✠ The Chronicle

✠ ENTRY XLVI

The loop's biggest single-day swing yet

an owner-directed rewrite repeals thirty-five accumulated design laws and replaces them with three pillars, THE PATH lands card upgrades and deck tiers on top of it, and a fresh reduced-nightly measurement shows the late-campaign win rate climbing from the pre-rewrite 4% floor to 28% — directional, not yet confirmation-grade. Per-item content work splits out of /forge into five standing adjust-* stewards. Breadth and deploy are green.

MECHANICS

THE BIG NUMBERS REWRITE — thirty-five laws repealed, three pillars stand in their place

What changed

A ground-up rewrite of the combat objective model. Bigger numbers become a design pillar (a starter hit is 6-9, a Saint-rank finisher 45-70+, uncapped payoffs reaching 100-300 in a fed deck). Only three constraints survive: the 5/5/5 aspect-thirds split, every card carrying a FREE line, and one tray roll per threat phase. Direct damage becomes a first-class verb competing on merit with status play and the alt-wins — nothing is protected, and there is no governing objective function any more (CQI, rank bands, count pins, and the status-engagement floor are all gone). Landed via feat/big-numbers-overhaul, PR #274 (handoff prompt + survey) and PR #276 (the engine + a new 112-card library, per-theme modules, and 157 rescaled enemy threat cards).

Why it changed

T's direction: "I want to see bigger numbers." Every prior governing curve had accreted into thirty-five overlapping laws that constrained design space without making any single play feel more consequential — the rewrite trades the curve for three pillars and lets everything else compete honestly.

commits 3a634f51 · 7b84c8bb · 82ac7bb4 · ed6b1be3 · 9caaa431 · 88559ce5 · 85dff127 · 7481a919 · eed · a · PR · 276 · 69c54c24

CONTENT

THE PATH lands — card upgrades, deck tiers, apocrypha, CONDEMN rescale

What changed

Models the progression axes the harness was previously blind to on top of the new engine: card upgrades, deck tiers, apocrypha cards, and a CONDEMN rescale.

Why it changed

The big-numbers engine needed a progression shape sitting on top of it before the matrix could measure anything representative of a full run rather than a single flat-power snapshot.

commits 5a326c55 · 515ac4d9

INFRA

Content lifecycle splits out of /forge into five standing stewards

What changed

Per-item content — cards, equipment, enemies, keywords, NPCs/dialogue — moves off /forge (now scoped to maps/continents/ events/art) onto five new adjust-* skills, each running a structural audit (reachability, duplication, domination, wiring honesty) with a mandatory kb-query research step before any create/update and an archive-never-silent-delete retirement path. plan/CONTENT_LEDGER.md tracks last-pass state per category, read by a new rate-limited /march gate (§3b: >=15 commits or >=36h per category, stalest first). No category has run a first pass yet.

Why it changed

T directly requested the loop be able to create new enemies, cards, and content generally, not just new world/map content — the single /forge verb had no per-item lifecycle (nothing ever retired a card or enemy), so five stewards replace it with create/update/ retire authority per surface.

commits 052c7a32 · (PR · 278

MECHANICS

Phase 77 (crash capture) and Phase 78 (W5 art sourcing) ship

What changed

Phase 77 adds in-house crash capture on mobile — global error/promise-rejection handlers plus a next-launch crash prompt, closing a six-week-old unpromoted owner ruling that two independent user crash reports had made urgent. Phase 78 researches open-source art candidates for the 9 W5 enemies (2+ CC/open-source options per enemy, license + provenance + fit rationale, all URLs live-verified) as a standing sourcing process — a research-and-present phase, nothing wired yet.

Why it changed

Phase 77: closes standing technical debt now backed by real crash reports. Phase 78: replaces ad hoc per-enemy art picks with a repeatable process, per T's instruction during the W3 oversight walkthrough.

commits 6249b19b · f391db7b · ae · 77 · 6abf8ea1 · 67be65e0 · 2afe17f1 · ae · 78

INFRA

combat-round-e2e widens to boss, withdraw, and non-starter-deck axes

What changed

Adds ENCOUNTER_KIND=boss and WITHDRAW=1 knobs to the crash-strict combat harness, plus a boss+non-starter-deck CI pass and a dedicated withdraw pass in both verify workflows.

Why it changed

Closes a plan/CRITIQUE.md [HIGH] row: 30 seeded runs never reproduced a reported mid-combat crash because CI only ever drove a standard foe, the mercy branch, and the fresh 4-card starter deck. This closes 3 of the 4 suggested-fix axes; the level-up-out-of-victory path is still untested and the original crash remains unreproduced — the row stays Pending.

commits 6529212a · adfd2761 · Ce · 277

BALANCE

Fresh measurement: late-campaign win rate climbs from a 4% floor to 28% post-THE PATH

What changed

The SessionStart hook flagged the standing baseline stale by one mechanics-source commit (THE PATH, 515ac4d9). Re-measured via the reduced-nightly pass (30 runs/cell, commit b826b2d4). Stage-level terminal outcomes across the 144-cell matrix: early 62.6% win / 6.4% defeat, mid 66.9% win / 16.2% defeat, late 28.1% win / 53.2% defeat, impossible 0% win / 91.7% defeat (as designed). The late figure is a marked jump from the 4%-win, 4-of-6-cells-at-0% cliff that /expand pass 11 cited from the pre-THE-PATH baseline earlier today.

Why it changed

THE PATH's progression axes (upgrades, deck tiers, apocrypha) were specifically meant to give a fed deck late-game reach; this is the first measurement taken after they landed.

Note: Reduced-nightly is directionally honest, not confirmation-grade — this is not a close of the late-campaign candidate in plan/PHASE_CANDIDATES.md, just evidence for whoever revisits it that the input data changed under it since it was filed. CQI is not reported here: THE BIG NUMBERS REWRITE repealed it as a governing score (b826b2d4 retires the last doc references), so this digest reports terminal outcomes only, per current doctrine.

commits b826b2d4 · baee · ee · + · dc · eee

✠ The marginalia

Measurement freshness

Was STALE at session start (baseline 5a326c55, one mechanics-source commit behind THE PATH). Re-measured this tick via npm run baseline:regen -- --runs=30 --confidence=reduced-nightly → now FRESH at commit b826b2d4. See the [balance] card above for the headline numbers.

The night's watch — what ran while nobody looked
when (UTC)verboutcome
2026-09-01 21:06–21:24march → iteratesuccess — combat coach copy rewritten off retired status-dominance doctrine
2026-09-02 00:19–00:33march → iteratesuccess — playtest.md rewritten off the retired ten-preset campaign
2026-09-02 04:22–09:38oversight (manual)W3 loop-call ratified; Phase 77 + Phase 78 promoted; baseline regenerated
2026-09-02 04:42–05:00march → iteratesuccess — defeat screen's dead vertical gap filled
2026-09-02 12:09consolidate (manual)consolidate: 2026-09-02
2026-09-02 13:13–22:36owner session (branch feat/big-numbers-overhaul)THE BIG NUMBERS REWRITE + THE PATH merged via PR #274, #276
2026-09-02 16:39–16:59march → ship-a-phasesuccess — Phase 77 (crash capture) shipped
2026-09-02 21:06–21:10marchsuccess, no-op
2026-09-03 00:19–00:32march → ship-a-phasesuccess — Phase 78 (W5 art-pass sourcing) shipped
2026-09-03 04:40–04:53march → expandsuccess — pass 11, 3 candidates filed
2026-09-03 08:22PR #278 merge (owner session)content lifecycle split — five adjust-* stewards
2026-09-03 11:22–11:47march → iteratesuccess — combat-round-e2e widened to boss/withdraw/non-starter axes
2026-09-03 12:57docs commit (manual)CQI retired from doctrine text

All march runs since the last digest (2026-09-01) succeeded; no crashes, no cancelled runs. The two largest changes this window (the big-numbers rewrite and the content-lifecycle split) shipped through owner-directed sessions on feature branches, not ordinary loop ticks.

The measure — did the fighting change?

None from this digest's own meta-loop rail this window — no starved queue or mistuned gate surfaced in the pulse. Separately (not this rail): /expand pass 11 filed three candidates today (lexicon doctrine registration, the late-campaign cliff, the fixed arena backdrop) — see Queues now. The late-campaign candidate's own input data has since moved (see the [balance] card); that reconciliation is for whoever picks the candidate up next, not this digest.

What is still unfinished
  • Build plan: 0 open phases ([ ]), 9 partial ([-]), 3 skipped, 124

shipped through Phase 78.

  • plan/AUDIT.md: 79 rows under "Pending" (~45 resolved inline, ~34

genuinely open, two of those marked RESOLVED but not yet swept).

  • plan/CRITIQUE.md: 52 rows under "Pending" (~28 resolved inline,

~24 genuinely open); last full pass still 2026-08-29 (pass 30), unchanged this window.

  • plan/PHASE_CANDIDATES.md: 52 rows under "Pending"; /expand pass

11 filed three new ones this window (lexicon doctrine registration, the late-campaign cliff, the fixed arena backdrop), all awaiting /oversight promotion.

  • plan/CONTENT_LEDGER.md: new this window, all five adjust-*

categories read "never" — no first pass has run yet.

  • No open triage:needs-user or loop:do issues.
  • Deploy: clean — HEAD green on verify-mechanics.
  • Nightly breadth (e2e:minigames): ALL PASS — hazard,

combat-encounter, encounter-routing, exploration/combat-prelude round trip, the upgradeable-dice flag-on stack, and the widened 10-run sandbox+live combat-round-e2e suite (now covering the new boss/withdraw/non-starter axes) all green, no crashes. Ran longer than usual this tick — THE BIG NUMBERS REWRITE's larger decks and boss-stage content.

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.