guts/web
type-two 975e1a4ec3 [lane B+C] The combo meter starts paying, allies stop counting as kills, and par stops
meaning "miss nothing"

Audit item 13 — the systems the levels were authored against, none of which worked.

THE COMBO MULTIPLIER DID NOT MULTIPLY. balance.js's own comment cites the GDD line "score x style
multiplier"; what shipped was a 2.5 s window and a counter. The HUD printed "x7 chain",
comms shouted about it at 5, the music added a layer at 3 — and combat/index.js added the flat
base score every time. Every par.score in every level was authored against a lever that did not
exist, which is exactly why they all sat at 75-94% of the un-multiplied ceiling. Now step 0.12,
cap 2.5 at a 14-chain: reachable in L4's ileum and nowhere else, because a cap you hit every
fight is not a skill expression. Verified in the running game: six chained floaters pay
100/112/124/136/148/160 where they used to pay 600 flat.

ALLIES COUNTED AS KILLS. combat/index.js did an unconditional score.kills++ on enemy:die, and
the ally branch of enemies.js emits enemy:die with score 0 — so shooting a lactobacillus reef,
the one act this game's whole reputation loop exists to punish, ticked your PATHOGENS tally up
by four. The pause card and the operative record told you that you were doing well at the exact
moment the biome decided otherwise. enemy:die carries `foe` now. Verified: seven deaths, six
kills.

THE PAR CHECK WAS COUNTING A THIRD OF THE GAME. scoreBudget() summed spawn events and nothing
else — not the 500 points a biopsy sample is worth, not the several thousand a boss puts in the
room. Four of six levels therefore shipped a permanent "par.score exceeds the kill budget"
warning, and a warning that is wrong two thirds of the time is one everybody learns to scroll
past. It counts pickups and boss rings now (nodes once per phase — a floor, since a phase
re-rings for as long as the fight lasts and no honest ceiling exists). Six levels, zero
warnings, and the ratios that were 94/75/52/59/44/32% of a fictional budget are now real.

AND THE CHECK THAT SHOULD HAVE CAUGHT L1. The only par test asked whether par exceeded the
ceiling. L1 sat at 94% of it: par the tutorial and you had to kill all eight foes, take all
seven orbs and all three samples — miss ONE sample and the medal was gone. That is a
clear-sheet requirement wearing par's clothes, in the level that exists to teach. par.score is
1900 now (71%), and the validator warns above 85% so the next one gets caught by a machine.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-26 22:19:34 +10:00
..
assets [lane D] The last two pathogens: the seeker and the turret get meshes 2026-07-26 19:39:19 +10:00
dev [lane A] MODELBEAST dresses the molecules: 5 matcaps, and the canal picks space-filling 2026-07-16 21:31:24 +10:00
js [lane B+C] The combo meter starts paying, allies stop counting as kills, and par stops 2026-07-26 22:19:34 +10:00
vendor [lane F] Round 0 scaffold: docs, contracts, lane charters, bootable stub world 2026-07-16 01:28:35 +10:00
index.html [lane F] Round 0 scaffold: docs, contracts, lane charters, bootable stub world 2026-07-16 01:28:35 +10:00