✠ The Chronicle

✠ ENTRY LIV


A story was written and cleared the same day

T's attended session wrote THE TALLY (a full narrative canon in Cormac McCarthy's voice), then hours later answered its own open deletion-scope question with the wide reading — THE BLANK PAGE wipes all story law, canon included. Two /oversight sessions drained seven phases (89-95) off the back of it, a real combat bug got fixed (automated witnesses were accepting mercy offers they should have fought through), a hands-on playtest found four bugs including one that blanks the screen for every returning player — already filed as issues #337-340, still unpicked — and 24 of 24 march ticks stayed green throughout.

CONTENT

THE TALLY, then THE BLANK PAGE — a narrative canon written and cleared in one session

What changed

An attended /world-spec session (T present, 2026-09-17 20:55) wrote THE TALLY: T's own instruction was "delete the story overview and write the most emotionally damaging story for this game using the voice of Cormac McCarthy instead of my story." The session found that no shipped line needed rewriting — the Herald's mercy-line, the Gate-Clerk's dead office, the river-court reading a boy as a ledger line were already telling this story; the new story-bible.md just named what was already on the board, and proposed W-02 (the Capital's payoff, Phase 84) on top of it. The commit flagged one open question: did T's "all mentions and extrapolations" mean a narrow deletion (just the overview) or a wide one (every dialogue node and retired spec)? Less than four hours later, an attended session answered it directly: THE BLANK PAGE (2026-09-18 00:31) — "remove ALL law about what the story is... an unidentifiable X as the first/main character" — deleted story-bible.md and W-02 too, the same day they were written. content/story/README.md now states the clearing outright: no arc, no premise, no canon, no name or figure for the player. Shipped narrative (~186 dialogue nodes) stays live as legacy text, explicitly not canon. plan/2026-09-17-story-outline.prompt.md is the lift condition: a new story-overview.md, built event by event, when someone runs that session.

Why it changed

T's own call, twice in one window — first to replace the story, then to remove the concept of a fixed story entirely. Not a loop decision either time; both are attended sessions with T's verbatim instruction quoted in the commit.

commits 5e14166b · E · A · 30201990 · 0102C01 · e · a38d6721 · c0a8d092 · (PR · 328 · E · BA · AE · 0ccb128f · 55d2b6c7 · efcaf1da · e · adff · prompt)

MECHANICS

A combat bug: automated witnesses were folding to mercy offers they should have fought through

What changed

Both the playtest matrix and the CLI's --auto mode unconditionally accepted any capitulation (RELENT) offer — no automated run could ever fight through one, even with the enemy at single-digit HP. This silently converted near-certain kills into mercy endings, inflating the matrix's mercy share for every policy except mercy-seeker/chaos, and let The Incompleteness's impossible-stage ceiling probe leak through a mercy exit 5 times out of 6 instead of holding. Fixed by giving every sim policy its own capitulationChoice: 'accept' | 'continue' stance — only mercy-seeker (mercy is its identity) and chaos (the noise floor) accept; every other witness now plays through the offer toward its own win condition, per the "alt-wins compete on merit, none protected" doctrine. Landed alongside FLURRY (Phase 90), a new multi-hit enemy keyword ("clause by clause"), first carrier Guild Knife.

Why it changed

the sim's job is to measure the game honestly; a policy that can't decline mercy can't tell you whether a fight is actually beatable head-on.

commits 87be9d91 · (PR · 330 · 39e2915e · ae · 90 · F

INFRA

Two /oversight sessions drain seven phases off the back of the story work

What changed

PR #324 refilled the phase queue and closed a stale ruling (axio-query now re-checks corpus freshness on every call instead of once). PR #325 — THE GROWTH FLOOR — gave the growth mandate its own tick budget and shipped small gaps directly: Phase 89 (die faces + HUD chrome get the art-direction-coherence pass), Phase 91 (amber-CI tick recovery — a deploy-URL comment floor), Phase 92 (a march job-ceiling decision: one phase per invocation, never chain), Phase 93 (card-base reconciliation — GameState.flags wired through initializeCombatEncounter, with a same-window follow-up fix stopping the dev combat sandbox from leaking real flags), Phase 94 (critique:drive's artifact wipe scoped to its own subpaths, not the whole tree), and Phase 95 (the combat primer now teaches the signature rune rail — one of the UI fresh-eyes six).

Why it changed

standing /oversight job whenever the phase queue needs a refill and there's scored, owner-ratified work waiting.

commits e20ac4b0 · (PR · 324 · d0b3ed2f · (PR · 325 · 6137ca5b · 3f50d663 · d7cc91d7 · e0107985 · e52efd89 · a438d344 · 2aa77567

CONTENT

Lifecycle rotation — five stewards, mostly zero-diff, one keyword backfill, one filed gap

What changed

adjust-keywords pass 13 backfilled UNMOVED, a turn-shape face word that had printed on six vigil/apocrypha cards with zero atlas row and zero mobile glossary entry since it shipped — structurally invisible to the card-face-honesty guard because that guard can only flag words it already recognizes. adjust-npcs pass 12 staged the Village Healer and did hygiene on the Coastal-Village unstaged-NPC backlog. adjust-cards pass 13 fixed a stale library-header comment (said 128 cards, the tree has 130 — EVENTIDE's two carriers were never counted in). adjust-equipment pass 13 found zero-CREATE/UPDATE/REMOVE on the data itself but filed a real gap: consumables have no cantrip-style secondary effect or HP-conditional scaling, unlike the corpus's usual anti-hoarding levers — needs a new Consumable field plus engine wiring, so it's a PHASE_CANDIDATES.md row, not a solo ship. adjust-enemies pass 13 re-derived every structural signal fresh and found Phase 90's FLURRY already fully wired end-to-end — nothing left for the steward to do.

Why it changed

standing rotation cadence; one real backfill plus one well-evidenced gap filing across five passes is the rotation working, not stalling.

commits 7bc53f82 · 900f4858 · 30e2e116 · 46942361 · 181a84e2

BALANCE

Baseline re-measured — the capitulation fix, not a regression, explains the late/impossible dip

What changed

Session start flagged the standing baseline (3ed75185, reduced-nightly, 2026-09-17) stale by 5 mechanics-source commits. Re-measured this tick at f79207c9 (30 runs/cell, reduced-nightly). Stage win rates: early 97.2% (was 97.3%), mid 95.4% (was 96.8%), late 74.2% (was 80.6%), impossible 12.5% (was 35.0%). The late/impossible drop traces cleanly to the capitulation-stance fix above, not a content regression: at impossible stage, greedy/blind went from 26-of-30-capitulate + 4 defeat to 0 capitulate + 30 defeat against The Incompleteness — exactly the "ceiling probe now holds instead of leaking through a mercy exit" the fix's own commit message predicted. chaos and control-lock (the two policies whose capitulation stance didn't change) hold steady at 67% and 23% respectively. CQI stayed flat to slightly up at every stage (early 0.65, mid 0.69, late 0.72, impossible 0.74) — the fights read the same, they just resolve through combat more often now instead of through mercy.

Why it changed

directional evidence only, per this digest's own reduced-pass caveat — the numbers moved because the measurement got more honest, not because anything got harder or easier to play.

commits f79207c9 · baee · ee · this · c

✠ The marginalia


The night's watch — what ran while nobody looked
when (UTC)verboutcome
09-17 10:30–10:39marchsuccess — critique pass 39, no findings
09-17 12:41–13:00march → adjust-enemiessuccess — pass 12, zero-diff
09-17 14:32–14:47marchsuccess — tooling fix, baseline:check ignores content/test-only commits
09-17 16:32–16:40marchsuccess, no-op
09-17 18:35–18:46marchsuccess — tooling fix, axio-query re-checks corpus freshness; oversight (PR #324) begins
09-17 20:30–20:55march / oversight (PR #324)success — Phase 89 shipped; THE TALLY (narrative canon) lands
09-17 22:27–22:53march / oversight (PR #325)success — Phase 90 (partial, FLURRY) shipped; S-01/S-02/C-01 purged; THE BLANK PAGE (PR #328) lands
09-18 01:02–01:16marchsuccess — Phase 91 shipped; combat-playtest session begins
09-18 02:38–02:49marchsuccess — critique pass 40, no findings
09-18 04:33–04:42marchsuccess — Phase 92 shipped
09-18 06:45–07:08marchsuccess — Phase 93 shipped, same-window CI leak fixed
09-18 08:36–08:58marchsuccess — Phase 94 shipped
09-18 10:30–10:44marchsuccess — Phase 95 shipped
09-18 12:39–12:44marchsuccess, no-op
09-18 14:30–14:35marchsuccess, no-op
09-18 16:30–17:10march → adjust-keywordssuccess — pass 13, UNMOVED backfilled; capitulation-stance fix (PR #330) merges
09-18 18:33–18:39marchsuccess, no-op
09-18 20:29–20:59march → adjust-npcssuccess — pass 12, staged Village Healer; bug-hunt (PR #336) merges
09-18 22:27–22:36marchsuccess — critique pass 41, 1 finding (1 low)
09-19 01:01–01:07marchsuccess, no-op
09-19 02:36–02:41marchsuccess, no-op
09-19 04:32–04:59march → adjust-cardssuccess — pass 13, header count fix
09-19 06:41–07:00march → adjust-equipmentsuccess — pass 13, zero-diff, filed consumable gap
09-19 08:33–08:51march → adjust-enemiessuccess — pass 13, zero-diff

24 of 24 march ticks succeeded this window, plus the two /oversight sessions and the combat-playtest session that ran alongside them. No /fix-ci dispatches were needed.

The measure — did the fighting change?

None from this digest's own meta-loop rail this window — 24 of 24 march ticks succeeded, the queue never ran dry (two /oversight sessions kept it fed ahead of the drain), and the nightly breadth check ran green. No gate/cadence/ceiling mistuning observed. The capitulation-stance fix above is a measured combat-sim bug fix already shipped through its own PR, not a proposal.

What is still unfinished
  • Build plan (plan/steps/01_build_plan.md): 0 open phases ([ ]), 1

blocked (Phase 84, needs an attended session — see Needs You), 140 shipped.

  • plan/AUDIT.md: 30 rows pending (headers under "Pending" minus

resolved-but-unmoved [x] bookkeeping rows).

  • plan/CRITIQUE.md: 22 rows pending; 1 standing [HIGH] — see Needs

You above.

  • plan/PHASE_CANDIDATES.md: 2 scored rows pending (consumable

anti-hoarding, summoner/multi-hit archetype).

  • 4 open GitHub issues from the bug-hunt playtest (#337-340, severity

high/high/med/low) — see Needs You above.

  • No open triage:needs-user or loop:do issues.
  • Deploy: clean — HEAD is a content-lifecycle tick, no gated workflow

triggered yet (grace window) as of session start.

  • Nightly breadth (e2e:minigames): GREEN — all legs (hazard, combat,

encounter-routing, exploration/combat roundtrip, upgradeable-dice, 10 sandbox+live combat-round runs) passed, no crashes.

  • Measurement: was STALE at session start (baseline 3ed75185, 5

mechanics-source commits behind); re-measured this tick via npm run baseline:regen -- --runs=30 --confidence=reduced-nightly → now FRESH at f79207c9. See the [balance] card above.

The nightly entry also carries a panel addressed to the maintainer — open questions, permission grants, things only he can rule on. It is held back from this page: it is correspondence, not a report. The unfinished work it tracks is published above, in plain words.