Found during an outside review of the /queen game and the publishing chain (2026-09-09).
The facts
- The live t27.ai still builds from this monorepo.
publish-website.yml in ghashtag.github.io runs on a 15-minute cron and "replaces assets/ with a build from trinity" (see 5b8ccb28349edf0055631079f2ff2e901b804b8b). Nothing in that workflow has pointed at tri-27 since tri-27 was created on 2026-09-03.
- tri-27 claims to be the site. Its README says "the publisher in ghashtag.github.io needs a one-line change to which repository it checks out" — a plan that was never executed.
- The game code diverged immediately. tri-27's
apps/website has only the Sep-3 state of Queen.tsx/Queen.css (last touched cd1ad56). Everything the game became afterwards — QueenComb.tsx, queenHud.ts, the qa contract suite (queen-pick/placement/foundation/castle/windows/honey/bar/round/void/honesty/stale-contract.mjs) — exists only here in trinity.
- tri-27's own gates guard a stale copy.
check:queen in tri-27 validates a Queen.tsx that the site no longer serves, and check:api has printed SKIPPED since the split. The CI there watches a museum.
- The game docs acknowledge this without resolving it: "the local checkout is 62 commits behind" (docs/game studies, 2026-09-03).
Why it matters
- A gate that validates a copy nothing serves is a green lie of exactly the kind the repo's own honesty rules prohibit.
- Any fix to the live /queen page must be committed here — but a reader of tri-27 would reasonably commit it there and see it never ship.
- The 2026-08-21 incident (eleven blog posts silently dropped by the unattended publisher) showed what happens when the publishing chain has more moving parts than it has checks.
Options (owner decision)
- A. Finish the extraction: move the current game code into tri-27, point
publish-website.yml at tri-27, make trinity's apps/website a pointer/deletion. One publisher, one source of truth.
- B. Reverse it: tri-27 becomes docs/game + specs only (its game docs are ahead of trinity's), trinity stays the sole code source; tri-27's CI contract gates move here or are deleted.
- C. Keep both deliberately: then tri-27's README must stop claiming to be the site, and its
check:queen must be removed or re-pointed — as it stands it validates a page nobody is served.
Which way the owner picks is a decision; that there are two copies and one gate is a fact, and it is recorded here so it stops being ambient.
Related work today: stray file apps/website/src/#### The most powerful Verilog design en deleted from both copies (tri-27 00cfa10, trinity 6e5f1f5). Live supervisor outage (502 on /queen/status, /queen/public-board) tracked in tri-27#1 — needs a Railway restart, not a commit.
Found during an outside review of the /queen game and the publishing chain (2026-09-09).
The facts
publish-website.ymlin ghashtag.github.io runs on a 15-minute cron and "replaces assets/ with a build from trinity" (see 5b8ccb28349edf0055631079f2ff2e901b804b8b). Nothing in that workflow has pointed at tri-27 since tri-27 was created on 2026-09-03.apps/websitehas only the Sep-3 state ofQueen.tsx/Queen.css(last touched cd1ad56). Everything the game became afterwards —QueenComb.tsx,queenHud.ts, the qa contract suite (queen-pick/placement/foundation/castle/windows/honey/bar/round/void/honesty/stale-contract.mjs) — exists only here in trinity.check:queenin tri-27 validates aQueen.tsxthat the site no longer serves, andcheck:apihas printed SKIPPED since the split. The CI there watches a museum.Why it matters
Options (owner decision)
publish-website.ymlat tri-27, make trinity'sapps/websitea pointer/deletion. One publisher, one source of truth.check:queenmust be removed or re-pointed — as it stands it validates a page nobody is served.Which way the owner picks is a decision; that there are two copies and one gate is a fact, and it is recorded here so it stops being ambient.
Related work today: stray file
apps/website/src/#### The most powerful Verilog design endeleted from both copies (tri-27 00cfa10, trinity 6e5f1f5). Live supervisor outage (502 on /queen/status, /queen/public-board) tracked in tri-27#1 — needs a Railway restart, not a commit.