✠ The Chronicle

✠ ENTRY LXIV


The game starts checking its own rules and numbers

The automated builder that writes most of this game's code paused for about a day to rebuild how its own changes reach the game, then picked back up. Once it did, it built three new self-checks, all still waiting on the maintainer's review: a tool to measure how hard every early fight actually is, a checker that catches content nobody can reach, and the game's first written rulebook for what a card is allowed to do. None of the three are live yet, so there is nothing new to see in the game itself tonight — but each one already turned up something worth knowing, and those are below.

INFRA

A one-day pause rebuilds how changes reach the game

What changed

The automated development loop that ships most changes to this game paused for about a day. When it picked back up, changes no longer land directly: each one now opens as a reviewable, named change for the maintainer to approve, with at most three waiting for review at a time. The pause also lifted a standing restriction that had kept the automation from authoring new game content on its own.

Why it changed

the restriction and the direct-landing process both dated from the project's "revamp" — the period that rebuilt combat, the world and the enemy roster from scratch. With that period over, the team decided new work needed a human checkpoint before it reaches the game, rather than landing unreviewed. Nothing a player sees changes by itself; it changes how every future change gets there.

commit 47c0e79e

INFRA

A tool built to measure how hard every early fight actually is

What changed

A new measuring tool plays two different card decks — the plain starting deck and the blood-cost "Blood Price" set — through every one of the game's 26 Act One fights, at three different character builds each, and records how often each deck wins and how close to death each fight brings the player. It isn't switched on yet; that needs the maintainer's review first, then one manual run to produce the first reading.

Why it changed

since the combat rebuild, there's been no reliable way to say whether an early fight is too easy or too hard — judgment has rested on playing it and guessing. An early, unofficial test run already turned up something worth a look: the Blood Price deck won every single one of its test fights, while the plain deck's win rate swung from 20% to 95% depending on the build. That's a finding for the designer to weigh, not something the tool changed on its own.

commits cdba0b85 · (on · PR · 505 · aa · ee

INFRA

A checker built to catch game content nobody can reach

What changed

A new checker reads every enemy, item, relic, character and quest the game knows about and asks two questions of each: does it point at something that no longer exists, and can a player actually reach it through ordinary play? Its first run found zero broken references and one unreachable piece of content: a ring called the Bleeding Signet, which exists in the game's data but which nothing currently gives to a player outside the developer menu.

Why it changed

content that nobody can reach has always been invisible by definition — nothing was watching for it before, so it could sit unnoticed indefinitely. The Bleeding Signet finding is filed for the designer to decide whether to grant it somewhere or cut it; it is not fixed by this change.

commits aaaa8571 · (on · PR · 507 · aa · ee

INFRA

The game's first written rulebook for cards, checked by machine

What changed

A draft rulebook now exists for what a card is allowed to do — how complex a card can be for its rarity, a floor on how simple a "filler" card must stay, a template for the newer "Global" effect cards, and a rule for which keywords need a supporting mechanic. Every card in the game is now checked against it automatically; a future card that breaks a ratified rule fails that check before it can ship. Running it against today's cards found 18 existing Blood Price cards already over their rarity's complexity limit, and five keywords whose rule needs a decision — both filed for the designer's next card session, not fixed here.

Why it changed

until now, what a card was "allowed" to do lived only in the designer's head. Writing it down and checking it by machine means a future card can be caught before it ships, instead of after.

commits bd6fe0fc · (on · PR · 509 · aa · ee

✠ The marginalia


The night's watch — what ran while nobody looked
window (UTC)what ranoutcome
10-12 19:14 – 20:42T's attended session: phase R11 ships — ends "revamp mode," adds the loop:pr review flowshipped
10-12 21:10 – 21:18merge; bearings updated: the dispatcher and nightly digest re-enableddone
10-12 22:34 – 22:42critique pass 80no findings
10-12 22:48 – 22:55T's attended session: three phases queued (R12a measured-truth, RH1 roster-health, CS1 card-spec)queued
10-13 01:16 – 01:42dispatcher tick: phase R12a builtopened for review, PR #505
10-13 01:49T answers the 33-question card-spec ballotfiled for CS1
10-13 02:55 – 03:07dispatcher tick: phase RH1 builtopened for review, PR #507
10-13 04:44 – 05:16dispatcher tick: phase CS1 builtopened for review, PR #509
10-13 06:57 – 07:01dispatcher tickno-op — review queue already full (3 of 3)
10-13 08:48 – 08:52dispatcher tickno-op — same reason
10-13nightly browser breadth leg (10 full combat rounds, sandbox and live, five seeds each)all pass, no crashes
10-13deploy state checkclean — HEAD's changes are docs/plan-only, no gate pending
10-13public-site publish decisionheld — tonight's work is process and internal tooling only
What is still unfinished
  • The automated builder that writes most of this game's code paused for

about a day to rebuild how its own changes get reviewed — a human now approves up to three waiting changes at a time, instead of them landing automatically. Nothing you'd see in-game from this by itself.

  • Three new checking tools were built and are waiting on the maintainer's

review, none live yet:

  • one that will measure how hard every early fight actually is — its

first unofficial pass found a newer card deck winning every single test fight, worth a second look;

  • one that catches game content nobody can currently reach — it already

found a ring, the Bleeding Signet, that exists in the game's data but can't currently be earned;

  • and the game's first written rulebook for what a card is allowed to

do, now checked automatically from now on.

The measure — did the fighting change?
  • Still open from prior digests, unchanged this window: collapsing the

five-screen-deep "one centred column" layout pattern into one shared width cap and test (score 6.5), and running the root npm test suite in CI so a red guard on main can't go unseen again (score 6.0). Neither is new; neither moved this pass.

  • No new mistuned gate surfaced. The two no-op ticks tonight (06:57,

08:48) are the review cap working as designed — three PRs were already waiting — not a starved queue.

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.