- THE CUT is judged on the PLATE now. It read cutter.last (any cut anywhere,
even on a piece never served) and kinds() stripped the piece suffix, so a
whole uncut tomato scored the same as a cut one — M2 was decorative in live
play. Sando.cutStats() now reports how much of the SERVED filling met the
cleaver; score = cutFrac x (0.6*quality + 0.4*evenness). Serving whole never
blocks a serve, it just zeroes THE CUT and caps near B.
Measured: same sando, cut pieces S(10.0) vs served whole B(7.39).
Fairness trap caught mid-build: evenness must be measured WITHIN a kind, not
across kinds (a katsu slice vs a tomato slice is the recipe, not a crime) —
that silently cost both weeks a grade until fixed.
- 無限 MUGEN: endless orders from a seeded Rng, layers 3->6, clock 50s->16s,
and the anger meter NEVER resets between orders — the only way a run ends is
the strings getting cut. No scene restart between orders. Judge escalates
samurai -> Shogun (5) -> the ONI (12, new MODELBEAST portrait set). Own
leaderboard board; run-over card posts the score. Title cycles 表/裏/無限.
- Spawner gives one whole food per two required pieces (one cut = two halves).
- setDay now explicitly resets the meter instead of relying on scene restart.
- Battery unchanged: 87f / 8.41u / 10-10 / cv0 clean / stack both / disaster
CUT / cleaver CUTS; weeks SSSSSSSS and AASSASSS; mugen escalation verified.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
dfff55fd1c