✠ The Chronicle

✠ ENTRY LXXI


Every run is about to begin with a choice of fighter

Character selection makes its first appearance: a run will soon open on a screen where you pick who you are, though no one is on the roster yet to pick. Equipment gets a real third slot built just for relics instead of one vague catch-all, and a relic now tells you outright which lane of cards it unlocks while you wear it. A region's elite enemy can come back after you rest there — same fight, same reward, until its one relic has dropped, then it's gone for good. Every card's old "(max +n)" print is gone from the two live families; a cap only appears from now on if a designer calls for one by name. And the card-vote ballots stopped pitching a colour for new cards at all — the designer picks the cards first, and colour gets split evenly afterward. Underneath all of that, the project's own testing tools got sharper: the fresh-eyes playtester now drives its own browser instead of borrowed tools, a full lap of Act 1 is proven node-by-node by the engine itself, and the iOS beta pipeline can now submit a build to TestFlight on its own.

MECHANICS

Character selection makes its first appearance

What changed

the engine now has a real idea of "a character" — a portrait, a starting relic, two starting card lanes, three starting lane cards and a base stat split — and a new run built from one wears its relic and deals its cards automatically (phase CH1a). The card-reward pool during a run is now the character's own two starting lanes plus whatever relics' lanes you currently have worn, rather than just the worn relics alone (CH1b). A fighting-game-style select screen exists now too: a carousel of square portraits, locked or unlocked (CH1c). And the project's own measured-truth report gained a "reward lanes by character" section, so a look at the numbers can show exactly what cards a given character would see (CH1d). Nobody is on the roster yet — the character list is still empty, and the two phases that put a real person there (T authoring the first one, then wiring it into new games) are waiting on a third card family existing to choose from.

Why it changed

T's ruling at the character-selection pitch (D108): "each run should open with an identity, and late-run lane control should differ from the start." This is the first piece of a from-the-ground-up answer to "who am I playing," not a cosmetic skin on the run that exists today.

commits 4a9b16eb · d24cd6cd · bf9f07ec · 931bfd3d

MECHANICS

Relics get a real home, and wearing one promises its cards

What changed

equipment now has three honest slots — one Weapon, one Armor, and three for Relics — replacing the old single "Accessory" slot that relics, trinkets and rings all had to share (EQ2a); saves made before this change carry over automatically. Every screen that used to say "Trinket" or "Accessory" now says "Weapon," "Armor" or "Relic" (EQ2b). The Bleeding Signet's own detail view now names its lane outright ("Lane: Blood Price, while worn"), and every other lane relic says the same about its own lane (LR1a). Elite and quest fights can now actually hand you one of those relics as a drop (an 8% chance on an elite win), instead of the one starting ring being the only way to ever see one (LR1c). The fight log now records which relics and signatures you had worn into a fight, feeding a new signature-usage report, and a signature refused for a reason the game can explain now logs that reason too (LR1g). Riding along: a long-standing bug meant every fresh run started with a ring already worn for you; now a fresh run starts with none, and any relic the game hands you afterward wears itself automatically if a slot is free (SG1).

Why it changed

a card reward has drawn from your worn relics' lanes since 10-25's reward-rule change, but until this window a relic was just one undifferentiated slot that never told you what wearing it would do, and the only relic in the game was a hardcoded freebie. This window gives relics a real place to live, a way to actually earn more of them, and the plain-language promise ("Lane: X, while worn") that tells you why putting one on matters.

commits f0c8e8b4 · cce4cf34 · 2553b458 · 01f02f53 · e6e20d82 · c74ca526 · ca102a2b · e7a1033c

MECHANICS

A region's elite can fight you again after you rest

What changed

resting at a region's rest spot now brings its elite enemy back for another round — same foe, same level, same full reward (XP and your pick of cards, or an upgrade). Each time you beat it, it rolls its own chance (8%) to drop its one relic; once that relic has dropped, that elite is retired for good and stops coming back. Old saves keep working unchanged, no conversion needed.

Why it changed

T's own task for this phase: "elite enemies respawn after a rest node is triggered." It turns a region's elite from a one-and-done fight into a real, if capped, source of repeatable XP and card picks — the brief that scoped it was explicit that this is meant as a farm, with the relic side kept finite on purpose.

commit db22fcc4

BALANCE

No more printed caps on cards, and ballots stop pitching colour

What changed

every "(max +n)" that used to print on a Blood Price or Elementalist card's upgrade line is gone — a card only carries a cap now if a designer calls for one by name (the two existing SACRIFICE caps stayed, by T's own call). One line whose only way of growing on upgrade was its cap (Repaid in Kind) now grows by a flat amount instead, and its paid effect became twice the damage you've taken. Separately, the card-voting ballot tool dropped its colour pitch entirely: it no longer proposes or tints a job colour for a candidate card, and /lane-creator now proposes splitting a new family's colours evenly only after the cards themselves are picked, for the designer to adjust by hand.

Why it changed

two blanket rulings from T (2026-10-26). On caps (D109): "remove the '(max +n)' from every card. If there should be a max, I'll call it out." On colour (D110): "a card's colour is ONLY what die pays for them and is not categorized otherwise... I pick the cards, then afterwards, it can split the colors up evenly." Both are rules set once, in T's own words, rather than re-litigated card by card or vote by vote.

BEFORE

The bp m 02b card before the change: every

AFTER · 27 OCT

The bp m 02b card after the change: every
bp-m-02b — reprinted; card bp-m-10a — reprinted; card bp-p-06d — reprinted; card bp-p-07b — reprinted; card el-p-06b — reprinted (5 of the affected cards; the rest changed the same way)

commits 22f70be0 · b2bd7bb5

UI

The board and the reward screens read more plainly

What changed

the combat board now spells out more of what it's doing: your own INFERNO reads as yours with its per-turn loss shown, an incoming hit that BLIZZARD will cut says so on its intent line, an all-MISS turn's dead tray still shows END following through, and every number that floats off a hit now has its own readable timing instead of arriving in a blur. Elsewhere: the reward cache's "TAKE A CARD" offer now shows the actual card face before you commit to it, instead of asking you to pick blind; grabbing spoils takes two taps instead of several, and a map node's gold total shows right on its panel; and a tap-to-open glossary now explains Conviction, dice, piles and the phase ledger right on the board, reachable by a sighted player without leaving the screen.

Why it changed

each of these closes a small but real "wait, what just happened" gap — a hidden INFERNO tick, a blind card pick, a buried gold total, game terms with nowhere to look them up. None of them change what a fight or a reward does; they change whether the screen is honest about it.

commits 17459ddf · 4b512e40 · 8e6aa246 · 76af5f19

INFRA

The project's own testing tools get sharper

What changed

the fresh-eyes playtester agent now drives its own dedicated browser through a shared session tool instead of borrowed Playwright tools, so the same four-way swarm it already ran stays intact on a cleaner foundation (PH1a, PH1b). The written playtest guides for hazards, combat and the minigames are now checked by a test against the actual ids and modules they describe, so a guide drifting from the real game gets caught instead of silently going stale (PH1c). A new script walks a complete Act 1 crossing — all four regions, door to exit — and proves every kind of event the region promises actually gets hit along the way, catching both code defects and filing content ones it can't fix itself (AW1a). And a companion test checks that every map node's own blurb actually matches the event that node triggers (AW1b).

Why it changed

each of these is the project checking that its own checks are trustworthy — a playtest guide that quietly drifted from the game, or a map node whose blurb promised the wrong fight, are exactly the kind of gap that would otherwise only surface when a player hit it first.

commits 8e9717b3 · d795b3b4 · 698884e0 · 087b8b0b · 07b52130

INFRA

The iOS beta pipeline can submit itself

What changed

the app now has real iOS build configuration and a written runbook for enrolling, building and shipping a TestFlight beta (IO1a); the preview-build workflow can now take a "submit" input that, only for an iOS beta build, writes the App Store Connect API key to a one-time file and runs the build with auto-submit turned on, so a beta build can reach testers' phones without a manual App Store Connect upload (IO1c).

Why it changed

gets the project from "a build exists" to "a build reaches a tester's phone" without a manual step in between, ahead of the iPhone enrollment decision still waiting on T (plot PR #803).

commits 6428ad76 · 74b39661

INFRA

Measured truth adds the Elementalist's own numbers

What changed

the nightly measured-truth report now samples two Elementalist variants directly — a 20-of-30 draw from its own cards, and a 20-of-60 draw mixed with other families — rather than only ever reporting its fixed 20-card test deck, and every cell now also reports how much VITAE a deck lost to its own INFERNO ticks (self-burn), separate from damage taken from the enemy.

Why it changed

the Elementalist is the first lane whose own cards can hurt the player who's playing it; without a self-burn column, a death from your own fire and a death from the enemy's hit would read identically in the numbers.

commit 768dc7bf

✠ The marginalia


Measured truth

> measured at ef477c5ac, 2026-10-27, 300 runs, seed 1; compared with b1197e6 (2026-10-27)

Doors, win rate / lowest-VITAE p10:

fightlevelgrey evengrey body-mindgrey bodyblood-price evenblood-price body-mindblood-price bodyelementalist evenelementalist body-mindelementalist bodyblood-price+half evenblood-price+half body-mindblood-price+half body
bw-17/death-beholder (door)L4 v L121% / 0.0027% / 0.0068% / 0.00100% / 0.50100% / 0.53100% / 0.54100% / 0.89100% / 0.89100% / 0.89100% / 0.57100% / 0.64100% / 0.65
cw-17/death-beholder (door)L9 v L250% / 0.0083% / 0.0099% / 0.37100% / 0.58100% / 0.62100% / 0.65100% / 0.90100% / 0.90100% / 0.89100% / 0.64100% / 0.71100% / 0.72
bc-15/death-beholder (door)L14 v L386% / 0.0095% / 0.23100% / 0.55100% / 0.63100% / 0.70100% / 0.71100% / 0.90100% / 0.91100% / 0.90100% / 0.69100% / 0.76100% / 0.77
ld-16/death-beholder (door)L18 v L482% / 0.00100% / 0.29100% / 0.57100% / 0.65100% / 0.73100% / 0.72100% / 0.89100% / 0.90100% / 0.89100% / 0.72100% / 0.78100% / 0.78

No drift past the thresholds.

The night's watch — what ran while nobody looked
windowwhat ranoutcome
10-25 daytime–eveningdispatcher ticksthe Elementalist and its two sampled families joined measured truth, with a self-burn column (EM1); a full Act 1 walk test was built and proved every region's event kinds are reachable (AW1a)
10-25 nightdispatcher ticksthe board started naming your own INFERNO, the BLIZZARD cut and float timing (BC1); the long-standing no-starting-ring bug was fixed (SG1)
10-26 early morningdispatcher ticksthe playtest session tool shipped, then the playtester moved onto it (PH1a, PH1b); the playtest guides started being checked against the real tree (PH1c)
10-26 morning–middaydispatcher ticksiOS build config and the TestFlight runbook shipped (IO1a); the cache started showing its card before the pick (RW1b); the iOS beta pipeline gained auto-submit (IO1c); the spoils pick dropped to two taps (TQ1)
10-26 evening, T attendedT, attended + dispatcher ticksT ruled no printed caps on any card (D109) and that a card's colour is only which die pays for it (D110); node blurbs started being checked against their events (AW1b); both rulings shipped the same evening (NC1, CL1)
10-26 evening–nightdispatcher tickssignature tracking landed in the play log (LR1g); the board glossary shipped (GL1); the relic slot replaced the old Accessory slot (EQ2a); the Signet started naming its lane (LR1a); the Defender's full lane-creator session (pitch, 90-card ballot, verdicts, sims, T's numbers call, revisions) finished and was plotted as 6 new phases (DF1), the first to need the third-lane slot that character selection (CH1e, CH1f) was waiting on
10-26 night–10-27 early morningdispatcher ticksthe character engine and save landed (CH1a); every screen started saying Weapon/Armor/Relic (EQ2b); the character-aware reward pool shipped (CH1b); the select screen shipped (CH1c); the measured report's reward-lanes-by-character section shipped (CH1d)
10-27 early morningdispatcher tickselites started coming back after a rest (ER1); quest and elite fights started actually dropping lane relics (LR1c)
tonightunattended /critique passes 101-10322 captures each pass, 0 nav trouble, 0 console errors; one LOW a11y finding (pass 101, a WITHDRAW sub-line's contrast) and one LOW process finding (pass 103: RW1b's only player-facing screen, /cache, isn't in critique's 11-screen set, so the change shipped unseen by any pass)
tonight/expanddid not fire — gated on no phase being pickable, and this window never ran the queue dry; still at pass 40, unchanged since the last digest
tonightnightly browser breadth leg (e2e:minigames)all pass, no crashes — 5 live combat-round seeds played to the round cap
tonightmeasured-truth table re-readone new run since the last reading (CH1d's, carrying the new reward-lanes section); no drift past threshold
tonightdeploy state checkclean — CI green on HEAD c674974 (verify-mechanics, verify-mobile both succeeded)
tonightpublic-site publish decisionpublished — a major release (character selection's first screen, elites that return after a rest) and a run of UI/QOL changes (the cache's card reveal, the two-tap spoils pick, the board glossary, the board's own-INFERNO and BLIZZARD cues)
What is still unfinished
  • A character-select screen exists for the first time, though no one is

on the roster to choose yet — that waits on a third card family (the Defender, now in design) to give a real choice between lanes.

  • Equipment now has a dedicated Weapon, Armor and three-Relic slot layout

instead of one shared catch-all, and a worn relic now says outright which lane of cards it unlocks.

  • A region's elite enemy can now be fought again after a rest at that

region, for the same reward, until its one relic has dropped — then it's gone for good.

  • Every printed "(max +n)" cap is gone from the game's two live card

families; a cap only shows up from now on if the designer calls for one specifically.

  • The card-vote ballots no longer pitch a colour for new cards at all —

colours get split evenly by hand after the cards themselves are chosen.

  • A third card family, the Defender, finished its full design

conversation (a ballot, verdicts, number sims, revisions) this window and is now queued to be built.

  • The combat board explains more of what's happening on it (your own

burning, an incoming hit about to be reduced, a dead turn's resolution), and a tap-to-open glossary now explains the game's own terms.

The measure — did the fighting change?

None this pass. The pulse was the heaviest yet — 346 commits and 136 merged PRs in about two days, zero stuck or left-open PRs, and three scheduled /critique passes catching only two LOW findings between them. /expand correctly stayed quiet the whole window: its gate only fires when nothing is pickable, and T kept the build-plan queue full by /plot-ing new work (DF1, NC1, CL1, CH1, ER1 among them) faster than the loop could run it dry. Nothing in the pulse suggests a gate, cadence or ceiling is out of tune.

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.