Auto Warz — War Layer Amendments¶

Document Type: Project Management — change specification Version: 3.4 Status: Proposed. Nothing in this list has been applied. Last Updated: 2026-09-20 Author: Solo Developer Related Documents: [31_War & Combat System Design](<../01_Design Documents/Systems/31_War & Combat System Design.md>) · [32_The Gate & War Logistics](<../01_Design Documents/Systems/32_The Gate & War Logistics.md>) · [33_Mothership & Fleet Design](<../01_Design Documents/Systems/33_Mothership & Fleet Design.md>)

Purpose¶

The war layer design session of 2026-09-03 settled the top of the dependency chain: what war is, what decides it, how supply reaches it, and what persists. Three new documents carry that design.

This document lists every change the new design requires in existing documents, with the proposed replacement text where it is short enough to state. It exists so the propagation pass is a checklist rather than a rediscovery.

Apply in the order given. Item 1 changes vocabulary that later items depend on.

1. `05_Reference/04_Data Dictionary` — vocabulary¶

1a. The word "session" is overloaded¶

01_Master GDD and 08_Research Tree Design use session to mean the whole playthrough. Other documents use it to mean one sitting. Lock these four terms:

Term	Definition
World	The persistent universe. One galaxy seed plus all changed state: factories, planets, ownership, credits, wreckage. This is the save file. Identified by an ID, never by its name — two players can both name a world "Tarn".
Session	One continuous run of a world, from load to quit. A sitting. No game state belongs to a session.
Account	The player across all worlds. Level, cosmetics, Engineering Research, saved designs, faction reputation, motherships.
Planet	A player's own territory inside a world. Where their factory lives.

1b. New terms¶

Term	Definition
War	A separate shared instance hosted by a player or a server. The only shared place.
Gate	The building in a world that links it to a war. One gate, one concurrent war.
Forward Depot	A faction's shared warehouse at a war. The destination of every friendly gate.
Supply umbrella	The area within which a mothership or depot can deliver. Equal to drone round-trip range.
Command range	Maximum operating distance of a ship from its mothership, beyond which it is out of contact and must return. Per class. *Renamed from control tether, which did not say what range it meant.*
Contribution	Goods delivered through a gate to a Forward Depot. A permanent personal record.
Adrift	A ship disabled by critical module loss.
Derelict	An abandoned adrift ship. Salvage for anyone.
Recovery	Bringing a stranded or crippled mothership home.

1c. Measurement vocabulary¶

Three kinds of number, never combined. Applies to every readout in the game, not only the editors.

Kind	Test	Example
Efficiency	Has a denominator: output per unit spent	fuel efficiency, ammunition efficiency
Power	Has units over time	offensive power, in kills per minute
Coverage / capacity	A fact, spatial or counted	armour coverage per face, storage capacity

Efficiency needs a denominator. Power needs units. Coverage is a fact. Never combine across the three.

Forbidden: armour efficiency — armour has no output, so no denominator exists — along with design efficiency, overall rating, and any single number standing for a whole design. A single aggregate is rankable, and anything rankable gets optimised until every player builds the same thing.

Sweep existing documents for misuse. See [34_Ship & Mothership Editor Design](<../01_Design Documents/Systems/34_Ship & Mothership Editor Design.md>) → What the Editor Shows.

1d. Deprecated words¶

Add to the deprecated list, with replacements:

Deprecated	Use instead	Why
"dead in the water"	adrift	Nautical. There is no water.
"towed"	recovered	Nautical.
any nautical direction or idiom	plain or military equivalents	The game is set in space

2. `01_Design Documents/Content/08_Research Tree Design` — the two-tree split¶

This is the largest amendment. Two changes.

2a. Factory Research is world-scoped, not session-scoped¶

The document currently says research "resets at the start of every new session" with "no cross-session persistence". Under the locked vocabulary that reads as losing the tech tree at bedtime, which was never the intent.

Factory Research is world-scoped. It resets in a new world. It survives a logout.

2b. A second tree exists¶

Split research in two, by what resets and what accumulates.

Tree	Scope	Unlocks	Resets
Factory Research	World	Capability: buildings, recipes, logistics tiers, vertical levels	New world
Engineering Research	Account	Capacity and efficiency on ship and mothership designs: armour density, reactor efficiency, umbrella radius, command range, drone throughput, tank capacity	Never

Rationale: what you rebuild each session resets; what travels with you accumulates. The factory is rebuilt in every world, so its tech resets. The mothership travels, so its tech travels.

Guard rail, to be stated in the document:

Account research improves capacity and efficiency. It never unlocks capability.

~~The infinite nodes move to Engineering Research, which is their natural home: they scale a rate rather than unlocking content, so they never gate anything.~~

⚠️ Withdrawn 2026-09-05, when the document was actually read. All five infinite nodes raise factory rates: RP rate, turret damage, Genesis speed, module effects, extraction rate. Moving them to an account-scoped tree would convert a long grind into permanent cross-world power, which [00_Canonical Design Facts](<./00_Canonical Design Facts.md>) §20 forbids. They stay in Factory Research and stay world-scoped. Engineering Research has its own uncapped lines; they are a different thing that happens to share a shape.

3. `01_Design Documents/Content/15_Ship & Army Points Design` — APR becomes a rate¶

3a. Delete the win-probability formula¶

Remove win_probability = player_apr / (player_apr + enemy_apr) and the pre-commit probability screen built on it.

3b. Redefine APR¶

APR is a rate, not a probability input: ships destroyed per tick, and supply consumed per tick doing it. Attrition maths produces the outcome.

3c. Closes six open items¶

OI-202 to OI-206 (and any sibling rows) are all questions about the removed formula. Move them to closed-items.md with the resolution "formula removed; APR redefined as a rate in 31_War & Combat System Design".

3d. Reconcile the class list¶

Adopt the seven classes and their attrition roles from [33_Mothership & Fleet Design](<../01_Design Documents/Systems/33_Mothership & Fleet Design.md>): Fighter, Frigate, Cruiser, Carrier, Freighter, Mining Vessel, Scout.

This resolves the disagreement with 02_Scope & MVP Document line 188. The Scout survives as the class defined by not needing supply, which makes the first ship a player launches a non-combat ship — as 02_Scope always had it.

4. `01_Design Documents/Content/14_Win Conditions & Faction Allegiance` — war terms¶

4a. Add the war terms sheet¶

Cross-reference [31_War & Combat System Design](<../01_Design Documents/Systems/31_War & Combat System Design.md>) → War Declaration & Terms rather than restating it.

4b. Victory conditions¶

Three, all requiring combat: Hold, Depot, Attrition. Record explicitly that a Contribution victory was considered and rejected, because if delivering alone wins, the dominant strategy is never to fight.

4c. Factions persist and level¶

Faction upgrades funded by the faction share of kill credits, persisting across wars, capacity only, never capability.

4d. Mercenary path¶

Update: a mercenary brings their own gate and fights for someone else's cause. They are not a player without production. Pay is a share of salvage plus reputation, since supply is free to them under depot free draw.

5. `04_Project Management/00_Canonical Design Facts` — new principles¶

Add as named sections following the §12b precedent.

5a. §16 — Supply Sets Magnitude, Position Sets Distribution¶

Supply decides how much damage a fleet deals and how long it lasts. Position decides who takes it. Position must never be more decisive than supply. If it becomes so, the game has quietly become a battle game and the factory has stopped mattering. Test every new combat mechanic against this line.

5b. §17 — Manual First, Automation Is Built¶

Every system has a manual mode that works and an automation path that costs something physical (hull space, research, buildings). Automation buys back attention, never power. Manual remains optimal because a player can see the situation and rules cannot; what automation buys is the ability to be somewhere else.

Instances: Genesis and belts, the Router, station priority, mothership supply allocation, fleet movement, gate delivery.

5c. §18 — Area-Based Supply¶

Power on the planet is area-based rather than port-based (§3). Space supply is area-based for the same reason: a mothership or depot projects an umbrella and any friendly ship inside it draws. Helping an ally is geometry, not a transaction.

5d. §19 — Disabled, Never Deleted¶

A ship is disabled, never deleted. Critical module loss makes it adrift; abandoned, it becomes a derelict and salvage for anyone. The mothership is crippled or stranded, never destroyed, on the same anti-softlock grounds that make Genesis indestructible. Only hardcore mode deletes.

5e. §20 — Capacity Crosses, Stock Does Not¶

Advantage carried between worlds is capacity, never stock. A mothership carries its design and its hull between worlds and arrives empty. Account research improves capacity and efficiency and never unlocks capability. Faction upgrades follow the same rule.

5f. §21 — The Channel Is Physical¶

The path from factory to front is a physical channel with finite throughput and real latency. It is never a transfer menu. The gate satisfies this: it has throughput, latency, a power draw and a single destination.

5g. §22 — World, Session, Account¶

Restate the vocabulary from amendment 1 here, since it is load-bearing and this page is read first.

A world is the data. A session is the world in motion. A world is identified by an ID, never by its name.

5h. §23 — Value Is Fixed, Return Scales With Presence¶

A delivery is worth the same to the faction however it arrived. The delivering player's personal return scales with presence and risk, never with how automated the delivery was. Scaling reward by automation tier would punish a player for automating and contradicts §17. Contribution is recorded at face value and tagged by how it arrived; the reward is what scales.

5i. Extend §12b — Closed Loops Are Always Lossy¶

Add two instances to the existing table:

Mechanic	Recovers	Would otherwise mean
Battlefield salvage	A fraction of a derelict's build materials	A war that feeds itself, and a factory that stopped mattering
Field refining aboard a mothership	Less per unit than planet-side processing	The planet becomes optional

5j. Correct §13 — Data & Tooling is stale¶

§13 still names SQLite as the source of truth for items, buildings and recipes. OI-136 and OI-137 record that the SQLite plan was dropped. Update §13 to describe the actual pipeline, or mark the row superseded.

(Unrelated to the war layer, found during this session. Recorded here so it is not lost.)

6. `01_Design Documents/01_Master GDD` — layer 3 and vocabulary¶

6a. Replace the Layer 3 stub¶

"RTS-style fleet combat and territory control. AI factions + optional PvP" becomes a summary of [31_War & Combat System Design](<../01_Design Documents/Systems/31_War & Combat System Design.md>), with the key sentence stated once:

Space is a war decided by supply. War must return a diagnosable failure back into the factory.

6b. Vocabulary pass¶

Every use of session meaning the whole playthrough becomes world.

6c. Add the gate to the macro loop¶

The macro loop currently ends at "Claim Territory → Unlock New Planets → Scale Further". Replace with the gate loop from [32_The Gate & War Logistics](<../01_Design Documents/Systems/32_The Gate & War Logistics.md>).

7. `01_Design Documents/Unwritten/` — retire two stubs¶

Stub	Action
`06_Combat & War System Design`	Superseded. Move `31_War & Combat System Design` into `Systems/`. Delete the stub, update every inbound wikilink, and update the `mkdocs.yml` nav.
`05_Spaceship Design & Building System`	Partly superseded. The editor half is now `33_Mothership & Fleet Design`; construction remains `30_Ship Construction System`. Delete the stub and repoint links.
`07_Space Exploration Design`	Still unwritten. Leave the stub.
`08_Economy & Progression Design`	Largely answered by the three new documents (contribution, credits, faction growth, persistence). Decide whether it still needs to exist or whether the stub should be retired with a pointer.

⚠️ Per the folder's own README, check the destination folder for a prefix collision before moving, and do not rename on the way out unless a collision forces it.

7a. `06_Outpost/04_Progression & Rewards` — the trickle gets a fourth destination¶

The document already establishes that completing an outpost locks in its output and unlocks "a passive resource trickle that runs continuously", routed to one of three destinations: self-supply, another player's request, or the open market.

7a.1 Add a fourth destination¶

A Forward Depot. Goods arriving there count as contribution exactly like anything that crossed a gate.

This makes Outpost Mode a legitimate way to support a war without touching your factory.

7a.2 State it as a rate, not a stock¶

Outpost progression is account-scoped and worlds are separate, so an account-level pile of goods would breach the new CDF rule capacity crosses, stock does not. Take the vault's own word literally:

An outpost grants a delivery rate, not a stockpile
Nothing accumulates while the player is away. There is no banked pile to draw down
The rate is pointed at one destination at a time

What crosses worlds is then a rate the player earned, which is capacity, and the principle holds.

7a.3 Note for `01_Outpost Mode Overview`¶

The three-mode table already gives Sandbox no contribution, to prevent seed exploitation, and gives Seasonal a boost of up to x2 in week 1. Decide whether that boost applies when the trickle is pointed at a Forward Depot. Logged below.

7b. Step 2 outputs — the editor¶

[34_Ship & Mothership Editor Design](<../01_Design Documents/Systems/34_Ship & Mothership Editor Design.md>) is written and closes OI-344 and OI-345. Three existing documents need amending to match.

7b.1 `15_Ship & Army Points Design` — delete the APR build display¶

The Real-Time APR Display During Ship Building section shows a single APR number while designing. Remove it. APR is now a rate, and a single design score is exactly what the editor document forbids: the editor shows consequences, never a score. Replace with a pointer to the readout specification in [34_Ship & Mothership Editor Design](<../01_Design Documents/Systems/34_Ship & Mothership Editor Design.md>) → What the Editor Shows.

7b.2 `30_Ship Construction System` §10 — the editor is no longer unwritten¶

Two rows in Not designed here now resolve:

Row	Now
The ship editor (OI-344)	Written. `[34_Ship & Mothership Editor Design](<../01_Design Documents/Systems/34_Ship & Mothership Editor Design.md>)`
The shared component list (OI-345)	Owned by the editor document

Ship component recipes (OI-343) remain open and remain the prerequisite.

Also add the draft versus buildable distinction to the Shipyard's design picker: drafts appear greyed with their failing rule, so a design is never silently missing.

7b.3 `24_Module System Design` §2.1 — add a do-not-harmonise warning¶

The document states that a module's position on a building carries no mechanical meaning. That stays true. But ship components are placed by layer or by cell precisely because position decides exposure.

Add a warning that the two systems differ deliberately, with the reason: nothing shoots a building. A future consistency pass that unifies them would delete the ship editor.

7b.4 Retire the last Unwritten stub for ships¶

Unwritten/05_Spaceship Design & Building System is now fully superseded: construction by [30_Ship Construction System](<../01_Design Documents/Systems/30_Ship Construction System.md>), the editor by [34_Ship & Mothership Editor Design](<../01_Design Documents/Systems/34_Ship & Mothership Editor Design.md>). Delete the stub, repoint inbound links, update the mkdocs.yml nav.

7c. Step 3 outputs — the economy¶

[35_Economy Design](<../01_Design Documents/Systems/35_Economy Design.md>) is written and closes OI-342. Two existing documents need amending.

7c.1 `13_Player Levelling System` — levels grant Engineering Points¶

Replace the mechanical bonuses (+2% supply chain contribution per ten levels, +1% extraction per twenty, +10% passive RP at level 80) with Engineering Points, the currency that buys account research in [35_Economy Design](<../01_Design Documents/Systems/35_Economy Design.md>).

Cosmetics stay, since achievement exclusivity is already a promise the game makes.

No conflict with the existing rule that Factory Research slots are unlocked only via the Research & Science branch: Engineering Research is the other tree.

7c.2 `08_Research Tree Design` — Engineering Research is flat¶

Beyond the world-scoping correction in amendment 2, record that the second tree is a flat catalogue of upgrade lines, not a branching graph. Prerequisites exist to sequence content, and Engineering Research has no content to sequence. Flat structurally enforces the rule that account research never unlocks capability.

7c.3 Not amendments, recorded so they are not re-litigated¶

A consumable property on ResourceDefinition was rejected. The same substance is both recoverable and consumed depending on use, so consumption is a property of the action. Same class of error as the removed ResourceTier. Derive the view from the recipe graph if it is wanted.
Contribution does not feed Research Points. RP sources stay as they are.
Restoration pays in materials, contribution and reputation, not credits. Credits are war-scoped and Restoration is the peaceful path.

7d. `28_Restoration - Third Endgame Path` — reconciled, v0.2¶

The August draft assumed a persistent shared galaxy in which one player's abandoned wreck became another's content. Worlds are private and parallel, so that loop does not exist. Applied directly to the document rather than left as a proposal:

Change	Effect
Wrecks are procedurally generated in the player's own world	Works in single player from hour one, needs no content pipeline, and closes OI-124 (consent, anonymisation and attribution for harvesting player wrecks)
New §3a generate the history, not the layout	The mechanism for producing a factory that reads as badly built by a person: simulate a bad builder who only ever applies local fixes and never refactors
New §3b generation cost, six rules	Generate once and persist · generate at Scanner reveal in the background · compute pollution analytically rather than simulating years of it · budget the steps by severity · validate only the core, so the old core works and the patches half-work · strict determinism, or the Seasonal leaderboard is meaningless
Restoration stays inside a world	Never touches a gate, since gates are war slots and Restoration is the peaceful path
Payment is materials, contribution and reputation	Never credits. Credits are war-scoped, and paying peace in war currency makes peace a kind of war.
§6a.3 rewritten around seeds	A seed is four bytes and identical by definition, which is what a leaderboard needs. Stronger than harvesting, not weaker.

Still open in that document: the RECLAIM naming collision in §6, where Outpost Mode already uses the word at storey scale.

APPLIED — Pass A, 2026-09-05¶

These are done, not proposed. Listed so the checklist does not read as outstanding.

Amendment	What landed
1a, 1b, 1c vocabulary	`04_Data Dictionary` v0.5: §0 Scope Vocabulary (world, session, account, planet, war), §4b War Vocabulary, §4c Measurement Vocabulary. Six deprecations added, including control tether, the nautical idioms, armour efficiency, and session meaning a playthrough.
7b.2 doc 30 §10	The ship editor and component list rows now read as resolved, with OI-344 and OI-345 marked closed. New §9a Drafts and buildable designs: the Shipyard picker greys drafts with their failing rule.
7b.3 doc 24 §2.1	Do-not-harmonise warning added, with the reason stated: nothing shoots a building.
7, 7b.4 Unwritten stubs	Converted to redirects rather than deleted, and the folder README updated with a state table.

Pass B, applied 2026-09-05¶

Amendment	What landed
5a to 5h Canonical Design Facts	v1.7. New §16 to §23: supply sets magnitude and position sets distribution · manual first, automation is built · area-based supply · disabled never deleted · capacity crosses but stock does not · the channel is physical · world, session and account · value is fixed but return scales with presence, carrying the efficiency/power/coverage rule as its corollary. §12b extended with battlefield salvage and field refining. Revision History renumbered §15 to §24; §15 is now a deliberate gap.
5i stale §13	Marked STALE in place rather than deleted, naming OI-136 and OI-137 and the three reference pages that replaced the SQLite plan. Row kept until the replacement pipeline is described.
3, 7b.1 doc 15	v0.2. Win probability formula and modifier table deleted; OI-202 to OI-206 close with it. APR redefined as a rate. Real-time APR build display removed. Scout added as a seventh class, resolving the clash with `02_Scope` line 188. Fleet Power recorded as the admission-control rating. Faction fleet scaling retained with a note that it needs rebasing.
7c.1 doc 13	v0.2. Levels grant Engineering Points, replacing the percentage bonuses. Vocabulary corrected from session to world and account.
4a to 4d doc 14	v0.4. World model corrected — worlds are private, wars are shared, gates are the link — replacing the Galaxy Is Shared section, which the new architecture contradicted outright. War terms and the three victory conditions added, with the Contribution victory recorded as rejected and why. Factions persist and level, capacity only. Path C, Restoration added to the endgame. Mercenary path corrected.

Pass C, applied 2026-09-05¶

Amendment

What landed

2a, 2b, 7c.2 doc 08

v0.6. Amended in place per the document's own §0.3 protocol, branch listings untouched. New §0.4: research corrected from session-scoped to world-scoped at every site, the §0.1 status row struck through and pointed at the correction, and the GDScript comment reworded. The second tree recorded as a comparison table: Factory Research is world-scoped, branching, unlocks capability, paid in RP; Engineering Research is account-scoped, flat, raises rates only, paid in Engineering Points. The flat shape is stated as the structural enforcement of account research never unlocks capability. Research Point sources confirmed unchanged, and contribution confirmed not to feed them.

Correction inside 2b

The claim that infinite nodes move to Engineering Research is withdrawn, see the struck paragraph above. They stay world-scoped.

All ten amendments in this checklist are now applied. What remains under §9 and §10 is follow-up work, not the checklist.

Pass D, applied 2026-09-05 — the progression arc¶

Not part of the original ten. This is the Step 4 design session landing, recorded here because it amends four existing documents.

Document	What landed
NEW `Content/36_Progression Arc Design`	v0.1. Progression is the ladder of bottlenecks, not the reward schedule, with the milestone test: a beat must change what you are optimising, not just how much. The five Phase 1 beats (First Assembly, First Ascent, First Blackout, First Restoration, First Launch), ceremony on two of them only. Hour 1 / 10 / 100 as run, fit, change. Phase 2 beats sketched. The three endgame paths staggered. Levels never gate world content as an absolute rule.
`02_Scope & MVP Document`	v0.4. Three phases collapsed to two, because the ship editor serves space and war alike and inter-planet logistics is war logistics with a different destination past the gate. First ship launch moved into Phase 1 as its finale, using one prescribed hull, not the editor. Restoration added to Phase 1 as its long tail. Progression & Goals rewritten around the five beats, replacing the host-set session goals system. Dev-hour totals rebased; the total is unchanged.
`Content/14_Win Conditions & Faction Allegiance`	v0.5. Three-milestone arc superseded by the ladder. Milestone 2, First Territory Claimed, withdrawn: the node-and-zone section is rewritten so territory lives inside a war instance only and passive accrual has exactly one source, the Outpost trickle. Master Rill demoted to the capstone of Path A. PvP flag reworded: it governs which wars a world may enter, never the world itself. Phase 3 references renumbered. Header version corrected, it had been left at 0.3.
`Content/13_Player Levelling System`	v0.3. Levels never gate world content, with the §20 rationale and the Fleet Power admission-control exception.
`Systems/35_Economy Design`	v0.2. Are Phase 1's three sinks enough without the war answered: not alone, and the answer is not more sinks. Restoration supplies demand rather than waste.

Raised: OI-346 to OI-354. OI-354 is the one that matters: nine further documents still describe a three-phase plan, and the renumber is mechanical but not blind, since deferred to Phase 3 changes meaning rather than just its number.

Pass E, applied 2026-09-05 — the phase pass¶

Closing OI-350 and OI-354. The second one was not a renumber, which is why it was logged as read each, do not sed it.

OI-350, the release cadence. Phase 2 ships as two releases: the Space Update (editors, the gate, other planets, inter-planet supply chains) and the War Update (1.0) (wars, factions, multiplayer, mobile). The gate goes in the Space Update, because it is the channel out of a private world rather than a war door: in the Space Update it leads to space, in the War Update it also leads to wars. Recorded in 02_Scope & MVP Document v0.5. Stated there as a release boundary that must not become a design one, since the whole point of collapsing the design phases was to build the editor once.

OI-354, the phase pass. Reading the nine documents found the word phase running on two unrelated axes, plus a third ad-hoc one:

Axis	Owned by	Numbers
Release phase	`02_Scope & MVP Document`	Two. Phase 1 Early Access, Phase 2 Full Release
Development phase	`ROADMAP`	Eight, phase 0 to 7, each holding `M<phase>.<n>` milestones
(ad hoc)	`06_Vertical Layers & Core Power System`	A feature rollout on neither axis

A find and replace would have corrupted ROADMAP and the milestone IDs. M2.4, M3.1 and the rest are an interface: referenced across the vault and read by tooling. Development phase 2 is work inside release phase 1, so renumbering it to match would have been wrong in the other direction.

Document	What landed
`05_Reference/04_Data Dictionary`	v0.6. New §0a Phase Vocabulary, distinguishing release phase, release, and development phase, and stating that development phases are never renumbered. Bare Phase N deprecated, along with milestone meaning a progression beat, which is now beat.
`ROADMAP`	Headings renamed to Development Phase N with a warning box explaining the axes and the do-not-renumber rule. Opening line corrected: it had said Phase 2 Core Factory, with Phase 1 Planet Layer complete, mixing both axes in one sentence and claiming a release phase was finished.
`06_Vertical Layers & Core Power System`	Phase Scope rewritten as Build Order, stating plainly that the whole vertical system is release Phase 1 content and pointing build order at `ROADMAP`. It had read as deferred content, which contradicts First Ascent being the second beat and the first ceremonial moment in the game.
`01_Master GDD` · `09_UI-UX` · `01_World Bible` · `03_Monetization` · `02_Risk Register` · `01_Technical Architecture` · `02_Server Architecture`	Genuine release references repointed at the Space Update or the War Update. Monetization's price ladder rekeys cleanly, since its two steps were always the two releases. `01_Master GDD`'s Claim Territory corrected to Hold Ground In War.

Raised OI-355, HIGH. 02_Risk Register carries a 2026-02-22 decision reading full online multiplayer in Phase 1, while 02_Scope lists multiplayer under explicitly NOT in Phase 1. That is roughly a development phase of work, and the source is genuinely ambiguous: the 2026-02-22 row may have meant a development phase. Flagged in both documents, not resolved, because picking a side is El's call.

Pass F, applied 2026-09-05 — co-op moves into Phase 1¶

Closing OI-355, the contradiction Pass E raised and deliberately did not resolve.

The evidence decided it, not a preference. ROADMAP has always sequenced development phase 4, Multiplayer, before development phase 5, Early Access Launch. The February decision row in 02_Risk Register said the same thing. 02_Scope was the outlier, and it was corrected rather than the other two.

The root cause was the third overloaded word in a row, after session and phase:

Term	Meaning	Ships in
Co-op	More than one account in one world, building the same factory	Phase 1
War multiplayer	Many separate private worlds into one shared war instance through gates	The War Update

Co-op needs one world and more than one pair of hands. It needs no gate, no war and no shared instance, which is why it can ship years earlier.

Document	What landed
`05_Reference/04_Data Dictionary`	v0.7. New §0b Multiplayer Vocabulary. Bare multiplayer deprecated where the form matters.
`01_Design Documents/02_Scope & MVP Document`	v0.6. Co-op added to Phase 1 Platform & Tech and to the Definition of Done. Only war multiplayer stays on the explicitly NOT in Phase 1 list. Scope risk rerated Medium to High: co-op is release-blocking now, not deferred.
`04_Project Management/02_Risk Register`	Decision row resolved in its own favour and dated. T2, T4 and D3 renamed to co-op. D3 given a Phase 1 answer, since its stated mitigation (role specialisation across Planet, Space and War) does not exist in Phase 1: the factory has to be big enough that two people are not standing on the same belt.
`04_Project Management/ROADMAP`	Development phase 4 renamed Co-op Multiplayer, with a note that it ships inside release phase 1 and that war multiplayer is development phase 7.
`02_Technical Documents/01_Technical Architecture Document`	Session-discovery question reworded to co-op, warning box removed.
`01_Design Documents/Content/36_Progression Arc Design`	v0.2. New §7a Beats in Co-op. A beat is world state and fires once for the world; everyone present takes the XP; a guest carries nothing in and the XP out. No new rule was needed — it falls out of §7's world-state versus account-state split.

Nothing in the design had to change to allow co-op. Beats are world state, research is world-scoped with a shared slot pool, levels never gate world content, and credits are war-scoped so Phase 1 co-op has no economy question at all. Four rules written for unrelated reasons all held.

Pass G, applied 2026-09-05 — no content quantity targets¶

Raised by El while auditing the Phase 1 list, after data/buildings.json was found to hold 24 buildings against a written target of "~20 to 30 buildable structures at launch". Her call: there is no target, the design decides how many there are, and that is how many there are.

Recorded as a principle rather than a one-off edit, because a count deleted from one list grows back in the next one.

Document What landed

00_Canonical Design Facts v1.8. New §23a, Count The Content, Don't Target It. A count in a planning document becomes both a floor to pad up to and a ceiling to argue against. Content is counted in the owning file, never targeted. §23a.1 excludes constraints the build must satisfy, which stay as numbers; §23a.2 keeps counts in the owning document only, the vault's usual one owner, others reference. Header version also corrected, it had been left at 1.2 while the history reached 1.7.

02_Scope & MVP Document v0.7. Six content targets removed: buildable structures, research nodes, raw resource types, achievements, music tracks, reachable systems. Each now names the system and points at its owner. Warning box added at the head of the Phase 1 list so they cannot grow back.

Kept deliberately, as numbers: 60fps with 500 entities, co-op with 2 to 4 players, fun for at least 10 hours, no desync in a 2-hour co-op session. Those are promises the build makes, not facts about how much content exists.

Pass H, applied 2026-09-05 — the phasing pass¶

Every design document now carries a Phase: field. 47 files across 01_Design Documents/ and 06_Outpost/, up from one.

That was the actual defect. For six months the only phase assignment anywhere was a hand-written feature list in 02_Scope, written before roughly fifteen of the documents it was meant to cover existed. Recorded as a convention in 04_Data Dictionary §0a so it cannot regress: a system can no longer be designed without someone deciding when it ships.

El's calls, applied:

System	Phase
Trains and rail	Phase 1
Organic farming and biomass	Phase 1
Metallurgy and alternate recipes	Phase 1
Outpost Mode (7 documents)	Separate track, off the phase axis entirely
Achievements	Phase 1, no launch count targeted (§23a)
Gauntlet Planet	Space Update

Split assignments, where a document genuinely spans releases: 30_Ship Construction System (Phase 1 for the Shipyard and the prescribed hull, Space Update for the computed-bill system), 32_The Gate & War Logistics (Space Update for the gate, War Update for war logistics), 35_Economy Design (Phase 1 for materials and Engineering Points, War Update for credits and contribution), 08_Research Tree Design (Industrial branch Phase 1, Exploration in the Space Update, Military in the War Update), 14_Win Conditions (beats in Phase 1, factions and wars in the War Update).

02_Scope v0.8: the Phase 1 list reconciled with reality. It was missing ports, pipes, vertical levels, power as a system, modules, waste handling and planet defence. Three Phase 1 beats and two of Phase 1's three economy sinks depended on systems that were not on the list. The storage line was also still using the deprecated word containers.

Raised OI-356. 12_Gauntlet Planet Design said its modifiers were session-scoped, which under the locked vocabulary means the player loses them at bedtime. Corrected in place, and that makes three documents with the same defect, so the rest should be swept rather than waited for.

Pass I, applied 2026-09-05 — fun is the tiebreak, and the editor hauls first¶

Two of El's statements turned into design, rather than being agreed with and lost.

"Phase 1 needs to be at least fun to play, if not there is no point."

00_Canonical Design Facts v1.9, §23b, Phase 1 Earns Phase 2. Written as a decision rule, not a sentiment, because it is the tiebreak whenever scope is contested:

When a system is contested, the question is not how much work it is. It is: is Phase 1 fun without it? If no, it is in Phase 1, and the estimate moves.

It is load-bearing for two reasons the document states: Phase 1 is the only thing anyone pays for, and Phase 2 is three and a half to four years away, so Phase 1 carries a hundred hours and a price with nothing above the planet at all. §23b records this as why trains, farming and metallurgy are Phase 1 despite being deferrable on paper.

§23b.2 bounds it deliberately. It is not a licence to add content or to skip the trade-off: the estimate still moves and the cost is still written down, so a slipping date is a known consequence rather than a surprise. It also does not reach past Phase 1 into pulling the Space Update forward.

"We need the ship and mothership editors to have the ship designed to go from planet to planet and to automate the supply chain."

34_Ship & Mothership Editor Design v0.3, new section: The First Ship Is a Freighter. The document was drafted war-first, and the priority was the wrong way round: the editor opens in the Space Update, where there is no war yet.

The first ship a player designs has to reach another planet, carry something useful and come back. The first mothership has to run a supply chain without them. Neither job involves a weapon.

So the editor must be good at hauling before it is good at fighting. Solved first: cargo volume, fuel and range, loading and unloading at both ends, mass against thrust, and the mothership as a route with an automation ladder. Solved later: weapon mounts, armour coverage, combat class profiles, escort layering, retreat doctrine.

It is the same editor, not a cut-down one. The Space Update ships frames, layers, validation and the full catalogue. What changes is which readouts have to be right on day one. The stated test: a player who never joins a war should still find the editor worth using, because a better freighter is more cargo per trip, less fuel and a shorter turnaround, which are factory problems wearing a hull.

Pass J, applied 2026-09-05 — planet threats¶

New: Systems/37_Planet Threats & Defence Design v0.1. The first real specification of the system 04_Factory calls "the core Phase 1 tension" and had described in half a page, ending in a pointer to a document that no longer covers planets.

The principle: threats put a price on "more", and each family prices a different kind, so you cannot get big without drawing something whichever way you get big.

Three documents disagreed and all three were half right:

Document	Drew them by	Now
`04_Factory`	Expansion into territory	Kept, indirectly: expansion puts your smog next to them
`26_Pollution` section 6	E-waste and power signature	Kept, and its robot-world argument widened to the whole game
`05_World Generation`	Per-planet factions	Kept, reinterpreted as machines

El's calls: two families; natives drawn by pollution, ferals by power signature and e-waste; everything is a machine (Aquatic Swarmers are water-adapted machines); territory kept; destruction is real and takes the contents, no spill, but leaves a wreck that keeps its identity and footprint; repair cost scales with damage; offence is in.

Two consequences worth naming. The wreck rule is 28_Restoration at building scale, the same verb at two scales, learned on your own factory before you meet a generated derelict. And §7, There Is No Quiet Level, is unauthored: per-level pollution meeting power signature means burying dirty industry hides it from the natives at the cost of severe corrosion, while clean sky generation hides from the natives and builds a beacon for the ferals.

Document	What landed
`04_Factory & Resource System Design`	Alien Defence System marked superseded, with what changed stated in place. Its framing of the Phase 1 tension is kept because it is still right.
`05_World Generation & Planet Types`	Warning added: the alien factions are machines, and each planet's faction names its Natives only.
`26_Pollution, Waste & Organic Systems Design`	§6 marked extended: argument accepted and widened, pollution is now a trigger, and the family needs a new name.

Closed: OI-091 (scavengers appear in peaceful worlds, because peace is earned rather than toggled) and OI-197 (aliens are not a fifth faction, but who left these here now has somewhere to go). Raised: OI-357 to OI-360, of which OI-360, ammunition and weapons, is the next deep dive and OI-358 and OI-359 are HIGH because §5.1 and §7 depend on them.

Pass K, applied 2026-09-05 — the phase boundary corrected, and underground answered¶

Two of El's calls, one structural and one small enough to look small.

"Space layer is not phase 2. Only war is phase 2."

I had inverted her earlier statement. When she said "how do we move from planet to planet, we would need to be able to build ships and do supply chain between planets", she was giving a reason the editors belong with the planet layer. I read it as a reason space and war belong together.

The seam is war, not the planet surface.

Space is the planet layer reaching further: you cannot travel between planets without building ships, you cannot automate a supply chain between them without the editors, and a freighter route is a factory problem played the same way and judged by the same numbers. War is the only thing that is a different game — shared instances, opponents, attrition, and a supply chain being consumed by someone shooting at it.

	Phase 1	Phase 2
Content	Everything except war	War, and only war
Hours	~1,500–2,200	~800–1,200

The total is unchanged; the boundary moved. Because Phase 1 at that size is years rather than a release, there are now three releases across the two phases: the Launch release and the Space Update (both Phase 1, both Early Access), then the War Update (1.0).

First Launch ends a release, not a phase. In the Launch release you do not follow the ship. In the Space Update you do.

Document	What landed
`02_Scope & MVP Document`	v0.9. Release philosophy and the phase table rewritten. Phase 2 Release Cadence becomes Release Cadence, three releases. Phase 1's in-scope list split into the Launch release and the Space Update, with two separate not in lists so the release boundary is visible without becoming a design boundary. Phase 2 is war only.
`36_Progression Arc Design`	v0.3. §2 rewritten as Two Phases, Three Releases with the seam argument. §3 renamed The Launch Release Ladder. §5 split into a Space Update ladder (Phase 1) and a War Update ladder (Phase 2), which makes the character change legible: every Space Update beat is a factory problem in a new place, and the War Update beat is the first time the problem is another player. §6 paths keyed to releases, one opening in each.
`04_Data Dictionary`	v0.9. §0a phase values corrected. Permitted `Phase:` values are now Launch release, Space Update (both Phase 1), War Update, Outpost separate track. Development phase 4 noted as co-op shipping in the Launch release.
`34_Ship & Mothership Editor Design`	Wording: the editor opens in the Space Update, still Phase 1, and the War Update is a whole phase away rather than a few months.
`ROADMAP` · `01_Master GDD`	Phase references repointed.
33 design documents	`Phase:` fields repointed. 22 Launch release, 3 Space Update, 2 War Update, plus the split assignments.

"Nothing lives underground, but machines can travel via shafts, dig, and go underground."

37_Planet Threats & Defence Design v0.2, §7 rewritten as Underground Is Late, Not Quiet, closing OI-359. Three triggers and three routes, each from a rule that already existed:

Trigger	Route	Family
Air pollution at their border	Overland, at the frontier	Natives
Ground contamination leaching downward (`26_Pollution` §3a.2a)	They dig, up through the floor	Natives
Power signature	Down the player's own shafts, above all the Power Conduit	Ferals

Three consequences. Burying dirty industry hides the air but not the ground, which splits smelting and dumping into two separate decisions where they had been one. The bill arrives late, because leaching is slow, landing after the player has built enough down there to care. And every shaft is a door, so shaft count becomes a defensive decision as well as a logistical one: a level with no conduit is unreachable by ferals and also unpowered, which is the honest version of safety.

Pass L, applied 2026-09-06 — weapons and ammunition¶

New: Systems/38_Weapons & Ammunition Design v0.1, closing OI-360.

The principle: defence is a production line, not a purchase, measured in items per minute rather than turret count. A turret is a smelter that shoots.

The finding worth keeping. Two turret families were already latent across three documents: 04_Factory lists a Basic Turret at 15 PU and a Laser Turret at 40 PU, Branch 6 says the basic one eats Metal Plates, and 37_Planet Threats makes power signature one of the two things that draws the ferals. Put together:

Building an energy grid to fight the ferals makes more ferals come.

It deepens with tier, because Canonical §4 says Machine Mk draw rises faster than output. The stable answer is to use each family against the threat the other exhaust attracted.

El's two answers combined better than either alone. All buildings tier and ammunition tiers are not the same mechanism, and separating them resolves cleanly against Canonical §4:

	Tiers by
The turret	Upgrade kit, in place, preserving modules and ports. Base Mk is the reload, Machine Mk is the rate of fire and what it can chamber
The ammunition	Research, which unlocks a recipe that requires a Mk-N turret

You never rebuild a defence line to improve it, and it improves while you are not looking at it: research a better round, produce it, and the wall built ten hours ago hits harder untouched.

One ammunition ladder, shared with ship weapons. Drawn from goods the factory already makes, so defence competes with expansion for the same items, and the Phase 1 ammunition line is the War Update's ammunition line — not an analogy, the same items. Closes half of OI-343.

Document	What landed
`24_Module System Design`	The warning "nothing shoots a building" reworded, since `37_Planet Threats` made it literally false. Its point survives: position on a building carries no combat meaning. Turret module meanings added.
`37_Planet Threats & Defence Design`	§11 repointed at doc 38. Ferals fly recorded in §7.

Raised: OI-361 (Branch 6's Heavy Turret and Reinforced Wall must become Mk levels, and the audit note defending them withdrawn), OI-362 (the ladder's upper rungs, and damage versus rate), and OI-363, the second pass El asked for: items, buildings, recipes and flows one at a time. That last one is where OI-342, OI-343, OI-347 and OI-362 all get their numbers, and it is design work rather than a documentation pass.

Pass M, applied 2026-09-06 — Branch 6, and what a rung buys¶

OI-361: both become Mk levels. 08_Research Tree Design v0.7, Branch 6 amendment §0.5, written under that document's own §0.3 protocol rather than as a rewrite.

The 2026-08-30 audit note is struck through and withdrawn: "Reinforced Wall and Heavy Turret are distinct buildings rather than Mk levels, so they are legitimate research unlocks."

Node	Was	Now
Advanced Fortification	Reinforced Wall + gate automation	Gate automation only
Turret Upgrade II	Heavy Turret (2×2, distinct building)	Hull Plating ammunition, requiring a Turret at Machine Mk2

The general rule this exposed, and it reaches past Branch 6:

Under Canonical §4, no building is ever a research unlock at a higher tier. Research unlocks recipes; a recipe may require a Mk-N building; the Mk comes from a kit. Any node anywhere in the tree that unlocks a bigger version of an existing building is stale by construction.

Branch 6 is where that first showed up. The 2026-08-30 audit covered the other branches and found nothing of this shape, so no sweep is raised, but the rule is now stated where the next reader will meet it.

OI-362: a rung buys rate of fire, range and spread. Never damage.

Ammunition does not hit harder. It hits more often, further out, and wider.

Recorded in 38_Weapons & Ammunition Design v0.2 §4.2, with four consequences:

Nothing goes obsolete. A Mk1 turret on Metal Plates is slow and short-ranged and still kills what it hits.
Every rung raises consumption, so researching one without scaling the line leaves the player worse off than before they researched it. That is defence is measured in items per minute in its purest form, and it is not a trap to soften.
It couples the two tier tracks. A faster round drains the magazine faster, demanding a higher Base Mk, so one research node cascades into a production demand and a construction demand. Neither is a build-menu click.
Range and spread are geometry. A rung changes where turrets belong, not only how many.

Spread is the anti-swarm answer, which points it squarely at the ferals, who arrive drawn en masse to a signature rather than pushing a frontier the way natives do.

Left open, and flagged in both documents: whether spread needs a cap, and whether Turret Upgrade I's flat +30% range, +20% fire rate should shrink now that the ammunition ladder owns those same two axes, since the ammunition must stay the dominant source or choosing a round stops mattering.

Pass N, applied 2026-09-06 — the space layer¶

New: Systems/39_Space Layer Design v0.1. The last layer in the game with a "NOT WRITTEN" banner where its design should be. Unwritten/07_Space Exploration Design is now a redirect, not deleted, since documents still link to that name.

The principle, extending 19_Transport & Logistics §2's rule that no carrier is an upgrade of another:

The freighter owns the only way between planets. Space is the fourth carrier, not a second game. Its cost is round-trip time and fuel, so a second planet is a batch relationship, never a flow. You stockpile at both ends, and the size of those two stockpiles is the design problem the layer hands the player.

A correction of mine: the gate does not lead to space. 02_Scope v0.9 claimed it ships in the Space Update as "the channel out of a private world". Two facts kill that: 05_World Generation gives every world its own galaxy, so another planet is inside your world and you fly there; and 32_The Gate is explicit that a gate is a war slot, powered down between wars.

The gate ships in the War Update, and First Gate Powered and Path B move with it.

The peaceful alternative, a shared hub where worlds trade before any war exists, is filed as 00_Ideas/05_Idea - The Trade Hub rather than assumed.

El's call, and the centre of the layer:

The game does not prescribe a settlement pattern. It prices them, and the player chooses.

Mines, one deep factory, two or three specialised mains, or a factory on every planet, all legal, and which is right depends on the seed. §4 lists what prices each pattern — round-trip time, fuel, power not crossing planets, per-planet threats, per-planet vertical levels — and how each one fails in a way the player can see coming. The counterweight that makes it a choice rather than a trap is that research is world-scoped: the one thing free to spread.

The orbital station is repair, hangar and refuel, a service node off the gravity well. Not a factory, which makes it a consumer with its own supply problem — the same shape as a turret or a war.

Document	What landed
`02_Scope & MVP Document`	v0.10. Gate moved to the War Update with the correction stated in place; Space Update contents rewritten.
`36_Progression Arc Design`	v0.4. First Gate Powered and Path B moved to the War Update. New First Site beat in the Space Update, whose bottleneck is where the factory goes.
`Unwritten/07_Space Exploration Design`	Converted to a redirect with a table of where each question went.
NEW `00_Ideas/05_Idea - The Trade Hub`	The gate's possible peaceful destination, with what it would buy and cost.

Raised OI-364 to OI-366. OI-364 is Genesis, which El asked to take as its own conversation: every settlement pattern in §4 depends on what the builder can do, so it is now the largest undecided thing in the vault. OI-365 restates the Simulation Level question with new urgency, since a remote planet must still run honestly or a second planet is a diorama.

Pass O, applied 2026-09-06 — power ships, and the ferals become optional¶

El: "power can be shared via batteries that can be shipped from one planet to another."

That settles 25_Power System Design §5.3, Interplanetary Power — NOT DESIGNED, which had listed three candidates and recommended filing them all to 00_Ideas. Cargo wins; beamed power and a direct grid link are dropped.

The principle, extending 19_Transport & Logistics §3.6 to a third container:

Power crosses the same way a fluid crosses onto a belt: in a container. A barrel holds a fluid. A cache box holds a solid. A Charge Cell holds energy, and a charged cell is an ordinary solid item.

The shipped object cannot be the Accumulator, and the numbers say so. §5.5 sets it at 500 PU·s, which is five seconds of a 100 PU deficit, deliberately, because it is a buffer and not a power source. Five hundred of them on a freighter deliver about fourteen minutes of a modest 300 PU site, after a round trip measured in minutes. So the Accumulator stays exactly as specified and a separate Charge Cell does the shipping. Same relationship a building's internal buffer has to a Silo.

The rule that stops this deleting the power game:

You can ship the stock. You cannot ship the capacity.

A cell is a fixed amount of energy that runs out; a generator is a rate that does not. Imported power bootstraps, covers a night, or carries a permanent draw at high freight cost, and never replaces generation, because the freighter's throughput is the ceiling.

Two payoffs, and one of them rescued something I had broken.

A new site is viable before it is self-sufficient. Land, build drills, run on imported charge, stand up generation later. My §4.1 line, power does not cross planets, every site needs its own generation, had quietly made settling anywhere all-or-nothing: a working power plant before anything produced. This is what makes the settlement freedom in 39_Space Layer Design §4 actually playable.
Hard planets stay hard without being closed. The Rogue Planet has no star, dead solar and the lowest core rating, called genuinely hard mode. It can now run on imported power at the price of a permanent freight burden. A cost is a better hard mode than a wall.

The ferals become an opt-out, deliberately¶

El confirmed the reading, and then made it a design stance: "To power using the imported cell if they don't want to deal with the enemies. They have a choice."

26_Pollution §6 attracts the ferals to large generators and dense grids. That means generation, not consumption. Power draw is silent. So:

A player who does not want to deal with the ferals can choose not to generate locally. Generate on a rock with nothing worth defending, ship the charge to the site you actually care about.

The threat is relocated and paid for, not removed, which is what keeps it honest:

Still paid	Why
A permanent freight route	Cells in, forever, competing with everything else on the manifest
The ferals still come, to the generating site	You picked somewhere you do not mind. They did not go away
The natives are untouched	They come for pollution, and a factory pollutes whatever powers it
Freighter throughput caps the site	You cannot import the capacity

That is the shape every good option in this game has: a real answer, at a real price, with a failure mode you can see coming. A player who takes it and then loses the route finds out what it was worth.

Document	What landed
`25_Power System Design`	§5.3 rewritten from NOT DESIGNED to DECIDED, with the Accumulator sizing problem stated in place so nobody re-proposes shipping them.
`19_Transport & Logistics`	§3.6 retitled and extended to three containers: anything that is not a solid crosses only by packaging, power included.
`39_Space Layer Design`	v0.3. New §4a, §4a.4 rewritten as the opt-out, §4.1 pricing row corrected, Charge Cell added to the vocabulary.
`26_Pollution, Waste & Organic Systems Design`	§6 confirmed: signature means generation, not draw, with the consequence stated.
`37_Planet Threats & Defence Design`	v0.3. Same clarification in §4, pointing at the opt-out.

Raised OI-367, the Charge Cell specification, HIGH and blocked on the data pass. Worth noting in that row and repeated here: its numbers decide how expensive the opt-out is, so it is not a convenience item.

No open item raised for the signature question — El confirmed it in the same message, so it is settled design rather than a decision waiting.

Pass P, applied 2026-09-07 — Genesis¶

New: Systems/40_Genesis Design v0.1. Closes OI-364, OI-217 and OI-135. Genesis had 284 mentions across twenty documents and half a page describing it.

The diagnosis:

Genesis is the least automated thing in an automation game.

Every placement, upgrade kit, repair and deconstruct went through one manual unit, and Branch 7's answer was "builds 25% faster". Making a bottleneck faster is the least interesting answer available.

The defect: two rules were wearing one coat.

Rule	Verdict
Indestructible	Kept. Canonical §19's anti-softlock reasoning holds
"Always exactly 1, cannot be replicated"	Dropped. Never the load-bearing half, and the cause of everything

El's calls:

One Genesis per player. She found the problem herself: "in co-op there is nothing for the other person to do." Under the old rule, three of four players could not build.
Bots do the work. "That's the point. If bots do the work, Genesis doesn't need to do it."
Genesis upgrades, and research points at it.
Genesis talks and makes sounds.

What co-op shares and what it does not. Shared: the factory, the grid, Factory Research and its slot pool, and construction materials. Personal: your Genesis and your build queue. Materials being shared is kept as a feature — two builders drain the plate line faster, which is a conversation rather than a rule, and role division then happens because people choose it.

"Supervisory role" answered, and the answer is nothing. 04_Factory asserted Bot Programming II makes Genesis oversee the bot fleet and never said what that meant. Bots take work from the build queue and so does Genesis. No supervision layer and no bot management screen, because inventing a management system for the thing meant to end management would undo the point.

Upgrades separated into three mechanisms, the same reconciliation Branch 6 got the day before:

Research unlocks capability. Kits and modules buy capacity.

So Branch 7 loses four flat-percentage nodes and keeps every node that unlocks a verb. Upgrades are world-scoped, never Engineering Points, which means a guest's Genesis is exactly as capable as the host's — they bring their skill and their designs, not a better robot.

Genesis should be the voice of the notification system. Three Phase 1 beats depend on the game telling the player something, and First Blackout is dead unless the shed tier and the deficit are named out loud. Personal per player in co-op, so four people do not share one nagging voice.

Document	What landed
`04_Factory & Resource System Design`	Genesis section superseded, count corrected to one per player, the supervisory assertion answered in place.
`08_Research Tree Design`	v0.8, §0.6. Branch 7 amended: research stops selling speed. Four percentage nodes deleted; Mining Upgrade's unlock half survives as Genesis Multi-Resource Task.
`39_Space Layer Design`	v0.4. §8's Genesis deferral resolved: Genesis founds, bots continue, so §4's settlement freedom holds.

Raised OI-369 to OI-372. OI-370 is the one to look at: §5 of the new document, founding a site requires Genesis, is the only call Claude made rather than El, and it is flagged as such in the document itself. It is what gives interplanetary travel a decision, and the simpler alternative (a freighter drops a self-unpacking bay) costs the space layer that moment.

Pass Q, applied 2026-09-07 — three additions to Genesis¶

All three from El, and each one fixed something.

1. Bots are faster. Genesis never stops.¶

"If bots run out of energy, Genesis can still do the build."

	Bots	Genesis
Speed	Faster, and parallel	Slower, and alone
Energy	They run out. The bay needs power	Never stops. Slows on a dead grid
Needed	When there is volume	When there is nothing else

This is the best thing in the document, because it connects the power system straight to construction. When the grid sheds, production stops and the bot bay stops with it. Genesis keeps working by hand.

So during a blackout the player is repairing their grid with one pair of hands, exactly as it felt in hour one. A blackout demotes you back to manual.

That is what makes First Blackout (36_Progression Arc Design §3.3) land as a beat rather than a notification: the factory does not merely pause, the player's capabilities regress until they fix it.

Genesis itself never runs out — grid when there is one, plus a reserve that always trickles back, so it slows and never idles. Canonical §19 again: a builder that can be switched off by the failure it exists to repair is a softlock with extra steps.

2. §5 confirmed, and broader than I wrote it¶

"Genesis has to be around for any build if there are no other bots."

That was the one call I had made rather than El (OI-370, now closed). She confirmed it and generalised it, which is better:

Genesis builds wherever bots cannot. Founding a planet is simply the most obvious case.

The freighter-drops-a-self-unpacking-bay alternative is dropped.

3. To hear a planet you are not on, build a satellite¶

"For Genesis to notify, they have to set up a satellite first."

This fills a hole 39_Space Layer Design §4.3 had left open and I had not solved: the UI has to make a second planet legible without being there.


With a satellite	The planet reports. Its Genesis speaks to you, alarms reach you, you can open its panel and queue work
Without one	Silence. You find out when you arrive

The satellite gives you the panel, not the hands.

Three things it earns. It is consistent — Canonical §17, remote awareness is built rather than granted, and it is the Scanner's sibling. It prices settling widely honestly: many sites means many satellites, and an unmonitored site is genuinely quiet, which is a cost rather than a rule against sprawl. And it makes coming home meaningful — arriving at a site you have not heard from in an hour is a good moment, and a game that tells you everything all the time does not have it.

It also gives §7.2's unattended-planet rule teeth: a remote site can genuinely stall, because its bots run on power and there is no Genesis there to carry on. With a satellite you would have heard.

Document	What landed
`40_Genesis Design`	v0.2. New §4.0, §5 confirmed and retitled Genesis Builds Where Bots Cannot, new §7.1 for the satellite.
`39_Space Layer Design`	v0.5. §4.3 and §7.2 answered.

Closed OI-370. Raised OI-373 (specify the satellite, HIGH — not cosmetic, it is what prices settling widely) and OI-374 (how long a bot lasts on a charge, HIGH — those numbers decide how bad a blackout gets, and the blackout must be felt without being fatal).

Pass R, applied 2026-09-08 — fluid packaging and carriage¶

New: Logistics/41_Fluid Packaging & Carriage Design v0.1. Closes OI-272, OI-278 and OI-297.

El's correction, and it is the one that made the system extensible:

Molten metal is still a liquid. fld_liquid_metal has phase = LIQUID; it is a liquid at 850°C. So the question was never is it liquid at ambient, it is can the container hold it at temperature.

Which gives the whole design, from her barrel-tier idea:

The container ladder mirrors the pipe ladder, because it is the same physics.

	Pipes	Containers
Ambient	Standard Pipe	Barrel
Hot	Insulated Pipe	Insulated Barrel
Cold	—	Cryogenic Barrel (Space Update)

A player who learned Liquid Metal needs Insulated Pipe already knows it needs an Insulated Barrel. The cryogenic tier has nothing to carry until Gas and Ice Giants supply Methane Ice and Ammonia; it is specified now so the ladder is complete.

OI-272's trap avoided two ways. The insulated tier is roughly the same research and material commitment as Insulated Pipe, so it is a different shape of infrastructure rather than a bypass. And:

A barrel line is throttled by the machines at each end, not by the belt in between.

Two machines, two power draws, two footprints and a return loop. A pipe has none of those, so a pipe wins at volume automatically and barrel capacity becomes a free parameter tunable for feel rather than a balance lever. That is a good place for a number to sit.

Her decompression question, answered: no. A sealed container is sealed and the vacuum outside it does not care, and a pressurise-depressurise stage at every planet boundary would tax every route with no decision attached. But the question pointed at something real:

Liquids go in barrels. Gases go in cylinders. Pressure is to a cylinder what temperature is to a barrel.

That is where the pressure question belongs, and it matters because gases come from the Gas and Ice Giants.

Her tank wagon improves the carrier picture rather than complicating it:

Rail is the only carrier that moves fluid as fluid without a pipe.

It keeps doc 19's no carrier is an upgrade of another rule intact, gives rail a genuine fluid advantage, and does not step on barrels because rail cannot reach another planet. The plumbing precedent was already there: the Diesel locomotive's fuel platform takes a fuel pipe in.

Also decided: barrels are a loop and Canonical §12b makes it lossy, so a barrel line carries an ongoing steel cost; flammable barrels are a fire risk, which makes routing a decision and means a fire costs steel and product together; tanks store, barrels move; the fluid drone is deleted; and barrelling needs two input media, confirming OI-057, which the port model already supports since medium is per port.

Document	What landed
`19_Transport & Logistics`	§3.6 extended to four container families with the barrel ladder named; carrier table and axis row updated for the tank wagon.
`07_Pipe & Fluid System Design`	Pointer added, noting `min_temp` is now load-bearing for which container a fluid needs, not only pipe blockage.
`22_Trains & Rail`	New §4.2a, the tank wagon, with the Diesel fuel-platform precedent.

Raised OI-375 to OI-378. OI-375 is OI-297's surviving half (the drone-making building). OI-376 is the one with a real design choice left in it: whether an Insulated Barrel cools on a timer, which would force casting to stay near smelting but can strand a train.

Why the stubs were redirected, not deleted¶

Deleting them breaks 16 inbound wiki links across the vault, and repointing those needs a working shell to find them all. Converting each stub to a redirect page achieves the important half immediately — the vault stops claiming these documents are unwritten — while every link keeps resolving and mkdocs build --strict keeps passing.

Deletion is a follow-up, logged below, to run when a shell is available.

Pass S, applied 2026-09-10 — the ambient rule, and what it deleted¶

Logistics/41_Fluid Packaging & Carriage Design to v0.2. Closes OI-377.

El's rule, which replaced the one Pass R had just written:

What state a substance is in at ambient decides how it travels — because that is the moment it is encased.

At ambient	Travels as
Solid	An ordinary item. No container at all
Liquid	A barrel
Gas	A cylinder

Pass R's framing (can the container hold it at temperature) was not wrong, it was the second question, and asking it first let the ladder grow a rung for every temperature band anyone could imagine. The ambient rule cuts that off. Temperature is still the axis a rung moves along; it is no longer the axis that picks the family.

Two containers deleted, and neither is a loss:

The Cryogenic Barrel. Methane and Ammonia are gases at ambient, so they are cylinder cargo. The Cryogenic Cylinder rung owns them, which is what closes OI-377: the overlap existed because two families were being selected by a temperature question, so one substance could satisfy both. Selecting by ambient state makes the answer unique by construction.
The Insulated Barrel, because of the next item.

Molten metal stops being cargo. It is solid at ambient, and ref_metal_plate already exists as its shipped form, so nothing had to be invented:

You cast it, and you ship the metal. Molten metal exists only between the smelter and the caster.

Which sharpens Insulated Pipe's job from the pipe for the hot fluid into the reason smelting and casting have to be neighbours. That is a layout pressure rather than a material cost, and layout pressure is the currency this game trades in.

El's question — does anyone transport molten in real life — changed one deletion back. They do, but only by rail and only over kilometres: torpedo ladle cars between a blast furnace and a steelworks. Cooler melts travel further precisely because they have less heat to lose, which gives the underlying rule:

How far something can travel molten scales inversely with how hot it is.

And the shape rule that saved the wagon while condemning the drum: molten cargo travels as one large vessel with real thermal mass, because what kills it is surface area against mass. A barrel is the worst possible shape for that. So the insulated tank wagon survives, short haul only, and it cannot be stretched into interplanetary shipping, which means molten metal never leaves the planet needs no separate prohibition anywhere.

"Mk" was refused for containers. Canonical §4 makes Mk a building concept: two tracks, applied in place by universal kits, preserving modules, ports and contents. A container is an item with none of those. Naming a container "Mk3" would have been the fourth time in this vault one word was given two jobs. Containers tier as named rungs, the way pipes already do.

Simulation Level, and the rule that came out of it. §10 gives packaging a three-level table — ratings bite at 2, wear and fire at 3 — but §10.1 is the part that generalised:

The empties loop is baseline at every level.

Abstracting it away passes doc 19's existing test (adds detail, never changes the mechanic) and still breaks the design, because the loop is one of the four costs that make a pipe beat a barrel line. Remove it and barrels become strictly better than pipes at the level new players start on. That produced Canonical §24, below.

Pass T, applied 2026-09-11 — two canonical rules, and the air¶

00_Canonical Design Facts to v2.0. Two new sections, Revision History renumbered to §26.

§24 — a Simulation Level never changes what is optimal. Fidelity scales, balance does not. §24.1 gives the applied test (what would a player do differently at each level) and the worked verdicts. Its most useful entry is a feature that failed:

Wind. Directional pollution is more interesting at a higher level and costs nothing at a lower one, which is exactly the shape of a good fidelity feature. It fails anyway, because a player with wind on puts farms upwind and a player with it off does not. Those are different games.

§25 — the world is shared and nothing in it has a per-player owner. El's correction: there is no this player's building or that player's building, it's a shared world that the host owns. Written down because the consequence is a whole family of ideas ruled out before they get designed — per-player factory statistics, in-session contribution scoring, ownership locks, and the who-built-this tooltip that is the thin end of all of them. Genesis stays per player and is not an exception: it is agency, not property.

26_Pollution, Waste & Organic Systems Design to v0.3. New §3b, closing OI-089.

Air is a coarse grid, per level, with weighted-falloff sources, diffusion, sinks and a small natural decay. Radius falloff was rejected for one reason with four faces: it stores nothing, so pollution could never arrive, never accumulate, never be cleaned up, and switching the factory off would clean the world instantly. 37_Planet Threats & Defence §5.1 already requires air to arrive where the natives live, so the grid is not a fidelity preference, it is what the threat design depends on.

The best thing in it was not planned. A small natural decay per cell produces an equilibrium, and the equilibrium is a free budget: below some emission rate the planet handles it and nothing stirs. That protects the first hour, which Canonical §23b requires, and it replaces more is worse with get back under the line, which is an instruction rather than a mood.

Wind ships in the Launch release, for everyone, per §24. It gives pollution a downwind tail, which poisons a farm downwind and leaves an identical upwind farm clean, and it makes threat direction predictable — so defence becomes a decision about where, instead of a tax on the whole perimeter. ⚠️ The price is real and is now a tracked debt: wind is invisible, so it obliges a persistent indicator and a tutorial beat naming it before the first farm goes down (OI-381).

§3b.4 folds §3a into coefficients. Underground's nowhere to vent becomes zero diffusion and zero decay rather than a special case in the code, so the level rules stop being a branch.

§3b.5 splits the haze from the overlay, which were about to become one thing. The haze is weak and always on; ⚠️ it must stay well below obscuring the factory, or it is switched off on the first bad day and takes the feedback with it. The overlay carries real intensity and shows pollution, sinks and native zones together, because showing only the cloud makes it a status light rather than a planning tool.

Two rows got their substrate rather than their answer. OI-358 (what pollution reaching territory means numerically) now has a stored per-cell amount to threshold against, and §3b.2's decay means a woken zone can be calmed, which that row asked the design to preserve. OI-365 (Simulation Level on an unattended planet) is answered by turning two dials — bigger cells, slower ticks — rather than by writing a second, fake simulation, which is what honest over cheap was asking for.

Pass U, applied 2026-09-11 — interface and feel¶

Five interface documents, all of which predated everything they exist to communicate. They were written in March and April; threats, pollution as a grid, wind, Genesis, the space layer and the whole progression ladder came after. Closes OI-372 and OI-381.

The sentence none of the five had:

The interface has two jobs. Tell me what my factory is doing wrong, and mark the moments. Everything else is chrome.

A panel list has no way to say no to a twelfth tab. That sentence does, and it also names what was missing: the documents did the first job well and the second job not at all. "First ship launched. The galaxy is watching" was an Info toast — the same visual weight as "Session autosaved". The finale of the Launch release had the presentation of a save confirmation.

Two new tiers that are not urgency, which is why they sit outside Info/Warning/Critical rather than above it:

Tier	Says
Primer	Something is new
Moment	Something happened that matters

El's answer on Genesis: vocalisations. Non-verbal sounds — a curious chirp, an alarmed warble, a flat descending tone when something dies — keyed to tier and category rather than to individual messages, so a new notification costs a table row and no audio work. Voice acting was rejected for a solo developer: budget, a localisation pass per language, and a recording session for every notification added later, which makes the list expensive to extend in exactly the way this system must not be.

The gain nobody was looking for:

A sound has a location. Genesis is somewhere.

So notifications pan and attenuate from Genesis's position, and the player learns where it is without looking. ⚠️ Two guard rails: throttled harder than toasts with Critical always winning the slot, and never sound-only.

Notifications have range, and the silence is the product. 40_Genesis Design §7.1 requires a satellite to hear a remote planet, which makes this system diegetic rather than omniscient. Taken strictly: the planet you stand on always notifies, a remote planet without a satellite is silent, and arrival delivers the backlog. Softening it into "something may be wrong somewhere" would sell no satellites and would be worse than either extreme.

Notifications become structured events rather than strings, because v0.1's power copy — "Grid 2 is drawing more than it produces" — is useless during a cascading shed, which is the entire First Blackout beat. The new copy names the tier dropped and the deficit, which a string cannot carry unless the event carries it first. §5.1 works that beat through all three purposes at once:

When	Tier	Player gets
First grid deficit ever	Primer	"Power has priorities. A grid that cannot pay drops its lowest tier first."
The grid sheds	Critical	"Grid 2 shed Tier 3. Deficit 47 PU."
The beat completes	Moment	First Blackout marked, logged, achievement unlocked

The rule that makes primers teaching rather than nagging, and the one that pays OI-381:

A primer fires before the lesson costs you, not after. Not when the farm withers — when the farm is placed.

El's answer on the tutorial: leave it, and revisit after everything is finalised. Recorded as the tutorial teaches the hand, primers teach the game, with the full revisit deferred to OI-383 because doing it now means doing it twice — the data pass has not yet settled what the first hour contains.

Ceremony stayed a flag, not a tier. 36_Progression Arc Design §3 already marks only First Ascent and First Launch as ceremonial, and this pass did not second-guess it. Inflating all five would flatten the two that matter: a Moment that happens five times is a notification tier, and a Moment that happens twice is an event.

Beats became achievements (El's addition), which fell out cleanly: a beat is world state, an achievement is account state, so everyone present when it fires unlocks it — the same rule that already governs the XP.

And that exposed a real contradiction. 11_Achievements Design Document v1.0 said production achievements are per-player, each player's output tracked individually. Under the new Canonical §25 that is not merely hard, it is impossible: there is no partition to measure, because a smelter is not anyone's smelter. Replaced with a general test:

Ask whose is it. If the question has no answer, the achievement is shared.

17_HUD Panel Flap had its own stale defect. The Pollution tab's overview listed Water Contamination — percentage of the fluid network affected, a framing 26_Pollution §3 had explicitly withdrawn: contamination lives in the ground and water table, not in the player's pipes. Corrected, and the overview now leads with emission against absorption — the two rates whose crossing is the equilibrium, which turns more is worse into get back under the line.

The wind debt is paid in three places: the indicator in the centre top-bar panel beside the planet name (because wind belongs to the planet), the primer on first farm placement, and the wind direction drawn in the pollution overlay.

Pass V, applied 2026-09-13 — the ship economy¶

30_Ship Construction System to v0.2, 34_Ship & Mothership Editor Design to v0.4. Started as the data pass and turned into a design session, because El's start from the top meant the ship, not the raw resources.

That reframing was the right one and it changed the method:

The ship is the bill of materials the whole game exists to satisfy. Work down from it, and every resource has to earn its place by appearing in that decomposition.

Bottom-up asks what should this ore be called. Top-down asks does this ore need to exist at all.

A defect caught immediately. §9's Shipyard tier table listed six classes and omitted the Scout — a class added to 15_Ship & Army Points Design four days after this document was written, and described there as the first ship a player ever launches. The Launch release finale was a hull no Shipyard could build. Mk1 now takes Scout, Fighter and Freighter, and the two-per-tier symmetry is abandoned, because the cheapest yard should make the cheap hulls.

What the Scout turned out to be, from the fiction rather than from a table: the player is a command intelligence with no body, who has only ever seen through things they built. So the first ship is the first time their senses leave the planet, which is why a scout rather than a fighter or a freighter, and why the Space Update is you following it.

El corrected an over-assertion: I had declared the Scout one-way. It is not.

It comes back if you gave it the fuel to.

Which is much better, because it turns the finale into a trade made in the editor — a bigger tank is a shorter reach and the ship comes home, a smaller tank sees further and never returns. Range is a round-trip decision, and it is the same tank-against-everything tension the editor is built on. The same mistake as the settlement question: a rule put where a decision belonged.

The editor moved to Phase 1, which it always was. The confusion was mine: Phase 1 contains two releases, and saying Phase 1 when meaning the Launch release made the editor look like Phase 2 content. The Launch release has one prescribed hull because there is nowhere to send a designed ship; the Space Update has the editor because now there is.

The internal model El proposed, and the objection I withdrew. Wires for power, pipes for fuel, designed once and saved as a blueprint. I had argued against conduits — internal plumbing makes designing a freighter an evening's work — and the blueprint is the answer:

Nobody has to do it twice. The cost is paid once and amortised across every ship built from that design, which is the deal factory blueprints already make.

Heat stays a constraint rather than a third network: hot parts have a radius and what sits inside it suffers. A coolant loop in every ship is one system too many, and coolant already has a job in power generation.

Blocks. A blueprint can be revised; a ship remembers which version it was built from. New builds take the current design, existing ships keep theirs and keep working, which is Canonical §19 pointed at a ship.

⚠️ Called Block, not Mk. Canonical §4 makes Mk a building concept, and ships stand beside buildings with real Mk levels. This would have been the fourth word in this vault doing two jobs.

Refit, and the rule that priced it. El asked whether the overhead should be fixed per part or fixed overall, and then asked the question that killed the whole idea of an overhead: instead of upgrading, I can decom and build a new one, might be cheaper. The answer was not to tune a fudge factor until the exploit closed. It was to notice:

Refitting and decommissioning are the same operation at different depths. Both mean opening the ship up, so the only question either asks is what survives being pulled out.

Discrete parts come out whole. Continuous material — wiring, piping, plating, structure — is destroyed being cut through. From that one rule: refit cost is the new parts plus what you cut through, decommission return is every discrete part and none of the material, and decom-and-rebuild can never undercut a refit because it destroys all the continuous material twice over. When the two prices converge, you genuinely are building a new ship.

It also produced a design axis nobody authored: a ship can be built to be refitted, keeping changeable parts shallow and accepting a slightly worse ship now so every later upgrade is cheap.

The yard's store, which made decommissioning interesting. El: the shipyard will have an internal storage system, it can be full, and it can use the items in the storage too. So:

Decommissioning is not a refund. It is logistics.

Parts land in the yard as crates and have to be dealt with. §5.1's shortfall rule then means pulling a reactor from an old ship stops the next build asking for one, with no new rule at all. ⚠️ The store is chest-like, not silo-like — one general mixed capacity — because Canonical §7's one-substance silos would make it is full, empty it meaningless across a dozen part types.

Wrecks. El first said destroyed is destroyed, then corrected it: a destroyed ship leaves a wreck that must be salvaged if they want it. Which gives a cleaner line than the one I drew:

A destroyed building is restored in place. A destroyed ship has to be gone and got.

A lossy subset of parts, no condition tracking, salvaged by a payload rather than a class (the catalogue already has a Salvage Module), priced by distance, and unfarmable because decommissioning is strictly better than blowing up your own ship. It also extends Restoration off the planet, and it makes the game's opening scene — a survivor picking through The Wreck, with Genesis described as a salvage and construction bot — the first instance of a verb the player uses forever.

One thing deliberately not changed, and reopened instead. 34_Ship & Mothership Editor Design gives escorts a layer model and the mothership a freeform grid, arguing that freeform on an escort would be decoration because only combat can judge an escort and that needs a damage simulation. This session produced an evaluator that is not combat — power and fuel routed through the hull — and ⚠️ the document already refutes itself: its mothership section computes directional exposure per face from a freeform grid with no simulation at all, which is exactly what the escort section says cannot be done.

That is El's call, not a pass edit. Logged as OI-386, HIGH.

Pass W, applied 2026-09-13 — one model, and everything is a hull¶

34_Ship & Mothership Editor Design to v0.5. Closes OI-386, which Pass V had opened rather than decided.

What was in question. v0.1 built escorts from concentric layers and the mothership from a freeform grid, and argued the split rather than assuming it: an escort is judged only by combat performance, evaluating geometry for combat needs a damage simulation, and 31_War & Combat System Design rules one out. So freeform on an escort "would be decoration".

Two things undid it, and only one of them is new.

An evaluator appeared that is not combat. Power on wires and fuel in pipes, routed inside the hull. A layout is valid or invalid whether or not anything ever shoots at it.

And ⚠️ the document already refuted itself. What the Mothership's Geometry Controls §1 computes directional exposure per face from a freeform grid with no simulation at all, and calls it "the mechanic that makes freeform building matter". That is precisely the thing the escort section said could not be done. The contradiction had been sitting in one file since 4 September.

Nothing the layer model taught was lost, which is what made the change safe rather than merely appealing. Every layer rule is re-derived from geometry:

Layer rule	On the grid
Weapons on the shell to fire	A weapon needs an edge cell
Every gun is a piece of armour you did not fit	An edge cell holds a gun or plating, never both
Only a deep class can bury a reactor	Falls out. A thin envelope has no buried cells

Layers survive as a derived read-out rather than a construction constraint, and the envelope improves on them: layers gave every class the same kind of decision, two to four rings, while an envelope gives design depth proportional to hull size.

El's call went further than the proposal. Not just one spatial model — all should be hulls, standardise it. A Scout and a mothership become the same kind of object at different sizes, which also collapsed:

Two data shapes into one. The slots[] form is deleted
Two validity lists into one, with rules that bite only on designs containing the part
Mothership frame tiers into three ordinary catalogue entries

Everything that made a mothership special stopped being a system and became a property of a hull: very large, many heights, one per player, accepts payloads nothing else fits.

The envelope is what stops free placement producing blobs. A hull class is a fixed outline that cannot be exceeded; inside it, build sparse or packed. Square cells rather than hex, because the factory is square, hexes do not stack into heights, and hexes are bad at the straight runs conduits need.

⚠️ Conduits occupy cells, and that detail is load-bearing: a fully packed ship cannot be wired, which is what makes envelope size and extra heights worth buying. If wires ran along cell walls for free, the envelope would collapse back into a capacity number. The open question of conduit throughput (OI-392) decides how hard that bites on the largest hulls.

The connectivity rules also gave an existing section content. Drafts and Buildable Designs drew a careful distinction between a saveable draft and a buildable design, at a time when almost any arrangement was valid. Wires and pipes give it real failures to report: drive core is not wired, no pipe run from any tank to the drive.

Two clarifications worth keeping. El asked what pay once meant, and the answer is that two currencies had been blurred: materials, which the factory spends on building, refitting and losing ships, and the player's time, which is what designing costs. A blueprint saves the second and never adds to the first — a ship built from a saved design costs exactly the parts it is made of.

And ⚠️ one inheritance flagged rather than carried: mothership frames were unlocked through account-scoped Engineering Research while hulls are gated by world-scoped Shipyard Mk. One catalogue puts both on one list, and an account-scoped gate granting world capability would rub against levels never gate world content. OI-391.

Pass X, applied 2026-09-15 — the two gates, and what a conduit carries¶

34_Ship & Mothership Editor Design to v0.6, 36_Progression Arc Design to v0.4. Closes OI-391 and OI-392, both opened by Pass W rather than answered by it.

OI-391 — El's reasoning settled it, and it is a phase boundary, not an inconsistency.

A ship is world content. The mothership is how you show up.

Planet-to-planet hauling, mining and scouting are things a player does at home, so a Shipyard is gate enough. A mothership exists only to carry a player into someone else's world, which makes it the same shape as the admission-control exception 36_Progression Arc Design §7.2 already allows: a war is a shared instance, not anybody's world.

⚠️ One line had to be amended rather than left standing. That table read Engineering Points gate rates and capacities on ship and mothership designs. Nothing else, which became a flat contradiction the moment Pass W folded motherships into the ordinary hull catalogue. Mothership hulls** are now named on the account side, and a new §7.3 records why. §7 itself is untouched and no looser: a guest still brings no buildings, no recipes and no research into a host's world.

⚠️ A side door was closed at the same time. A mothership is commissioned in your own world (OI-393), or a high-level guest could spend a host's materials on a hull the host cannot build, which is account progress arriving as capability inside somebody else's factory — precisely what §7 exists to prevent, reached by another route.

OI-392 — the answer was to stop inventing.

A ship's wires behave like the planet's power grid. A ship's pipes behave like the planet's pipes.

A power network's capacity is its weakest conduit. Exceed it and the player upgrades the tier or splits into two networks — a sum and a minimum, with no flow solver anywhere in it. A large hull therefore carries several internal grids rather than one spine, which is how a planet-side factory already behaves and exactly the pressure that should exist at that size.

Unlimited throughput was cheaper and was rejected on one decisive ground:

⚠️ It inverts the difficulty curve exactly where it should be steepest. One thin spine feeds a mothership, and a mothership has ample room for one, so the largest hull would be the easiest to wire. It would also flatten the refit axis, since cutting a single thin spine is always cheap, and reduce wiring to a formality repeated from memory.

The gain nobody was looking for, and the second time this session that reuse has paid:

A ship whose grid runs short sheds tiers in flight.

Exactly as a base does. The First Blackout lesson transfers whole: the player already knows what a shed looks like, and now it can happen to a freighter halfway to another planet, because they fitted one cargo hold too many.

Pass Y, applied 2026-09-20 — the hull model, and ships stop landing¶

34_Ship & Mothership Editor Design to v0.7, 39_Space Layer Design to v0.6, 19_Transport & Logistics amended.

Geometry did most of the design work, once one number was noticed.

internal = (w − 2) × (d − 2) × (h − 2)

⚠️ Nothing can be protected below height 3. A hard gate from pure arithmetic, which says exactly what v0.1's class-depth table was reaching for: a flat hull has no inside, so its reactor is always exposed and — pleasingly — it is always cheap to refit, there being nothing buried to cut through to. And since internal volume grows far faster than size, small hulls are all skin, large hulls have guts.

Hulls are named by size and shape, never by role. Before this, Scout was simultaneously a hull, a class and a job — the word-doing-two-jobs pattern, for perhaps the sixth time in this vault.

Name hulls by size. Name the blueprints by role.

Light · Medium · Heavy Lift · Heavy Line · Command I–III. Roles demote to stock blueprints, which means a new role is data rather than a system: somebody saves a design, rather than somebody models, balances and gates a hull. The catalogue drops from seven-plus to five, and ⚠️ the Space Update needs three hulls, not ten.

Heavy Lift and Heavy Line are the branch, and the numbers are the argument:

	Footprint × height	Cells	Internal	Edge
Heavy Lift	9 × 9 × 3	243	49 (20%)	194
Heavy Line	7 × 7 × 5	245	75 (31%)	170

Near-identical size, opposite ships, neither an upgrade of the other. Shape decides the ratio of what you can carry to what you can protect.

Command rather than Mothership, since a command ship is what a commander operates from and the player is a command intelligence. The fiction and the catalogue agree for once.

El's observation reshaped the economics of the whole space layer:

Getting to orbit is most of the energy. Going between orbits is comparatively little.

39_Space Layer Design §6 already half-knew it, noting that a ship landing to refuel "spends fuel landing and taking off again". Taken to its conclusion: ships never land. Three stages, three carriers — Mass Driver up, freighter hulls across, Drop Pod Bay down.

⚠️ The lift must not be a ship. A shuttle would collapse the fuel story back into the fleet and every hull would be designed around its worst leg, crushing the editor's real decisions under one number. As a building, reaching orbit becomes an infrastructure problem, which is the right shape for this game.

You do not fly cargo to orbit. You throw it.

The asymmetry is the point. Importing to a planet is cheap; exporting from it is expensive. Which pays off the cargo mass decision taken in the same conversation: if mass matters and the expensive leg is the lift, then refine before you lift, and a mining planet grows a factory rather than staying a field of drills. Nobody teaches that; a player watches the power draw.

The Space Elevator is the later upgrade, and it is the right shape for one: a Mass Driver is cheap to build and expensive to run, an Elevator is brutal to build and cheap to run. That is the belt-against-drone trade applied to the hardest corridor in the game.

Two rules written down that will otherwise be violated by accident.

Nothing is declared, everything is derived. Only the envelope and the parts are authored; class, speed, range, capacity, survivability and APR weighting are all computed. ⚠️ It deletes most of 15_Ship & Army Points Design's class table (OI-394), and it forbids in advance any hull with a typed-in speed or a role with a declared cost tier.

Mass is not unit_size (OI-395). A barrel of foam and a barrel of lead occupy the same space at very different weights, and Canonical §12a already owns unit_size for Silo capacity. A separate mass property, and a consequence worth stating: range stops being a number and becomes a curve, empty to full, which fits shows consequences, never a score.

And rail cargo weight got a guard rail rather than a free pass. Level 1 ignores load, levels 2 and 3 do not — but ⚠️ Canonical §24 forbids a Simulation Level changing what is optimal, so: cargo weight may scale fuel consumption, and must never be large enough to change which carrier a player picks. OI-397 owns proving the numbers stay inside it.

8. New file — `05_Reference/14_Pitch Lines.md`¶

Sentences that emerged from the design session and are worth keeping for the Steam page, trailer and store copy. Separate from the design principles, which live in Canonical Design Facts.

9. Proposed open-items rows¶

IDs are provisional — confirm the next free ID before adding, since the register's rule is that IDs are stable and never reused.

Area	Kind	Item	Priority
docs	task	Apply amendment 1: lock world/session/account/planet in the Data Dictionary and sweep the vault	HIGH
docs	task	Apply amendment 2: split research into Factory (world) and Engineering (account) trees	HIGH
docs	task	Apply amendment 3: APR becomes a rate; delete the win-probability formula; close OI-202 to OI-206	HIGH
docs	task	Apply amendment 5: add CDF §16 to §22, extend §12b, correct the stale §13 SQLite row	HIGH
docs	task	Apply amendment 7: retire two Unwritten stubs, repoint links, update mkdocs nav	NORMAL
war	undecided	Set attrition tick rate and the supply-satisfaction curve	NORMAL
war	undecided	Set module knockout probabilities and how exposure is computed from a layout	NORMAL
war	undecided	Decide whether faction AI fleets use the player supply model or an abstraction	NORMAL
gate	undecided	Set gate throughput, latency and power draw per Mk, and the cost escalation for gates 2 and 3	NORMAL
gate	undecided	Decide whether the direct belt feed carries a throughput penalty or only a predictability cost	NORMAL
mothership	undecided	Define the mothership kit: parts, hull footprints, assembly rules	HIGH
mothership	undecided	Set umbrella radius and command range per class	NORMAL
docs	task	Rename control tether to command range across the vault, and add both radii to the Data Dictionary	NORMAL
mothership	undecided	Set transit and operating fuel rates	NORMAL
depot	undecided	Set depot storage, umbrella radius, and replacement cost and lead time	NORMAL
economy	undecided	Set credit yield per kill and the faction/personal split ratio	NORMAL
economy	undecided	Define the faction upgrade tree (capacity only)	NORMAL
ui	task	Design the mothership supply panel, retreat doctrine panel, and gate stability readout	HIGH
ui	task	Design the in-faction contribution board	NORMAL
outpost	task	Add a Forward Depot as a fourth destination for the outpost trickle, and state the trickle as a rate rather than a stock	HIGH
outpost	undecided	Decide whether the Seasonal contribution boost applies to a trickle pointed at a Forward Depot	NORMAL
outpost	task	DECIDED: trickle deliveries give the faction full material value but a reduced personal return (research points, spoils share, XP and reputation). Hauled, routed and belt-fed deliveries all pay full return so automation is never punished. Write it into `06_Outpost` and the economy document.	HIGH
economy	undecided	Set the reduced-return multiplier for absent players. Well above zero	NORMAL
power	undecided	Set the accumulator drain rate for holding a gate open with the world frozen, which decides how much battery buys an hour of overnight supply	HIGH
ui	task	Surface accumulator-backed gate runtime before logout: "your fleet has N minutes of supply after you leave"	HIGH
economy	undecided	Confirm whether contribution feeds research points at all; the vault's RP sources are Science Drives and Passive Field Data	NORMAL
scope	task	Phasing pass: assign every war-layer element to a phase and milestone	HIGH
editor	task	Apply amendment 7b: delete the APR build display from doc 15, resolve doc 30 §10, add the do-not-harmonise warning to doc 24, retire the last ship stub	HIGH
editor	undecided	Set slot counts per layer per class	NORMAL
editor	undecided	Define standard pressure: the reference attrition rate defensive power is quoted against. Must be one constant across all designs or the numbers stop being comparable	HIGH
docs	task	Add the measurement vocabulary (efficiency / power / coverage, never combined) to the Data Dictionary and sweep existing readouts for misuse	NORMAL
editor	undecided	Set mothership frame footprints and mass per Mk	NORMAL
editor	undecided	Define the exposure computation: faces or rays, and how armour thickness maps to protection	HIGH
editor	undecided	Decide whether a commissioned mothership can be refitted in place or only replaced	NORMAL
editor	undecided	Decide whether escort designs can be shared between players like factory blueprints	NORMAL
data	task	Set component mass values, which drive the whole transit-fuel relationship	HIGH
ui	task	Design the two editor screens, including armour coverage per face as the mothership headline readout	HIGH
economy	task	Apply amendment 7c: Engineering Points into `13_Player Levelling System`, flat catalogue into `08_Research Tree Design`	HIGH
economy	undecided	Set per-building construction costs	HIGH
economy	undecided	Set Engineering Point yield per level and the catalogue cost curve	NORMAL
economy	undecided	Decide whether ground items from deconstruct overflow decay or persist	NORMAL
scope	undecided	Check whether Phase 1's three sinks (research, defence, pollution) give enough tension without a war	HIGH
restoration	task	Specify the bad-builder generator: expansion step rules, severity-to-step-count mapping, and the cheap core validation	HIGH
restoration	task	Make the wreck generator strictly deterministic from seed. Seeded RNG, no float ordering, no hash-map iteration	HIGH
restoration	undecided	Decide the RECLAIM naming collision with Outpost Mode at storey scale	NORMAL
restoration	undecided	Define Planet Health weighting across levels, and the severity scale by distance from the starting system	NORMAL
docs	task	Repoint the ~16 inbound wiki links away from the three Unwritten redirect pages, then delete the pages and their `mkdocs.yml` nav entries. Needs a shell to find them all	NORMAL

10. Deliberately unresolved¶

Recorded so they are not mistaken for oversights.

07_Space Exploration Design remains unwritten. The war layer does not depend on it.
Phasing. The war layer as designed is larger than the current Scope document plans for. That was accepted deliberately: design first, phase afterwards. Row added above.
Central account service. Account-scoped data cannot be verified without one. Admission control at the war is the mitigation. A funded community server would replace it.

Revision History¶

Version	Date	Changes
3.4	2026-09-20	Pass Y recorded: the hull model, and ships stop landing. Editor doc to v0.7, space layer to v0.6. internal = (w−2)(d−2)(h−2), so ⚠️ nothing is protectable below height 3 — geometry replacing a class-depth table. Hulls named by size and shape, never role (Light / Medium / Heavy Lift / Heavy Line / Command I–III), with roles demoted to stock blueprints so a new role is data rather than a system; the Space Update needs three hulls. Heavy Lift against Heavy Line is the branch: same size, 20% internal against 31%, neither an upgrade. Parts have footprints, keeping the largest hulls placeable. El's fuel insight ends surface landings: Mass Driver up, Drop Pod Bay down, Space Elevator later, ⚠️ the lift is a building not a ship, and the import-cheap/export-expensive asymmetry makes refine before you lift a discovery rather than a lesson. Two rules recorded: nothing declared, everything derived (OI-394 for the doc 15 fallout) and mass is not `unit_size` (OI-395). Rail cargo weight takes a ⚠️ §24 guard rail (OI-397).
3.3	2026-09-15	Pass X recorded: the two gates, and what a conduit carries. Closes OI-391 and OI-392. The gate split is a phase boundary — a ship is world content, the mothership is how you show up — so ordinary hulls stay on world-scoped Shipyard Mk and motherships on account-scoped Engineering Research, which is the admission-control exception rather than a breach of levels never gate world content. ⚠️ `36_Progression Arc Design` §7.2's nothing else line amended and §7.3 added, and a side door closed: a mothership is commissioned in your own world (OI-393). Conduit throughput inherits the planet's rules — capacity is the weakest conduit, exceed it and split or upgrade — because unlimited throughput ⚠️ would make the largest hull the easiest to wire. Free consequence: a ship sheds tiers in flight like a base.
3.2	2026-09-13	Pass W recorded: one model, and everything is a hull. `34_Ship & Mothership Editor Design` to v0.5, closing OI-386. The layer/freeform split is struck: wires and pipes are an evaluator that is not combat, and ⚠️ the document already computed freeform exposure per face for motherships, contradicting its own escort argument. Every layer rule re-derived from geometry, so nothing was lost; the envelope gives design depth proportional to hull size. El standardised further than proposed — all hulls — collapsing two data shapes, two validity lists and the mothership frame tiers into one of each. ⚠️ Conduits occupy cells, which is what makes the envelope mean anything (throughput is OI-392), and the mothership's account-scoped unlock is flagged as OI-391 rather than inherited.
3.1	2026-09-13	Pass V recorded: the ship economy. `30_Ship Construction System` to v0.2 and `34_Ship & Mothership Editor Design` to v0.4. ⚠️ The Shipyard tier table had omitted the Scout, making the Launch release finale unbuildable. Blocks replace Mk for design versions; the salvage rule (discrete survives, continuous is destroyed) prices refit and decommission from one idea and closes the decom-and-rebuild exploit without an anti-exploit clause; the yard's chest-like store turns decommissioning into logistics rather than a refund; a destroyed ship leaves a wreck to be recovered, salvaged by a payload rather than a class. The Scout gains an identity, and comes back if you fuelled it to — El's correction of a rule I had put where a decision belonged. ⚠️ The escort layer model reopened as OI-386 rather than changed, since the editor document argues against freeform for a reason this session partly dissolved.
3.0	2026-09-11	Pass U recorded: interface and feel. All five interface documents brought up to date, closing OI-372 and OI-381. The two jobs stated as the test for whether a panel should exist. Primer and Moment added as tiers outside the urgency ladder. Genesis vocalises rather than speaks, keyed to tier and category so notifications stay cheap to add, spatialised from its position. Notifications have range: a remote planet without a satellite is silent, and that silence is the satellite's sales pitch. Structured events replace strings, because a cascading shed cannot be expressed in the old copy. A primer fires before the lesson costs you. Tutorial left alone on purpose, full revisit deferred to OI-383. Beats became achievements, which exposed and corrected the per-player production rule against Canonical §25, and the Pollution tab's stale water-in-pipes framing was fixed at the same time.
2.9	2026-09-11	Pass T recorded. Canonical to v2.0 with §24 (a Simulation Level never changes what is optimal, with wind as the worked failure) and §25 (the world is shared, nothing in it has a per-player owner, ruling out attribution mechanics before they are designed). `26_Pollution` to v0.3 with §3b, closing OI-089: air is a coarse grid, falloff rejected for storing nothing, the natural decay turns out to be a free budget that protects the first hour, wind ships at launch for everyone with its invisibility logged as a debt (OI-381), per-level behaviour folded into coefficients, and the weak haze split from the planning overlay. OI-358 and OI-365 both gained substrate.
2.8	2026-09-10	Pass S recorded. `41_Fluid Packaging & Carriage` to v0.2 on the ambient rule: what state a substance is in at ambient decides how it travels. Cryogenic Barrel and Insulated Barrel both deleted, OI-377 closed by construction, molten metal stops being cargo and Insulated Pipe becomes the reason smelting and casting are neighbours, the insulated tank wagon survives short-haul on the torpedo-car precedent, "Mk" refused for containers as it is a building concept, and §10.1 held the empties loop baseline, which produced Canonical §24.
2.7	2026-09-08	Pass R applied, fluid packaging. New `41_Fluid Packaging & Carriage Design` v0.1 closing OI-272, OI-278 and OI-297. Molten is still a liquid, so the constraint is the container, not the fluid, and the container ladder mirrors the pipe ladder. Gases go in cylinders, and no decompression step at a planet boundary. Rail carries fluid as fluid in a tank wagon, the only carrier that does. Barrels are a lossy loop, flammable barrels burn, the fluid drone is deleted. `19_Transport`, `07_Pipe` and `22_Trains` amended. OI-375 to OI-378 raised.
2.6	2026-09-07	Pass Q applied, three additions to Genesis. Bots are faster, Genesis never stops — a blackout stops the bot bay and demotes the player to one pair of hands, which is what makes First Blackout land. §5 confirmed and generalised: Genesis builds wherever bots cannot. A satellite is required to hear a planet you are not on, filling the hole `39_Space Layer` §4.3 left open; it gives you the panel, not the hands. OI-370 closed, OI-373 and OI-374 raised.
2.5	2026-09-07	Pass P applied, Genesis. New `40_Genesis Design` v0.1 closing OI-364, OI-217 and OI-135: one Genesis per player (indestructible kept, exactly 1 dropped), bots are the builders, nothing supervises the queue, Genesis founds and bots continue, upgrades split into research for verbs and kits and modules for capacity, and Genesis talks. `04_Factory` superseded, `08_Research Tree` v0.8 Branch 7 amended, `39_Space Layer` v0.4. OI-369 to OI-372 raised.
2.4	2026-09-06	Pass O applied. Power ships between planets in a Charge Cell, settling `25_Power System Design` §5.3; the Accumulator is too small to ship and stays untouched. You can ship the stock, you cannot ship the capacity. Signature means generation, not draw, confirmed, so importing power is how a player opts out of the ferals — the threat relocated and paid for, not removed. `19_Transport` §3.6 extended to a third container; `39_Space Layer Design` v0.3, `37_Planet Threats` v0.3, `26_Pollution` §6 confirmed. OI-367 raised.
2.3	2026-09-06	Pass N applied, the space layer. New `39_Space Layer Design` v0.1: the freighter as fourth carrier, the gate corrected out of the Space Update and into the War Update, and settlement priced rather than prescribed. `02_Scope` v0.10, `36_Progression Arc` v0.4, the Unwritten stub redirected, the Trade Hub filed as an idea. OI-313 closed; OI-364 (Genesis) to OI-366 raised.
2.2	2026-09-06	Pass M applied. OI-361 closed: Branch 6 amended so Reinforced Wall and Heavy Turret are Mk levels, the audit note defending them withdrawn, and the general rule recorded that no building is ever a research unlock at a higher tier. OI-362 closed: a rung buys rate of fire, range and spread, never damage, which raises consumption, couples the two tier tracks and changes turret placement.
2.1	2026-09-06	Pass L applied, weapons and ammunition. New `38_Weapons & Ammunition Design` v0.1 closing OI-360: defence is a production line, kinetic and energy families with the signature trap, buildings tier by kit and ammunition by research, one ladder shared with ship weapons. `24_Module System` warning reworded, `37_Planet Threats` repointed. OI-361 to OI-363 raised.
2.0	2026-09-05	Pass K applied. Phase boundary corrected: the seam is war, not the planet surface. Phase 1 is everything except war, Phase 2 is war; three releases across the two phases. `02_Scope` v0.9, `36_Progression Arc` v0.3, `04_Data Dictionary` v0.9, plus 33 `Phase:` fields repointed. OI-359 closed: nothing lives underground, but natives dig and ferals follow the shafts, so underground is late rather than quiet.
1.9	2026-09-05	Pass J applied, planet threats. New `37_Planet Threats & Defence Design` v0.1: two machine families, pollution draws the natives and power signature draws the ferals, damage is a scale and destruction leaves a wreck that remembers. `04_Factory`, `05_World Generation` and `26_Pollution` reconciled to it. OI-091 and OI-197 closed; OI-357 to OI-360 raised.
1.8	2026-09-05	Pass I applied. `00_Canonical Design Facts` v1.9 gains §23b, Phase 1 earns Phase 2, the tiebreak for contested scope. `34_Ship & Mothership Editor Design` v0.3 gains The First Ship Is a Freighter: the editor opens where there is no war, so it must be good at hauling before it is good at fighting.
1.7	2026-09-05	Pass H applied, the phasing pass. A `Phase:` field added to all 47 design documents and recorded as a convention in `04_Data Dictionary` §0a. Trains, farming and metallurgy to Phase 1; Outpost Mode off the phase axis; Gauntlet to the Space Update. `02_Scope` v0.8 reconciled with the documents it had drifted from. OI-356 raised on the session-scoped sweep.
1.6	2026-09-05	Pass G applied. `00_Canonical Design Facts` v1.8 gains §23a, count the content, don't target it; `02_Scope` v0.7 loses six content quantity targets and gains a warning box. Build constraints kept as numbers.
1.5	2026-09-05	Pass F applied. OI-355 closed: co-op moves into release Phase 1, war multiplayer stays in the War Update. `04_Data Dictionary` §0b added; `02_Scope` v0.6 corrected as the outlier; Risk Register, ROADMAP and Technical Architecture repointed; `36_Progression Arc Design` v0.2 gains §7a Beats in Co-op.
1.4	2026-09-05	Pass E applied, the phase pass. OI-350 closed: Phase 2 ships as the Space Update and the War Update (1.0). OI-354 closed, and it was a vocabulary fix rather than a renumber: `04_Data Dictionary` §0a added, `ROADMAP` headings disambiguated with milestone IDs protected, `06_Vertical Layers` rewritten, seven documents repointed. OI-355 raised on the multiplayer contradiction.
1.3	2026-09-05	Pass D applied, the progression arc. New `36_Progression Arc Design`; `02_Scope` collapsed to two phases with the launch as Phase 1's finale and Restoration as its long tail; doc 14 to v0.5 with territory confined to war instances; doc 13 to v0.3; doc 35 to v0.2. OI-346 to OI-354 raised.
1.2	2026-09-05	Pass C applied, checklist complete. `08_Research Tree Design` to v0.6: world-scoping corrected throughout, the two-tree split recorded, Engineering Research stated as flat. Amendment 2b partially withdrawn: infinite nodes stay in Factory Research, since every one of them raises a factory rate and moving them would breach Canonical §20.
1.1	2026-09-05	Pass B applied. Canonical Design Facts to v1.7 with §16 to §23 and the stale §13 flagged, doc 15 to v0.2 with APR as a rate and OI-202 to OI-206 closed, doc 13 to v0.2 with Engineering Points, doc 14 to v0.4 with the world model corrected. Only amendment 2, the research tree split, remains.
1.0	2026-09-05	Pass A applied. Data Dictionary vocabulary, doc 30 §10 resolution and drafts section, doc 24 do-not-harmonise warning, Unwritten stubs converted to redirects. One follow-up row for the link repoint.
0.9	2026-09-05	Added amendment 7d: `28_Restoration` reconciled to v0.2 with procedural wrecks in the player's own world, the generate-the-history mechanism, and the six generation-cost rules. OI-124 closes. Four register rows.
0.8	2026-09-05	Added amendment 7c for the Step 3 economy document, and the control tether to command range rename across amendment 1 and the register. Five register rows. OI-342 closes.
0.7	2026-09-04	Added the measurement vocabulary as amendment 1c: efficiency needs a denominator, power needs units, coverage is a fact, and the three are never combined. Armour efficiency forbidden. Two register rows.
0.6	2026-09-04	Added amendment 7b for the Step 2 editor document: doc 15 loses the APR build display, doc 30 §10 resolves two rows, doc 24 gains a do-not-harmonise warning, and the last ship stub retires. Eight register rows added. OI-344 and OI-345 close.
0.5	2026-09-03	Presence redefined as a property of the player rather than the shipment. Added the logout interaction: a frozen world stops powering the gate, so accumulators buy overnight fighting time, and the trickle is the only channel that bypasses the gate. Three register rows added.
0.4	2026-09-03	Trickle reward decided: full faction value, reduced personal return, scaled by presence rather than by automation tier. Added CDF §23 and renumbered 5h to 5j.
0.3	2026-09-03	Added a register row on whether trickle-only contribution ranks equally with active hauling.
0.2	2026-09-03	Added amendment 7a: the Outpost Mode trickle is a passive source of material, gains a Forward Depot destination, and is stated as a rate rather than a stock. Two register rows added.
0.1	2026-09-03	Initial. Nine amendments across seven existing documents, two stub retirements, one new reference file, nineteen proposed register rows.