12 asset-marked materials, UV mapping with a scalar Tiling node, images
linked by relative path to Assets/Materials/PBR, previews generated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2K-JPG Color/Roughness/NormalGL (+Metalness, AO where the set has them),
generated MetallicSmoothness.png (R metallic, A 1-roughness), texture and
material .meta files, SOURCES.md. Unity side not opened yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul s1-s2 hold (aetheria-release:verdict:cut-cultmath-0-3-pin.s2). The generated galaxy re-rolls with the noise kernel.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The tutorial-seed scan for a zone with no genuine Lagrange candidate ran
once per test; it now runs once behind a Lazy. The leak test regains
Assert.NotSame(galaxy.Entrance, otherZone), and the scan's miss path uses
Assert.Fail (xUnit2020).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Body facts RF1-RF8, model rows, booster and consumable tables, and rationale for the
revisions admitted under follow-up rulings-2026-10-09-spec-refresh. Sweeps the stacking
and hoarding lines the 2026-10-09 rulings made stale.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the target's fit written out the jinking price is .383 against .440
coasting, but 200 rolled shots gave 93 hits jinking against 78 coasting: the gap
between the prices is smaller than the noise of two 200-shot volleys. 2000 shots
put the expected gap near four standard deviations; the assertion is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
EvRangeLaunch placed stage.Generated("Djinni"), whose fit follows the galaxy; at
CultMath 0.3.0 the generated target gave 95 hits jinking against 78 coasting.
The target now carries an explicit Talaria/cockpit/reactor fit, so the hit
comparison does not depend on generated content.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
EAC-7089 and EAC-2733 were generated-content names that CultMath 0.3.0 re-rolled
away. The no-Lagrange-candidate zone is now found by its property (planets that
are all rosette members, so no genuine candidate) by scanning seeds, and the
zero-natural-station zone by its isTutorial: false roll. Every assertion keeps
its claim; the candidate predicate is shared with the leak test's fixture check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The noise kernel changed at cultmath-unity-v0.3.0, so the fixed seed gives a
different tutorial. TutorialGalaxy_unchanged keeps its claim (a fixed seed
gives a fixed tutorial) against the regenerated shape. Does not pass the whole
suite: the RestoredHullsTests and the evasion test follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Directory.Build.props, Packages/manifest.json and the lock move to
cultmath-unity-v0.3.0 (614fd445) together. VerifyCultLibRevision matched the
CultMath ProjectReference by a backslash-spelled identity and was skipped on
Linux; it now matches by project file name. Does not pass tests: generated
content re-rolls with the new noise kernel, fixtures follow in later commits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The backfilled Large Drives differ in quality, so a held forward thrust is a net torque and TurnRate never settled to zero. Release the movement demand with the turn demand.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
AetherDb roles-backfill apply on master's blob 32a2e401: 12 designs and 6 products authored; a second apply is a no-op.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
AetherDrive.cs: master deleted the behaviour (2cfdc728); the branch only added a
TurnRate line to it. Deletion kept. Thruster remains the one rotation actuator
and feeds Entity.TurnRate. No sim stepping outside the clock on master
(mines presenter, Pirates AI, lot-product all step through Zone.Update).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Large Drive and Small Drive shipped with unnamed Quality terms on their flat stats; the command keeps those exponents. The reset stripped them, so the new exponent check read the Thruster precedent instead.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A non-flat stat must carry exactly one Quality term, named, declared, with a positive exponent; a flat stat may not name a role. The check is one function with its own synthetic-stat test.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
ExponentFor drew its pool from items.Except(targets), so growing the targets from six to twelve changed the exponents a rerun on master's catalog writes. The pool is now snapshotted once, from every design that declares roles and names a Quality term, before the loop mutates anything. Describe prints each named term's exponent, one test pins Autocannon's and LRMM72's exponents against the shipped blob, and one clears a single role of a two-role seller and expects only that role re-authored.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
ShippedCatalogRolesAreLive checks every design of every kind: each non-flat stat reads one declared role,
declared roles are read, sellers author each once, role-mapped designs are not all flat (the tool answers
which, via Program.HasRoleMap). RolesBackfillIsAuthored pins ratios and roles, with the Autocannon
Penetration, Mine Range and LRMM72 LockAngle asserts. RolesBackfillCommandTests resets the backfilled designs
on a scratch catalog, applies, and expects the shipped catalog back, then replays as a no-op.
EveryShippedProductHasAMaker. TheEnvelopeComesFromTheGear bounds each drive by the catalog's Large Drive
Thrust range; ShippedWeaponOperabilityTests reads the catalog's DefaultEntitySettings.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
RoleMapFor answers for the Mine Launcher (MineLayerRoles) so roles-migrate and roles-backfill no longer
disagree on it; BackfillMapFor is deleted. Count, MinRange, DamageSpread and the reactor Modifier join their
kind maps, and the backfill points a varying stat's unnamed Quality term at its role without touching its
range. DeathCluster, Flak Gun, GT 3K, plight and both MoveOnPros join the backfill set (CB-R1 forces them).
RolesBackfill(apply, root) mirrors PiratesFaction so tests can replay it on a scratch catalog.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
DrawAhead leaves a ship in a wormhole animation where the step put it. The clock is
one static instance; a zone entered inside a step ends that frame's stepping and the
clock resets between frames (SimClock.Reset throws inside Advance). DrawAhead.Time had
no consumer.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
EmojiFont.Build leaves TMP's emoji fallback list empty. EmojiCheck fails when that list is non-empty,
when the global fallback lacks Emoji.asset, or when a glyph is served by a font other than Emoji.asset
or the body font; reads the Pirates product names from the shipped catalog; renders the
emoji-test.txt corpus and reports what Twemoji lacks; counts only used atlas pages (TMP doubles the
array capacity, so trailing slots are null by design).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
1400 single code points (the 2 and 3 digit ones included), Twemoji lacks none; font sha256
unchanged (reproduced in python:3.13-slim on Yggdrasil).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Each fitted Talaria has its own lot's thrust, so Left and Right are the two flanks'
accelerations, not both the larger one.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
ShippedCatalogRolesAreLive skips gear with no PerformanceStat at all (Tractor Beam is a
thruster kind with none). TheEnvelopeComesFromTheGear fits the Longinus by name: the
generator's picks follow the item stream and every authored ProductRole draws from it, so
the generated hull stopped carrying Large Drives (Medium Drive and RevvITup instead). Its
forward bound is the drives' own thrust, since a Large Drive now rolls 0.8x to 1.2x.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Probes B1-B9, the owner answer (published AetherDb), the r5 authority map
and rationale from the imagination-addon-frame pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Output of `AetherDb roles-backfill apply` on origin/master 325ef4bc's catalog, run on
Yggdrasil at 54999a64. Dry run: Autocannon, LRMM72, SRMM72, Mine Launcher, Large Drive
and Small Drive gain roles and quality ranges; Autocannon (Zhestokost, Pirates),
LRMM72 (R&D), SRMM72 (DME), Mine Launcher (DME) and True North (Lightsail) author their
ProductRoles. A replay prints "already authored". At merge, take master's catalog and
rerun `AetherDb roles-backfill apply`.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
AetherDb roles-backfill gives Autocannon, LRMM72, SRMM72, Mine Launcher, Large Drive
(and Small Drive, which declared roles no stat read) live roles: each mapped flat
stat becomes a 0.8x..1.2x range read through a Quality term naming its role, and
every seller authors a ProductRole per role from its maker profile. MineLayerRoles
is new; the maker profile table is hoisted so pd-gear and this command share it.
ShippedCatalogRolesAreLive and RolesBackfillIsAuthored read the shipped catalog.
The catalog blob lands in its own commit.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Body facts CC1-CC12, rules CS-R1..CS-R4 and rationale from the
imagination-consumables pass; specs and questions are in the mind.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The corpus now derives from emoji-test.txt (sha256 pinned), keeps single code points of any digit
count, and lists the emoji Twemoji lacks. The script runs only on linux, Python 3.13, image
python:3.13-slim and records those in the lock; identical pins with a different font exit 1 and
leave the committed font and lock untouched.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
