✠ The Chronicle

✠ ENTRY LXVI


A full Act 1 run now ends at level 20

Two days of work: the leveling curve was retargeted so a full Act 1 clear takes a pilgrim all the way to level 20, live combat moved off the exploration map and onto its own full-screen route, and a brand new puzzle mechanic for hazard encounters took its first steps behind a developer-only preview. Alongside that, several small fixes queued from the last entry landed, and the loop ran its first self-grading pass on how well its own checks actually catch mistakes.

BALANCE

Leveling now runs the full length of Act 1, to level 20

What changed

The experience curve was retargeted so a complete Act 1 clear ends at character level 20 instead of wherever it landed before. Each level costs 10% more XP than the last up to level 20 (15% dearer after that, to 40, then 20%), and every level grants one stat point. Leveling up no longer tops off your life total (VITAE) for free. Existing saves keep whatever level they're already at under the new math.

Why it changed

the old curve topped out short of where a full playthrough should land a character; this retargets the whole arc so Act 1 has room to use all 20 levels the game's own stat-point math expects.

commits 6d7cf28f · ae · 1

INFRA

Live combat moves onto its own screen; the old overlay plumbing is gone

What changed

Combat used to render as a modal layered on top of the exploration map the whole time a fight was on screen. It now gets its own dedicated full-screen route, the same way the hazard and rest screens already do — no back button or swipe-to-dismiss while a fight is in progress. The older system that used to host combat as an overlay (its chain bars, its tab-bar locking, and the sandbox screen it was first built in) has been deleted outright, along with a new automated test that plays a fight in the Labyrinth start to finish to prove the new route holds up.

Why it changed

the overlay approach was legacy plumbing from before the game's last major rebuild; moving combat to its own route removes an entire layer of indirection that every future combat change would otherwise have had to work around.

commits 0e4c6432 · ae · C1a · 118216b6 · ae · C1b

MECHANICS

A new hazard puzzle takes its first steps, behind a developer-only preview

What changed

A new mechanic for hazard encounters is taking shape: a fixed puzzle layout (a starting hand, a predetermined reserve, dice, a goal per route and a difficulty band) played once with no randomness, with an internal "solver" that proves each layout is actually winnable and sits in its intended difficulty band. Clearing one now has real stakes: XP, a loot roll, a full-completion bonus, and scaled life loss (with a lasting "scar") on a weak finish. Optional thresholds can unlock an extra die, an extra card, or turn blank cards into authored ones. Three hand-built example puzzles (easy, medium, hard) are playable right now from a HAZARD PUZZLES board in the developer menu, each against its own countdown clock.

Why it changed

this is new content under active construction, authored and checked piece by piece; it is not yet reachable from a normal playthrough and is being proven out in the developer menu first.

NEW · 17 OCT

The devpuzzle board screen after the change: A new mechanic for hazard encounters is taking shape.
devpuzzle-board — the new HAZARD PUZZLES board in the developer menu, showing the three demo puzzles and their clocks.

NEW · 17 OCT

The devpuzzle outcome screen after the change: A new mechanic for hazard encounters is taking shape.
devpuzzle-outcome — the outcome panel a puzzle run shows on completion.

commits 4e645f76 · b8460de7 · 93e9433a · 7daad066 · 24cec61c · 37e3712a · ae · B3aB3e

UI

The combat screen's phase clock gets easier to read

What changed

The small readout at the top of the fight screen that tracks which phase, round and turn you're on used to abbreviate "ROUND" and "TURN" down to single letters at a tiny size. It now spells out "PHASE · ROUND · TURN" in full, at a larger size.

Why it changed

this was a fresh-eyes finding from the project's own unattended testing pass (critique pass 88) — a readout a player actually needs mid-fight was too small and too cryptic to read at a glance.

BEFORE · 15 OCT

The combat encounter screen before the change: the small readout at the top of the fight screen that tracks which phase.

AFTER · 17 OCT

The combat encounter screen after the change: the small readout at the top of the fight screen that tracks which phase.
combat-encounter — before/after/diff of the fight screen's clock readout.

commit 0dc39188

UI

A refused "pay with life" is now announced out loud, not just shown

What changed

Some cards let you pay their cost with your own life total instead of a die. When the game refuses a payment that would have dropped you below 1 life, it already flashed the life-total ring red — now it also announces that refusal to a screen reader, the same as every other refusal in combat.

Why it changed

a visual-only refusal reads as nothing happening at all to anyone using a screen reader; this closes that gap for players who rely on one.

commit 6974aa46

MECHANICS

A hazard scar now mends at any rest, not only at an inn

What changed

A lingering injury ("scar") picked up from a bad hazard outcome used to only heal when resting at an inn specifically. It now mends at any rest — inn, camp, or otherwise.

Why it changed

a full manual playthrough of Act 1 (the "RC2 walk") flagged that the scar's own in-game wording never actually promised "only at an inn"; the fix brings the behavior in line with what the game already tells the player.

commit 3d953081

INFRA

Sentry learns to read performance and crashes on real devices

What changed

Preview builds (the test APKs built ahead of a real release) now report slow or frozen frames, app startup time, and a timing span for every card drag in combat, to the project's error-tracking service. Production and web builds send none of this. Crash reports from those same preview builds now also name which build they came from and decode their own stack traces automatically instead of showing raw, unreadable addresses.

Why it changed

this gives the team real performance and crash data from device-like builds ahead of time, rather than finding out about a slow screen or an unreadable crash report after a release.

commits 3686d84d · ae · F3 · f91ed7a0

INFRA

The loop runs its first self-grading pass

What changed

The project's automated build-and-ship loop built a set of tools to check how well its own safety nets actually work. One tool tabulates every automated pull request merged in the last several days against which checks ran and which were skipped. A second replays each of those pull requests' new tests against the code as it stood before the fix, to catch any test that would have passed even without the fix it was supposed to prove (4 of 58 test files currently would have). A third deliberately plants small bugs in the core fight-resolution code and measures how often the existing automated tests actually catch them (76.3% of the time, right now). A fourth ranks which fixes would have caught the most for the least effort, and sets concrete targets for the next round.

Why it changed

the team wants to know whether its own quality gates are actually catching mistakes before shipping, rather than assuming they are; this is the first real measurement of that.

commits 86eea99d · 3c0dc312 · 15a2fa76 · dd6f39c1 · f51fa926 · ae · E1aE1d

✠ The marginalia


The night's watch — what ran while nobody looked
window (UTC)what ranoutcome
10-14 18:48 – 10-15 04:41dispatcher ticksPF3 merged (#549); the HUD phase/round/turn readout fixed
10-15 05:55 – 11:05dispatcher ticksthe SACRIFICE-floor announcement and the any-rest scar mend both shipped and merged (#559, #562)
10-15 14:23 – 21:36a dense run of dispatcher ticksCR1 plotted and shipped both halves (#569, #573); LV1 plotted and shipped (#571); EV1 plotted; EV1a shipped (#565); CR1c skipped (no data to act on)
10-15 16:42critique pass 891 LOW finding (a tense mismatch in on-screen rest copy)
10-16 01:27 – 07:40dispatcher ticksEV1b and EV1c shipped (#576, #579)
10-16 02:48critique pass 901 LOW finding (a dev fixture's level-15 preset banks more VITAE than that level can reach)
10-16 08:19 – 13:32a dense run of dispatcher ticksB3 plotted and shipped all five phases (#581, #584, #587, #588, #592); EV1d shipped (#583); RV1 (a PR-review pipeline) plotted across three briefs; the dev menu's collapsible sections and the map's wheel-zoom fix merged (#590); the dev menu trimmed (debug-only presets and leaves removed, #594)
10-16 10:37expand pass 362 candidates filed (dev presets drifting off the new level curve; promote the isolated-worktree shell lesson to a reflex)
10-16 16:42critique pass 911 LOW finding (the map's on-screen gesture hint still says "drag · pinch" after the wheel-zoom fix, on desktop)
10-16 20:35expand pass 37no new candidates; two existing ones rescored up as repeat lessons kept recurring
10-16 20:38 – nowsix further dispatcher ticksno-ops — the open-PR review queue has been full (3 of 3) since 18:46 on the 16th
tonightnightly browser breadth leg (hazard, combat-encounter, encounter-routing, exploration-combat-roundtrip, combat-round — sandbox and live, five seeds each)all pass, no crashes
tonightmeasured-truth table re-readno new run since the last digest; last reading unchanged, 0 drift
tonightdeploy state checkclean — HEAD's changes are docs/plan-only, no gate pending
tonightpublic-site publish decisionpublished — combat's move to its own screen, plus this window's readability and accessibility fixes, are UI/QOL changes a player sees or feels
What is still unfinished
  • Leveling was retargeted: a full Act 1 clear now reaches level 20, with

one stat point per level and no free life-total refill on level-up.

  • Combat now opens on its own dedicated screen instead of layering over

the exploration map, and an older, overlay-based way of doing that has been removed entirely.

  • A new puzzle mechanic for hazard encounters took its first real steps —

still a developer-only preview, not yet part of a normal playthrough.

  • The fight screen's phase/round/turn readout is easier to read, and a

refused "pay with life" is now announced out loud as well as shown.

  • A hazard scar now heals at any rest, not only at an inn.
  • The loop ran its first self-check on how well its own safety nets

catch mistakes: right now its planted-bug test catches about 3 in 4 defects in the core fight code, and a small number of recent automated fixes turned out to have tests that would have passed anyway. Nothing was changed as a result yet — these are measurements, filed for the team to act on.

The measure — did the fighting change?
  • Highest-scored open proposal (6.0): a handful of developer-only test

presets and fixtures still bank more stat points, or carry starting gear, than the current rules actually allow since the level curve changed — proposed as one guard-test phase so future rule changes can't strand them silently again.

  • A recurring lesson (one awkward shell workaround needed inside an

isolated work area) has now repeated 11 times across five different loop verbs; a proposal (score 5.0) would promote it to a standing reflex read at the start of every tick, instead of being rediscovered by hand each time.

  • Two smaller proposals were rescored up this window as the same issues

kept recurring: a deploy-check script that reads as "hung" on a docs-only change (now bitten three times), and the same isolated-work shell lesson 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.