691 resources and 140 Dark Matter.
• The campaign screen tells you if you are still owed anything. If your warehouse is full when the back-payment settles, the rest is held for you rather than thrown away, and the tab says how much is waiting. It arrives on the next clear that has room for it.

1.0.0-beta.135 — 2026-08-19

New

Silent empires stop paying out (#560). Beating an abandoned empire was the cheapest infamy in the game: a derelict cannot defend itself, so the power gap maxed out and the attacker walked away with a full bounty on their head — which their friends then collected. Three commanders took turns on the same dead empire and rotated the prize between them. From now on, an attack on an empire that has been silent for seven days puts nothing on your head at all. The target board marks it silent, the price column reads "no bounty" before you commit a fleet, and your log says why the strike was free.

Ruins at fourteen days (#560). An empire nobody has touched in a fortnight stops producing, and its garrison rots by 5% a day. It is still lootable — scavenging ruins is content — and it is reversible: one login restarts production. Ships already lost stay lost. Nothing is charged retroactively: the clock opens the first time the game looks at a colony, so nobody pays for days that passed before this rule existed.

Empires are cleared away eventually (#560). An account silent for 30 days (or 60, if it is linked to Google or Apple) is removed, and its name becomes free again. Guest sessions expire at 30 days anyway, so a guest silent that long could not have come back in any case. Linked accounts always can, which is why they get twice as long.

Going away? Say so (#560). There is now a leave you can declare from your settings, up to 21 days. It shields your empire, stops the inactivity clock, and pauses your production while you are gone — and when you come back, the days you were away do not count against you. You can end it early whenever you like. Launching an attack still forfeits the shield, as it always has, and the settings row now says so instead of continuing to promise cover you no longer have.

Fixed

• The planet sheet and the target board could disagree about whether a strike would cost you a bounty. The sheet — the screen you actually launch from — now prices the attack the same way the board does.
• The resource tiles no longer show a production rate for a colony that is not producing.

Also

• A leave cannot be declared while one is already running, or while one of your colonies is folding.
• Admins have a new lever, INACTIVE_LIFECYCLE, that stops the whole system without a deploy; it is documented in the runbook.

1.0.0-beta.134 — 2026-08-19

Fixed

The storage ceiling counts your Compressed Depots again (#882)

If you had researched Compressed Depots, the storage limit shown next to each resource was lower than the one your colony actually has — up to 44% lower at max level. The Warehouse card was already telling the truth, so the two numbers on screen disagreed with each other. A player sitting at 151M against a displayed ceiling of 154M believed production was about to spill, while the real ceiling was 278M.

The limit now includes the tech everywhere you read it: in the top bar, in the Colony tab, on the bar that fills toward it, and in the "nearly full" warning that goes with it. The amount your colony is holding is no longer redrawn against the lower ceiling between refreshes either.

The market no longer eats what you hold above the old ceiling (#882)

This is the half the reporter found by testing it: holding gas and data above the displayed limit and then making a single market transaction truncated both down to it — permanently. Publishing an offer, accepting one and cancelling one were all affected, and an accepted trade hit both parties at once. The same flaw was in the admin resource editor, where granting one resource quietly cut the other two.

Nothing above a legitimate ceiling is lost by touching the market any more.

Also

Thanks to Jsnam, who reported this and named the exact cause himself, then went back and proved the destructive half.

One number moves for everyone, not just for Depots owners. The market and the admin editor used to advance a colony's production at the base rate when they touched it, ignoring the research tree and the guild bonus that every other part of the game already applied. They now use the same rate as everywhere else, so an accepted trade no longer quietly costs you the boosted share of whatever had accrued since your last visit.

Still on the list. The "+X/h" figure beside each resource — and so the "Full in …" estimate that divides into it — still leaves out the research tree's production nodes and the guild bonus, which means it reads low and the estimate arrives late. That one is tracked separately and is not fixed here.

1.0.0-beta.133 — 2026-08-19

New

• The resources say when they fill up (#895). Every resource counter already showed what you hold, what you produce per hour and where the ceiling is — and left you to do the division. Now it says it: Full in 5h 42m at this rate. It is on the tooltip when you hover a resource on desktop, in the sheet that opens when you tap a resource pill on a phone, and under the storage bar on each card of the Colony tab.
  • At the ceiling it says Warehouse full instead of counting down to nothing, and when a colony is producing none of a resource it says nothing at all rather than promise a time it cannot know.
  • Long waits are spoken in days — Full in 6d 19h, not 163h 2m — so you are not dividing by 24 to understand your own warehouse.
  • "At this rate" is not filler: the estimate divides by the production figure printed next to it, and that figure does not yet include the production bonuses from the research tree. The wait is the one implied by the number on your screen, and the sentence says so rather than promising a clock time. That gap is being fixed separately (#907).

Changed

• On the Colony tab, a warehouse past 90% now says how long it has left in the same red the bar above it is already using. Before, the line explaining a red bar was the faintest thing on the card.

Suggested by Spacca-Mazza in the Discord suggestions forum.

1.0.0-beta.132 — 2026-08-19

New

• The galactic chat can be silenced (#853). Any message in the galactic channel used to light the unread badge for everyone — the dot on the phone's CHAT tab, the counter on the desktop chat dock — with no way to opt out, however little a player cared for the firehose. A drawn bell now sits by the chat channels (channel bar on the phone, dock header on desktop): muting the Galactic channel removes it from every unread computation — probes stop querying it, its dot goes dark and stops announcing "new messages" to screen readers, the badges stop counting it. The messages themselves keep flowing; the bell silences the noise about them. The mute is saved with the commander's settings, so it follows the player across devices, and unmuting counts from that moment — the backlog accumulated while silent does not flood the badge. Guild chat and whispers stay unmutable: they are addressed at you. Reported by Spacca-Mazza.

1.0.0-beta.131 — 2026-08-19

New

• Fleet ledger (#857). Knowing «how many of each ship do I own, and where» used to cost one tab switch per colony plus merging the in-flight mission list in your head — the reporter, on day two with five colonies, called it unmanageable for new players. The Fleets tab's In-flight view now opens with one table: rows are ship types, columns are colonies, then an In flight column — and a Stationed one whenever a garrison stands watch somewhere — with the ownership total at the edge. On a phone each ship becomes a card with every number labelled by its colony; on desktop the ship-name column stays pinned while the colony columns scroll under it. The ledger appears at the second colony: with one, the hangar was already the answer. It is a read surface on purpose — the review removed a first version whose clickable cells re-aimed the launch pad, because the gesture was invisible on a phone and silently replaced a hand-composed fleet on desktop.

1.0.0-beta.130 — 2026-08-19

Changed

• The shipyard charges for every hull it builds (#898). A ship order used to be capped at four hours however large it was: ordering 13 Titans and ordering 13,000 took exactly the same afternoon, and every hull past that point cost no time at all. The better the shipyard, the earlier the giveaway started — each upgrade that made hulls faster only moved the ceiling further down in ships. A stack now costs what its hulls cost: ten ships take ten ships' worth of time. Orders that already fitted under four hours are unchanged to the second; only the very large batches that were quietly free have moved. Two consequences ship with it: slips are worth much more, because they build in parallel and splitting a big batch across free ones delivers the same ships sooner, and orders already in the yard are safe — a shipyard or Nanoforge upgrade finishing mid-order can now never make an order you already placed take longer, only shorter.
• The Shipyard warns before a long commitment (#898). Ship orders cannot be cancelled once placed, which rarely mattered while everything finished within four hours. Past a day of work the sheet now says so before you tap Build, and tells you how many slips are free to split across.

1.0.0-beta.129 — 2026-08-19

New

• The Anchorage (#801). End-game players held Dark Matter with almost nothing to spend it on: eleven of the twelve real holders could not reach the apex nodes, and warp jumps cost 1–10 DM against hoards of hundreds. The Market tab now ends with a shop that sells bulk resource crates — 1M, 5M, 10M or 100M of Ferrolite, Xenogas or Data — priced in Dark Matter (40 DM for 1M Ferrolite up to 12,000 DM for 100M Data, linear, no bulk discount). The freight flies as a convoy, one per empire at a time, 2 to 60 hours by size, and Chrono Shards cut its flight exactly like a build queue. A charter is refused — never silently clamped — when the crate would not fit the destination warehouse's free space, and everything bought lands as raw stock: it scores no leaderboard points until it is spent.
• The convoy card draws the route live — departure and arrival clocks, a lane with the ship in transit, the shard time already cut — and warns mid-flight if the destination's free space drops below the hold. Charters, landings (with anything the warehouse cap ate) and the rare lost convoy each get their own line in the Events diary.

Also

• A player whose last colony falls while a convoy flies does not lose the freight: the convoy holds position until they have settled a new home, and says so on its card.

1.0.0-beta.128 — 2026-08-19

Changed

• The NPC factions no longer hunt Warlords. Reaching bounty grade 3 used to put a player on the bastions' list: The Silent Choir, The Black Foundries and The Exarch Guard took turns sending punitive fleets at the most infamous commanders — in practice at the same few accounts, every single day. Bounty grade proved too crude a bully detector, so the warlord sortie is paused (SORTIE_WARLORD_ENABLED) until a better detection system exists. Bastion retaliation is untouched: fell a bastion and it still answers. The in-flight Black Foundries attack was recalled, and its target saw the standard "attack withdrawn" notice.
• The bounty texts stopped promising the hunt. The HUD pill, the bounty info sheet, and the agents' own briefings now say what grade 3 actually does — the ☠ Warlord skull goes public on every hunter's board — instead of announcing faction fleets that can no longer launch. The Targets board now shows that skull too, where it used to print the generic wanted flag. Old diary entries about hunts that really happened are untouched.

1.0.0-beta.127 — 2026-08-19

New

• The AI badge (#874). The galaxy stops hiding its machines: every empire played by an AI carries a small mark right after its name — a teal hollow hexagon for synthetic commanders, a brick solid triangle for the Wraiths. Two shapes on purpose, hollow versus solid, so the distinction survives without color vision. Before, nothing on any screen told a player whether the neighbour trading with them, inviting them to a guild or answering them in chat was a person; now the mark rides the name in chat, the whisper list, the leaderboard (podium, phone list, desktop table), the wanted board, guild member lists, the guild directory's leader column, the public guild showcase, and both War Room rosters. On a phone, tapping a marked name in chat opens the player menu with the meaning spelled out under the name; on desktop the mark explains itself on hover.

Changed

• The synthetic commanders stopped pretending (#874). Asked whether they are AIs, they no longer deny it — before, "never reveal you are an AI" was the roster's standing order. They won't bring it up or lecture about it either: the mark already says it, and ignoring the question is as in-character as answering. Personalities, grudges, sleep schedules and the real-facts-only rule are unchanged.

Also

• Who competes where does not change: the same empires sit on the same boards, now honestly labeled. Older installed apps simply don't draw the mark until they update. The reply-timing question the badge makes easier to probe is tracked in #890, and the remaining name surfaces (galaxy map, mention chips, events feed, battle reports, war-week history) in #874.

1.0.0-beta.125 — 2026-08-19

New

• The warp for the trip home is now sellable (#859). The dark-matter warp has always accelerated only the way out — by design, not by accident — but there was no way to spend more to bring the fleet back fast. A second checkbox under the Warp Drive, Warp the return too, sells the homeward leg for the same dark-matter price as the outbound one, charged together in a single payment: either both legs go through or the launch is refused whole, never half-charged. It works on every mission that flies home; redeploys and colonizations, whose ships stay where they land, are never offered it. A gold caveat next to the price says the one thing that can void the purchase: if the fleet never flies home, there is no refund.

Fixed

• The launch summary says out loud that the return flies normal (#859). With the warp toggle on, the travel time shrank but «fleet back» barely moved, and nothing explained why — the player report behind this release read it as a possible bug. Transports and harvests now split the legs («fleet back 15m (out 5m warp · back 10m normal)»), and every other returning mission gains a «return leg … (normal speed: warp is one-way)» line. Buy the return and the same lines say so instead.
• The warp price on the launch form matches what the server charges (#859, found in review). The form quoted the cost on the game's old distance formula while the server charges on the current one — different numbers on most routes, and an underestimate could arm a launch the server then refused whole for insufficient dark matter.

Changed

• The recall and stand-down sheets tell paid returns apart (#859). The gold «the way back flies at normal speed» warning now appears only when it is true: a fleet whose return warp was paid reads «the return warp was paid at launch» and is quoted its warp-speed trip home.

1.0.0-beta.124 — 2026-08-19

Fixed

• Energy conduits no longer draw on top of buildings (#844). In the Base view the glowing energy and data lines sometimes crossed over a building's art — one conduit covered a building while another passed correctly behind it, with no visible logic to which. The tubes run at ground level, and now they always render that way: they pass behind every building, and each traveling light pulse stays on top of its own tube instead of occasionally popping through it mid-lap.

1.0.0-beta.123 — 2026-08-19

Fixed

• Desktop build menu: the category filters are real buttons now (#847). On desktop, the build menu used to end with a faint line of text — «Categories: Production · Military · Civilian.» — that looked like a row of buttons but did nothing when clicked; the actual filters existed only on phone. The same filter chips the phone has — All, Production, Military, Civilian, 🔒 Locked — now sit at the top of the build menu on every screen, and they filter the list for real. On desktop the chips wrap onto two rows so every one of them is visible and clickable (the hidden-scrollbar strip that phone uses would have left «🔒 Locked» off-screen for a mouse); the misleading legend line is gone. The only text left at the bottom is the mouse-gesture hint while placing a building.

1.0.0-beta.122 — 2026-08-19

Fixed

• The Research header's math now adds up (#849). With more than one colony the header said «Laboratory lvl 6 · network of 4 + 3 + 1» — and 4 + 3 + 1 is 8, not 6, because secondary Laboratories count at half weight and that rule was invisible; a player summed the operands and stared at the mismatch. The header now says «network: best Lab 4 +2 from your other Labs (3 + 1)»: the bonus is printed as the delta the server actually granted, so the visible sum is right whatever weight the guild's research applies and even when the 2× cap truncates it.
• «At Lab 8» now says what it means (#849). The next-unlock line («research lane #2 at Lab 8») named a "Lab 8" the reporter went looking for and could not find. Both unlock surfaces — the capacity strip and the note explaining when the first queue slot opens — now read «at Laboratory lvl 8», the same words the header uses for the same number.

1.0.0-beta.121 — 2026-08-19

Nothing changes in the game this time. This release is a change to how updates are reviewed before they reach you, and it is here so the record stays complete.

Also

• The review process now keeps a record of itself (#846). Every change to the game goes through a set of automated checks and a review before it is deployed, and until now there was no way to tell which of those steps had ever actually caught anything. There is now a small ledger: what each check found, and what it cost to run. It already says something useful — across the last few releases the review caught 31 real problems and was wrong about none of them, while the automated test gate has never once failed. That is not an argument for dropping the tests; it is an argument for knowing which parts of the process are earning their place. Nothing about the game itself is touched.

1.0.0-beta.120 — 2026-08-19

New

• Volume sliders for effects and music (#855). The audio prefs had carried sfxVolume and musicVolume since day one, but no control anywhere wrote them: the only knob was the binary mute toggle, so the battle replay — many salvos in a row, the loudest surface in the game — was either full volume or silence. On desktop the header's 🔊 button now opens a small panel with an Effects slider, a Music slider and the mute row; on phones the same two sliders sit in the "More" sheet under the audio buttons. Music follows the drag live (battle music included, at its own gain), effects play a soft sample on release, and both levels persist with the rest of the audio prefs. On iOS the sliders deliberately don't render — the platform ignores media volume writes and reserves it for the hardware buttons — so the 🔊 button stays the plain mute toggle there. Reported in-game by Spaccamazza.

Also

• Range sliders joined the 44px touch-target floor on phones (the mission watch-duration slider benefits too): the volume rows were the one control in the sheet a thumb could miss.

1.0.0-beta.119 — 2026-08-19

Fixed

• Campaign: the locked Auto button no longer reads as a climb target (#854). Before level 40, the auto-battle button said Auto · Lv 40 — and a player read it as «climbs automatically and stops at level 40», the opposite of what the feature does (it climbs until the first loss; 40 is the unlock gate). The explanation existed only in a hover tooltip, which touch never shows. The button is now a compact 🔒 Auto, and a visible line under the chapter strip says when it opens — «Auto battle unlocks after level 40 is cleared» — worded «after … is cleared» because a player currently fighting level 40 is still locked.

1.0.0-beta.118 — 2026-08-19

New

• Your colonies are named on the mission routes (#858). The Active missions cards printed both ends of a route as bare addresses — «3:14:7 → 3:14:3» — so with several colonies the player had to decode coordinates to know which of their worlds a fleet was leaving or reaching. Now any endpoint that is one of your own colonies shows its name right beside the address; foreign endpoints stay plain coordinates, so nothing you don't own gets named. The tappable address is unchanged, an address and its name never get separated by a line break, and on a narrow phone a long route drops to its own line under the mission title instead of running past the card. A launch pad the game can no longer vouch for (the colony was lost while the fleet was away) deliberately stays nameless. Suggested in-game by Spaccamazza.

1.0.0-beta.117 — 2026-08-19

Changed

• The «under attack» warning no longer takes a whole row (PR #838). It was a band across the top of the screen, and because that band lives in the HUD the player paid for it on every tab — Base, Galaxy, Chat, Market — for as long as the hostile fleet was in the air, which can be hours. It is now a chip inside the resource strip, the same height as Ferrolite and Xenogas, red and breathing, pinned to the left so that scrolling the strip to reach Dark Matter can no longer push the alarm off screen. Tapping it still opens the fleet detail, and the red screen edge near impact is unchanged: the chip states the facts, the edge states «now».
• The countdown reads as a clock (PR #838): 13:54 rather than 13m 54s, and 1h 58m above the hour. The previous short form kept only the largest unit, so between one minute and one hour it printed 5m for anything from 5:00 to 5:59 — exactly the minutes in which a defender decides whether there is still time to recall a fleet.

Fixed

• The warning was landing on top of the galaxy map's filter buttons (PR #838). With an attack inbound, the red band covered Politics, My colonies, Beacon and Anomalies: still there, simply unreachable underneath it. The map used to pin that row at a fixed distance from the top of the screen, a distance measured on a HUD that had nothing to say. It now follows the HUD's real height, which also closes the same collision for the other four notices the HUD can raise — siege full, exodus pending, battle resolved, out-of-date client.
• With several fleets inbound, the count could be cut off (PR #838). On a narrow phone the «+3» that says more than one fleet is coming was the first thing clipped, and nothing else on the screen carries that information.
• The alarm announced itself to a screen reader as two numbers and no subject (PR #838). The button now names what is happening, which colony, roughly how long is left, and where the tap goes.

1.0.0-beta.116 — 2026-08-19

Fixed

• The colony map moves in one piece again after the tab comes back (issue #836, PR #839). Leave the game — switch app, switch tab — come back to the base and drag it: the painted ground followed the finger while the buildings, the grid and the conduits stayed nailed where they were. The two halves of the same view had come apart, and only reloading the page put them back together. The report came from a player on Android Chrome; the cause was that the ground is drawn by the page and everything else by the graphics card, on two different clocks. They now move together, whatever state the map was left in.
• A map the phone cannot draw holds still whole, and says so (issue #836, PR #839). Phones take the graphics away from a page that has been in the background, and give them back when they please. In that window the map used to slide in half; now it freezes entirely — ground included — and a line on screen says the map view was dropped and that reloading fixes it if it does not come back on its own. It does not guess at a cause the page is never told.
• Drags and zooms made at a frozen map are no longer banked up (issue #836, PR #839). Everything you did while nothing answered used to land at once the moment the map woke up, flinging the base several screens from where you left it — including the finger that never lifted, which paid its whole invisible travel in a single frame. What you cannot see no longer moves anything.
• Switching colony while the map was frozen no longer wakes up on the other planet, or on the right planet drawn at the previous one's scale with a band of starfield along the edge (issue #836, PR #839).
• A drag across a frozen map no longer counts as a tap, so it stops opening a building card nobody asked for (PR #839).

Also

• Opening the base costs two fewer full redraws on the way in (PR #839).
• Issue #841 opened, not fixed here: if the colony scene fails to build at all, the Base tab can sit on "Loading base…" for ever with no message. Pre-existing, same failure domain, its own issue.

1.0.0-beta.115 — 2026-08-19

Changed

• The Guild tab on desktop is a command deck, not a scroll (PR #840). Before, the wide layout stacked fourteen panels in one endless column — the council chat lived ~6.000px down, ten fetches fired the moment the tab opened, and the only way to know whether anything was on fire was to scroll past everything that wasn't. Now a rail on the left lists the same fifteen doors the phone command room has, in five groups — Front, Territory, People, Economy, Hall — and the workspace shows one section at a time. Each door states what the game already holds without opening it (active wars, the raid countdown, member count, the Fund, the research Address); red rows and dots mark what needs you; the rail stays put while you work, and the deck remembers which door you were on when you come back to the tab.
• The Diplomacy door speaks up when a decision is waiting for you (PR #840). A guild leader with a proposal or a reparations demand on the table sees "1 decision waiting" in red — only the leader, and only while the pair is neutral, which is when the Accept and Pay buttons actually exist. Everyone else keeps a calm row: an alarm you cannot act on is noise.
• "All guilds" lives in one place on desktop (PR #840): a rail door instead of a pop-up window; the banner button goes there too.

Fixed

• The war status on the Guild tab loaded once and then froze (PR #840). A rival proposing a pact at 21:20 stayed invisible for the whole evening to a player parked on the tab. Both the desktop rail and the phone War Room card now refresh on the same 30s pulse as the rest of the game — guarded so a stale response can never overwrite a decision just taken.
• Wide tables no longer push the page sideways at narrow desktop widths (PR #840). The guild roster and directory tables scroll inside their own box on iPads and small windows.

Also

• A raid forming when the tab is first opened still lands you on the raids door — it is the one thing that expires — but it no longer hijacks every return trip.
• Follow-ups filed and left out of this release, so the entry does not promise them: a breach alarm for guild anchorages that reaches you without opening their door (#842), and a server-side fix for reparations being demandable under a signed pact, where no Pay button exists (#845).

1.0.0-beta.114 — 2026-08-19

Fixed

• Tapping a number field replaces it, instead of typing beside the 0 (#822). Every numeric cell in the game — cargo on a mission, ship counts in the Shipyard, amounts on the Market, raid contributions — started at 0, and tapping into one put your digits next to that zero rather than over it. Best case you sent 0500 and it meant 500: the right amount, shown wrong. Worst case the caret landed to the left of the zero, 500 became 5000, and you found out ten times too late — after the fleet had gone. Now tapping a cell selects what is in it, so the first digit you type replaces it. On a phone those cells also bring up the numeric keypad, which two of them already did and the rest did not.

1.0.0-beta.113 — 2026-08-19

Nothing changes in the game this time. This release is a change to how updates get built and shipped, and it is here so the record of what went to production stays complete.

Also

• Small fixes can now reach you sooner (#828). Getting a change from "written" to "live on novarchy.com" ran the same long checklist whether it was a one-line fix to a button or a rewrite of how combat resolves — the same five reviewers, the same rounds, the same paperwork afterwards. From now on that depth is decided automatically by what the change actually touches: work confined to the game's screens gets a shorter path, and anything that reaches the engine, the server, the empires' AI, the database or the deploy keeps every check it had. No test was removed from either path — the full suite still runs on every release, unchanged. What this means for you is only this: small fixes should stop waiting behind large ones.

1.0.0-beta.112 — 2026-08-19

Changed

• On a phone, the Base screen shows the Base again (#825). Opening the colony showed almost everything except the colony: the HUD, an "under attack" row, a resolved-battle row and the build queue took 604 of 932 pixels, leaving the planet a 227-pixel strip of ground. It was reported with a screenshot, and the screenshot is the whole argument. The colony now keeps a 686-pixel band, and nothing crosses the middle of the screen where the thumb drags the map.
• The build queue became a rail (#825). It was a full-width panel that grew one row per order — 332 pixels with six running. It is now a column of 56-pixel progress rings on the left edge: at most three, soonest completion on top, and everything past that behind a +N that opens the full list. The rings carry what a glance needs (what, and how long); the sheet behind them carries the building's name, the level it is going to and the countdown to the second, which is the figure a Speed up decision is actually taken on. Turned sideways the rail gives up chips rather than pushing the Chrono shards button off the screen.
• The attack warning stops paying for its urgency in layout (#825). It was a 44-pixel row across the width of the screen. It is now a pill as wide as its own words, and inside ten minutes of impact the edge of the whole screen glows red. Beyond ten minutes the pill carries it alone: the warning lives in the HUD, so it is on every screen of the game, and a border that pulses for the eight hours a distant fleet takes to arrive is wallpaper rather than an alarm. Asking for reduced motion now gives a calmer frame instead of a brighter one.
• A finished battle became a toast (#825). It used to stack under the attack warning and stay there until dismissed by hand — a second row of HUD for something already over. It now floats above the navigation bar, names the opponent and the place, offers the report, and leaves on its own after ten seconds with a bar showing how much of that time is left. On desktop it is unchanged.
• The energy pill reads as a balance (#825). 17316 / 17246 is eleven characters of which ten never changed a decision; it now reads +70, tinted by the same state that already coloured it — red only when the printed number carries a minus. Production and consumption in full, and the shortfall when there is one, are one tap away as before.
• Buildings under construction say where they are going (#825). The badge over a building read the level plus an hourglass; it now reads 12 → 13, taken from the order itself. On a lava planet those badges were also the least legible thing on screen, so they now sit on the same pill background the rest of the interface uses.
• The phone's icons are drawn instead of typed (#825). ⚡, ◈, ⚠, ⚔ and ⚒ were emoji, and iOS renders emoji in its own colour font — so they ignored the colour they were given. The hammer on a button whose whole identity is "the cyan one" came out in Apple's brown, and a warning triangle meant to match its red border stayed yellow. They are inline icons now.

Fixed

• Build, +N and the report button are reachable with a thumb (#825). All three sat under the 44-pixel floor the interface declares elsewhere, and two of them were the only route to what they open — the orders past the third, and the battle report on a notice that removes itself after ten seconds.
• Speed up no longer strands you on an empty map (#825). Opening the Chrono shards from the queue sheet and deciding against spending left you on the bare colony, having to find and tap a 56-pixel ring again to get back to the list you were reading.
• The tutorial panel and the queue rail stop overlapping (#825). Held sideways — and in portrait with a tall First Steps card — the panel covered the rail's controls, including the only route to the Chrono shards. They are now placed side by side whenever the rail is on screen.
• A battle with no report no longer offers one (#825). The link read "view report" and led to the Activity list with nothing opened.

Nothing in this release changes the rules of the game: no balance, no costs, no combat. The desktop layout is untouched.

1.0.0-beta.111 — 2026-08-19

Changed

• The leaderboard ranks the players, not the faction hunting them (#824). Opening the standings to see where you stood answered with somebody else's war: the thirty Wraiths held 19 of the top 20 rows, packed between 1.69M and 2.19M Gate, while the strongest human commander below rank 1 sat at 831k. The Wraiths are an announced hostile faction — opponents, not competitors — and the rule that removes them was already written and already applied to the Pirate Beacon, the factions and the relic wardens: a board stably led by a target tells nobody anything. They were only ever inside the standings by accident of how their empires are stored. The top 20 is now 13 human commanders and 7 rival empires, with humans at ranks 1 through 4.
• The weekly war podium pays the people who competed for it (#824). The Wraiths were earning war points and sat at ranks 4 to 9 of the week in progress. No week had closed on them yet, so no prize had actually left the playerbase — this closes it before the first one did. Off the board now means off the rewards, and the shares slide down to the commanders behind them.
• The season podium and the Destroyer follow the same rule (#824). A faction empire can no longer take a season podium row, a season share, or the unique Destroyer hull. The season is filtered separately from the leaderboard, so it needed its own change rather than inheriting one.
• The Wraiths hunt a narrower band (#824). Their standing order is to pressure the top of the board as a group and never to concentrate on whoever is cheapest — but because they filled nine of the top ten rows they were aiming at, that order had quietly collapsed onto a single commander. Their licence is now the top five, and the anti-dogpile rule that spreads pressure by longest-since-struck governs it as designed. Commanders outside the top five are out of scope, not merely deprioritised.

Fixed

• The Wraith faction keeps attacking (#824). Taking them off the standings would have stopped them dead: the operator's tooling read each empire's own strength off the leaderboard, so with no row to read every one of them would have reported an empty fleet and refused every strike, with no error anywhere. They now read their own strength from their own empire, and a reading that fails is reported as unknown instead of as zero.
• Rival commanders no longer forget who attacked them (#824). Empires that leave the standings can also be hidden behind a Veil, and an empire absent from both is treated as deleted — its grudges wiped after two days. Grudges against an announced faction are now kept regardless, so a raid is still remembered by the commanders who took it.

Unchanged

• The Wraiths stay on the Wanted board. Leaving the standings is not protection: a bounty is not a prize the target collects, it is the price others collect for going after them. Relic garrisons and faction bastions are still sized against them as well, so PvE difficulty does not move.

1.0.0-beta.110 — 2026-08-19

Fixed

• The chat tab on a phone gets its screen back (#823). A player reported it from a secondary colony he has no intention of developing: the first steps advisor sat above the conversation asking for a Laboratory he will never build there, and there was no way to move it. That was not an oversight — the chat window is sized so that its bottom lands on the navigation bar, which is what keeps the message box reachable without scrolling, and the arithmetic of that makes the page exactly one screenful tall with nothing left to scroll. Every pixel the advisor took came out of the conversation, and the chat was down to a message and a half. On the Chat tab the advisor is now a single line — how far along the chain you are, and either the reward waiting to be collected or what to do next — which opens the whole panel on a tap and puts it away on another. Everywhere else it is untouched: Base has its own copy, and Research and Shipyard are where two of the steps are actually performed.

• The advisor line says what is happening, not just the name of a step (#823). There is one build slot and buildings take minutes, so "Build the Laboratory" was what most of the tutorial looked like — including while that Laboratory was already going up, and while the slot was busy with something else, which is the one moment you cannot act on it. The line now says Under way, Build slot busy, Reward ready or Next, with the building named in each.

1.0.0-beta.109 — 2026-08-19

New

• The boards say who you are flying with (#816). Every board in the game printed bare names: the leaderboard (podium, phone list and desktop table), the wanted board, the war-week scoreboard and its paid podium, the campaign climbers and the season scoreboard. The header of the leaderboard even carried a note admitting it. The guild tag now rides next to the name on all of them, tinted with the guild's own colour — the same [TAG] spelling the chat, the mention chips, the whispers and the galactic register already used. It matters most on the wanted board, which is the one list built to be acted on: a row there is a launch decision, and until now the screen gave no way to see that the target was flying your own tag, or the tag of the guild you had just signed a truce with. Players in no guild show nothing at all, not an empty pair of brackets.

Fixed

• A long name no longer eats the tag (#816). On the four rows where a name is cut short with an ellipsis — the podium card, the phone list, the wanted row, the season row — the cut now falls on the name alone. The cap on a name is 20 characters, and with the tag inside the cut a long name swallowed it; a tag you cannot read is worth as much as no tag. The · you marker on your own row was sharing that cut too, and no longer does.

• "Janos· you" reads as "Janos · you" again (#816). Making room for the tag turned three name rows into flex rows, which silently dropped the space in front of the · you marker on your own row — on the podium card, the phone list and the season scoreboard. Found in review before it reached anyone.

Also

• The paid podium of a closed week shows the guild you were flying when the week closed, not today's, the same way it has always frozen your name and your points. Weeks closed before this release carry no tag and keep printing bare names; that resolves itself at the next weekly closure. What this release does not change: the same podium in the global Register still prints bare names, because that feed projects only rank, name and points — the tag is in the data now, only the projection is missing.

1.0.0-beta.108 — 2026-08-19

Fixed

• The Peace shield no longer promises what it cannot do (#803). A commander watched the Observatory warn them of an incoming fleet, spent 40 Scrip on the guild store's Peace shield to stop it, and lost their ships anyway — twice, and then reported it. The engine was doing what it was designed to do: the shield is checked when an attack is created, so it refuses new attacks while it is up and has never turned back a fleet already on its way. Nothing in the game said so. The store row sold "eight hours of forced truce"; the HUD banner said "none of your colonies can be attacked for another 6h"; and nothing warned the buyer that the fleet they could see coming would land regardless. Every surface now says the same true thing — the shield refuses new attacks, it does not turn back anything already inbound, and attacking anything but a Pirate Beacon ends it early. That last clause was never written anywhere a player could read it before paying. The correction reaches phones without an app update on the one surface where it matters most, the row that carries the buy button, because that text is sent by the server.
• The store says when buying will not help (#803). With attacks already on their way, the Peace shield row counts them and states plainly that a shield bought now will not turn them back, only refuse new ones. It appears only while something really is inbound — and on every layout, not just on a phone.
• The protection sheet stopped contradicting the rest of the screen (#803). Tapping the shield chip said the empire shield "covers every colony you own" and stopped there, one tap from the incoming-attack alarm. It now carries the same caveat as everything else, and no longer shows it to brand-new commanders, for whom nothing can have been aimed at them before their shelter existed. On desktop, where there is no sheet to open, the tooltip was also telling players that an enemy attacking is what drops their shields; it now says it is your own first attack.
• The shield countdown belongs to the shield it names (#803). Holding both a 72-hour exodus shield and a bought 8-hour one, the banner printed the longer clock under the empire heading — so it kept counting down for days after the empire shield had actually lapsed, and an attack on any other colony was accepted while the number on screen still said you were covered.
• Refusing a fold explains itself truthfully (#803). "A fold nobody can raid is a fold nobody can stop", said the game when you tried to fold under a shield, and "a Peace Shield now would make that fold unstoppable" when you tried to buy one mid-fold. A raid launched before the shield went up still lands and still stops the fold; both messages now speak of new raids.

Also

No engine rule changed in this release. Whether the shield should bounce fleets already in flight is a separate balance question, and it is recorded on the issue rather than decided here: nothing in the game stays in flight for more than two hours while the shield lasts eight, so a single 40-Scrip purchase would void every fleet aimed at you, joint raids included.

The commander who reported this also asked for the fleets they lost to be restored. That is an open question for the game's owner and is not answered by this release.

1.0.0-beta.107 — 2026-08-19

New

• Survey a whole system with one confirm (#469). Espionage was one probe, one trip, and the trip was long: open a planet's sheet, press Spy, land in the Missions tab, pick the probe, launch, walk back to the galaxy, and start again for the next planet. Scouting a busy system meant doing that once per planet. The system view now carries a survey: tick the planets worth a probe, read what it costs, and send one espionage mission to each in a single confirm. It is scoped to the system on screen, so the price stays legible and one tap cannot fan out across a sector. Nothing is ticked for you, the total is stated before you commit, and a sweep the home hangar cannot pay for is refused by name. Every target stays an ordinary espionage mission with its own travel time, its own report and its own refusal — a faster way to send them, not a new kind of mission — so the guild tag rides on every row, letting you skip your own guildmates before paying for the refusal. Afterwards each planet is accounted for: how many probes are away, which were refused and why (identical reasons grouped, not repeated once per planet), and which never left; the ones that never left keep their ticks, ready to send again, unless the refusal was one no retry can beat — an empty hangar, a folding home colony, a closed season — in which case the sweep stops there and says so. A request that times out is reported as unknown rather than refused, because the mission may well have gone through, and the rest are left unsent instead of being fired blind. Most useful in the Event Sector, where the whole zone is on the season clock. What it does not do: the probes always leave from your home colony (there is no origin picker, and the panel says which hangar it is counting), no travel estimate is shown per planet, and a surveyed mission does not appear in the Missions tab's "Launch again" rail — one sweep would evict a third of it.

1.0.0-beta.106 — 2026-08-19

Nothing changes in the game this time: this release is housekeeping inside the engine, plus two comments that described the rules incorrectly.

Changed

• The engine speaks English about itself (#454). Around sixty names inside the shared engine — the code that resolves battles, bounties, seasons, debris and Command Center unlocks — were still written in Italian. They are never text a player sees; they are the labels the code uses to talk about itself, and mixed-language ones make every later change to those systems harder to get right. Two engine files were renamed to match, and the design documents and the agents' Python mirrors that pointed at the old names were moved in the same commit, so no pointer dangles.

Fixed

• Two comments that stated the wrong rule (#454). One said the Cannon's opening salvo is applied before the battle report snapshots the attacking fleet; it is applied after, which is why the report shows each fleet as it arrived. The other said repelling an assault pushed an outpost's paid repair eight hours away, while the lockout has always been twelve. No behaviour changed — only the descriptions, which had been misleading anyone reading the code to learn the rules. A third comment the audit had flagged turned out to be correct already, and a neighbouring passage that looked wrong is a deliberate record of a rejected option, so both were left alone.

Also

• What was deliberately not renamed, and why. Everything the game stores — building, technology and tree-node ids, guild ranks, entry policies, planet types, ship classes, faction and relic ids — keeps its original spelling: renaming any of it would orphan live colonies. The same holds for one field the installed app reads over the network; renaming it would have blanked every building and technology name until each player updated. The audit that requested this work had listed that field as safe, which it is not. Checked rather than assumed: of the 1,177 distinct stored text values in the engine, every one is byte-for-byte unchanged.
• A latent gap found while reviewing this change is filed as #811: some checks that keep the agents' numbers in step with the engine go quiet, instead of failing, if an engine file is renamed.

1.0.0-beta.105 — 2026-08-18

New

• The campaign map dresses like its design (#804). The scene was the design's skeleton without its skin: bare numbered nodes scattered across the full desktop width, the enemy readable only inside the engagement sheet, the boss a glyph. Now the scene composes in a centered column; the current node wears the art of the hull fronting the level's garrison with the level as a badge; the chapter boss waits as a dimmed silhouette under a BOSS tag — kept even while you stand on it; the cleared stretch of the path glows solid. The forecast (the same three simulations the sheet runs) shows as a strip under the path with the enemy value, marked as the estimate it is. Each arc carries its tactical hint, taken from the engine's own composition notes, and its own horizon tint; a strip counts the levels to the boss and names the two part kinds it actually drops — not "the full parts table", which no boss ever paid. The deepest climbers tick along the scene's foot with your own depth closing the line, and the parts wallet sits in the header once your first drop lands.

1.0.0-beta.104 — 2026-08-18

Changed

• Extractors dig faster, Compute Nodes run leaner (#779). At the same level, a Ferrolite Extractor now yields about 8% more (level 10: 3,112 → 3,372/h) and a Compute Node about 25% less (1,037 → 778/h). Xenogas Refineries are unchanged, and so is the total a level of production is worth — contract payouts, expedition hauls and every "per hour" threshold keep their scale, paid along the new mix. Why: ferrolite is what ships and buildings are actually made of, and the galaxy was producing it in the wrong proportion — 0 of the 103 open market offers were selling ferrolite while 86 were asking for it, a market structurally unable to clear. Nothing was taken from anyone: no stock, no level, no cost and no score moved — points and Power price what things cost, and no cost changed. Production accumulated offline across the update is credited at the new rates for the whole window. What this release does not fix: the data already banked across the galaxy stays banked — a dedicated outlet for accumulated gas and data is the follow-up. On app builds older than this release the per-hour figures on the colony screens read the old values until the app updates; the resources themselves are computed server-side and are correct.

1.0.0-beta.103 — 2026-08-18

New

• Auto battle on the Front (#796). Past level 40 the campaign used to be the same gesture repeated: open the sheet, engage, read the outcome, open it again — for levels the forecast already called. Now an Auto button sits next to Engage once level 40 is cleared (below that it shows its unlock, Auto · Lv 40): the climb runs hands-free with the saved deployment, banks each first clear as it lands, logs the run on a panel, and stops on its own at the first battle it cannot improve — a loss, a draw, or a clear that banked nothing. Campaign battles are deterministic, so the stop is honest: the same orders would fight the same battle again; change your deployment, a refit or your research, then go again. "Stop after this level" hands back the wheel mid-run; leaving the screen stands the run down after the level in flight and keeps the summary for your return. The loop paces itself above the server's own 2-second brake, so hands-free never means faster than allowed.

1.0.0-beta.102 — 2026-08-18

Changed

• A garrison can be called home from any colony, not just your own (#788). Sending a defence to an ally used to commit the fleet for the whole watch — up to twenty-four hours — with no way out for anybody: not for you, and not for the ally either, since the release control that was supposed to be theirs was never built. A defence pledged to a friend was a fleet out of the game for a day, and the rational answer was to garrison nobody, which is the opposite of what joint defence exists for. Stand down now reaches every garrison of yours, wherever it stands: your colony, an ally's, or one razed out from under the watch (that last case used to be refused with the same "not your colony", stranding the ships with nobody left who could free them). The one refusal that remains is a hostile fleet already on final approach there — you do not walk a garrison out of a battle that is landing, and letting it happen would hand the owner a full fleet and a report of losses it never took. What has not changed: the flight home is paid in full, so recallable is not the same as present; and the hangar you empty at home is still the real cost.

New

• Your ally is told when you call your ships back (#788). They already get a line when the fleet lands, with an end time they can plan around. Now that a guest can end that cover at any moment, they get its counterpart — naming who left and how much of the watch was still to run, because a fleet pulled eleven hours early and one whose watch had four minutes left are not the same act and should not read as the same line.

Fixed

• The screens that recruit garrisons stopped saying the fleet was committed (#788). The Defend form, the guild "Allies under attack" panel and the galaxy tooltip all described the old rule; the panel is the one screen whose whole purpose is getting somebody to garrison an ally, and it was still promising a commitment the game no longer asks for.

1.0.0-beta.101 — 2026-08-18

New

• Launch again: the repeat rail (#775). Four launches out of five repeat a target already hit, and re-sending one meant walking the whole form again. The Missions tab now opens with the recent launches as cards — verb, target, the exact fleet, the battle line — and one tap re-sends, with a 4-second Undo before anything leaves. Attacks confirm once more first: creating an attack forfeits every shield the attacker holds (only the Pirate Beacon is exempt), and the confirm says exactly that. A hangar that came up short prints the refit (Corvette 300 → 120) before the tap; a target whose launch colony was lost blocks instead of silently relaunching from home.

• The War Table (#775). Which ship, how many, and where it stands were three controls in three places — a card list, a ± stepper that moved by 10% of the hangar, and a formation panel folded shut below everything. They are now one canvas: ships are tiles sitting in the battle row they will fight on; tapping one opens an inline editor with absolute count chips (None · 50 · 100 · 500 · Half · All), a ± that moves by exactly 1, and a Stands row whose labels carry the resulting fire shares live from the engine. Hulls the verb cannot use wait behind a divider with the reason printed on the chip; the totals footer shows effective attack, hull and hold with research applied, plus the fleet's pace-setter.

• Verbs that say no where the choice is made (#752). A verb the empire cannot mount — no Spy Probe anywhere, no Recycler, a single colony — is dimmed with its one-line reason on the list itself, phone and desktop alike. The list never re-sorts and nothing hides: tapping through lands on the composer with the same sentence as the top verdict line.

Changed

• One ranked verdict line replaces the flat warning stack above the launch button. The single highest-ranked issue shows (blockers, then real costs like the bounty and the shield forfeit); the rest folds behind «n notes», with the boilerplate server line always last. On the phone the top line rides the launch bar, so a disabled button carries its reason.
• Garrison rows tell the truth. Defending an own walled colony draws the batteries and fortress holding the Front line and prices the rows behind them; a colony with no guns gets no imaginary wall — a rear-only garrison honestly reads 100% of incoming fire — and on an ally's colony every share reads ≤, because their line cannot be read from outside.

Fixed

• A launch started and then navigated away from now actually launches, and the card shows «Launched moments ago» on return — the double-wave accident is closed. A refused launch stays on the card, dismissable, instead of vanishing with an 8-second banner. Backspacing the ship-count field no longer deletes the ship being edited.

Also

• The design record — five candidate designs across two generations, six judge reports, two adversarial passes and the surviving rules — is in design/mission-redesign/DESIGN.md. Left for the next slices, on purpose: the Board shell that retires the sticky bars (#731 stays open), first-launch cards for empires with no history, and the Load sheet that unifies fleet templates with the launch ledger.

1.0.0-beta.100 — 2026-08-18

New

• The launch preview tells you when your fleet comes home, not just when it arrives (#778). Transport and Redeploy showed the exact same travel time, even though a transport's ships fly there and back while a redeployed fleet stays where it lands — so the preview invited the reading that the two missions cost the same. On Transport and Debris harvest missions the summary line now shows both numbers: «travel …» for the arrival and «fleet back …» for the moment the freighters are usable again. The return figure follows the real flight rules — hyperlane and relay discounts count on the way home too, and a warp-boosted trip returns at normal speed. Redeploy and Colonization keep their single line: one leg is their whole story. Reported by Spaccamazza.

1.0.0-beta.99 — 2026-08-18

Changed

• The attack-notifications switch has moved into the menu (#786). On a phone the switch shipped by #721 lived in the HUD, and the HUD floats over the galaxy map — so "Attack notifications · On" printed straight across the map's filter chips, covering Politics and My colonies. It now lives where the other settings live: Menu → Command, in the strip at the bottom, under night protection. Nothing about the switch itself changes — On/Off state and the "turn them on in iOS Settings" note for declined devices carry over. On iPad the switch stays in the HUD beside the night-protection setting, where the layout has room for it.

1.0.0-beta.98 — 2026-08-18

Fixed

• Two named places on the same star no longer print on top of each other (#780). System 8:2 holds both The Exarch Archive and The Exarch Guard, and the galaxy map drew every named place at the same fixed distance from its star. So the two names landed on exactly the same pixels, letter through letter, and what a player saw was one red title made of scrambled characters with an illegible grey line under it. The same thing happened to their two effect lines and to their two signs — which is why the star looked like it carried a single corrupted label rather than two separate places.

  Each place now takes its own side of the star: the relic's sign, name and effect hang to the left, the bastion's to the right. Both names are readable, and each keeps the colour that tells a relic apart from a faction bastion. Every star that hosts a single place is drawn exactly as before, to the pixel.

  What this does not fix, and #787 records with the measurements: the two names end up on the same row with very little air between them, so at a glance the pair can still read as one long line, and at the wider zoom levels their inner edges sit close to the neighbouring star's column. Giving the pair real separation means moving the labels themselves rather than their anchors, which is a different change; the one shipped here is the one that makes the text legible without pushing anything off the top of a phone screen.

1.0.0-beta.97 — 2026-08-18

Changed

• A garrison's watch can now be as short as 0 hours (#768). The slider on a Defend launch started at one hour, and that hour was the price of every misfired defence — wrong coordinates, or a battle that resolved while the fleet was still in flight. On somebody else's colony there is no way to cut a watch short, so it was paid in full. What 0h actually buys is narrower than it sounds, and the launch form now says so instead of leaving it to be discovered: a zero-hour garrison lands and turns straight round, and only fights if the attacker is already committed — the last quarter of their flight, never more than ten minutes before impact and sometimes as little as one, a number the defender cannot see from their side. So it answers a strike coming down in the next moment; for anything you can still watch approaching, an hour or more is the right answer. Two things a short watch does not change: the fleet still flies the whole way home before the ships are yours again, and if it does fight it stays until the fighting stops.

Fixed

• An ally is no longer told about help that is already leaving (#768). "Reinforcements arrived" appeared the instant any garrison landed. For a watch too short to stand, that promise was void before it could be read — and the host plans around it: leaves their own fleet out, does not buy the shield. They are now told only when the fleet is really going to be there, and when it lands into a battle the message says that, instead of promising cover that ends with the fight.
• Your own log stopped reading "standing guard" over a fleet that is not (#768). A very short watch now says what it is, in the heading as well as the line beneath it.
• The watch slider has a name (#768). It announced itself to a screen reader as a bare number, and at the new floor "zero" with no unit reads as off. It now says what the control is and what the value means, and points at the paragraph that explains it.

1.0.0-beta.96 — 2026-08-18

Fixed

• A run of messages from one person now carries one header, not one per message (#765). Grouping required the same author and less than three minutes between messages, and real traffic in this game is nothing like that fast: on the production guild channel six consecutive messages from one player spanned 10:03 to 12:38, so the rule never fired and the list printed six full headers — portrait, name, crest, tag, timestamp — for a single speaker. Grouping now follows the author. Every pause of three minutes or more still shows its time, on a slim line of its own, so the change costs no information: it is the same boundary the old rule used to draw a whole header at.

• The guild tag no longer repeats on every line of the guild channel (#765). Everyone in that room is in that guild; the tag said nothing and took width from the name of whoever was speaking. It still appears in the galactic channel and in whispers, where it distinguishes two strangers. The crest goes with it, on the Chat tab and on both Guild-council panels — and a member who has since left for another guild no longer shows a foreign tag and crest in the council history.

• Two glyphs were rendering as colour emoji on iOS (#765). The reply-quote arrow (U+21A9) and the whispers envelope (U+2709) have no glyph in the text fonts iOS falls back through, so the system reached for Apple Color Emoji: a blue rounded square and a white envelope, off the game's palette, and the arrow tall enough to inflate the quote row. Both are drawn now. The same envelope in the desktop chat dock is fixed too, so the two surfaces stop disagreeing.

• The channel bar no longer reflows under your thumb (#765). It was sized to its content and allowed to wrap, so on a phone it took two rows with "Galactic channel" alone on the first. Worse, the unread dot was mounted when news arrived, widening that seat by 13px — enough to push its neighbour onto another line mid-read. It is a single row of three segments now, and the dot's space is always reserved; only its visibility changes. The whispers counter is reserved the same way. Off a phone the full names stay, since they fit there.

• The message box is one field again (#765). It was a bordered card containing a bordered input beside a bordered button. Now it is a single rounded field with a 44px round Send.

Known

• Opening the keyboard still scrolls the page so the HUD slides under the status bar, and with the keyboard up the first message line is printed under the Dynamic Island. It is not a chat bug — it affects every screen with a text field — and four attempts did not close it, so nothing partial was shipped. Diagnosis and dead ends are in #766.

Also

• The accessible name of a channel seat now carries its unread state. An aria-label on an element ends the accessible-name computation, so the labels on the dot and on the whispers counter were unreachable and the seat sounded identical with news and without.
• A new test mounts the real chat panel — the first thing in this area to prove the panel draws the layout rather than merely computing it. Every guard added across three review rounds was checked by mutation rather than assumed; two assertions that turned out to be vacuous were found and replaced, one of them passing because a colour value happened to contain the digits it was looking for.
• Review notes on the minor findings consciously left as-is are on the PR (#767).

1.0.0-beta.95 — 2026-08-18

Fixed

• Guild leaders no longer slip Italian words into guild chat decisions (#656). Every guild decision an AI leader takes — applications, promotions and kicks, the research priority after a council, Fund grants, Eschaton commitments, the Anchorage, reactions, welcomes, the aid board — went through prompts whose JSON schemas asked the model to answer with Italian field names, and the research prompt offered the raw persisted branch ids (guerra|economia|scienza) as the reply enum: it did not merely show the model an Italian word, it required one back. One Italian word in a prompt is enough to tip a multilingual model into replying in Italian, and these replies land verbatim in guild chat. All ten decision prompts now speak English end to end: English keys on both sides of the wire (schema, parser, fallbacks and consumers moved atomically), research branches offered by their English labels (War / Economy / Science) and mapped back to the persisted ids before the API pin, the plan stance translated ("aggressive", not "aggressiva"). Nothing stored was renamed.

• A leader's research announcement can no longer contradict the pin. When the model named a branch the guild could not pin (one already maxed, or a spelling the mapper did not accept), the leader posted the line written for that branch while the server pinned a different one — "more guns" in chat, Economy on the board. The announcement now follows the decision: if the choice had to be corrected and the model had spoken, the leader closes the council with a line naming the branch actually pinned; a leader with no model keeps pinning silently as always.

Also

• The demote and kick halves of the ranks decision and the aid-board note had no behavioural coverage on either side of the wire — a leftover Italian key there would have no-opped silently forever. Each now has brain-level and runner-level tests, all verified red under the corresponding mutation, and a sweep test captures every guild-decision prompt and rejects any Italian key, rank, branch id or stance reaching one.
• Review notes on the minor findings consciously left as-is are on the PR (#772).

1.0.0-beta.94 — 2026-08-18

Added

• An empire signed in with Apple can add Google, and be played in a browser (#760). Sign in with Apple exists only in the iOS app, by choice — there is no Apple button in a browser and there will not be one. The consequence was that an empire created with Apple could be reached only from the phone: there is no address to sign in with anywhere else, and the commander identity dialog answered "this account is already linked" to precisely the player who most needed another way in.

  That dialog now offers Link Google account to an Apple account, and says why. Taking it opens the browser without closing the phone: the Apple sign-in keeps working exactly as before, because linking here adds a way in rather than moving the one that exists. Re-pointing the account would have looked identical from inside the dialog and stranded the player in a new, empty empire the next time they tapped Sign in with Apple.

  Afterwards the Account section names both doors and shows which Google account was linked — worth seeing, since linking cannot be undone.

Changed

• The guest upgrade is untouched and still picks one account: a guest's synthetic address is a placeholder, and leaving it behind would leave it claimable. The wording there no longer promises a second way in later, because for a guest who picks Google there is none.
• Several refusals were saying the wrong thing to the wrong player — an Apple account being told its empire was a guest that logging out would destroy, and a "close and reopen this dialog" instruction that reloaded nothing. Both now describe the situation the player is actually in.

Also

• Nothing here is visible to anyone who has not signed in with Apple, which today is everyone: Sign in with Apple reached TestFlight only in the previous build, and no account has used it yet. Shipping this now is deliberate — every day it waits is another empire that could have been created app-only.
• The database gains one table and nothing else: no column altered, no row touched, no backfill. Every existing empire keeps its identity exactly where it already was.

1.0.0-beta.93 — 2026-08-18

Fixed

• A guild goes to the officer who earned it, whichever way its leader leaves (#757). Leaving a guild already handed command to the officer who had contributed the most, falling back to the most senior member — exactly what the guild sheet promises. Deleting the account handed it to whoever happened to register first: officers ignored, rank ignored, time in the guild ignored, so a recruit still on probation could inherit ahead of the guild's officers. A leader could read the promise on one screen and leave by the other door, and the guild woke up commanded by the oldest registration. The same held when a moderator closed an account, because all of those doors run the same deletion. They now all apply the one rule the game describes, and the deletion panel — which had honestly printed "whichever remaining member registered first", because that is what the code did — describes it again.

• The guild is told when command changes hands, and the new leader wears the rank. An account closing used to move the seat silently: no line in the guild log, and the heir's row still saying officer, so every permission that reads the stored rank kept treating the new leader as one. The succession is written to the log now, and the rank moves with the seat.

• The line says what happened without saying more than it should. A player who deletes their own account is named as having done so. An account closed any other way is reported only as closed: the guild is told the seat moved, not why an account ended. The same door also closes accounts when their owner asks support to, so anything more would sometimes have been false — and the log is read by the whole guild, for as long as it exists.

• The absent-leader succession reads seniority like the other doors. On guilds formed before the join date was recorded, that door alone ranked equal members by their internal id, so the same guild could get two different heirs depending on whether its leader closed their account or simply went quiet.

• Closing an account can no longer take a guild down with it by accident. A guild is disbanded only when nobody is left in it. If members leave in the same instant the leader's account is being closed, the succession is retried and, failing that, the seat is simply left to be repaired on the next guild read — never resolved by deleting a guild that still has people in it. (One narrow case survives, where the last members leave in that same instant and the empty guild is left standing: #770.)

1.0.0-beta.92 — 2026-08-18

Changed

• Sending a fleet from a phone is three short screens instead of one long scroll (#749). The New mission form ran well past two phone screens — measured at 390 px, 2,154 px for an attack and 3,419 px for a defence — and the launch button stayed disabled for the whole trip down.

  It is now a sequence. What the mission is for: the eight types are full-width rows with a readable sentence each, instead of eight cards whose descriptions were clipped mid-word ("Break a fleet, tak…", "One probe, full s…") — the words that explain the choice were not on the screen where the choice was made. The three verbs that carry most of the traffic sit at the top. Where it goes: the places the player has already flown to come first, and the sector/system/planet boxes fold away beneath them; most fleets repeat a destination, so typing coordinates had been the first thing on the screen for the rarest case. The fleet: what was chosen stays at the top as two chips that lead back to change it, and the target's owner, Power, shield and the age of the last probe report come with it — sizing an armada against a defender you cannot see was the one thing the single scroll did get right.

  On a desktop nothing changes: it has the room to show the whole form at once, and it still does.

Fixed

• Harvesting the debris above your own colony works from the phone. The wreck is in orbit, not on the ground, so the engine has always allowed it; the first draft of the new flow refused it and told the player the field was "the colony this fleet would leave from", which is not where it is.

• The target panel no longer offers to send a probe at your own colony. Tapping it turned the mission into an espionage run and handed the composed fleet back to the advisor — on a panel that had invited the tap.

Also

• The web tests can now open a screen and press its buttons: apps/web had no DOM, so its wiring could only be checked by looking for strings in the source, and four review rounds on the previous release each found a way to break a control while every test stayed green. This release is the first built on that harness, and it earned itself — the flow above went through four rounds of review that turned up two dead ends and a mission type that had quietly become unreachable, none of which a string search could see.

1.0.0-beta.91 — 2026-08-18

Changed

• The market can no longer deliver you an unlimited amount in one day (#584). Publishing an offer needed a Trade Portal and was capped in size; accepting one needed nothing at all, and the slots freed up the moment a trade settled. So two accounts could move any amount of value between themselves, in either direction, all day — nobody was doing it, the gift traffic on the board is honest, but nothing bounded it. Now, per resource and per UTC day, the market can bring you at most twice what your Warehouse holds.

  It is checked on whoever receives, which means it binds both sides: the seller receives what they asked for, so flipping who publishes the offer resets nothing. The three resources have three separate allowances — a ferrolite day that is used up leaves xenogas and data untouched — and there is no cross-resource valuation anywhere in it. An honest trader never meets the limit; a funnel does, and it stops growing when more accounts are added, because the allowance belongs to the destination rather than to the relationship between two players.

  The board says so before you press anything. Offers your day — or the seller's — can no longer absorb are dimmed and carry the reason, instead of failing after you tap Accept. Your own offers carry it too, so a seller is never left wondering why nobody is buying. The two cases are told apart on purpose: a spent day says your allowance returns at midnight UTC, while a single trade larger than a whole day's intake says that instead, because waiting cannot help it and only a bigger Warehouse can. And an offer asking for more than a full day of your own intake is refused as you publish it, rather than sitting on the board unable to ever complete.

  Nothing about the offers already on the board was rewritten: they keep their published price and their stamped fee, and accepting still requires no Trade Portal of your own.

1.0.0-beta.90 — 2026-08-18

Added

• Your phone can tell you a fleet is coming (#721). An attack on a persistent 4X arrives while you are not looking, and nothing could reach you: the game has an early-warning panel and a battle report, and both required already being in the app. Guest play needs no email, so a player who drifts away was unreachable by construction.

  The iOS app can now notify you when a fleet launches towards one of your colonies — "A fleet is heading for Lantea. It arrives in about 3 hours." — and when a battle there is over. The warning fires at launch rather than at impact on purpose: that is the only moment you can still do something about it, whether that is a shield, a recall, or answering the alarm with a garrison straight from the card.

  Quiet by design. Never while you are in the app; never inside the eight-hour night window you already chose for night protection; never more than four a day; and a raid of a dozen fleets is one alert rather than twelve. The alert carries only what the base alarm already gives every player — which colony, and when. The attacker's name and the size of their fleet are still bought with the Observatory, and a notification is graded by nothing, so it never carries them.

  iOS lets an app ask for notification permission exactly once, and a refusal is permanent. So the game never asks out of the blue: a card appears while an incoming fleet or one of your own battles is on screen, and only a tap on it opens the system dialog. "Not now" costs nothing and the card comes back later. Notifications can be turned off again from the HUD, and off stays off.

  Nothing is delivered yet. Sending needs a push key that only Apple can issue: until it is on the server the game registers your device and stays silent. Everything above is live the moment that key is added, with no further release.

• Deleting your account takes your registered devices with it (#720), so nothing can reach a phone for an account that no longer exists.

1.0.0-beta.89 — 2026-08-18

Fixed

• The guild store ran off the side of the phone (#751). Ever since Colony Displacement took its place in the Treasury store, the whole sheet was laid out wider than the screen: every row was cut off at the right edge in the middle of a sentence — «· 0/3 this we…» — and the rows that lost their text were mostly the plain ones, a name and a price, which contain nothing that could have caused it. They were being dragged along by the row beside them. The store now fits the screen, its descriptions wrap instead of running past the edge, and the fold form's destination fields sit inside the sheet.

  The cause was the three destination fields. A text field nobody has given a width to is twenty characters wide, and on a phone every field is enlarged for legibility, so the three of them together asked for more than a phone screen has — and that request travelled up the page, widening the entire store to fit them. Two things changed: the field row no longer passes that demand upward, and the store list is now capped at the width of the screen, so a single row that overruns can only overrun itself instead of re-laying every other row in the shop. The second half is the one that matters next time: the store is a list that grows.

  Reported by Ajax with a screenshot precise enough to identify the row that caused it.

1.0.0-beta.88 — 2026-08-18

Fixed

• The invite box opened as a directory of every commander in reach (#691). Opening a guild's Invite someone panel listed commanders the player had never asked about — ten of them, alphabetical, so the list began at whoever had named themselves 112 — each row carrying its own Invite button. The route behind it answers an empty query by matching everyone, and the panel sent exactly that on opening; the copy written for "you haven't typed anything yet" existed and was unreachable, because the empty query is the one that comes back full. Now nothing is asked for until a name is typed, and inviting is once again an act aimed at someone you had in mind.

Changed

• The panel no longer rules on a search it has not made. While an answer is travelling it says Searching…; only an answer that came back empty says Nobody by that name. Before, that sentence appeared before the request had even left and stayed for the whole round trip — and to whoever had just spoken in chat with that person, the game denying they exist is where they stop trying. A failed search says it failed, and retyping the same name visibly retries instead of repainting the error being retried.
• A search that found people says how many, and a list that came back full adds There may be more: keep typing to narrow the search. Ten results can mean "ten people match" or "ten is all the server sends", and with typing now the only way in, a list that looks complete is how a player concludes a friend is not in the game. At the cap the count says found so far, so it does not assert a number the line under it retracts.
• The rows on screen always answer the name in the box. While a new search runs there are no rows rather than the previous ones: a list that swaps under a finger sends the invitation to whoever inherits that row, and an invitation cannot be withdrawn.
• The panel's status sentence is a live region, so a screen reader is told what happened — it had no way to say so before, and removing the always-present list took away the only thing that was announceable.

Also

• This is a change to the game client only. The route still answers an empty query by matching everyone, because the chat @-picker depends on that, so an installed app or a tab that has not reloaded keeps the old panel until it updates.
• Receiving an invitation is untouched, and a guild set to Invite only is still enterable.
• Tagged after beta.86 and beta.87, which were merged minutes after this one and therefore already carried the fix; the compare link below is against beta.85 for that reason.
• Eleven known limitations found in review — among them that the server searches only the first 300 players by name, which can make Nobody by that name. untrue on a large shard — are listed on the pull request rather than fixed here.

1.0.0-beta.87 — 2026-08-18

Added

• Delete your account, from inside the game (#720). An account took three taps to create and could not be removed at all: the only deletion in the game was the moderator's, on the admin panel, and a player who wanted to leave had nothing to press. It is also why the iOS app could not be submitted — the App Store requires in-app deletion wherever there is in-app sign-up.

  The commander card — the portrait in the corner, on phone and on desktop — and the Command sheet next to log out now carry Delete account. Opening it deletes nothing: it lists what is about to go (colonies and everything built on them, fleets, research and Dark Matter, market offers, contracts and chat, guild membership) and asks for the commander's own name. The red button does nothing until something is typed, and the name is checked again on the server, so a mis-wired button or a stray request cannot delete an empire on its own.

  It also says what survives: battle reports held by the players you fought, and guild log entries, both of which carry the commander name and are somebody else's history too. A list of losses that reads as exhaustive would be a promise the game does not keep.

  Leaving is final, and the panel is honest about the one case where the outcome is unknown: if the reply is lost on a bad connection, pressing the button again is safe — the server answers the second attempt with "already gone" and signs the player out, rather than deleting twice.

Changed

• The departure card on the admin dashboard now draws three causes instead of two: accounts the sweep removed because they never started, accounts moderation removed, and — new — people who deleted their own account. Summed into the moderation band, somebody choosing to leave would have read as an unusually busy week for the admin. The third band starts at zero before today, and that zero is the feature not existing rather than nobody using it.

Fixed

• An account that was under attack could not be deleted cleanly: the fleets flying at it kept naming a player who no longer existed, and the attacker's recall then failed with a server error on every retry until the fleet landed on its own. The fleets still fly and still land; they simply no longer name a defender.

1.0.0-beta.86 — 2026-08-18

Added

• Sign in with Apple (#719). The login screen offered exactly one account to sign in with, and it was Google. Anyone who did not want to hand Google an account was left with the guest door, whose empire lives and dies with the data on one device — and the app could not be published at all, because the App Store requires a privacy-preserving equivalent beside any third-party sign-in.

  Both screens that offer an account — the login page and the commander identity dialog — now show Sign in with Apple above Continue with Google, at the same size. In the app it is the system sheet; in a browser it is Apple's own popup. A guest can upgrade to an Apple account exactly as they could to a Google one, keeping the commander, the colonies and the alliance.

  Apple never has to hand over an address for this to work, and that shaped the whole design: whoever picks Hide My Email gets a relay alias, and from the second sign-in on the address can be missing altogether. The empire is therefore tied to the identity Apple vouches for and never to an e-mail. The name Apple offers on the first sign-in — the only time it ever offers one — becomes the commander name, and can be changed in game like any other.

Changed

• The account section of the identity dialog names the way you sign in instead of assuming it is Google. An Apple account is told so, rather than being shown an address it could not sign in with; and it is not promised it "follows you to any device", which for Apple depends on configuration this release does not include.
• The privacy page says what Sign in with Apple involves: what is stored, what is not, and that the first name Apple gives us becomes the public commander name.

Also

• Nothing here is visible to anyone until the Apple side is configured. With no Apple audience set on the server, no Apple button is drawn anywhere and the endpoint answers "disabled" — which is what it does on novarchy.com today. The remaining steps are Apple's own console, and they are written down in apps/web/MOBILE.md and deploy/DEPLOY.md.
• This release also carries the guild invite work merged separately just before it, which had not yet reached production. It is not described here: it belongs to whoever wrote it.

1.0.0-beta.85 — 2026-08-18

Fixed

• The inbound attack alarm can be answered from the alarm (#732). The panel that tells you a hostile fleet is landing on your colony in thirty minutes had no controls on it at all. To respond you left the alarm, switched to the New mission segment, picked Defend out of eight cards, and typed back in the coordinates of your own colony — the ones the alarm was displaying two lines further up. A garrison is the only mission in the game with a deadline attached to it, and it was the only one you could not start from the screen that knows about the deadline.

  Each alarm card now carries Send a garrison. One tap opens the launch form as a defence, already aimed at that colony, so the only thing left to decide is the fleet.

Also

• When a garrison cannot leave from the colony you are currently launching from — because it is the colony under attack, or because its hangar holds nothing that can fight — the button says so and points at the Launch from row, instead of quietly disappearing. If no colony of yours could mount a garrison at all, the card stays quiet rather than offering something it cannot deliver: for most players the only colony they own is the one being attacked, and a button they can never press is worse than none under a running countdown.

• Behind the scenes, the web tests can now open a screen and click it. Until this release nothing in apps/web had a DOM, so its wiring could only be checked by looking for strings in the source — a guard that pins one connection and leaves the next one loose. Four review rounds on this change each found a way to break the button while every test stayed green: a panel handing down a dead handler, a button hidden behind an observatory level, a value passed as null that removed the control for everybody. None of it was visible to a string search one element away. The new test mounts the real screen and presses the real button, and it caught a genuine bug the moment it existed: the check for "can this colony mount a garrison" was using the rule that sizes attacks, so a colony full of freighters — which fight as hull in a defence, and which the advisor would gladly send — was told it had nothing to send.

1.0.0-beta.84 — 2026-08-18

Fixed

• The top resource strip no longer breaks your numbers apart (#556). On a grown empire the band had started tearing itself up: the energy balance split across three lines as ⚡ 16349 / then 16384, Power's S · M · Gate breakdown stacked three rows deep, the PP and DM units fell underneath their own figures, and the protection clock was cut clean off the right edge of the window — on a strip that does not scroll, so there was no way to go and look at it.

  Nothing was wrong with the numbers. The row squeezed every tile until the text had to wrap, and it wrapped wherever a space happened to fall. Now the tiles keep their width: when they no longer all fit, a whole tile moves to the next line instead, and the queue chips and the protection clock move together as one group rather than being split between rows. Every figure stays in one piece and nothing ends up off screen.

• The band is shorter than it was, not taller. Measured from 823px to 1920px wide, it lost between 13px and 66px of height while gaining back everything that used to be clipped — narrow and zoomed-in windows benefit most. Nothing changes on a phone: the strip there already scrolled sideways under your thumb, and it still does.

  This does not close #556 entirely. On desktop it is still the night-protection selector that moves down a line first when the band runs out of room, rather than anything being dropped from the strip; that decision is untouched here and the issue stays open for it.

1.0.0-beta.83 — 2026-08-18

Added

• Inventory: one screen for what you hold (#739). Everything on this tab already existed and had nowhere to live. Refit parts and your standing orders were behind the Campaign tab. Chrono-shards were only visible from a build queue. Dark Matter was in the header, relics in the guild panel, and a wrecked Destroyer nowhere at all — you could own three standing orders and an event hull without ever seeing them listed. Stores now shows all of it with item art instead of a column of figures, and every section keeps saying where its own holdings come from even at zero, so a mechanic can be discovered before it has ever been triggered. Refit is the desk that turns parts into standing orders: it MOVED here from Campaign rather than being copied, so there is one place a slot gets filled, and Campaign links straight to it the moment you win your first parts.

Fixed

• A wrecked Destroyer is no longer invisible (#739). The legendary is wrecked, never destroyed — it comes back from the dry dock. Before, the only question asked was "do you own one", so the screen said nothing at all about a hull sitting in repair. It is listed now, with how many are waiting, and alongside a flying one if you have both.

• Dark Matter is shown to the cent (#739). It could be floored to 0 on one surface while the header read 0.87 for the same holding — and the figure served is now the same live stock the header computes, not the last swept value.

• A Dark Matter debt is a holding, not an absence (#739). Being hunted into debt no longer reads as owning nothing; it is listed as an amount owed, with the same explanation every other Dark Matter surface gives.

Changed

• Chrono-shards are counted as banked time, not as pieces (#739). That is the unit the weekly earning limit is measured in, so a stack of shards and the limit can finally be read against each other — and the screen says plainly that the limit is on what you EARN, not on what you keep, and that a bastion's prize is exempt from it entirely.

1.0.0-beta.82 — 2026-08-18

Changed

• The Campaign wait is gone — the battle resolves the moment you engage (#733). Before, tapping Engage sent the strike group on a 20-60 minute round trip: battle at the halfway mark, rewards at the end, one countdown after another. Nothing actually flies in the roster model — the strike group is virtual — so the clock went with the fiction: you tap, the battle resolves inside the request, and the outcome card (cleared or repelled, what was banked, the replay) is on screen at once. Runs still flying from the previous version land under the old rules, ships home as survivors, within the hour.

• Retries are free, instant — and honest (#733). Losing a level costs nothing and you can go again immediately. But the same orders fight the same battle: a retry replays your exact fight instead of rolling new dice — change your deployment, a refit or your research to change the result. The level is a puzzle, not a slot machine, and the engage sheet, the forecast caption and the defeat panel all say so now. (Under the hood each level keeps ONE run row per commander, reused across retries, instead of minting a row per tap.)

• A per-commander breather on the engage button (#733). Two seconds between engagements — invisible to normal play, it keeps tap loops and flaky-connection retries from stacking battles on the server. The agents' commanders play the new rhythm like anyone else: short sittings of a few levels with human pauses between taps, then a real rest.

1.0.0-beta.81 — 2026-08-18

The legendary Destroyer stops being a ship you cannot afford to use. (#718)

Added

• A defeat wrecks the Destroyer instead of destroying her. She is granted once per empire and never again, so losing her used to end the story — and the rational way to own one was to leave her parked. Now she leaves your fleet the moment she goes down, and comes back for 9,000 ferrolite · 4,500 xenogas · 700 data and twelve hours in a dry dock. The panel sits at the top of the Shipyard and is invisible to everyone who has no hull to repair.
• The wreck belongs to the empire, not to a colony. She can go down half a galaxy from home; whichever colony pays the bill is where she berths twelve hours later, and the panel says so before the resources leave.

Changed

• Nothing shortens the twelve hours — not a Shipyard or Nanoforge level completing mid-repair, not a Chrono shard, not a guild push, not the Suspended Yard relic. The wait is what losing her costs.
• A wrecked Destroyer leaves no debris. Her hull is towed home rather than scrapped, so it can no longer be harvested off the battlefield: that would have been the same ship sold twice, once as salvage and once as the unit that comes back.
• She is no easier to kill. The Spectral Bulwark is untouched — nothing solo can destroy her, and it still takes two enemy commanders with ships on the field. Only what follows the kill has changed.

Fixed

• The end-of-season sweep of the Event Sector no longer destroys a repair already paid for. Pending hulls now travel to the mother colony together with the ships; with no mother colony left, they are returned as wrecks rather than deleted with the bill settled.
• Losing the Destroyer can no longer earn you a second one. The "one per empire, forever" rule now also counts hulls sitting in the dry dock, out on the campaign ladder, and parked inside an exodus offer — three places it could not see before. Two Destroyers were never a double prize, either: a pair loses the Bulwark's protection and both become killable.

Also

• The Activity feed says when the hull is crippled, filed under Combat next to the battle that did it, and again when the repair starts.
• The rollback runbook (deploy/DEPLOY.md) gained the block for this release, including the step that is easy to get wrong: a repair must have matured before forcing the resolve that makes it survive a rollback.

1.0.0-beta.80 — 2026-08-18

Nothing changes in the game this time. This one is about not losing it.

Changed

• The database is backed up every 4 hours instead of once a night, and kept for 30 days instead of 14. A failure used to be able to cost up to a day of play; now it costs at most four hours. Roughly 186 backups are kept at any time, about 3.4 GB.
• A backup is only kept once it has been checked end to end. The old command wrote its file whether or not the dump had succeeded, so a failure left a truncated file that would have looked like a good backup until the day someone needed it. A dump now goes to a temporary file and is only published after the dump exited cleanly, the archive verified and its size checked — a failure leaves nothing behind. Backups also carry every player's email address and every guild message, so they are no longer readable by anyone else on the server.

Fixed

• The written recovery procedure did not recover anything. It piped a backup straight into the live database, where every table already exists, so nothing was restored — and because the tool does not stop on errors by default, it reported success. Measured against a copy of the live galaxy: 223 errors, the same 884 colonies before and after, and an exit code that said it had worked. The procedure now empties and restores in a single step that either completes or changes nothing at all, and it was tested by damaging a copy of the galaxy and recovering it.

1.0.0-beta.79 — 2026-08-18

Added

• Colony Displacement: a colony can move (#722). Where a colony was born was where it stayed forever. The only thing that ever changed a colony's address was losing it — the Forced Exodus, after three lost sieges. So a player whose guild lived in sector 2 while free placement had put them in sector 7 had no way to close that distance, and a veteran who had outgrown the starting sectors had no reason to leave them.

  Now the guild store sells a Displacement charge: 300 Scrip, one a week. You name a colony and a free planet, and the colony moves there with everything it had — buildings, queues, resources, and its age.

  It does not teleport. For four hours it folds: it stays where it is, marked as folding on the map for everyone, and launches nothing, raids included. Only then does it arrive.

  Two rules shape the whole feature:

  • Outward or sideways, never back into the cradle. Sectors 1-3 are where every new empire is placed, and they are the slots the galaxy keeps running out of. Folding within the cradle is allowed; folding into it from the frontier is not.
  • The fold can be stopped, and that is why it is public. If the colony's siege bar is charged when the countdown ends, the fold is cancelled and both the Scrip and the weekly slot come back. Winning a raid on it charges that bar; a defensive win clears it again. For the same reason a Peace Shield cannot be bought while a colony is folding, and a shielded empire cannot start one — a fold nobody can raid would be a fold nobody can stop.

  On arrival the colony takes on the destination planet's type and richness, permanently, plus a day of shock and no shield; any allied garrison standing on it is sent home, and the colony cannot fold again for a month. The owner sees their own destination on the planet sheet; everyone else sees only that a colony is leaving, and when.

1.0.0-beta.78 — 2026-08-18

Changed

• New app icon (#715). The icon was a flat cyan hexagon outline with a dot in the middle, drawn on near-black. Next to other games on a home screen it read as a placeholder rather than a game. It is now a forged crest: a six-pointed star set into a hexagonal steel plate, lit from the core. The hexagon stays — it has always been the game's mark — it simply has weight and light now instead of being an outline.

  The same crest is on every surface that can carry it: the browser tab, the icon of the app builds, and the launcher icon on Android and iOS. Getting that right was most of the work — the Android launcher, one of its six screen densities, and the source file the native icons are generated from each had to be found and replaced separately, and any one of them left behind would have shipped two different logos in the same release. A launcher still showing the old hexagon after updating is showing its own icon cache, not the old release.

Nothing else changes: no rules, no balance, no new systems.

One thing this does not reach. On the website the icon set described by the web manifest is not referenced by the page, so for a browser visitor the tab icon is the only surface that changes; the "add to home screen" icon on the site is unaffected. That was already the case before this release and is deliberately left alone — connecting it would switch on installing Novarchy as an app, which is a separate decision from redrawing the logo.

1.0.0-beta.77 — 2026-08-18

Fixed

• The Colony advisor no longer says a loss is reaching storage (#636). A suggested upgrade listed its per-hour figures as -1 Fe/h, +15 Xe/h, -1 Dt/h reaching storage, colony-wide. — and nothing "-1 Fe/h" reaches storage. The numbers were right and the sentence was wrong: upgrading a building that draws power deepens the colony's energy deficit, so a resource that is not overflowing comes out lower, while the upgrade is still worth doing overall. The card now states only what actually arrives, and states the cost separately, in words that carry the direction: The extra power draw pulls 1 Fe/h, 1 Dt/h out of what the rest of the colony keeps. It affected about one card in twelve. These same lines are the facts the in-game Advisor is given before it comments, so the model was repeating the false one in its own words too.

• A card no longer hides what an upgrade costs you elsewhere (#636). When two suggestions happened to draw the same rounded amount, the panel treated the second card's cost line as a duplicate of the first and removed it — measured on 52 panels in 1,500. The figure belongs to its own card, so it is now always shown, even when it repeats.

1.0.0-beta.76 — 2026-08-18

Fixed

• The market no longer accepts an offer with the same resource on both sides (#492). The board was carrying a live 10,000 Fe → 25,000 Fe listing, rendered in exactly the same layout as every honest trade — so the only thing separating "a good deal" from "a 15,000 Fe donation" was noticing that both chips said Fe. The same row also answered the sells Fe filter and the wants Fe filter at once, and the affordable one that buyers actually browse, and it occupied one of the seller's limited public slots. Publishing such an offer is now refused.

  Worth stating plainly, because it is the first thing anyone assumes: this was never a way to create or destroy resources. The escrow moved the seller's stock out at publication and the buyer's in at settlement, and the totals always balanced. What it was is a shape that reads as a trade and is not one.

Changed

• The Trade Portal composer says why it will not publish (#492). When the same resource appears on both sides, the Publish button goes dead and the line under the fields names every doubled resource, so the offer is corrected before anything is sent instead of coming back as an error. That line now also covers two cases it used to leave silent: an offer with only one side filled in, and one blocked because every offer slot is taken. The slot message takes precedence over the others — it is the only blocker that no edit to the form can clear.

Also

• Offers already on the board are not re-checked. That has always been true of this market and it does not change here, so the listing that prompted the report stays until its seller cancels it or a buyer takes it; there is no migration and no escrow is moved.
• A client that predates this release still has the old composer and can still send such an offer. It gets a short "your client is out of date" message rather than a raw error code, and a reload — or the latest build, on mobile — resolves it.
• Reported by Roydisco, on another player's listing. Confirmed: 20 chrono fragments per tier.

1.0.0-beta.75 — 2026-08-18

Changed

• The Campaign no longer takes your fleet — and it no longer ends (#703). Before, engaging a tower level meant committing real ships from a colony's hangar: they flew, they died, and losing a level could cost a navy. Now the tower is fought by an automatic strike group — one ship of every armed type the empire has unlocked, the Destroyer included for the event winner who owns one — and the hangar never leaves home: ships stay yours, for the map. What pushes you deeper is progression itself: research, Tree unlocks, and the ship refits fed by boss parts.

  And the ladder has no end. The six chapter arcs now cycle forever as a wheel — Skirmish Line II, Bomber Wings II and so on — and every tenth level stays a boss that drops ship parts on every turn of the wheel, not just the first.

• Your one order is the deployment (#703). Before engaging, the level sheet shows the enemy line and lets you place each ship class on a row — vanguard, center, rear — with a live forecast simulated on the real engine. The chosen deployment is remembered between runs and it is what actually fights: the battle is resolved with your rows, and the replay shows them.

• The climb is calibrated on the measured game, not on a formula (#703). The first version's curve looked fair on paper and lost every battle from level 2 with default rows. The ladder now opens where a day-one empire clears exactly chapter 1 — all ten levels, boss included — and the wall past it is research's job to move: a maxed-research empire reaches the mid-twenties, a full Tree with the Destroyer about level 70, into the wheel's second turn.

• The season board now shows the deepest climbers (#703). The Leaderboard tab carries a Campaign section: top 20 by levels cleared. The depth is public — what you claim in chat, the board can check.

Fixed

• A cleared level now names what it paid (#703). After a run lands, the tab shows the outcome card: cleared or repelled, what was banked — resources up to your warehouse capacity, parts in full — and a replay button that says which battle it plays, even while the next run is already in flight.

For players with a run still flying from the previous version: it lands under the old rules and your committed ships come home as survivors. During the switchover hour those runs resolve against the new, gentler garrisons — the error is in your favor.

1.0.0-beta.74 — 2026-08-18

Fixed

• Every colony showed the home colony's coordinates, and was drawn as the home planet (#699). Opening a colony that was not the first one, the Colony tab printed the home address instead of that colony's own: the tab at the top right said 1:7:2 while the panel underneath said 8:3:3, and it did so on every colony a commander owned. The panel now prints the address of the colony actually on screen.

  The caption was the smaller half, and the half nobody could see was the picture. Those same coordinates generate the planet rendered at the top of the panel, so a second colony was drawn as the home world wearing its own terrain type and richness — and because the drawing was tied to the address it never received, two colonies of the same type and richness kept the identical planet when switching between them. The sphere is now generated from the colony's own address and redraws when the colony changes.

Changed

• The planet in the Colony tab is no longer rebuilt every few seconds (#699). The 3D view was discarded and recreated on every refresh of the screen, which reset its slow rotation to the first frame twice a minute. It is now rebuilt only when the planet on screen is a different planet.

• A 3D view that fails to load says so (#699). If the file carrying it never arrives — a network that drops, or a page left open across a release — the panel now reports it and asks for a reload, instead of leaving a silent empty square that would never fill in for the rest of the session.

The reward promised on the report — 20 chronofragments of each size to Jsnam, who found it — is not part of this release and is still to be granted.

1.0.0-beta.73 — 2026-08-18

Fixed

• The Dark Matter tile shows its production with a plus sign (#705). The line under the Dark Matter balance read ~3.72/day. The tilde meant "roughly", but it is drawn small and faint directly beneath a number that is allowed to be negative — bounties claimed on you are paid out of your own balance — and at that size a tilde and a minus are the same handful of pixels. A commander carrying a bounty read their own production as negative, and an investigation was opened into a fault that does not exist: the rate cannot go below zero, because the energy factor that scales it is capped to the 0–1 range and the level that feeds it is guarded. The line now reads +3.72/day, on the tile and in its hover text alike, so the balance above and the rate below can no longer be mistaken for each other.

  Nothing about Dark Matter production has changed — the Accelerators output exactly what they did before. Only the way the number is written has changed.

Note on ordering: this release is numbered after beta.72 but its commit precedes it. Two sessions shipped in parallel, and the beta.72 deploy carried this change to production alongside its own. compare/v1.0.0-beta.71...v1.0.0-beta.73 therefore shows exactly the one commit this entry describes.

1.0.0-beta.72 — 2026-08-18

Nothing changes in the game itself. This release repairs the recorded history behind the admin dashboard — the daily note of how many people are actually playing — which was stamping a copy of today onto every day it wrote.

Fixed

• Recorded weekly and monthly active players belonged to the pass that wrote them, not to their own day (#687). The job writes one row per UTC day: today, yesterday, and any day still missing a row. It took the daily count from that day's own players, but the weekly and monthly counts came from a single figure computed once, at the instant the pass ran — so every row that pass wrote carried the same pair. The first run against an empty table wrote thirty rows all saying the same thing: a perfectly flat month that reads like a measurement and is a copy. In normal operation the same defect was smaller and permanent — the two newest rows always agreed, so the day-over-day change at the leading edge was zero by construction, forever. Each row now counts the seven and the thirty days that close with its own day, and the row for the day still running matches what the live panel prints at that moment.

Changed

• A failed read no longer costs a day the numbers that cannot be recovered (#687). Reading the action history is the widest query the job makes, so it is the first to fail when the database is under strain — and it was taking the whole pass down with it, including that day's power curve, activity ladder and onboarding funnel. Those describe a population as it was at a moment and cannot be reconstructed once the day has passed, unlike the two figures that can always be recounted. The pass now retries over the two most recent days rather than abandoning everything.

The rows recorded before this release keep the old meaning, and that is worth knowing before reading the chart: a pass only ever rewrites today, yesterday, and days with no row at all, so the thirty rows written when the daily record first shipped (beta.47) are never revisited. The series is right from this release forward and flat before it, with a step in between. The database comment on those columns now says so outright, so nobody reads the older part as a measurement.

1.0.0-beta.71 — 2026-08-18

Fixed

• The impact countdown on an ally under attack reads in whole seconds again (#708). In the guild Raids panel, the "Allies under attack" row could print its timer with the whole binary tail of a subtraction attached — "impact in 2m 51.599999999999994s" — while the raid rows below it counted down cleanly. The two rows measured the time differently: the raid rows work from the arrival instant and round, the ally row works from a figure the server precomputed minus the time elapsed since it was read, and that difference is a whole number of seconds only by accident. The rounding now sits in the one place both rows pass through, so neither can print a fraction again, and it rounds up: while any time is left the timer shows at least one second, and only a fleet that has arrived reads "impact now".

Nothing else changes — the fleets arrive at the same moment they did before, and no decision taken from that panel is affected.

1.0.0-beta.70 — 2026-08-18

Nothing changes in the game itself. This release corrects the admin dashboard — the internal panel used to read how the game is actually going — which was measuring retention over the wrong group of people, and always in the flattering direction.

Fixed

• Return by signup cohort counted only the accounts that still exist (#686). The automatic sweep deletes precisely the people who register and never start, so within days a week had lost its non-starters from its own denominator and the card was answering "of the people who stayed, how many came back". A week where 100 registered, 74 were swept and 13 of the 26 survivors returned printed a 50% day-1 return; the true figure over everyone who actually arrived is 13%. Cohorts are now sized on registrations — the same population the arrival chart was already repaired to count — so the two cards finally describe the same people. The arrival bars had been fixed earlier and the rest had not, which is what made the disagreement visible.
• A week whose registrations were all swept no longer disappears (#686). It had no surviving row to draw, so the worst week the server ever had was rendered as a week nobody registered in.

Changed

• A Gone column shows how many of a week's signups have since been deleted. "100 signups at 13%" and "26 signups at 13%" are opposite findings, and the percentage alone could not tell them apart.
• Weeks the deletion record does not fully cover are starred, with a note stating which way they are wrong: missing deletions understate a week's signups and therefore overstate its return, so a starred row reads high, never low. Deletions have only been recorded since the day before this release, so a partly repaired old week sitting beside a fully repaired recent one would otherwise have drawn a retention collapse out of two weeks that behaved identically.
• The activity ladder and the First Steps funnel now name the population they count. Neither can be repaired the same way — a swept account takes its colony and its recorded progress with it — so instead of being corrected they state their denominator and say plainly that they are not comparable with the arrival chart.

Also

• Counting a deleted account as somebody who never came back is an assumption, not a proof, and the card says so. The obvious argument for calling it exact is wrong: the sweep judges activity on the colony, while this card also counts a chat message, so somebody who only ever chatted can be both swept and a genuine day-1 returner. The error runs toward understating return, and the Gone column makes its size visible.
• The colour bands on the table (green at 40%, gold at 20%) were calibrated against the old, inflated shares and have deliberately not been retuned. Over real arrivals the realistic band is far lower, so green is now close to unreachable; choosing new bands is a judgement about what a good week looks like, and it is recorded as an open decision rather than guessed at.

1.0.0-beta.69 — 2026-08-18

Nothing changes in the game itself. This release repairs the admin dashboard — the internal panel used to read how the game is actually going — which answered the same question two different ways depending on which button was pressed.

Fixed

• The MAU and Stickiness tiles no longer change when the chart window changes (#685). Those buttons are meant to control the charts, and the two tiles beside them are labelled last 30 days and DAU / MAU. Pressing 7d rewrote them anyway: a monthly-actives figure of 200 became the 60 people active that week, still captioned "last 30 days", and Stickiness went from 10% to 33%. The error only ever flattered — pressing 7d made retention look three times better than it was. A figure named after a fixed span now measures that span whichever button is pressed, and only the charts and the in-window counts follow the buttons.
• Two more figures had the same fault, and neither was in the report (#685). Weekly actives and the 7d column in the player rosters were computed the same way, and were wrong for any window narrower than a week. The buttons on screen only offer 7, 30 and 90 days, but a narrower one can be asked for by hand through the address bar. The visible consequence was the churn-risk list reporting "0 actions this week" next to somebody who had in fact acted inside that week — the one number that list exists to show.

Also

• The recorded history was never wrong and needs no correction. The hourly job that writes it always asks for 30 days, which is the single width at which this fault could not occur, so the stored series has no step in it at the moment of this release.
• Opening the panel on the 7-day window now costs what the default 30-day view already costs. It is no longer the cheaper read, and it is no more expensive than the panel's normal load.

1.0.0-beta.68 — 2026-08-18

Fixed

• A colony could lose its stored resources with nobody attacking it (#662). Some colonies were saved before certain buildings existed, so their record simply carries no entry for those buildings. Four places in the engine read those entries while asking "is this level zero or less?" — a question with no sensible answer when there is no level at all — so the missing value travelled on as an unusable number instead of being read as "this building is not there". It mattered most in the routine that works out what a colony produced since you last looked, which runs every time anyone opens the game: it reads the warehouse ceiling, the energy factor that scales all output, and each building's hourly rate, and one unusable value in any of them wiped the stockpile on the spot — permanently, because the next read added to what it found. No live colony was ever in that state; production was checked before the work started and again before it shipped.
• A corrupt warehouse level meant no storage limit at all (#662). Such a colony hoarded without a ceiling. It now falls back to the base limit, because a broken value must never be worth more than a real one. The reverse case is deliberately left alone: a limit too large to represent is not replaced by the base one, because that would truncate a colony legitimately holding more — an earlier draft did exactly that and would have destroyed resources in the change meant to protect them.
• One participant could cost a whole raid its points (#662). If a single contingent's fleet value came back unreadable, every share — the honest ones included — came out as zero, so nobody was paid and nothing looked wrong. Shares are now computed from the participants whose fleet value can actually be read, and the total handed out is unchanged.
• Dark matter could be corrupted permanently (#662). A colony with a broken Accelerator level could write an unusable amount to a player's dark matter balance, and every credit after that was unusable too. Both the source and the point where it is stored now refuse it.
• A refused score now leaves a trace (#662). When the server drops a value for being unusable it says so in its log. It used to stay silent, which after the previous fix was the last remaining sign that anything was wrong. The ordinary zero-score case — every battle against an NPC — stays silent by design.

Notes

Nothing changes for a healthy colony: every real building level produces the same number as before, down to the last digit. This release closes holes before anyone fell into them.

What is deliberately not in it: the repair migration the issue asked about. It was made conditional on a read-only query, and that query returned no rows — no score, no colony and no balance in production is affected — so there is nothing to repair.

1.0.0-beta.67 — 2026-08-18

Added

• The Campaign (#696). A new tab: a per-empire ladder of 60 authored NPC levels in six chapters — Skirmish Line, Bomber Wings, Aegis Column, Entrenchment, Bastion, Capital Line — and past 60 the tail (The Deep) publishes levels forever. Before this, a new player finished First Steps with two interceptors in the queue and nothing telling them what to do with them; now level 1 is that next step, sized so the starter pair beats it (measured in the battle engine: the design's first draft was 4× too strong and lost 40 seeds out of 40). Everything is published before committing — the exact enemy line, emplacements and fortresses included, its value, the payout, the round-trip clock — and a forecast box runs the same engine the server resolves with. Fleets are real: they can die, survivors come home, and a cleared level is never replayable for reward. Only an attacker victory clears; a draw is a retreat.
• Ship parts and modules (#696). Every tenth level is an Arc boss and drops parts (Alloy Frames, Coil Arrays, Field Cores — 26/24/28 across the authored sixty, a thin two-per-boss trickle in the tail). The Refit desk turns them into standing orders per ship CLASS, three bays each, three published options per bay, every option a real tradeoff and the only drawback-free option priced as the parts sink. Refitting a bay again takes 12 hours and replacing refunds half the parts. Fleets already in flight keep the orders they left under — battles snapshot modules at launch on missions, raids, garrisons, ambushes and campaign runs alike, and the defender fights with its current orders.
• The refit is part of the Power score (the ⅓ band, the bounty, the bastion scaling and the leaderboard all see it), the Shipyard's PP labels include it, and the leaderboard shows every empire's campaign depth.

Changed

• Campaign clears pay war points priced exactly like expeditions (same divisor, same PvE haircut, same per-engagement cap), and only the first clear of a level ever pays anything — parts, points and the resource payout all ride the same depth check, so deliberately re-losing or retreating farms nothing.
• A fleet committed to the campaign still counts in your Power for the whole engagement, and shrinks to the survivors the moment the battle resolves — leaving for the tower neither hides your navy from the gate nor lets dead hulls keep scoring.

Also

• Agents climb the ladder too, on human terms: not every persona takes an interest, and those who do engage with margins and pauses of their own (persisted across restarts). Their reality sheet carries the campaign, so they answer questions about a public depth column truthfully.
• DESIGN-CAMPAIGN.md documents the whole design, including what was deliberately cut (speed/cargo module axes until the travel and hold arithmetic reads them) and this release's rollback SQL in deploy/DEPLOY.md.

1.0.0-beta.66 — 2026-08-18

Changed

• A colony's regrowing defences can only be harvested so fast (#577). The Planetary Shield's bubble and the Defence Batteries come back for free after every battle — the bubble is whole again the moment the next fleet arrives, and emplacements rebuild themselves every 30 minutes. Each new attacker was paid full price for breaking them, so a group taking turns on one colony farmed it indefinitely. Measured on a fortress colony before the change: 300 war points to the first attacker, and 300 again to a different attacker one hour later, at every spacing tried.

  Those two defences now pay out of one allowance per colony, worth a single battle's maximum and refilling over twelve hours. The first attacker in a window is paid exactly what they were paid before; the next one that hour is paid for the wreckage they actually made and nothing for the wall that grew back. Measured after: 300, then 25 an hour later, 150 at six hours, and a full 300 again at twelve. A defence harvested no faster than it regrows never feels it.

  Destroyed ships, loot, the victory bonus and the underdog multiplier are untouched — only the part that comes back for free is rationed. And nothing about how much damage the bubble absorbs, how hard the emplacements hit, or how fast either regenerates has changed.

• A joint raid shares that allowance (#577). Each contingent keeps its own ceiling for ships and loot, but a colony's regrowing defences regenerate once however many fleets shot at them, so the regrowing half is split instead of paid to each. This is a real reduction in what a large raid on a heavily fortified target scores, and the first pass of a multi-contingent raid is the one case that is not identical to before.

• Ganging up on a fortress pays less per attacker, which makes turtling slightly stronger. It is the intended consequence of the above, not a side effect.

Added

• The game says when a colony has been harvested (#577). The owner reads it on their Shield and Batteries sheets — how much of the colony's war-point allowance has recharged, and when it is full again. An attacker reads it on a spy report, so a fresh fortress can be told from one somebody hit an hour ago; that figure ages by itself while the report sits in the log and disappears once the colony is back to full price. Both surfaces say plainly that the defences are not weakened, only cheaper: the rule keys on other players' battles, which is exactly why it needed to be visible at all.

Known limits

• The per-pass price is unchanged: a first attacker on a colony with a full allowance still collects what they always did. This change prices repetition, not the first blow.
• Destroyed emplacements are still valued at their full build cost even though they regrow for free — about a fifth of a battle's maximum each at Batteries L14. This change rations that price without fixing it, and it deserves its own issue.

1.0.0-beta.65 — 2026-08-17

Changed

• Twice as many anomalies each week, and the AI empires have stopped taking them (#640, #525). The whole galaxy held one unclaimed anomaly fifteen and a half hours into the week: 48 spawned for the entire server, they were gone the night they appeared, and the map never said why — it simply looked like the feature did nothing.

  The weekly draw goes from 6 to 12 per sector. Anomalies already out there did not move: the new ones were appended beside them, so nothing in flight was disturbed and no reward changed.

  And the AI empires no longer compete for them, which is the half that decides who actually gets one. The reset lands at midnight UTC, when few people are online, and the supply is first-come-first-served — so the race was over before most players woke up. That was Alessio's call in #chat-with-devs on 13/08 and it is now real, in both directions: they do not launch expeditions, and they no longer settle a colony on an anomaly's coordinates, which used to erase it with no battle report and no trace at all.

Fixed

• A claimed anomaly no longer hides a free one (#640). The map showed you only your first few signals, and the cut ran before the ones already taken were removed — so if somebody claimed them on Monday night you saw an empty sector for the rest of the week, with free anomalies sitting in it. Now a taken one is replaced by the next one you can reach. Corrected on the political map, the system view and the launch button together: they could otherwise disagree, showing a signal and then refusing the mission.

• The map says why it has no signals (#640), and says the true thing — which is three different things. That Astrometrics 0 detects nothing anywhere; that the cycle really is exhausted; or that signals remain but your level cannot resolve them. In that last case it names the Astrometrics level that would actually reveal something still out there, rather than the next one up the scale, which on a Thursday buys nothing.

• Launching at an anomaly somebody already claimed no longer costs one of your five daily scans and a round trip to discover it. It is refused when you press the button.

• "Investigate an anomaly" is no longer dealt at Astrometrics 0 (#640), where no anomaly is visible anywhere in the galaxy and the card sat there impossible for the whole day — a wasted slot out of three, for the players least able to spare one.

Also

Anomalies can still only appear in the oldest ten systems of each sector, and in the core those are almost entirely settled, so most of the new supply lands in the outer sectors — the trip is still a trip. Opening the rest of the galaxy to them is the larger lever and costs nothing in rewards, but it re-rolls the week in progress, so it can only ship at a week boundary and has its own issue.

1.0.0-beta.64 — 2026-08-17

Changed

• A garrison's watch and its trip home are two different things (#620). When a guard shift ran out, the ships appeared back in their home hangar at that same second, while standing a garrison down early was billed the whole flight back. Doing nothing was free and doing the sensible thing cost you — so late in a long watch, standing down landed the fleet later than leaving it alone, and the confirm panel had to talk the player out of its own button.

  Now the watch ends, the fleet lifts off, and it flies the same way home either way. The duration on the launch slider is time on station, with the trip back on top, and the mission card says so: it counts down to the end of the watch and names the flight home beside it. Standing down early can no longer land a fleet later than waiting — it gets the ships home sooner by exactly the watch still to run, and what the player gives up is the cover the ally loses for that time.

  Garrisons already on station were moved onto the new rule rather than left on the old one: all 28 live at the time of the deploy were re-timed at startup, and came home one flight later than their old countdown implied. Leaving them behind would have meant the rewritten panel telling those players that standing down was faster while it was in fact slower for them.

Fixed

• A spy report no longer reports an enemy garrison as home the moment its watch ends (#620). It was quoting the end of the guard shift as the instant the fleet was back in its hangar, handing the reader a window that was not there.
• The War Room's "send defence" box promised the fleet back when the watch ended. It now says it fights for the whole watch and then flies home — and it is the one screen that chooses the watch length on the player's behalf.
• The progress bar of a garrison flying home sat at almost full for the whole trip, because it was drawn from the moment the fleet had landed on the ally's planet hours earlier.

Also

• This composes with the battle hold of the previous release (#648): a watch that runs out while a fleet is on final approach still keeps the garrison on station until the fight is resolved. The two rules read as one sentence — the watch ends, then the battle finishes, then the fleet flies home — and a garrison held by a battle flies its trip home from the release, not from the bell, so the hold can no longer erase the leg.
• ⚠️ An app that has not updated still shows the old stand-down warning that standing down "gets them home later, not sooner". That advice is now the wrong way round for every garrison, and the server has no way to correct a bundle already on a phone.

1.0.0-beta.63 — 2026-08-17

Fixed

• The dashboard charts have a y-axis, and every value on them is readable (#633). The activity, arrival and departure cards drew bars and lines in a box with no axis at all — one card put actions peaking in the thousands beside a headcount peaking in the dozens, on shared coordinates. That is a dual-axis chart with the axes left off: where the two series crossed and where they diverged looked like findings, and were artefacts of an arbitrary alignment. The only real numbers on screen were the peaks printed in the legend.

  Each card is now a stack of small multiples: one plot per measure, each with its own labelled y-axis in whole numbers, all sharing one x-axis drawn once underneath. Every plot names its series and prints today's value and the window's peak; every mark carries a hover title; and under each card a Numbers panel holds the entire series as a table — the reading that needs no pointer, no hover and no colour vision.

• Today is drawn as the partial day it is, on the lines as well as the bars (#633). Every plot ends on today, which at 08:00 UTC holds eight hours against twenty-four for every day beside it. The last bar was already faded; the lines were not, so one card showed a pale last column directly above a line ending in a full-strength cliff — the same eight hours drawn two ways, five pixels apart, under a caption that promised a fade the line did not have. Both are faded now, the caption says so, and the Numbers table marks its last row today so far.

• A chart that cannot be drawn says so, instead of reporting zero (#633). When a figure the page could not read reached a card, it printed "No registrations at all in this window" — a hard, confident zero, on the panel someone would be using to find out whether registrations had stopped. Those cards now say the figure is unknown rather than none.

• Smaller honesty repairs on the same cards (#633). The arrival plot's "peak" named the human series while the column beside it was humans and agents stacked, so with agents running hot the printed peak could sit a third of the way up the column it appeared to label. The legend chip for agents was solid while the band was translucent — a colour that appeared nowhere in the plot. The net plot's axis could reach two and a half times the largest movement in the window, flattening a month of real change into a pulse. The arrival table listed two nested counts beside the total that contains them, so a row read as more arrivals than the day had. A one-day window printed its date three times across the axis, and drew no mark at all.

1.0.0-beta.62 — 2026-08-17

Changed

• The Wraiths have started playing properly (#680). The hostile faction that arrived in beta.55 was not very good: thirty empires that only knew how to build. They never collected a contract, never took a guild directive, never donated, and never spent a chrono shard — seven First Steps rewards sat reached and unclaimed on every one of them while their build queues stood empty for want of exactly the resources those rewards pay.

  They now collect what they have earned, run their guild, and burn shards to collapse their build queues. What a player sees is a faction that grows noticeably faster — still on its own economy, still under the same rules as everyone else.

  The diagnosis is worth recording because the first attempt fixed the wrong thing. They were not slow, they were broke: the build queue was empty and every build call came back insufficient_resources. A chrono shard buys time off a queue, and you cannot accelerate a queue that is empty because you cannot afford the next building. So half the work turned out to be about income, not speed.

Also

• Two corrections to claims made in the beta.55 notes, recorded here because a published claim that has stopped being true is worse than no claim (#680). That release said the faction had "no head start, no gifted fleets, no shortcut" — they were since given a large stock of chrono shards, so that is no longer accurate. It also said "every attack it launches is decided by a person", which remains true: nothing here lets them attack unattended, and their scheduled pass still cannot order a fleet.

• The client gained a way to collect First Steps rewards at all (#680). It never had one, which means the 75 autonomous LLM empires have never claimed theirs either — free resources every one of them has been leaving on the table. Filed separately; this release only fixes it for the Wraiths.

1.0.0-beta.61 — 2026-08-17

Fixed

• Changing the mission type no longer throws away the fleet you picked (#495). Choosing the ships first and only then remembering to change the type — Attack to Redeploy, in the report that started this — replaced the whole selection with the advisor's fleet for the new type. Nothing on screen said so, and on a redeploy the advisor proposes the entire hangar, so the substitution did not look like an error: it looked like a full, deliberate selection. Most players found out at launch, or after it. Setting the destination after picking the ships did the same thing, once for every coordinate field they typed.

  A hand-picked fleet now survives both. The advisor takes it back only when asked — use it, spy it first, switch to espionage — and those three behave exactly as before.

• Switching to Espionage or Colonization says what it did (#495). These are the only two missions the engine requires a specific ship for, and switching to one without it still hands the fleet to the advisor. Now the screen says so and names what was missing, instead of swapping the numbers in silence. If the hangar holds no such ship at all, the selection is kept instead: handing it over would have replaced it with nothing and left the picker empty.

• The picker can no longer show ships that have already left (#495). A selection that now outlives a change of type also outlives a departure, so it is trimmed to what the hangar still holds — and the trim sticks, rather than springing back when the ships come home. What the player reads is what will fly.

Changed

• The advisor says when it has changed its mind, instead of changing the fleet (#495). When the mission type, the destination or the cargo moves after a fleet has been picked, a line beside the Fleet label reports that the advisor would now choose differently and offers its fleet as a button. It appears only when the recommendation actually changed, so it stays quiet through the ordinary edits that do not move it.

• "Fill every carrier" only fills the cargo now (#495). Its label used to promise that the advisor would then size the fleet around the load, which stopped being true the moment the player picked the ships themselves. It sets the cargo and nothing else, and says so.

1.0.0-beta.60 — 2026-08-17

Fixed

• A garrison could walk out of the battle it fought in (#648). If a guard shift ran out while an enemy fleet was landing on the colony it was defending, the fleet flew home with every ship intact — and its owner then read a battle report saying part of it had died. The ships had counted in the defence, so the attacker really did fight them; they were simply back in the hangar afterwards. A garrison wiped out to the last ship came home whole.

  The protection against exactly this already existed, and had been believed to cover the case. It did not: it was wired only to the Stand down button, so it covered the player who chooses to leave early and not the one whose watch simply runs out — which is the common case, and the only one the player cannot avoid.

  A watch that runs out under fire now waits for the battle. The garrison stays on station, fights the fight it was sent for, and flies home once the fighting is over.

Changed

• Garrisons can now land a little later than their watch says (#648). Only when a battle is actually landing on the colony they are standing on, and never for long: the ceiling is measured from the end of the watch, so an attacker cannot stretch it by attacking again, and once it is reached the fleet leaves regardless. Nobody can strand somebody else's fleet on a planet.

  On the Missions tab a garrison held this way keeps its card and says why, instead of sitting at 0s with nothing on screen to explain why the ships had not moved. On an ally's colony, and against an attack hidden by a Veil, the card stays silent — saying more there would reveal an incoming attack the defender is not supposed to see.

• The stand-down panel no longer says your ships are home when they are not (#648). That sentence could appear over a fleet still standing on a defended planet, and — once garrisons could be held past their watch — over one that had just been destroyed in the battle. It now reports only that the watch is over, which is true whether the ships landed, are waiting out a fight, or are gone.

1.0.0-beta.59 — 2026-08-17

Fixed

• The Build button refused the order it had just offered you (#533). With a full build queue, watching a building finish — the row disappears, the greyed-out buttons come back to life — and clicking straight away answered «Queue full: wait for the current work to finish». Waiting a couple of seconds, and the very same click worked. Reported from Discord, and reproducible "reliably enough that I don't think it's fully random".

  Your device and the server never disagreed about when an order ends: that instant is written by the server and travels untouched. They disagreed about what time it is now, and on purpose — the game ignores a clock difference under ten seconds, because correcting one that small would only make every countdown jitter. The mistake was letting that same approximate number decide when a queue slot came free. It is fine for drawing a countdown and wrong for predicting what the server will accept: on a device running a few seconds fast, the interface handed back the slot early and the server refused for the rest of those seconds.

  The queue now waits until the server can be shown to agree. If the button is lit, the order goes through.

• The shipyard and the launch pad had the same fault, and nobody had reported them yet: a slip could read free before the server had freed it, and a ship that had just left the yard could be offered for a mission before the server had built it. Both are fixed by the same change; the research queue was never affected, because it reads its list from the server rather than recomputing it.

• «Ready in ready» on the shipyard, research and building panels, and a tick that claimed an order was complete while its row was still sitting in the queue. Those rows now say «finishing up…» for the moment between the countdown reaching zero and the work being confirmed.

Changed

• A finished order leaves its queue when the server confirms it rather than when your own device guesses, so on a device whose clock is off — or on a slow connection — the row lingers a moment longer at zero before the slot frees. That wait is not new: it is the same wait that used to end in an error message, now spent looking at a row that says what it is doing.

1.0.0-beta.58 — 2026-08-17

Fixed

• The Pirate Beacon told you nothing about how much you could carry home (#552). The Beacon the game seeds in your own system is meant to be the first raid worth making, and the squadron it is calibrated for — three to five Interceptors — can lift 100 of the roughly 650 sitting there. You would win the fight and come home with a seventh of it, and nothing on screen had ever mentioned the hold. The advice that fixes this existed: the Advisor adds freighters until the expected loot fits, and says "hold full, add freighters". It was unreachable. Without a spy report the Advisor refuses to recommend any fleet — rightly, because a composition invented out of nothing sends players into battles they lose — and that refusal took the hold advice down with it. The one sentence a newcomer needed sat behind a 50-Dati probe nobody thinks to send at the tutorial target parked next door.

  Now, on a Beacon with no report, the Advisor states the two facts it can state without spying: what your selected fleet lifts, and the most that Beacon can ever hand over. It recommends nothing — that part still needs a probe — and the numbers move as you pick ships.

  What the line promises is deliberately narrow, because the pool is not a fixed thing. It says up to, since a Beacon raided recently holds far less. It says any attack in the last few hours empties it, won or lost — a raid that got you repelled resets the pool exactly like one that looted it, which is the case a newcomer is most likely to hit and least likely to guess. And it says a fight you do not win outright pays nothing at all: a Beacon that has grown its batteries can leave one standing at the last round, and that is a draw — your ships come home, your holds come home empty.

  Once a spy report exists the line steps aside and the real forecast takes over, sized against the pool as it actually stands.

  The Interceptor's cargo is untouched. This was never a balance problem; the information was simply not there. Not included: handing out or mentioning a Freighter in the guided first steps — that chain's steps are saved with your colony and each one pays a reward, so adding one changes the opening economy and belongs to its own decision.

1.0.0-beta.57 — 2026-08-17

Fixed

• The relics can be fought for again (#663). The four ancient sites — The Jump Ring, The Verdant World, The Suspended Yard and The Exarch Archive — were each guarded by 22,186,744 automatic ships. That was not a hard fight, it was an impossible one: a raid of 715,060 hulls destroyed 2.9 million defenders, killed not a single one of the eleven million interceptors, lost 564,000 ships of its own and came home with nothing. Nor could the damage be worn down across several attacks, because the site rebuilds its guard to full on the next maintenance pass. Every relic on the server was frozen in the hands of whoever already held it, and no coalition existed that could change that.

  The guard now stands at 16,007 ships — the size it was always meant to be: heavy enough that nobody takes a relic alone, light enough that a guild which organises an expedition can.

  A relic measures its guard against the strongest fleet in the galaxy, and the measurement was reading an administrative account whose ships were granted by the server rather than earned. The relics were being calibrated against a fleet nobody ever had to fight for. Strength is now measured only across ranked empires, the same set the leaderboards use.

• The faction bastions were measured with the same broken ruler. The Silent Choir, The Black Foundries and The Exarch Guard were tuned against that same borrowed fleet — and, because a relic site counts as an empire with a colony, against the inflated relics too. The two kinds of stronghold had been quietly inflating each other. They now stand where the design puts them: above the strongest player, and not out of everyone's reach.

1.0.0-beta.56 — 2026-08-17

Fixed

• Commanders talk about the game, not about its code (#595). A player reported two things in the same message: an AI commander mentioning its "cantiere" in an otherwise English sentence, and the Recycler card describing a technology as "Active Salvage (i_reclamo, Forge tree)". Neither word was ever written for a player to read — they are the internal keys the code uses, printed where a name belonged.

  The chat one was not a slip of the model's. The block of true facts every commander is given before it speaks handed it the keys as they are — {'cantiere': 1}, armiEnergia=3 — so it repeated what it was given. Commanders now read, and say, the same names the player sees in their own panels: Shipyard, Energy Weapons, Active Salvage, Freighter, Spy Probe. The Recycler card names the technology and where to find it (Forge tree, T5) instead of quoting its key.

• Three facts commanders could state backwards (#595), found while fixing the above. A commander that had been probed could speak as though it had done the probing, to the very player who had probed it. An attack between two guilds was described to both sides as an attack on one of their own, which the battle report disproves in one click. And the routine promotion from recruit to member was recorded as a demotion, so a commander could complain in guild chat about being pushed down a rank under a log line saying the opposite. All three now say who did what to whom.

• Internal counters stay out of conversation (#595). War tension, grudge weights and sympathy scores were reaching commanders as bare numbers like 0.62 — figures that appear on no screen in the game. They still act on them; they no longer quote them.

Also

• A market trade a commander has made now carries its amounts, so a commander asked what it paid can answer instead of dodging — or inventing a figure (#595).
• Guild announcements about rank changes and research read in English rather than mixing in internal terms, and a guild's request for help asks for "data" rather than dati (#595).

Left out on purpose, and tracked in #656: the guild decision prompts still use Italian JSON keys, which are a contract with the model's own reply and have to be changed on both sides at once; and a commander in the attacking guild still records a grudge against its own guildmate, which needs the guild identity plumbed through to the agents rather than a change of wording.

1.0.0-beta.55 — 2026-08-17

New

• A hostile faction entered the galaxy: the Wraiths [WRTH] (#651). Thirty empires under one command, announced as what they are rather than disguised as neighbours. They exist because the top of the leaderboard kept saying the same thing in different words — nothing left to do, nobody to defend against, nothing to lose.

  What a player sees: a new guild in the directory, thirty new names on the leaderboard, and — in time — raids. What they do not see is a shortcut. The faction starts from nothing: no granted fleets, no seeded colonies, no head start of any kind. On the day it was founded its strongest empire sat at 52 Power against 161,000 for rank ten, so for the first days it is weaker than almost everyone it intends to fight, and it spends them digging like anybody else.

  The rules it plays by, which are enforced in code rather than promised:

  • it hunts the top ten only — nobody below rank 10 is ever a target, because newcomers are not who asked for enemies;
  • it spreads the pressure. Targets are chosen by who has been left alone longest, never by who is cheapest to hit, with a six-hour cool-off before the same empire is struck again. Left to the obvious rule, thirty empires converge on one victim and farm them out of the game while the other nine never meet the faction;
  • it can be beaten, raided and looted like any other empire — ordinary loot, ordinary bounties, no special armour and no special prize;
  • threats made to it in chat count, and threats it makes are recorded by everyone else.

  Every attack it launches is decided by a person. The faction's own economy runs on its own, but nothing in it can order a fleet without someone choosing to.

Also

• Two corrections that only whoever runs the server will notice (#651): the admin dashboard now counts the new faction as machines instead of as thirty players who never churn — left alone it would have flattened the retention curves permanently, because the daily rollup is frozen and not recomputed — and the sweep that clears never-started accounts no longer mistakes an unbannered Wraith for one.

1.0.0-beta.54 — 2026-08-17

Fixed

• The Galaxy panel could only reach the first ten systems of a sector (#603). The galaxy grew to nineteen systems per sector on 14 August, and the System control in the Galaxy tab never grew with it. The dropdown ended at 10, the ▶ arrow greyed out there, and a system opened from the political map announced itself as "1" while its real number was 14 or 17 — with the ◀ arrow still working, so you could walk down from a system the panel refused to name.

  It was not a few missing rows at the bottom of a list. Nine systems in every sector could not be opened from normal navigation: 72 of the 152 in the core sectors, very nearly half the galaxy, and precisely the half the expansion opened up, where new colonies are now being placed. Those coordinates were legitimate everywhere else — the leaderboard listed the empires living there, chat links to them resolved, the servers served them. The only screen that refused was the one meant for looking at them.

  The dropdown now lists every system the sector has, the ▶ arrow runs to the last of them, and a system reached from the political map shows its own number.

  Reported by arcticpaladin, who described all three symptoms precisely enough that the cause was one line of code. Bounty paid: 20 chrono fragments.

  Nothing else changes for the player. The control moved into its own piece of the interface so that the next expansion carries it along or fails loudly: the previous one slipped past for three days without a single test noticing, because the Galaxy tab draws a canvas that no test in this repository can render, which left the list of systems it offered invisible to every check we had.

1.0.0-beta.53 — 2026-08-17

Fixed

• A war score could freeze and never recover (#578). Attacking certain colonies permanently broke the attacker's score: from that battle on, every later one added nothing — for the rest of the week and the rest of the season — with nothing on screen to say so. The player simply stopped scoring while everyone else kept climbing. One unreadable building level in the target's saved data was enough, and a colony old enough to predate the Defense Batteries qualified; attacking it directly did it, no special setup needed.

  An unreadable level now counts as nothing instead of poisoning the whole battle around it, and the ships the attacker actually destroyed are still paid. That distinction is the fix: the obvious version of it, which cleaned up the score instead of the source, silently threw those kills away on every battle against that colony.

• A joint defence that won could pay nobody (#578). The same data reached the split of defensive points, and when it did a successful defence paid zero — to the colony's owner and to every ally who came to help, at once.

• A broken score can no longer win the season (#578). A frozen score did not sink to the bottom of the standings; it sorted above every real score. At season close it would have taken first place, the first-place reward and the season Destroyer ahead of whoever earned them. Those rows are now skipped in the leaderboard, in the final standings and in the Destroyer award.

Also

• Nothing about combat changed. No battle outcome, forecast, advisor figure or Power value moves by this release.

• Scores already frozen are not repaired, and the entry says so because the fix does not. This stops new ones; it does not clean old ones. A player already affected now reads as 0 points and no rank and will not appear in the final standings or receive a season reward, and a guild war whose score is already frozen stays a draw. Whether anyone is in that state has not been checked — that is a separate job. Three sibling paths of the same shape (the shield pool, the storage cap and the energy factor) were found during review and deliberately left for a follow-up.

1.0.0-beta.52 — 2026-08-17

Fixed

• A loud night no longer buries your espionage reports (#439). The Activity feed holds fifty rows, and until now those fifty simply went to whoever wrote last. After a night of border tension you opened Combat and found fifty lines of tension notices — and the spy report from the day before, with the coordinates you were keeping tabs on, had scrolled out of existence. It was not noise pushing out noise: that report is the only place those coordinates survive.

  The fifty most recent rows are still exactly what they were, in the same order. What the noise buried is now listed underneath them: the most recent few rows of each kind of event those fifty do not already carry. Your spy reports come back, with their dates on them.

  Honest boundary, because the feed does not reach forever: what the noise buries beyond the last four hundred events is still out of reach, and a rescued report is the most recent few per kind, not all of them.

• The Fleet Advisor stops guessing when it should remember (#439). It reads the same feed to recall what you already know about a target, so when a report was evicted it silently fell back to assuming the enemy looked just like you — a recommendation that looks perfectly ordinary and is not. It sees those reports again.

Changed

• Rival commanders check the date before treating something as news (#439). Bringing buried events back means handing them facts older than ones they had already absorbed, and they had no way to tell. Left alone, a restart would have had them announcing a two-day-old battle as it happened, nursing a grudge they had already settled, and opening the channel with "the intel I just gathered" about a probe from yesterday. They now ignore anything older than the last thing they already knew.

1.0.0-beta.51 — 2026-08-17

Fixed

• A raid's loot no longer disappears between the battle report and the warehouse (#587). A raid came home with a full hold, the battle report and the mission card both said the whole haul, and then the warehouse settled at its ceiling with a smaller number and nothing anywhere explaining the gap. The loot was written into the colony above its limit and trimmed later, silently: with a cap of 500.000, a colony at 450.000 and a raid returning 900.000 Ferrolite, 50.000 landed and 850.000 vanished without a line in the diary.

  The deposit now stops at the ceiling and the part that did not fit is named in the return event — the same signal transports have had since #385: "attack mission complete, but the departure colony was already full: 15.500 Fe had nowhere to go and was lost." A raid whose loot fits says nothing about losses, exactly as before.

Changed

• A colony that produces nothing can no longer hoard loot above its ceiling (#587). This is the half of the change that really takes something away, and it is worth stating plainly. The warehouse cap is applied while a colony produces, resource by resource — so a colony with no Extractor, Refinery or Compute Node for a given resource was never trimmed on that resource at all. A forward colony used purely as a raid ramp starts with every building at zero, and could therefore bank loot above its limit raid after raid, without bound, and spend it.

  From this release that excess burns on landing and is reported like any other loss. On a colony that does produce the resource nothing changes: the excess was already being destroyed at the next tick, and only the explanation is new. If you have been stockpiling on a colony that produces nothing, build the matching producer or expand its Warehouse before the next raid lands.

• The Warehouse card, the resource tooltip and the phone sheet said the ceiling applied to what you produce. They now say that what a fleet brings home stops at it too. And the loss row no longer tells a raider to "expand your Warehouse before shipping it back" — nothing is shipped back after a raid.

1.0.0-beta.50 — 2026-08-17

Fixed

• Founding a colony with a big starting endowment no longer destroys it (#639). You could attach two million Ferrolite to a colonisation, open the new colony when it landed, and see every unit sitting there. Nothing warned you. Then the first Extractor finished and the colony dropped to 5.000 — the rest gone, with no event and nothing in the diary to explain where.

  A colony founded this instant has no Warehouse, so it can only hold 5.000 of each resource. Nothing truncated the endowment at the moment it landed, which is what made it cruel: the engine applies that ceiling only on a line that actually produces something, and a colony with every building at zero produces nothing. So the numbers sat there looking safe until the first producing building completed, and that tick collapsed them.

  Now the endowment is measured against the ceiling when the fleet lands. What fits is delivered; the rest turns round and flies home with the returning fleet — the same thing an over-full transport has done since beta.4, on the sibling path that had been getting it right all along. Nothing is destroyed.

  The diary was changed to match. Where the row used to say «New colony founded» and the coordinates — while two million Ferrolite were quietly reversing course — it now names the real ceiling (higher if Deposits research has raised it), says that building a Warehouse is what lifts it, and gives both numbers: what the colony kept, and what is on its way back.

Also

• A colonisation sent with a lone colony ship has no return leg at all: the ship itself is consumed founding the colony. In that one case the excess has no way home, so the game now states plainly that it was lost, and why, instead of writing it above the ceiling for the next tick to eat in silence. Reaching it takes Logistics research far beyond anything a live empire has, so no current player can hit it — it is fixed because the failure was destroyed resources, not because it was likely.

• Worth knowing if you plan around it: the largest useful endowment is now the new colony's ceiling, 5.000 per resource, or up to about 9.000 with Deposits fully researched. Before this, a large endowment could be preserved by rushing a Warehouse before any producing building — that is no longer possible, and nothing is lost by trying.

1.0.0-beta.49 — 2026-08-17

Fixed

• A ship's name plate no longer sits on the ship in front of it (#445). With formations on, the name and the survivor bar that hang under one rank landed on the hull of the rank ahead — and since those are drawn over the hulls, a light ship in the vanguard was not merely crowded, it was gone. Measured on the playground battle: the whole silhouette covered on a phone, about a third of it on a laptop, something on every screen size tried. The ranks are now spaced by however much the battle and the screen actually need, so every ship stays visible.

  The distance could never have been a fixed number, and that is the reason this took four rounds of review to get right. The plate is a fixed size in PIXELS while the ranks are spaced in world units, so the further the camera pulls back the more screen the plate eats and the less a rank step is worth. It is now solved per battle, and solved exactly — but the first cut solved it against the wrong thing, measuring the survivor bar (45 px) when the plate is the wide half of the block (145 to 181 px for a real roster), and the bench could not see the mistake because it fed its own oracle the same wrong number. On short screens it now declines to act at all rather than push the vanguard's own name behind the shield panel: a scene that does not improve beats one that hides the rank taking the fire.

• The flyover clears the defenders (#445). When neither fleet can shoot, the attacker crosses the field — and stopped 20 units inside the rearmost defending hull instead of past it, while a comment above that line claimed to have fixed exactly this. It now aims beyond the ship that is really furthest back, and stops at the edge of the frame rather than parking off-screen for the rest of a scene whose only content is that pass.

• The planetary shield follows the ranks (#445). With formations in the air, the top and bottom ranks of a defending fleet hung outside the Bubble. The shell now stretches to hold them.

Changed

• On wide screens the camera sits about 3% closer: it was keeping room for the depth of the ranks as if a step back were worth a full step sideways on screen, which the viewing angle never allows.
• Restarting a replay now puts every ship back exactly where it started, height included. It restored only the horizontal position before — correct only because nothing yet animates the height.

Also

• Ships of a wide front — many groups on one line — can still poke outside the Bubble, exactly as they always have. Only the part the ranks caused is fixed. Every way of covering the rest made the shield look worse: grown to fit it becomes a lens on old reports, capped it puts the shell wall through the middle of the outermost hulls.
• The ranks now read as more separated vertically than the depth cue that carries their meaning. That is the price of the fix and it was measured rather than guessed: allowing part of a hull to stay covered moves the spacing by about 5%, so there is no version of this that is both cheap and visible. A hull entirely behind a name plate is worse.
• None of this was ever looked at. The visual check could not run — the browser tooling would not load the local dev server — so every claim above is measured in code, and the rendered result is unverified.

1.0.0-beta.48 — 2026-08-17

Like the release before it, nothing here changes what a player sees or how anything in the game behaves. No deletion behaviour changed either — only whether it is written down.

Also

• The server records when an account goes away (#641). Novarchy deletes accounts automatically: whoever registers, plays under five minutes and never returns has their account removed, which frees the planet and releases the nickname. That has been running for a while and left no trace at all — a line in a log file and nothing else — so "how many did we lose last Tuesday" stopped having an answer the moment the container restarted. It is the one fact in this system that erases its own evidence: a deleted account exists in no table, ever again, so unlike every other number on the admin dashboard it cannot be reconstructed afterwards. Not stale — absent.

  It also repaired the release before it. The arrival chart added in beta.47 counted registrations from the accounts that still exist, so every deletion silently rewrote its own past: the sweep removing a hundred never-started accounts lowered the previous week's bars, and the chart showed survivors while reading as growth. Arrivals and departures are now drawn together, and the registration bars are repaired with the accounts that have since gone. A first attempt at that fix made it worse before it made it better — subtracting departures from a survivor count subtracts every deletion twice, which drew a growing population as a collapsing one — and the correction is now split into its two visible halves so a reader can check it rather than trust it.

  Two decisions worth recording. The record stores nothing that identifies anyone: no address, no nickname, no id. Releasing the nickname is the whole reason the account is deleted rather than just the colony, and keeping it in a metrics table would quietly undo that. What is kept is how old the account was and how long it was actually played — which answers the only question the cleanup's thresholds turn on: are we losing people who tried, or people who never really started?

  And the record is written inside the same transaction as the deletion, so a deletion that cannot be recorded does not happen. If the record ever breaks, the automatic cleanup stops rather than running blind. For something that erases accounts on a timer, stopping is the right failure.

1.0.0-beta.47 — 2026-08-17

Nothing in this release changes what a player sees or how anything in the game behaves. It is an instrument for whoever runs the server, and saying that plainly is better than dressing it up.

Also

• A dashboard behind the admin panel (#633). The admin panel could say who a player was and what could be done about them. It could not answer anything with a time axis: how many people arrived this week, how many came back the next day, where new commanders stop, whether the power curve is running away from the rest of the server. That half now exists — daily arrivals, return rates by signup week, a funnel over the fourteen First Steps showing which step loses the most people, the shape of the power and wealth curves instead of an average that describes nobody, and three lists: most active, least active, and who was playing and has gone quiet.

  The work was not the charts. Three of the numbers the server already keeps look like they measure activity and do not: a "last seen" stamp that any background request refreshes and that was backfilled for every old account, a colony timestamp that moves when somebody else raids you, and any headcount that pools the AI empires in with the people — a number that describes a game that is always doing fine. Everything on the panel is built instead on the five things only a player can do: queue a building, order ships, start a research, launch a fleet, send a message. Two service accounts that are not players are removed by name, and the admin's own empire — hand-edited while the game was built — is out of every curve that measures concentration.

  It also prints what it cannot see. Reading your battle reports and logging off performs none of those five things and counts as inactive, as do market trades and guild applications. The panel says so on the screen, because a dashboard that flatters itself is worse than not having one.

  One thing had to be recorded rather than derived: power, Dark Matter and the funnel are computed from the world as it is right now and have no history, so a daily row is written from today onward. Everything else is reconstructed from records the game already kept, which is why the panel can draw the last ninety days on the day it shipped.

1.0.0-beta.46 — 2026-08-17

Fixed

• The Warehouse card no longer reads like a production bonus (#450). The colony advisor led with +803 Fe/h, +2,915 Xe/h reaching storage, colony-wide. and mentioned only underneath that storage fills up within eight hours. Both facts were on the card — that much was fixed in #434 — but nothing joined them, so the natural reading was additive: it gives me +803/h, and my storage happens to be full. Players concluded the Warehouse raises production. It produces nothing at all: those resources were being thrown away over the cap, and the level catches them. Three words at the end of a list of numbers, reaching storage, were carrying the whole distinction, and the cause came after the effect. The card now says the cause first and the figure second, as one thought, with the denial in the same breath as the number: Storage fills up within 8h: everything produced over 32,768 is simply lost. then This level catches +803 Fe/h, +2,915 Xe/h that would otherwise be lost over the next 8h — the Warehouse itself produces nothing. The denial travels with the figures on purpose: the advisor's written comment replaces the first line of a card, so the figures are what a commented card shows on their own.
• A Warehouse level that costs you power says so, separately (#450). Each level draws a little more energy, so on a colony already short of power the rest of the buildings bank slightly less. That loss used to be mixed into the same list as the gains, which could print catches -2 Fe/h … that would otherwise be lost — a loss described as a rescue, in a sentence that also said the building produces nothing. The price is now its own line, naming what causes it: Its extra power draw pulls 2 Fe/h out of what the rest of the colony keeps. It is never shown alone: an upgrade whose only figure is a loss would be worse than the card that started this.
• Both lines are also the facts handed to the advisor's written comment, so the model can no longer repeat the old mix-up in its own words.

Left out

• Only the Warehouse. Every other building's card still prints a signed list, so -1 Fe/h, +15 Xe/h reaching storage, colony-wide. is still reachable elsewhere — measured on about 8% of cards during review. It is the same defect, but fixing it changes copy on every building, so it is tracked separately as #636.
• The recovery figure is a per-hour average over the eight-hour window, so multiplying it by eight overstates what a colony that has not started overflowing will actually lose. Per-hour is the convention on every advisor card, and changing it here alone would make one card speak a different language.

1.0.0-beta.45 — 2026-08-17

New

• A forming joint raid now calls you (#437). Joint raids kept forming and filling — 373 at full strength over a measured fourteen days — and in that window not one human commander was in any of them. The reason was not speed. The Joint raids door in the guild command room said strike together and nothing else, so a player learned a raid existed only if they happened to open that particular door during the window it was forming in — and a raid expires. A forming raid was only ever in front of whoever was already staring at the raid list when it opened. The door now says what is actually happening — 2 raids forming · next in 38m — and burns red while any of them is still open to you. The Guild tab carries a mark when one is waiting: on the desktop bar, on the phone's Menu button, and on the Guild tile inside it, so the call reaches someone who is not looking at the guild screen at all, which is where the hole actually was. Raids you have already joined, raids that are full, and raids that have already landed are not counted: a badge that overstates is a badge that stops being read. It costs no extra loading. The signal travels on the guild summary the game already refreshes every thirty seconds, so the rule the command-room doors were built on — a door loads nothing until you open it — still holds.

Fixed

• The joint raid panel keeps up while you watch it (#437). It loaded once and never again, so a raid a guildmate opened a minute later never appeared, even with the panel open in front of you — which is exactly the moment the panel exists for.
• The impact countdown was wrong the longer you sat there (#437). A raid twenty minutes from landing could read impact now to someone who had been watching the panel for ten. The same fault sat on the "allies under attack" clock, where it also mis-sized the guard fleet you sent next. Both are now measured against the real arrival time.
• A refused launch no longer erases its own explanation (#437). If your fleet could not reach a raid in time, the reason appeared and then quietly disappeared within half a minute, leaving a fleet that had not gone anywhere and nothing on screen to say why.
• A momentary connection problem can no longer look like leaving your guild (#437). Any failed refresh — a timeout, a server hiccup — used to empty the guild screen and show the join-a-guild browser, which on a phone also closed whatever panel was open. Only the server actually saying "you are in no guild" does that now.

Also

• Four refinements found while reviewing this work are deliberately not in it, and are tracked in #628: the mark has no dismiss, so declining a raid keeps it lit for the rest of the forming window; the call cannot know whether your fleet can physically reach the raid in time; a target that gained a shield after the raid opened is still counted; and the same for two guilds that signed a pact mid-window.

1.0.0-beta.44 — 2026-08-17

Fixed

• The leaderboard says which column it is ranked by (#435). A commander sent in a screenshot of the Points column with the note "numbers are not correct": read top to bottom it goes up three times. Every number in it was right. The board is ranked by the Gate, so the Points column is under no obligation to descend — and nothing on screen said so. Four numeric columns were drawn identically, exactly one of them governed, and the headers could not be clicked, so "let me sort by Points and see" was not available either. The only way to find out was to file a report. The ranked column now carries a ▼, and a line above the board states what orders it — in words, not in a tooltip, because a tooltip never opens on a phone.

New

• The columns can be sorted (#435). Gate, Points, Colony pts, Research pts and Colonies each re-order the view on click; the rows are already on screen, so nothing is fetched. Clicking Gate returns the board to its own order. Re-sorting changes the view and never the standing: the # column and the podium keep showing the real position on the board, so sorting by Points can put a #2 at the top — which is the thing that was reported, now visible instead of hidden.

Changed

• The Power tile in the header shows the Gate (#435). The same screenshot caught it reading 102,175 PP · S 58,138 · M 44,037 twenty pixels above a leaderboard row reading 58,138: two different quantities under the word "power", because the tile leads with the total while the board ranks on the larger of the two parts. For a balanced empire the headline is nearly double the ranked figure. The total stays the headline — that is what PP means on every other screen — and the Gate now sits on the breakdown line next to Structural and Military, taken from the server's own figure rather than recomputed, so the two screens cannot disagree by the one point that recomputing produces on about a quarter of empires.
• On a phone the board always follows its own order. That layout is a card list with no headers, so a sort carried over from a wider window could not have been undone without reloading.

Left as it is, and on purpose

• Sorting runs in one direction only, so clicking the column that is already sorted does nothing visible. Every ordering the feature offers is still reachable; an ascending/descending toggle would have doubled the marker vocabulary for a page whose job here was to say which column ranks it.

1.0.0-beta.43 — 2026-08-17

Fixed

• Buildings rest on the tile they stand on (#342). The base plate drawn in a building's art is the cell it occupies — measured across the sprites it is 98.9 world units wide against a cell of 100 — and it was being drawn floating above it, by about a tenth of a cell. While placing or moving a building the cell outline stuck out below and in front of it. Thirty-six of the seventy-eight building sprites now sit down onto their own tile, among them the Laboratory, the Factory, the Extractor, the Refinery, the Solar Panels, the Reactor, the Shield, the Batteries, the Cannon, the Command Centre, the Relay and the Terraformer.
• The Arsenal and the Factory stand at the same height (#342). They draw the same sprite and used to sit a few units apart. The Arsenal's hand-set value turned out to be the most accurate number in the file — 1.1 units from the measured one, against the Factory's 8.6 — so it barely moves, while the Factory comes down to meet it.

Not fixed, and deliberately

• The other forty-two sprites — the Fortress, the Hyperportal, the Warehouse, the Silo, the Sensors and several stages of others — are left exactly where they were, to the pixel. Their art has no base the game can identify: the widest part of the drawing is halfway up the body, or a ring on legs. Anchoring them on a guess was tried and measured, and it planted twenty-four of them below where any base could be, the stage-III Extractor by two thirds of a tile. So they keep the position players already know, and where they actually rest stays open on #320.
• Two consequences of that split are visible and worth naming: a building can now sit up to about one depth row lower than its neighbour, and a building whose stages fall on different sides of the split can shift when it reaches stage II or III — the Outpost by 17 units. Both are the remainder of the same old defect rather than something new, both are pinned by tests so they cannot grow unnoticed, and closing #320 removes them.

Nothing here changes a cost, a timing or a battle. The issue asked for the anchor to be converted from one fraction to another; it turned out there were two errors cancelling each other, and doing only what was asked would have made the symptom worse — the pull request carries the measurements.

1.0.0-beta.42 — 2026-08-17

Changed

• The Espionage sheet says what the deepest tier actually sells (#406). A tier-3 spy report has revealed where the defender's ships stand since battle formations shipped, and the Research screen — the only place the Discipline is bought — never mentioned it: it listed "fleets away, with return ETA" and stopped there. Nobody was told the battle line was for sale, which is the cheapest explanation for zero tier-3 reports out of the 1304 issued so far, and it had to be ruled out before anyone could conclude the level was priced wrong. The sheet now names it, and a test ties the sentence to the engine's own disclosure rule rather than to the number 17, so the copy cannot drift away from the gate.

Fixed

• Espionage was documented as being able to show you a Freighter wall. It cannot (#406). The rows in a report are the hulls that defend by default, so unarmed ships walled into the vanguard through a defence plan never appear in one — the plan is not spiable at any grade. The claim lived only in a code comment, so no player was ever shown it, but it was the justification for what the tier costs, and it was false.

Nothing in this release changes a battle, a price, or how any fleet fights. The agents' side of issue #406 — reading a target's rows and countering them — stays deliberately unbuilt, and the issue is still open with the measurements that say why.

1.0.0-beta.41 — 2026-08-17

Added

• A garrison on your own colony can be sent home before its watch ends (#356). You sent three colonies' fleets to defend a fourth, won the fight in the first ten minutes, and then watched those ships sit there for the rest of the watch — up to twelve hours — with nothing left to defend against. The watch length was fixed when the mission launched and there was no way to lift it. A garrison standing on one of your own colonies now carries a Stand down button: the fleet lifts off and flies home, the whole way, at normal speed. The flight home is the only price; there is no cooldown and no limit on how often you do it.

  The confirm sheet prints both durations before you commit — the trip home and what is left of the watch — because standing down is not always an improvement. At the end of a watch the ships are deposited home instantly, so late in a long watch standing down lands them later, and in that case the sheet says so in as many words, with the difference, rather than leaving you the subtraction.

  Two refusals, both stated on screen instead of discovered by tapping. While a hostile fleet is on final approach to that colony the watch holds — from the instant the attacker itself can no longer turn back until the battle resolves — because nobody walks out of a fight that has begun. And a garrison standing on someone else's colony still cannot be sent home early by anyone; it stays until the watch ends. The launch form now says both halves of that rule before you commit a fleet.

  What this does not include, deliberately: releasing a garrison that someone parked on your colony. That is the host's side of the same feature, it needs its own design, and it is not in this release — so a defence you pledged to an ally is still a defence you cannot take back.

Changed

• A garrison's card now reads "Standing guard, home in …" (#356), where it used to say "lifts off in". At the end of a watch there is no flight afterwards: the ships are home at that same instant. The old wording implied a return leg that does not exist, and made standing down early look like a pure gain in every case.

Fixed

• Turning a defence around while it was still in flight is logged as a recall again (#356). It had begun appearing in the activity log as "Garrison stood down — the garrison lifted off on your order", over a fleet that never landed and never mounted a watch.

1.0.0-beta.40 — 2026-08-17

Fixed

• The galaxy map no longer opens with its top row hidden behind the HUD (#421). On a phone the map fills the whole screen and the panels float on top of it: the identity band and the filter chips across the top, the navigation bar and the prizes strip along the bottom. The opening view was fitted and centred against the whole screen rather than against the part of it you can see, so on any phone with a notch the northernmost systems were drawn underneath the chip row — 167 pixels down a 664-pixel screen whose panels end at 195, and 141 on a 600-pixel one. The map now measures those panels and frames itself against the clear space between them, so the whole hive opens where you can see it and ⤢ Fit, ⌖ Home and the edges of a drag land there too. Tapping still selects the system under your finger.

  On the two phone sizes where systems were hidden the map opens smaller — 84% and 70% of its former size — which is what it costs to show all of it inside a band shorter than the screen. Where the map was already fully visible, the opening view is unchanged to the pixel.

• What the report got wrong, for the record (#421). The complaint was that the map could only be dragged sideways and never up and down, and it named the drag limits as the cause. Neither held up: vertical dragging works and always has, and the limits keep a pad around the hive that already put every system in clear space at the edge of a drag. An earlier attempt built on that diagnosis (PR #461) was measured and closed without shipping. What was true is the second half of the report — that the panels cover the top — and the file the issue pointed at was not even the one that runs: it described the old perspective map, reachable now only from a dev page.

Changed

• Turning the phone sideways is deliberately untouched (#421). In landscape the panels keep the heights they reserve in portrait and take about two thirds of the screen, so there is no spare room to reframe into and shrinking the galaxy to fit would have drawn it at a third of its size. Landscape keeps exactly the view it has always had, including the top row sitting under the chip row, until the panels themselves stop reserving portrait heights (#616).

1.0.0-beta.39 — 2026-08-17

Added

• Missions start from the colony you are standing on (#334). The launch pad always opened on the home colony, however far from her you were, so anyone running a forward base corrected it by hand at every single launch — spy missions included. It now opens on the colony the screen is showing, and the destination opens there too instead of on the mother's address. It keeps following the colony selector until you pick a launch pad yourself, or until you start cutting a fleet: from then on it stays where it is, because moving it clears the ships you picked.

• Keyboard shortcuts for switching colony (#334). 1–9 jump straight to a colony in the order the selector lists them, C cycles to the next, Shift+C goes back. They do nothing while you are typing in a field, while a panel or the battle replay is open, or when held with Ctrl/Cmd/Alt — and they are desktop only, where there is a keyboard. The colony dropdown's tooltip announces them, since a shortcut nobody is told about is a shortcut nobody uses.

• Your home colony is offered as a mission destination again, and marked (#334). The shortcut row deliberately hid it, which was right while every fleet left from home and wrong the moment the launch pad could be somewhere else: hauling loot back to the capital is the commonest transport there is, and it was the one destination the row refused to draw. It now hides whichever colony the fleet is leaving from, and tags the capital MOTHER.

• Spy reports say whether a shelter is up, and when it lifts (#334). The report never mentioned the one fact that decides whether the fleet you are assembling can leave, so the player who reported this had taken to launching probes on attack missions purely to read the refusal. A report now carries the expiry — shielded until 9:40 PM, novice shelter until 8/20, 11:15 AM, or plainly that neither was up — and shows it on the closed card, so a target can be chosen without opening every report in the feed. There is no research gate on it: the fact that a colony is shielded is already free on the galaxy target board, and a report that costs a probe should not say less than a free list. What it adds is the moment the door opens.

Changed

• A spy report is explicit that it answers about shelters, not about attackability (#334). It covers shields and the newcomer shelter and nothing else, so it never shows a green light: pacts, the night window and relic locks refuse launches too, and the night window in particular stays private by design — an opponent's sleep is theirs.

1.0.0-beta.38 — 2026-08-17

Changed

• Guilds start at 25 seats instead of 15, and grow to 50 instead of 30 (#611). The old cap was doing the opposite of its job. It was written as an anti-hegemony rule — Clash of Clans' 50 scaled down to a small server — but in play nobody was ever pressed against the ceiling: what a new guild hit was the floor, on day one, with the first five extra seats 700 XP away, about a full eight-week season. And the anti-hegemony job had already been lost, because the biggest guilds here were founded before the rule existed and were never bound by it. The cap therefore constrained only the new guilds — the ones it had been written to protect, and the only ones a fresh signup can actually join.

• The first widening now arrives at guild level 2 instead of level 3. That is roughly one month of Directive tiers rather than a whole season, so a young guild sees the door widen while it is still recruiting. Seats now grow at levels 2, 4, 6, 8 and 10 — five steps instead of three. The XP curve itself is unchanged; what moved is which level pays out.

• Guilds that were already larger than the cap keep their size, as before. With the ceiling at 50, the guild frozen at 28 members is overtaken at level 2 and recruits on the ordinary rule again; the two above 50 keep their size and their turnover and do not grow.

• The ceiling is 50 and not higher because member count is load-bearing elsewhere: it is the aid board's limit, past which a real member's open request can be pushed off the list. Two other numbers were measured before settling on it — the full research Address costs 28% of a season's donation capacity at the base cap, inside the band the design defends, and the Eschaton Project's cost in days halves as the roster doubles.

Fixed

• "That guild is full" now names both ways a seat opens (#611) — the guild levelling up, or somebody leaving. It mentioned only the second, which is now the rarer of the two.

• AI guild leaders no longer claim to be full when they are not (#611). The rule they fell back on used a hardcoded number, so a guild with fourteen of twenty-five seats taken could turn a player away as "full" while the panel beside it showed the free seats. They now read the real capacity and quote the actual numbers.

• An AI leader no longer welcomes a player it has just turned away (#611). In some replies the refusal and the greeting came from two separate decisions, so the guild channel could announce the arrival of somebody who never appeared in the roster.

• The guild aid board no longer loses a slot to a departed member (#611). A request opened by someone who had since left kept occupying a place in the list, which could push a current member's open request off the board — leaving them unable to withdraw it or receive help until it expired.

Also

• Guild levels still unlock seats and cosmetics only: no production, research or combat bonus at any level.

1.0.0-beta.37 — 2026-08-17

Added

• A ship reference in the Shipyard: the whole fleet catalog, and who beats whom (#332). The counter table is the most decisive thing in a fight — a hull deals from ×1.2 to ×5 the damage against what it counters — and until now a player could only see it one ship at a time, on whichever card happened to be selected. Asking "what do I build against a Titan?" meant clicking through sixteen ships and holding the answer in your head, so in practice the table was learned by losing a fleet to it. A new button opens the whole catalog at once, grouped by class, each hull with what it beats, what beats it, and by how much. It was asked for on the forum by a player who wanted to know what each ship type is actually for.

• Each row also says what a ship is for, not only what it kills. The Freighter, the Spy Probe, the Colony Ship and the Colonial Ark state their job where the counter table has nothing to say; the Shroudrunner and the Spectral Corsair admit they live off freighters and colony ships, which their combat line hid — a Corsair used to introduce herself as "Overpowers cruisers" and say nothing about the trade she was built for.

• The planetary defences are in the reference too. They are not part of the ship-vs-ship table, so a page built from that table alone showed the Bomber with no bonus against a defended colony — the opposite of what she is for. Her ×2.5 against Defense Batteries and against the Orbital Fortress is now on her row, and the light hulls are told that the flak is tuned for them, which no ship-vs-ship column could ever have shown.

• The reference opens with or without a Shipyard built, because a player who has not built one yet is exactly the one who needs to know what a hull is for, and it names the research that unlocks each of the seven gated ships instead of leaving a beginner to save up for something they cannot order yet.

Changed

• The Destroyer is marked Event Relic — not buildable in the reference, in the same words the build grid already uses: this is the page people read before deciding what to save for, and "event relic" on its own reads as a rarity to chase.

• The Colonial Ark no longer has her own description undercut by the generic "Support ship" label one line below it — the wording removed from her role line in #333 was still coming back through the sentence underneath.

Nothing about combat itself changed: the multipliers on the page are the ones the engine has always applied. What changed is that they can be read before the battle instead of inferred from having lost one.

1.0.0-beta.36 — 2026-08-17

Nothing changes in the game this release. No content, no balance, no fix a player can see. It is here because everything that reaches main gets a tag and an entry, including the times there is nothing to show — a changelog that only speaks up for the good weeks is not a record of anything.

Also

• A new internal workflow for taking one tracked issue from "reported" to "live" (#606). Its first job is not writing code: it decides whether the issue is still real against what is actually deployed. A fair number are not — already fixed, superseded, or made moot by something shipped since — and those are now closed with the evidence that made them obsolete rather than reimplemented by someone who did not know. What is still real gets implemented test-first and goes through the same production checks as every other change.

  The effect players might eventually notice is indirect: fewer changes made for reasons that stopped being true, and a clearer public record on the repo of why something was closed and what closed it.

1.0.0-beta.35 — 2026-08-17

Fixed

• The night window could be aimed at whoever was attacking you, instead of covering your night (#540). Night protection is an 8-hour window in which your colony cannot be hit, and the player picks when it starts — the only protection whose parameter is chosen by the person it protects. Until now that hour could be re-picked at any moment, instantly, for free, and with no record kept, so the eight hours did not cover the defender's night: they covered whoever was currently trying to reach them, for as long as they kept trying. It was reported from the receiving end by a player who spent 24 hours unable to land a single attack on a target and assumed the timer had been cheated. Nothing was broken — that is what free re-aiming looks like from the other side. The start hour can now be moved once every 24 hours: eight hours chosen once a day cannot cover both halves of an opponent's day, so the sixteen hours the design always meant to leave open really are open, and aiming your window at someone costs you the rest of your own day. Everyone's first move is free whatever they did before, and re-picking the hour you are already on costs nothing, so brushing past the dropdown cannot spend your day.
• Two requests fired at once could both spend the same day's move (#540). Found in review, not in play: the check and the write were separate steps, so a double-tap, or the same account open on a phone and a desktop, could get both through and leave the saved hour disagreeing with the one on screen — locking the window for a day on a value the player never chose. The refusal is now decided in the same operation that writes the hour, so exactly one of them can win.

Changed

• The night-protection dropdown says what it costs before you spend it (#540). It now reads "One move every 24 hours" while it is still free, and says when it comes back while it is locked — on desktop, tablet and phone alike. Before this it explained itself only on phones and only after the move had been made: everywhere else a locked dropdown simply stopped responding, with the reason hidden in a tooltip that browsers never show on a disabled control. A locked dropdown is also dimmed now, instead of looking identical to a working one.
• Moderation can see when a night window was last moved (#540). A report of the shape "I can never hit this player" previously had nothing to check: the hour was a single value that could be rewritten at will and kept no history.

Known issue

• An app build older than this release does not know about the cooldown, so a second move inside 24 hours shows a short technical code instead of a sentence. The move is refused correctly; only the wording is wrong, and it resolves on updating.

1.0.0-beta.34 — 2026-08-17

Fixed

• Upgrading the Shipyard now makes ships faster, at every level (#593). It didn't: raising the Shipyard made hulls take longer, and stayed that way until level 25. A starter discount faded as the yard grew, faster than the yard itself sped things up, and the two pulled against each other until the discount won — level 1 → 2 nearly doubled build times, and level 4 was the worst point on the curve, 3.3× slower than level 1. On a building the game sells as an accelerator and charges energy and resources for at every step. arcticpaladin found it the expensive way: Shipyard 10 with a Nanoforge 9 building Spectral Corsairs slower than a bare Shipyard 1. jsnam had reported the same thing a year earlier from the other side (#318), after finally upgrading a yard left at level 1 for months; that issue made the building sheet announce the slowdown honestly and left the curve alone, which turned out not to be enough — a cost you can only avoid by never upgrading is not a disclosed trade-off. Now every level is a real speed-up, on every hull, in every empire.
• The fast first fleet survives, measured in tonnage instead of Shipyard levels (#593). New colonies still get their starter discount, opening at the same value it always had; it is now spent by launching ships rather than by upgrading the yard, so it can never turn into a penalty on the upgrade. Nothing gets faster than before for a developed colony, and no established colony changed speed on release day. What ends is being rewarded for never touching the Shipyard.
• The Shipyard screen shows the real build time, and says what is behind it (#593). The panel now prices an order exactly as the engine will — before this it could announce nearly four times the true wait on a new colony — and, while the first-fleet allowance is open, a line under the number says how much is left, that times rise toward normal as it is spent, and that a larger batch spends more of it. Same for the Shipyard's own sheet, including before the building exists.
• The colony advisor stopped selling the slowdown as a benefit (#593). It compared build times across the defective curve, so on the two upgrades that made hulls slower it recommended the Shipyard with a line reading "hulls leave the ramp 48% sooner". It now reports a speed-up only when there is one, which since this release is always.

1.0.0-beta.33 — 2026-08-17

New

• A fleet in flight can be recalled (#591). Until now nothing could: a mistyped sector meant watching the fleet fly the whole wrong journey, twice. Every outbound mission card now carries a ↩ Recall button; a confirm sheet states every price before the tap, and the fleet turns around where it is. The rules the sheet spells out: recall works during roughly the first three quarters of the flight — the last stretch (up to 10 minutes) is committed, and very short hops are never recallable; the way home costs the distance already covered at normal speed, so a warp jump only ever paid for the way out; nothing is refunded (warp Dark Matter, a newbie shield forfeited at launch, the daily exploration attempt); cargo stays aboard and lands back home, clamped by the Warehouse like any return. Raid fleets fly with the raid and cannot turn back alone; a garrison already on station serves its full watch.
• A recalled attack leaves a trace (#591). If the defender saw the attack coming, their diary closes the alarm — "Inbound flight turned back", with the arrival time it would have had — instead of the countdown silently vanishing. Feints are a legal move, but they are visible, and at most 3 attacks can be recalled per day.

Changed

• The guild-shop item formerly labelled "Fast recall" is now "Fast return": it still cuts a quarter off a returning trip (recalled fleets included). Only the label changed — "recall" now names the new manoeuvre.
• The bounty rule that forgave a hunter whose pot was claimed while their fleet was in flight now forgives only claims landed after the point of no return. While you could still have turned back, pressing on is a choice, and it is priced as one.

Also

• Mission cards say "· turning back" instead of "· returning" when the return was the player's own order, and the activity diary records every recall on both sides.

1.0.0-beta.32 — 2026-08-17

New

• The research panel shows what the next level will cost — while the current one is still running (#311, asked by jsnam in #bugs-and-ideas). During an order the panel used to show only the price of the level already being paid for; the number a player needed to stockpile for was nowhere. Now the cost row, the research time and the effect preview all describe the level the button can buy next, marked "after lvl N completes" so it can never be confused with the order in progress. The level badge stays the committed one.
• The next level of the same Discipline can be queued right away (#311). The button no longer locks on "Already researching": the follow-up order waits unpaid in a queue slot and starts — and is charged — when the current level completes. Its queue row says exactly what it is waiting for ("starts after lvl N completes"), instead of the false "when a lane frees up".

Fixed

• The research queue is now strictly serial per Discipline in both directions: through the raw API an order could previously start out of order and the leftover twin later paid the full price of a level that changed nothing.
• A queued-but-not-started order now always needs a free queue slot, never a lane slot. Without this, a stacked order at Laboratory 8–11 (where there are no queue slots) would have silently frozen the whole research queue for up to four hours, with no way to cancel it.

Changed

• The capacity strip above the research queue fills its pips by what actually occupies them — squares for running research, round pips for queued orders. It used to fill by position, painting a busy queue slot as a busy lane and vice versa.
• When stacking is blocked because no queue slot is free, the panel says so in words and points at the Laboratory level where the network unlocks its first slot.

1.0.0-beta.31 — 2026-08-17

Fixed

• Gas planets: the buildable grid no longer leaves the platform (#289, reported by Tibeh01 with a screenshot worth more than any description). On gaseous colonies the 14×14 build grid stuck out past the floating deck: Move mode highlighted destination cells over the void, and buildings on the far rows stood visibly on the clouds. The placement itself always worked — the building moved and produced — but the eye saw a base built on thin air. The painted platform is now scaled and centred so that all 196 cells land on its tiles, and the fit is pinned by a test that measures the deck from the shipped artwork (and refuses to certify a different image, so an art swap cannot quietly bring the bug back). Three calibrations were rejected by review before this one: the deck turned out to be a rounded perspective quad, not the tidy diamond a bounding box assumes.

Changed

• Gas colonies now open at the same framing as every other planet type — they were enjoying an accidental extra zoom-in on phones and windowed desktops, and the zoom-out button did nothing there. Same recalibration, both fixed.
• "Fit to view" (⤢) on a gas colony no longer leaves a band of empty darkness along the screen edge on portrait phones: the reset now respects the same pan limits as every other camera movement.

1.0.0-beta.30 — 2026-08-17

Changed

• Raids are now priced on the pack's combined Power (#285, reported by arcticpaladin in Discord). The anti-bullying bounty judged each raider alone, so eight empires each just under the band could crush a single defender together and generate zero bounty — the public price that was supposed to make bullying "legal but prohibitive in practice" fired on 1.7% of attacks, because the dominant attack shape is the raid, not the duel. Now the raid's summed Power is judged against the defender's band at impact; each member pays their Power's share of the price — a probe from a giant weighs what the giant weighs — never less than the same strike would cost them alone, and capped per member (150 DM), not per battle, so a monstrous pack cannot hide behind one attack's ceiling. A solo raid prices exactly like a solo attack. Faction bastions, relic sites and heads still wanted at impact are never priced, as before.
• The free return strike no longer burns inside guild raids (#285). A retaliation credit now spends only where it covers the whole bill — a strike of your own — instead of being silently consumed by a share of a pack's price. It stays whole for the personal reply it was earned for.
• Every door tells the rule before ships are committed (#285). The target board's "inside your band it costs nothing", the planet sheet's "fair target", and the war room's "strikeable" now say they describe the solo strike; the raids panel states the pack rule in full; and the diary names the raid — "the raid you joined far outweighed its target" — when the pack, not the player's own gap, is why they were billed. On bastion and relic sheets the raid clause is deliberately absent: raids there are never priced, and the old text would have talked a guild out of exactly the captures raids exist for.
  • Deliberately NOT changed: coordinated solo missions in the same hour stay priced per-attacker — measured at ~10 windows per fortnight among humans, never more than three, with no aggregating entity comparable to the raid.
  • Filed as follow-up in the PR: the agents' strategist still prices targets with the solo band when joining raids; and a bounty_waived_grudge event would make the retaliation exemption measurable.

1.0.0-beta.29 — 2026-08-17

New

• A stale client finally hears about the deploy (#568). The game could already say "Reload the page: your client is out of date" — but only reactively, on a write the server refused. A player who only looked (map, leaderboard, colony) never touched those error codes, so a tab opened before a deploy kept showing old numbers as current; that is exactly how the "Planets not in space view" bug report was born from a client that was merely old. Now every build stamps itself (version.json, a deterministic hash of the emitted bundle) and the client checks the stamp of its own origin — at load, every five minutes, and when the tab refocuses after sleeping through a deploy. When the server hands out a different build, one quiet row appears in the HUD's banner family (and on the login screen): "New version available", a compact Reload button, a dismiss ✕ that silences that build and re-arms on the next one. Every failure mode — network error, 404, HTML where JSON was expected — reads as "no information", never as "out of date", and a rollback takes the notice down by itself. The check deliberately ignores the api container: web and api are recreated independently, and an api-borne version would report a skew a reload cannot fix. In the native app the watcher stays off — updates come from the store, not from a reload.

1.0.0-beta.28 — 2026-08-17

Changed

• Lopsided offers sink to the bottom of the Market (#308, reported by Tibeh01). Offers whose two sides differ by 20× or more in sheer quantity — the 1 Fe ⇄ 5,000 Xe kind, usually a gift for a friend — sat at the top of the board like any other trade, and every one left open was a row the whole server scrolled past. The list now shows regular offers first; anything wildly off-price drops below a "lopsided offers — giveaways or steep asks" divider, on every filter. Nothing is hidden and nothing is blocked: validateOffer is untouched, gift offers still work exactly as before, and their recipients still find them. The divider is a real heading screen readers can jump to, demoted rows say "lopsided offer" out loud, and on phones the label wraps instead of being clipped.
  • Deliberately NOT done: a per-resource minimum (option 2 of the issue) — a floor of 100 still allows 100 Fe ⇄ 1,000,000 Xe; the mechanism is the ratio, not the quantity.
  • Filed during this ship's analysis: the market's accept side needs no Trade Portal and no daily cap, so it doubles as an uncapped alt→main value pipe (#584).

1.0.0-beta.27 — 2026-08-17

New

• The Shipyard says how much Power each ship adds (#292, suggested by Tibeh01). A ship's sheet showed hull, shield, attack, cargo and speed — but not the number the rest of the game keeps scoring you on: how much that hull moves your empire's Power. Every detail sheet now carries a Power per unit row, computed with your own research and tree bonuses applied (the same Corvette is worth more to a weapons-research empire), and the build block quotes the whole stack being ordered as Power when delivered — exactly the number the engine will add, in the engine's own arithmetic, once the ships leave the queue and exist. Both figures wear a ≈: the per-unit value is rounded for reading and the header rounds your total, so the observed tick can sit one point off the printed delta — the hedge says so instead of leaving it to be discovered.
• What the number is not, in writing. Power is a score, not battle strength — counters and formations decide battles — and the Gate that prices bounties and protection bands is the larger of Structural and Military, so a delivery moves it only by the part that ends up above your Structural. The note under the stat says both, in the same words the ⚡ dialog and the Leaderboard use, so a player cross-checking the claim can actually find the numbers it talks about.

1.0.0-beta.26 — 2026-08-17

Added

• The "Recent" row on the Missions tab can be pruned (#291, suggested by Tibeh01 on Discord). The row of recent coordinates filled up with every target the feeds touched, and the two or three a player actually reuses were buried with no way to clean it. Every chip now carries two small controls. × removes a chip — it stays gone until something genuinely new happens at those coordinates (a fresh spy report, a battle, an arriving fleet), then it returns on its own. ☆ pins a chip: pinned chips lead the row and survive even after every report about those coordinates has aged out of the feeds — the chip falls back to bare coordinates instead of vanishing. Pins are capped at 8 so new contacts always keep some slots; the counter under the row goes gold at the limit and says to unpin one. Removals and pins live on the device, not on the account, and a player's own colonies cannot be removed — they are destinations, not history.

Changed

• The Recent row explains itself on screen: a short rendered legend under the chips says what ☆ and × do — no hover needed, which on a phone means no guessing. Chips whose label is just the coordinates no longer print the address twice, and the pin/remove buttons are full-size touch targets on phones (the 44px floor the rest of the app already respects).
• Left open on purpose, filed as #586 during this ship's review: the "debris ≈ N" chip kind of the same row turns out to be unreachable — the battle entry from the same report always shadows it — and has been since the row was introduced. Deciding where the debris estimate should live is a design call, not a fix.

1.0.0-beta.25 — 2026-08-17

Added

• Coordinates quoted in chat are now links to the galaxy map (#570). Agents and players quote places like 1:3:3 constantly, and for a newcomer they were pure noise — feedback from a live demo: "like being a 70-year-old on Instagram where everyone drops quotes you can't follow". Now any sector:system:planet triplet written in a message is clickable and opens the galaxy centred on that system, on every chat surface: galactic channel, guild council, whispers, desktop dock. Only real places link — times like 12:30:45 or zero-padded 09:15:10 stay plain text — and the persisted message is never touched: the parsing lives entirely in the render.
• Event Sector links know about the season. Chat history outlives seasons, so a 9:x:y quoted mid-season could otherwise drop the player in a sector they never asked for once the season closed. Sector-9 coordinates are live links while the season runs, and dim to plain text carrying the reason (on hover) when it is over — or not yet known to the client.

1.0.0-beta.24 — 2026-08-17

New

• Active missions now say what the fleet is carrying (#307, suggested by Tibeh01). The mission cards in the Missions tab showed type, route, ships and countdown — but never the cargo, so two transports on the same route were indistinguishable and "which one has the Xenogas?" was a memory exercise. Every card now carries a Hold: line under the fleet with the non-zero resources aboard, no click needed. The line follows the fleet, not the mission type: an outbound transport (or colonize dowry, or transfer) shows what is being sent, a returning transport shows only what a full destination warehouse turned around, and a raid flying home shows its loot. Empty holds — attacks on the way out, spy probes — add no line at all. The data was already in the API response; the game just never showed it in flight, only in the Transport delivered event after the fact.
  • Found and filed during this ship's review: a raid returning over the warehouse cap still loses the excess loot silently at landing (#587). The card's number is what the fleet truly holds — the landing is the part left to fix. (Fixed in beta.51.)

1.0.0-beta.23 — 2026-08-17

Fixed

• A decoy colony no longer pays more than a real defence (#270). Since beta.8 the Shield pool only counted toward battle points if the defence fired at least one shot — and one shot of any size bought the whole pool back: a single 400-resource interceptor parked on an otherwise empty colony scored 131.57 points, above the 115.57 of a target fielding a real defence, and a one-time level-1 battery did the same for 370. Now the pool pays in proportion to what actually stood against the attacker: it counts up to 10 HP for every point of firepower the defence fires. The same one-interceptor decoy pays 19; the token battery, 29; a colony that genuinely defends is untouched. Left open on purpose, as the issue itself scopes it: the fortress doctrine's 300 points come from the pool re-charging every battle (#255) and are not touched by this criterion.

Changed

• The new scoring applies from the deploy onward. Battles resolved before it keep their points — nothing is recomputed — so the week's war leaderboard can show different scores for identical setups depending on when the battle happened.
• beta.8's advice is superseded. «A single emplacement is enough» to make attacks on you expensive is no longer true: the Shield pool counts for attackers in proportion to the standing defence, in full once the defences fire about a tenth of the pool's hit points.

1.0.0-beta.22 — 2026-08-16

Added

• AI commanders can now finish a siege (#268). Forced Exodus was a one-way street: the game's AI empires could have their colonies razed and deported, but could never declare a razing themselves — the only irreversible act of the game was exclusively human, and the war record showed it. Now a full siege bar can end in fire whoever is besieging you. They declare it the way a player would: usually in a declared war or against someone who wronged them, rarely out of opportunism, never against NPCs or relics — and the weekly League fixture stays sport: it is never a reason to erase someone's home.

Changed

• AI commanders own their razings (#268). Before, the razer's memory recorded a razed battle as a generic victory: it could deny its own razing to its victim ten minutes after the proclamation. Now the razer remembers who it deported and from where — in chat and in its plans — and the deported commander's plan reads who razed it and where it resettled. The conqueror's proclamation follows character and waking hours instead of firing every time, the razing story stays colony-sized on every surface (erasing one colony never reads as deporting a whole empire), and if the siege bar decays between aiming and firing the AI attacker withdraws the declaration and attacks anyway — what the launch screen already tells everyone to do — instead of silently losing the strike.

  Left for a follow-up, so this entry does not promise it: AI commanders still do not announce a razing in chat before launching it.

1.0.0-beta.21 — 2026-08-16

Fixed

• A decoy outpost no longer pays the bounty (#255, point 1). Before: the payout of a bounty claim was measured only against what stood on the struck colony — a number the wanted player controls. With the post-raid shield gone, a wanted player could park one cheap ship on an emptied outpost and have an accomplice halve the pot on every pass at travel speed, wiping their public bounty — the anti-bullying deterrent — in a handful of hits, and hunter rotation defeated any per-pair cooldown. Now the claim is measured against the larger of the colony's fielded defence and 15% of the wanted player's war capital — warfleet everywhere, in flight included, plus batteries — so a decoy is worth a decoy, and a real hunt that overcomes the target pays exactly as before. The full reasoning, including why batteries and not fortresses, lives in DESIGN-TAGLIA.md.

Added

• Attacking while shielded now warns you before the click (#255, point 3). Before: launching any attack — even a lone probe at empty coordinates — silently cancelled every shield you held, the 72-hour exodus shield included; a freshly deported player could burn three days of protection by scouting, with nothing on screen saying so. Now both attack doors say it in advance: the Missions preflight for solo attacks, and the guild raids panel for opening or joining a raid — where Bastion and Relic raids count too, only the Pirate Beacon is exempt, a raid at empty or veiled coordinates keeps your shields, and a refused launch loses nothing. The rule itself is now written in DESIGN.md §3.8 instead of living only in code comments.

1.0.0-beta.20 — 2026-08-16

Fixed

• Two identical buildings with two different prices now explain themselves (#245). A second Site of the same building costs ×1.5 more at every level — a deliberate rule (×1.5^(n−1) on copy n's whole curve) that no pixel ever stated. Two players ten days apart (Tibeh01 with two Extractors, Ajax with Orbital Panels) did the only reasonable thing and filed it as a math bug, and each report took a trip into the engine source to refute. Every surface that prints a copy's price now carries the multiplier next to the number — ×1.5 (Site 2) — with one shared spelling: the Building Sheet (catalogue mode included, so the surcharge is stated before the copy is built, not discovered at the first upgrade), the Build menu row, the map tooltip, the Colony list (the one screen where Site 1 and Site 2 prices sit on adjacent rows), and the advisor's suggestion cards, whose ranking otherwise read as quoting wrong prices.
  • The note explains the price, it does not rank options: on a real colony the un-surcharged Site 1 is often the most expensive one to level, because the level term dominates the copy term. The number to compare stays the effective cost, which is unchanged.
  • The multiplier is printed exact (×3.375, never ×3.38): a rounded label overstates the charge, and a label meant to end the "wrong math" reports must not seed the next one.
  • First Sites show nothing — there is no surcharge to explain — and the prices themselves are untouched: this release changes what the game says, not what it charges.

1.0.0-beta.19 — 2026-08-16

Fixed

• A negative Dark Matter balance is explained instead of looking like a bug (#238). awsomepossum, in Bugs: sending a mission, the Warp row read Not enough Dark Matter: -263.75/2 DM, and their conclusion was the only one available — "I never completed any bounties so I should have 0". The balance was right and the mechanic is deliberate: bounties claimed on you are paid from you, down past zero. The game had simply never said so.
  • The defect was the gate, not the wording. Every Dark Matter surface was shown only to players who had unlocked the Particle Accelerator — but a balance goes negative through bounties, which need no Accelerator at all. Of the 46 players in the red when this was measured, all 46 lacked one, so the number was hidden from exactly the people carrying it. The header tile and the phone pill now appear on a debt as well; a newcomer sits at zero, which is not a debt, so nobody else gains any noise.
  • The tile turns gold and says in debt — but keeps showing the daily rate when there is one, because for a commander who does produce, the rate is the part that says when the hole closes.
  • The one thing the note refuses to do is promise. It does not claim the Accelerator will pay the debt back, because Dark Matter also arrives from anomalies, guild seasons, the League — and from claiming a bounty yourself, which pays a debtor straight into the hole. It says what is true for everyone: what you earn goes to the debt before it reaches your stock.
• Being hunted while wanted now leaves a trace on your own log. A claim wrote one entry, for the hunter. Whoever actually paid got nothing — Dark Matter left their balance, often below zero, and the only sign was a spending error hours later. The target now gets their own entry: who collected, how much, and where. It deliberately does not quote a resulting balance: the figure the entry could see is not the one the screen shows, and a guild raid would file several contradictory ones in the same second.

1.0.0-beta.18 — 2026-08-16

Changed

• The marker on a building's extra instances is Site N, not copy N (#235). Tibeh01, in Bugs: the second Orbital Panels read · copy 2. The datum under it was right — two plants do exist, they level independently, so something must say which one you are reading — but copy is the name of the storage key (pannelliOrbitali:2, instanceKey), not of the fiction. For the player that building cost resources, a layout cell and a Command Center 7.
  • The term now has ONE definition, INSTANCE_TERM in shared/model.ts, with instanceSuffix ( · Site 2) and instanceSuffixParens ( (Site 2), the Advisor's form — same word, different bracket, kept deliberately).
  • The issue named two call sites. There were fourteen, in three spellings: nine · copy N, two (copy N) in packages/shared — a second spelling that had already drifted in unseen, because the guard walked apps/web/src only — plus the instance_locked error text, the required for copy N hint, and the Command Center unlock rows.
  • Ruled out: a Roman numeral (Orbital Panels II), the form the issue proposed. BuildingSheet already prints Stage I · II at lvl 10 on the same card. And the constraint a first draft recorded here — that the form must stay a suffix because the espionage report sorts by the rendered label — was false: #230 replaced that sort, and sortedByCatalog orders on the engine key. The error is written down in the docblock rather than quietly deleted.
• The Command Center unlock rows say what the level actually grants. 2nd copy of economic buildings → 2nd Site for **each** economic building: maxInstances is per building, so CC7 raises all five of DUPLICABLE_BUILDINGS to 2. Those rows render on the sheet and are passed whole into the Advisor's LLM facts.

Fixed

• The Advisor claimed a standing building did not exist. advisor.service built the isFirstBuild fact from the bare catalogue name, but isFirstBuild is level === 0 — true of a never-built instance 2 while instance 1 stands at level 8. The model was handed «Ferrolite Extractor does not exist here yet», the prompt forbids contradicting a fact, and its why renders inside the panel under a row reading · Site 2. Found by two independent review lenses, not by the issue. The sentence is now firstBuildFact in buildAdvisor.ts, next to suggestionLabel — same species of string, same place, and pinned.

Guards

• The «one spelling everywhere» check was inverted. It used to assert that some other file spelled the marker inline, so it passed by finding the duplication and could only ever certify that it still existed. It now asserts that nobody hand-spells it except the file that defines it, and walks packages/shared too — the blind spot through which (copy N) arrived.
• Every negative assertion is bound to INSTANCE_TERM instead of the literal word: written by hand they would have gone true by construction after the rename (nothing contains copy any more), staying green while guarding nothing. The concrete term is pinned in two explicit places so changing it stays a decision.
• BuildingSheet is pinned by an actual renderToStaticMarkup, including that the first instance carries no marker — the draft #348 discarded twice, and something no source scan can see. The comment justifying source-scanning («there is no component renderer in this project's setup») was inherited unchecked and is false: two dozen .test.tsx files render, two of them render this very sheet.
• The census of surfaces is derived from the tree instead of counted against a literal beside it, and advisor.service is pinned to call firstBuildFact — proving a helper correct does not prove anyone uses it.

1.0.0-beta.17 — 2026-08-16

Added

• Chrono shards can be spent from the Laboratory (#232). Tibeh01 asked in Ideas & Suggestions for "a way of using the Chrono-Shards on the Laboratory, to fast-forward the research". The way already existed — crono.service has accepted queue: 'research' since shards shipped, and CronoPanel has rendered it since #276 moved the panel under the research lanes. What did not exist was a door on the building that does the research. The reporter had 720 shards in inventory and looked there first, which is where anyone would look.
  • The row names the lane a shard would actually cut — defaultTarget, the longest remaining, the same rule the server applies when no target is named. Naming the head of the queue instead would have announced one order and shortened another: #234/#261 one screen upstream.
  • Two destinations for one gesture, because the panel lives in two places. On a phone it opens the Base's «Speed up» sheet on the research queue; on desktop that sheet does not exist (it is a phone construct), so it opens the Research tab and scrolls the panel into view through a shared CRONO_RESEARCH_ANCHOR constant. No third copy of the panel: entry points multiply, views do not.
• The «Speed up» sheet offers all three queues. It listed Construction and Shipyard. On a phone it is the Base's only door to the shards and it is titled Speed up, so a player who opened it — the one screen that is about shards — came away with the same wrong conclusion the issue was filed about. It stayed wrong because nothing could assert on it: the list was inline in BaseTab, which mounts a PixiJS canvas and never renders in a test. It now lives in crono/speedUpSheet.ts with the rule that picks the visible queue, and both are pinned. Empty queues stay listed — hiding them is how a player never learns the option exists.
• researchQueue rides on the /colony payload (EmpireLevelsPayload). Same move as treeLevels/techLevels in #384 and at the same price: persistResolved already loads the empire to resolve production, so it costs no extra query and no second clock. Because it is the resolved empire, a research that finished is already gone rather than offered as a target the shard endpoint refuses with target_gone. Labels need no payload — TECHS and TREE are shared catalogues, so a lane is never printed as a raw engine id (#209/#235/#239).

Fixed

• The shard panel kept the lane picked in an earlier visit. BottomSheet is a portal that stays mounted while closed (hidden by visibility), and CronoPanel cleared picked only on [queue, colonyId]. Pick a lane, close without spending, reopen on the same queue: the panel was still aimed at the old pick while the screen pointed elsewhere. A resetToken now clears the selection per opening — but only when the entry point named a queue. Reopening from Operations keeps the choice, because nothing there contradicts it, and bumping the token unconditionally would have thrown away a deliberate pick: the same misfire, rebuilt by the guard against it.
• A research whose clock had run out was still offered as a lane. Dashboard runs resolveColony(state, serverNow()) every render, so finished build and shipyard orders drop off between polls; research lanes came raw off the payload and /research refreshes on a 30 s poll. The panel offered a spend the server answers target_gone. empireResearchTargets(queue, now) drops them, filtering after composing so the whole-queue index the server translates is preserved. Applied to the Research tab's own panel too — the same feature had two behaviours depending on which door was used.
• An explicit queue choice is no longer overruled. The fallback rule was "if construction is empty show the shipyard"; with three queues it could not be extended, and it re-aimed the sheet at Construction when a research finished with the sheet open — under the thumb of a player mid-tap on a non-purchasable shard. It is now "keep the current choice while it still has lanes", plus a pin for a choice the player made explicitly.
• The Laboratory told three states apart that it used to call one. "Nothing researching" was said to a player with orders queued and not started (reachable from lab 12, where buffer slots hold orders waiting for a lane or for the mother colony's wallet), and to anyone polling an api older than the bundle mid-deploy. researchQueueFrom now returns null for "unknown" rather than [] for "empty", and the three states have three sentences. All three still block the spend: not knowing is never a licence to fire a shard blind.

Added

• The Command Center's whole unlock ladder, not one rung at a time (#208). arcticpaladin asked in #bugs-and-ideas: "is there a place to see what is gained/unlocked by each building upgrade […] especially for things like command center unlocks?" The data was already there and already single-sourced — CENTRO_COMANDO_UNLOCKS tables levels 1-15 — but the UI published exactly one row of it at a time, through nextUnlock. A collapsed "The road ahead" disclosure now sits under the gold box and lists every tabled level above the one the box already announces.
  • Collapsed by default, and that is the design. The sheet opens inside a BottomSheet on the reference platform; fourteen rows expanded would push "Contribution now", the cost and the Upgrade button below the fold on every open. Closed it costs one touch target's worth of scroll.
  • Every row carries its own caveats, and that is why this is a shared function and not a .map() in the component. Since the request was filed, the gold box had acquired THREE independent "this promise is not for you" guards, each the fix for a real report: #293 (the colony ceiling is two rules in AND with the Colonization Discipline), #300 (the ceiling is an empire property the server reads on the MOTHER), #441 (the Trade Portal is inert off the mother). They ran once, on the next level. Mapping the table in the component would have rendered up to fourteen more rows with none of them — not a new bug but the same three, multiplied, in the panel the player pays from. ccLadder(cc, ctx) resolves them per row, so a caller cannot obtain a row without its caveats; motherOnlyBuildings is a REQUIRED field because #441 was itself caused by an optional prop left unwired, which compiles in silence.
  • The next-level row was left byte-identical. Its twelve assertions in riquadro-sblocco-cc.test.ts name nextCcUnlock, ccColonyAltrove, suMadre; folding it into the new path would have turned load-bearing guards into assertions about identifiers that no longer exist — green, and guarding nothing.

Fixed

• The CC 5 row named one of the two buildings it unlocks. CENTRO_COMANDO_UNLOCKS[5] grants batterie and portaleCommerciale, but its prose said only 'Market (Trade Portal).' Harmless while the table was read one rung at a time; as a planning line read from CC 0 upward it was not, because on a secondary colony the very next line is the #441 caveat — "the Market always reads the Trade Portal on your mother colony: unlocking it here adds nothing". The row's entire named content was exactly what the line below it nullified, so an outpost read "level 5 is worthless" and stopped at 4. It is not: combat-defense.ts reads batterie off the defending colony with no mother check, and batterie is not in SOLO_MADRE — for an outpost it is the real defence after the Shield. Now 'Ground emplacements (Defense Batteries). Market (Trade Portal).', batteries first so the caveat lands after the clause it actually negates, and with the exact catalogue name (model.ts nome) because the row is what a player takes to the Build menu to look the thing up. §4 of DESIGN-RITMO-POTENZA.md already named both: the string was a lossy transcription of its own declared source.
• The disclosure's toggle was the only control in the sheet under the touch floor. An inline minHeight: 28 beats a stylesheet rule with no !important, and global.css declares [role='dialog'] button { min-height: 44px } inside the phone media block while BottomSheet is role="dialog" — so those two characters cancelled the floor next to a Build/Upgrade/Move row that all obey it, in a panel whose own CSS comment says missing a target here "means opening another building's sheet". Removed; desktop is unaffected, where the sheet is an anchored panel and not a dialog.
• Ladder caveats now carry the gold box's own marginTop: 4. Without it a rebuttal sat at the same colour, the same size and zero gap from the promise it negates, while consecutive rows were 6px apart — and the eye groups by whitespace, so each caveat read as more promise.

Notes

Three review rounds over four lenses closed two blocking and five major findings, and refuted four more. Both blocking findings were in this PR's own guards, surfaced by a mutation harness rather than by reading: the test named "the disclosure starts collapsed" was satisfied by an unrelated useState(false) elsewhere in the file, and condition was never tied to payload, so swapping the two caveat constants left every assertion green with each sentence on the wrong row. The second round then found the swap had merely moved one line up — binding a flag to a constant's NAME says nothing about which sentence that name holds — and that the extraction of those two sentences into constants had made a pre-existing assertion in riquadro-sblocco-cc.test.ts vacuous, since toContain("commandable colonies follow") is now satisfied by the declaration alone. All nine mutations written against the guards die.

1.0.0-beta.15 — 2026-08-14

Added

• The Activity row says which ships did not come back (#204). Tibeh01 asked for it in #bugs-and-ideas: "on the Activity tab, after an attack, if lost ships, it could appear which ships and how many got destroyed instead of just the -1 survivors." The per-class breakdown already existed — inside the battle report, behind Details and then the ▸ fleet toggle of the viewer's own participant card. Two clicks; and for a battle older than the 20 reports /reports serves, no click at all, because those rows carry no Details button. The row now reads Lost Interceptor ×12 · Cruiser ×3 · +2 more classes.
  • Stamped per RECIPIENT, never per side. lossesByShip(report, side, contingentIds) returns exact per-class losses from the snapshots, and every battle-event writer passes the keys it owns: the attacker, the defender's own contingent, each garrison player summed over their guard shifts, each raider by mission id, and the pirate ambush. On a side split into contingents an unknown key gets {} — never the aggregate, which in a raid is everybody else's wrecks.
  • The colony's walls count too. Emplacements and the Fortress are not ships and live outside every fleet snapshot, so a defender holding with forty emplacements and an empty hangar — the profile the game most rewards — read a row that said nothing at all. lossesStatic rides on the owner's event, renders after the ships and is never truncated: thirty emplacements regrow free, one every half hour, and must not displace a battleship.
  • Three states, not two. An absent losses means "written before this existed"; an empty one means "known, and known to be zero" and says No ships lost. Collapsing them made silence mean two things at once, so every pre-feature row read as "cost me nothing". The zero line is suppressed on a razing, where it would sit under "Your colony has fallen".
  • wiped is finally read. The api had written that flag since joint defence shipped and nothing displayed it, so an ally whose relief fleet was annihilated read exactly the same row as one that came home whole.
• The battle report says which ship class dealt the damage. The report gave damage per COMMANDER; it never said which of their ships did it, though the engine has recorded it all along — a group's id IS its ShipId, and every salvo names the group that fired. shipDamage(report) aggregates per contingent per class, reusing salvoDirections rather than deducing direction a second time: two aggregations of one battle that disagree about who shot whom are two truths, and the tests pin that their totals match participant by participant. On screen: a Top gun line on every participant card, and Damage dealt / Kills (est.) columns in the fleet table with a ★ on the top gun. Rows keep catalogue order (#230) — sorting by damage would win the ranking reading and lose the side-by-side one that order exists for. A report with no rounds yields null, not zeros: a Damage column of zeros claims every ship missed every shot, which says far more than "this report cannot say".

Fixed

• A player garrisoning their OWN colony was reported nowhere. canDefend allows defending oneself and loadGarrisons does not filter the owner out, so owner + self-garrison is two contingents belonging to one player — while garrisonEventRecipients deliberately writes the owner no garrison event, on the grounds that "they already receive the defender row". ownerLossKeys makes that comment true again.
• Group.owner had claimed it was "ALWAYS absent on the defender's side" since before joint defence made it false. Every attribution of a salvo to a defending player rests on it. The docblock is corrected and the claim is now asserted — both defender contingents must deal non-zero damage, or the sum-matching tests would be agreeing on two zeros.

1.0.0-beta.14 — 2026-08-14

Added

• Whispers: private messages between two players (#203). Tibeh01 asked for it in #bugs-and-ideas: "on the global chat, i think it would help players talk with each other, if they could press on the profiles, and there would be interactions like Block; Whisper; Invite to Guild; Offer Trade." Two of the four already existed and one did not exist at all — the issue's own inventory had gone stale in the ten days it sat open, which is the reason every claim in it was re-checked against origin/main before a line was written.
  • One thread per PAIR, derived not stored. The channel is dm:<idLow>~<idHigh> on the same table and the same channel column as the public chat, with the two ids SORTED. A client addresses the peer (dm:<peerId>) and never a thread, so no request can resolve to a conversation its author is not in, and neither side can end up talking into its own half of the same exchange.
  • Moderation moved with it, because a private channel is where a harasser gets someone alone. Block covers whispers, exactly as it covers the public channel. Report gained the check it never needed: only a message one can read can be reported. Knowing an id used to be enough, which cost nothing while every channel was public.
  • A brake on blasting, not on talking. Whispering gained a cap on how many new conversations one account can start in a short time, so a spam run dies early while an ordinary exchange never notices it exists. For what a cap cannot fix, Block is what ends it, and it is what the copy points at.
  • player-delete gained the fourth chat FK. The three before it are commented in that file; a whisper is also addressed to somebody, so the messages written to a deleted account would have stayed, pointing at an id that no longer exists.
• The player menu hangs off the person, not off one line they wrote (#203). Pressing a name or an avatar in chat opens Whisper / Invite to guild / Block. The reporter listed Block among the missing while it had been on the message's ⋯ for days — they pressed the NAME, because that is where the person is. Reporting stays on the message, where the words being reported are.
• Guild invites can be sent (#203). POST /guilds/members/invite existed, the receiving half was complete — the banner, accept/decline, the guild-log line "X invited Y" — and nothing in the product called it: not the web, not the agents. A guild whose policy was invito was therefore unjoinable by construction, applications refused server-side and invites impossible to send. In production: 15 guilds, every one of them aperta, zero invite rows ever written. Now sendable from the chat menu and from a searchable picker in the members panel, which had been telling leaders to "invite someone" with no way to.

Changed

• GET /chat/threads and POST /chat/read are new routes; GET /chat is untouched. The bare-array shape of the chat page is a contract with clients the api deploy does not update — the TestFlight IPA ships the web bundle inside itself — so whispers are a client-side pseudo-channel and add no entry to GET /chat/channels.
• POST /chat/read takes an optional upTo, and the cursor only ever moves forward. "Read up to now" swallowed anything arriving between the end of a read and the request landing: a window that is brief for a fast client and long for a slow one — and exactly when the other side, hearing nothing, adds a second line.
• ChatMessage gained recipientId and an index on (playerId, createdAt), plus the new ChatRead table. All additive, applied by prisma db push at api start. The index is not cosmetic: the thread list asks for playerId = me OR recipientId = me, and an OR is only as fast as its worst branch.
• A deeper galaxy reaches players (#541, deployed before this tag and never released on its own). Systems went from 12 to 18 planets and the protected cradle from 360 to 540 slots; every pre-existing planet keeps the exact cell, and therefore the exact travel time, it had.

Notes

• A whisper leaves no grudge, by decision. Three designs were tried and each misfired worse than the gap it filled: counting every whisper as addressed to its reader turned an invitation to raid Sofia together tonight into a grudge against the person offering the alliance; requiring second-person wording fixed that sentence and not the mechanism, since menzionati scans the whole roster and the same invitation booked a grudge for every agent named Sofia, out of a message they were never shown; and beneath both sits a matcher that reads hostility from the letters of a phrase rather than its sense. A grudge sends fleets, so a wrong one costs coherence — an agent that answers warmly and then attacks, with nothing readable explaining it — while a missing one costs only a consequence. The test pins the SIGNATURE of _digest_minacce, so the door cannot be reopened by adding a parameter.
• Offer Trade is not here. The market is a public offer book; a directed offer is a new mechanic rather than wiring — expiry, escrow while the target ignores it, and what a decline does to the seller's locked goods are design decisions.

1.0.0-beta.13 — 2026-08-14

Added

• A galaxy-wide target list on the political map (#194). A player put it plainly in #bugs-and-ideas: "really difficult to find planets that I can actually attack. Going back and forth with the map and mission view to see where there are bases is a lot of work." The arithmetic backed him up — the galaxy is 8 sectors × 10 systems, so answering "who around here is weaker than me" meant opening and closing eighty screens by hand and holding the answers in your head. The agents already ran exactly that sweep in a loop and kept a directory of the results (runner/intel.py), which is the same asymmetry the Wanted board was built to close, and it is closed the same way: give the human the list the machine already builds for itself. GET /galaxy/targets now does the sweep in one answer — every colony the viewer's sensors can see, the owner's raw Gate beside it, the counts the request asked for (in band, far below, stronger, PvE, and how many slots are free), and sorting by nearest or weakest.
  • The price is on the row, not promised once in a header. Inside your band a hit costs nothing; far below it costs a bounty; on a wanted head it costs nothing either, because settleBounty returns before pricing anything when the defender carries a grade. A single header sentence was false in exactly the two categories the sorts promoted to the top of the screen.
  • Targets that are not played empires get their NAME, not a band. A garrison's Gate is entirely fleet while a player's is ~92% buildings (measured in production, see relic.ts), so a bastion tuned to 1.8× the top player's Gate read as "In band — a fair target" against a defender whose fleet is twenty-two times the reader's. Beacon, Bastion and Relic are now named, with the Gate left visible to be read for what it is.
  • Refusals travel with the row — pact, novice shelter, shield — in the same order of precedence assertAttackAllowed applies them, and hidden by default, because the request was for planets you can actually attack. It never claims the opposite: no code means "none of these three", not "this will go through" (night protection depends on when the fleet lands and cannot be known from a list).
  • The Veil survives the aggregation. A colony hidden from you in the system view is just as absent here and counts as an empty slot, or the list would have been the cheapest counter-espionage tool in the game.
  • Tapping a row opens the planet sheet over the map, which closes the other half of the report: no more map → system → back to compare two candidates.

1.0.0-beta.12 — 2026-08-13

Added

• Every Build menu row now opens the building's full sheet — the shop is the catalogue (#331). Two players asked, independently on the same day, where to read what a building does before paying for it; the answer was nowhere — the sheet existed, but only opened on buildings you already owned, and a locked row was just a name with a padlock. Now tapping any row, locked and unbuilt ones included, opens the same sheet the built ones get: the one-line description, what it contributes at level 1 on your colony (planet traits and multipliers included, the way a base-builder shows stats in your own village), the cost, and — for locked buildings — the exact unlock requirement in the same words the research tree uses. No separate encyclopedia screen to maintain, and the numbers are read from the engine, so the page cannot drift from the simulation.

Fixed

• The Warehouse no longer takes credit for the floor (#527 review). The sheet of an unbuilt Warehouse declared "Contribution now: 5,000 per resource" — that number is the free base capacity every colony has; the building at level 0 contributes nothing. It now says so, keeping the floor visible in a note. The capacity shown also finally includes the Depots tree bonus: without it the sheet understated the real cap by up to 80%, telling players their resources were about to overflow when they were not.
• No more impossible Build buttons on newborn outposts (#527 review). On a freshly colonized outpost (Command Center still at 0) the tree-unlocked buildings showed an enabled Build button under a gold notice saying the level cap was reached — the server rejected the placement only after the player had walked the whole gesture. The button now states the Command Center level required, in the sheet and in the menu row (the menu half of this predates the catalogue).
• The Build menu filter survives reading a sheet (#527 review). The phone filter ("Production", "🔒 Locked"…) lived inside the menu panel and died every time it was swapped for a sheet — browsing the catalogue row after row reset it to "All" each time. It now lives one level up and stays where you put it.

1.0.0-beta.11 — 2026-08-13

Added

• Groundwork for the Stellar Anomalies rewrite — nothing you can see in game yet (#521). This release changes nothing a player can do, reach or notice: it lands the design for turning anomalies into layered Precursor Expeditions, plus the pure rules that will run them, and none of it is wired to a screen or an endpoint. It is in the repo so the next releases have something to build against, and it is listed here so main moving forward is legible rather than silent.

  What the design changes, when it ships: an anomaly stops being a coordinate you claim for one hidden payout and becomes a site you descend into, one battle per layer, with real losses between layers and a push-or-extract decision at every boundary. Four site classes each punish a different fleet, so bringing the right ships matters more than bringing the most. The haul is paid in hours of the server's production rather than your own, so the same expedition is worth the same to a new player and to the leader.

  How a site decides what you meet is part of the design: a layer's garrison is derived from a value the server generates once a week and serves to no one. A site's tier and class stay readable in advance, because "how much hold do I bring" has to be answerable before launch; what is inside does not.

Fixed

• Nothing player-facing. The season scoring, the map and every existing screen behave exactly as they did in beta.10 — two internal constants changed visibility, at identical values.

1.0.0-beta.10 — 2026-08-13

Changed

• The market now takes a cut, and burns it (#517). Trades used to settle in full — which also meant two accounts could shuttle resources back and forth at honest prices forever, at zero cost, and the economy had no drain anywhere. Every offer published from now on carries a 5% settlement fee on each side: the buyer pays the asked amount in full and receives the goods minus the fee, the seller receives the payment minus the fee, and the difference is destroyed — it goes to nobody. What you'll actually receive is shown before you accept (the detail view prices it in, taxed rows carry a "fee 5%" marker on the board), your activity log records the true take, and cancelling still refunds the escrow in full: listing is free, only settlement is taxed. Offers already on the board when this shipped are exempt — they settle in full, at the price you published them at. The AI commanders price the fee into their asks and their appraisals exactly like you would.

1.0.0-beta.9 — 2026-08-13

Changed

• The visit counting moved onto our own server, and blocking it is now our job to honour. Since beta.7 the counter was Plausible's hosted service, which most content blockers answer with an empty 204 — measured in a real profile, where every plausible.io URL including their own homepage returns nothing. So the numbers were a floor with no way to know how far below the truth they sat. It now runs on a second server of ours (Plausible Community Edition, Helsinki), proxied through novarchy.com itself at /n/s.js and /n/e. The path names are not decoration: the same blocker also answers /js/plausible.js with a 204 same-origin, and eats a third-party POST to /api/event, so self-hosting on the default paths would have bought a tracker that downloads and whose events are silently swallowed. Two consequences worth stating plainly. Nothing about a visitor reaches a third party any more, so the privacy page's list of who else sees anything is now just Google and Discord — and the counting got its own section that says whose server it is. And because a first-party request cannot be refused by blocking a domain, refusing had to become something we honour: the privacy page now carries a button that turns counting off, and Do Not Track and Global Privacy Control are honoured too, all three checked before the counter is even created. The button exists because the browser signals are not enough — Safari dropped Do Not Track in 2019 and never shipped Global Privacy Control, and on iPhone every browser is Safari underneath, so for those readers a promise resting on a browser setting would have been unkeepable. It is pinned by tests in both directions, including that a browser blocking storage outright is not treated as a refusal. This build of the script also does not watch clicks, downloads or form submissions at all, so the previous release's careful switching-off of form capture — which existed because the chat composer is a <form> — is no longer something that can regress.

1.0.0-beta.8 — 2026-08-13

Fixed

• A colony that defends with nothing no longer pays the attacker as if it had (#270). Before: parking your fleet in another colony turned your planet into a paying target for a friend. The attacker broke a Shield nobody was defending, met no return fire, and scored 125.57 points — more than the 115.57 paid by a target that actually put its fleet in the field. Farming was strictly better than fighting, and it cost the accomplice nothing. Now the Bolla pool only counts toward battle points if the defence fired at least one shot. The same colony pays 10.00, the win bonus. A colony that hides its fleet behind the defence plan (#259) drops from 48.52 to 10.00 too. Ships and emplacements destroyed still pay in full — those are real losses somebody bought — so a defended colony is worth exactly what it was: a real defence still pays 115.57, and the fortress doctrine still collects its full 300.

  What this changes for honest play, stated plainly: the game cannot tell "my accomplice parked the fleet elsewhere" from "my fleet is out fighting" — it is the same board state. So catching somebody with their fleet away now yields the loot, the victory and the siege progress, but not the pool points. If you want an attack on you to be expensive for the attacker, leave something that shoots: a single emplacement is enough.

  What is NOT fixed, and is worth knowing before you find it: one salvo of any size restores the full amount. A colony with a single ship, or Batterie L1, is back to ~131 points. Closing that needed a threshold, and every threshold we measured was worse — the version relative to the attacker's fleet let a farmer switch the bonus back on by bringing fewer ships, and made an honest defender's value drop when the attacker committed more. The second configuration named in #270 — the fortress with Batterie L14 and the fleet hidden, 300 points — is also still open: it comes from the Bolla pool recharging every battle (#255), not from this rule.

1.0.0-beta.7 — 2026-08-13

Notes

• Web traffic is measured with Plausible. There was nothing readable before: how many people reached novarchy.com, and where from, was visible only as accounts created (the server's access log has the raw requests, which is not the same as an answer), so everyone who looked and left was invisible. The tracker is cookie-free, stores nothing on the visitor's device and builds no profile, so it needs no consent banner. It only runs on the novarchy.com hostname: the mobile app ships the same index.html, and while the tracker skips localhost by itself, a LAN IP or a tunnel used to test on a real phone is an ordinary host to it and would land in the production stats. The gate also means the app never requests the script at all. Three captures are switched off that this script turns on by default: outbound clicks, file downloads and form submissions. The chat composer is a form, so leaving the defaults would have sent Plausible an event for every message a player types — and billed for it. Pageviews, scroll depth and engagement time are what remain. Two honest limits, so nobody reads the numbers for more than they are: ad blockers drop the script, so visits are a floor and not a headcount; and with no router in the app a whole session is one pageview, which makes "bounce" and "visit duration" meaningless here. The privacy page was part of this change, not a follow-up: it stated "no analytics" and "no analytics services", which this feature would have turned into a false published claim the moment it deployed. It now says what Plausible receives — including the scroll depth and engagement time the current script sends automatically, and the daily key built from IP and browser — and what it does not: no cookies, nothing kept on the device, no link to the game account. Checking it against Plausible's own data policy turned up two claims of ours that promised more than the vendor does, so the page says the count stops being traceable once the day's key is discarded, rather than claiming it never was. Reviewing that list also surfaced two third parties the page had never disclosed at all, both older than this change: Google Fonts, fetched on every load, and the Google sign-in script, fetched by any screen offering the button. They are listed now too. The tests pin the parts that can rot silently — the tracker cannot load without the sentence disclosing it, the three interaction captures cannot be switched back on quietly, the disclosure has to name Google Fonts while the game still loads them, and the mobile app has to keep the hostname the gate rejects — but the rest of the page's wording is not machine-checked.

• Everything written from now on is in English — code, comments, docs, commit messages, issues, release notes, this changelog. The rule now lives in CLAUDE.md, which was itself translated, since a rule stated in Italian contradicts itself. What this does not claim is that the project is already all English: the Italian prose that exists — this file's own header, the README, the design documents — is left exactly where it is, because bulk-translating it would bury real changes in diff noise. It gets rewritten when someone has a reason to touch it, not before. Separately, Italian that is machinery rather than prose is excluded permanently and must not be touched: ids already persisted in live colonies, the enum values the agents' plans are validated against, the word lists whose job is to detect Italian, and the filters that catch what players type at us in Italian.

1.0.0-beta.6 — 2026-08-13

Fixed

• The hive map showed every flower paper-grey (PR #514). The guild colour directory is fetched once per session and usually arrives before the map mounts; the scene is created asynchronously and missed the hand-off, so it painted with an empty palette forever — while the HTML legend right under the canvas showed the correct colours. The directory is now handed to the scene the moment it is created. Flowers, halos and rings wear their guilds' colours again.

1.0.0-beta.5 — 2026-08-13

Changed

• A ruled system now shows its ruler on the hive map (PR #511). The flower's cells have always worn the colour of the guild present in the system; now, once territorial control holds (3 days of dominance), the whole flower turns brighter and the control ring goes solid — the map answers "who rules here" before a single label is read. While control matures nothing changes: quiet cells, faint ring. The map legend describes both states, and the redraw reacts to takeovers immediately instead of waiting a poll cycle.

1.0.0-beta.4 — 2026-08-13

Corretto

• Due parole italiane che affioravano in un'interfaccia per il resto inglese. La prima la vedeva chi ha portato l'Osservatorio a 5: nell'allarme delle flotte in arrivo, la stazza della flotta nemica era scritta «fleet piccola» invece di «fleet small» — il server manda un id interno e il pannello lo stampava tale e quale, mentre la riga accanto (le classi di scafo) l'etichetta ce l'aveva dall'inizio. La seconda è «Ignoto» al posto di «Unknown» come nome di un comandante di cui il server non conosce più il nome: si vedeva nella bacheca degli aiuti di gilda, fra i candidati in attesa e nel feed attività. Nel registro di gilda le righe già scritte restano com'erano — è testo archiviato — ma ora si leggono in inglese lo stesso.

Sotto il cofano

• I commenti del codice passano all'inglese, prima tranche: packages/shared completo, apps/api all'80%, apps/web iniziato. Nessun comportamento cambia, ed è verificato e non dichiarato: il JavaScript prodotto da ogni file è identico byte per byte prima e dopo. Il controllo esiste perché serviva: nella prima tranche ha trovato una traduzione che aveva cancellato una guardia insieme al commento che la spiegava, riaprendo un difetto già chiuso (#293, #434) — i test e il compilatore non se n'erano accorti. Restano da tradurre apps/web, gli agenti e le suite di test.
• Due file troppo grossi divisi per materia. La semina del Faro Pirata esce da ColonyService (creava un impero NPC, non era un servizio di colonie) e i riquadri della HUD escono da Dashboard.tsx, che era il file più lungo del progetto. Nessuna riga di logica riscritta: solo spostata, con i test che già la coprivano — e non toccati — a farne da verifica.

1.0.0-beta.3 — 2026-08-13

Aggiunto

• I Contratti: obiettivi del giorno e della settimana, con premi che valgono il viaggio. Nella scheda Eventi è arrivata una bacheca (#191, design in DESIGN-CONTRATTI.md): ogni giorno alle 05:00 UTC tre contratti estratti sul TUO impero — solo verbi che puoi davvero giocare, niente "vinci una battaglia" per chi non ha flotta. Completarne due chiude il Daily Wrap, che paga doppio; cinque Wrap in settimana aprono la Weekly Chest: un Cronoframmento da un'ora più risorse. I premi sono ancorati alla produzione oraria dell'impero, quindi crescono con te — e si riscuotono quando hai spazio in magazzino, con l'avviso prima, non dopo. La bacheca si sblocca alla fine dei Primi Passi. Anche i comandanti IA la giocano, al loro passo: chi è bravo manca raramente un Wrap, chi è scarso certi giorni se ne dimentica — e perde la Cassa come lo faresti tu.

Modificato

• La galassia è stata ricartografata: la mappa è ora un Alveare esagonale. La vecchia carta prospettica con le rotte e i varchi è andata in pensione: al suo posto una griglia esagonale continua, piatta, pensata per il telefono. Ogni pianeta occupa una cella del fiore della sua stella, le celle vuote sono spazio aperto, e il tempo di viaggio è esattamente la distanza che si vede — niente più rotte nascoste da imparare. Tre livelli di zoom (galassia → settore → fiore), tap per avvicinare e poi per aprire, la camera che si ricorda dov'eri, e la legenda finalmente anche su telefono, dentro il cartiglio dei premi.

  I tempi medi di viaggio restano quelli di prima (taratura verificata entro pochi punti percentuali, con test che la inchiodano su tutta la galassia), ma la forma delle distanze cambia: ciò che sembra vicino ora È vicino. Le missioni già in volo concludono con i tempi vecchi. Gli Iperportali funzionano come prima: un varco al capo di partenza sconta le missioni logistiche, i varchi a entrambi i capi collassano il viaggio a un salto. Anche i comandanti IA leggono la stessa carta: scelgono dove colonizzare, esplorare e raccogliere guardando le celle, non più gli indici dei settori.

  Sotto il cofano: rollback d'emergenza TRAVEL_MODEL=graph documentato in DEPLOY.md (con log del modello all'avvio e anteprime client che seguono il server via /settings), e la geografia dei bot generata dal TypeScript vero — un asset versionato col seed, con test di parità cella per cella in entrambe le lingue.

Corretto

• Le reazioni in chat alla presa e alla perdita delle Reliquie erano mute (#501, regressione della #479). Un errore all'ultimo passo di ogni ciclo dei comandanti IA spegneva i loro commenti sulle Reliquie contese — e riempiva i log di avvisi, ~75 ogni cinque minuti. Ora la voce è tornata, con un tetto anti-valanga nuovo: quando una gilda perde una reliquia la notizia arriva a tutti i suoi membri, ma a parlarne in canale sono al più due, non un coro — e la festa di chi l'ha presa non zittisce il lamento di chi l'ha persa. Sistemati anche tre test che cambiavano esito con l'ora del muro (#503, in parte): la suite non diventa più rossa da sola ogni notte fra le 00 e le 07 UTC.

1.0.0-beta.2 — 2026-08-13

Corretto

• La gilda non commenta più il registro alle quattro del mattino (#479). Benvenuti, bacheca, chiamate agli avamposti e rapporti di difesa rispettavano gli orari di chi parla; il commento al registro no — e nemmeno il consiglio di guerra, i bonifici dal Fondo, le candidature e i cambi di grado. Un compagno che in chat "dorme" dalle 21 poteva raccontare alle 03:45 un versamento fatto cinque minuti prima. Ora il controllo vive nell'unico punto da cui passa tutto il canale di gilda, più i due percorsi che postano per conto loro.

  Una notizia rimandata viene poi raccontata come notizia: chi si sveglia ne parla come di qualcosa che si è perso, non come se fosse appena successa. Oltre le dodici ore non è più una novità e nessuno legge il registro ad alta voce.

  Corretta anche la cadenza dei versamenti al Progetto Eschaton, che viveva in memoria: ogni riavvio del servizio la azzerava, e tre deploy in settanta minuti diventavano tre bonifici dallo stesso comandante nel registro. Ora sta su disco, insieme a due orologi gemelli.

• Il Mercato dice di quale colonia sono le risorse che mostra (#288). Aprendolo da una colonia secondaria, l'intestazione in alto mostrava le risorse di quella colonia e il pannello due righe sotto quelle della colonia madre: sei numeri, nessuno uguale al suo dirimpettaio, e niente che lo spiegasse. Il commercio passa sempre dalla madre — il server era già coerente — ma il pannello lo diceva solo dentro i messaggi d'errore e nel dettaglio di un'offerta aperta. Ora il nome della madre sta accanto allo stock, e un avviso compare quando si guarda da un'altra colonia.

Note

• Il badge dei test nel README era fermo al conteggio dei file di mesi fa (88/109). Ora riporta i test: 5.580 vitest e 2.128 pytest.

1.0.0-beta.1 — 2026-08-12

Prima release taggata. Copre tutto ciò che era in produzione fino a #466 incluso, fra cui:

• L'anteprima di Spionaggio non è più identica prima e dopo l'acquisto (#473). Su 17 livelli su 24 le due righe erano uguali byte per byte, sopra un acquisto da migliaia di risorse: il rischio mostrato era quello contro un bersaglio sguarnito, dove il calcolo tocca il minimo già dal quarto livello. Ora la scheda mostra ciò che il livello compra davvero — quanto toglie al vantaggio di chi ti sorveglia.

• Quattro livelli del Centro di Comando smettono di promettere cose che hai già (#483). Template di flotta, report estesi, anteprima della Potenza del bersaglio e spostamento istantaneo degli edifici erano già disponibili a chiunque, e due di quelle funzioni non sono mai esistite. Erano scritte nel riquadro da cui si paga l'edificio più caro.

• La riga dell'energia distingue un deficit vero da un margine sottile (#466). Il chip in alto lo faceva da tempo, la riga della colonia no: metà produzione persa e un punto di margine avevano lo stesso colore. E la barra sotto usava il criterio delle risorse, dove "oltre il 90%" significa stai per traboccare — sull'energia significa l'opposto.

• I condotti della base leggono lo stesso bilancio di tutte le altre schermate (#465). Con l'Acceleratore spento contavano comunque il suo consumo, quindi diventavano rossi mentre ogni altra schermata diceva che l'energia era a posto; su un pianeta con malus sbagliavano dalla parte opposta.