Andrew Fribush

fig. 2.17.1: twelve real regressions put back one at a time: 8 caught on the first run, 10 after retargeting, 11 on the current run

fig. 2.17.1 · twelve real regressions put back one at a time: 8 caught on the first run, 10 after retargeting, 11 on the current run.
MUTATION COVERAGE: 11 OF 12 CAUGHT, THE ONE MISS AN EXPECTED REDUNDANTCLAMP; EARLIER RUNS 8 OF 12, THEN 10 OF 12ONE ROW PER REGRESSION; LIT CELL: A CHECK FAILED, HOLLOW: ALL PASSED. THE 1 STILLHOLLOW IN THE CURRENT RUN IS EXPLAINED IN ITS ROW.FIRST RUN · 8 OF 12 CAUGHTAFTER RETARGETING · 10 OF 12 CAUGHTCURRENT RUN · 11 OF 12 CAUGHTSUPPLY RESPONSE REMOVEDARBITRAGE OVERSHOOTSMISSED, THEN RETARGETED: THE RE-CHECK ALONE MOVED VOLATILITY 26% TO 20%;THE DAMPING ON MOVE SIZE HOLDS THE MARKETWELFARE IGNORES PRICEEXTRACTION IS FREEOVERRIDE COSTS NOTHINGCENSUS DOES NOT CORRECT BELIEFEXPLAINED, THEN CAUGHT ON THE CURRENT RUN: A SECOND PATH CORRECTS BELIEF;ERROR MOVED 7.9% TO 12.4%, THRESHOLD RESET 0.60 TO 0.34NO LOCAL PRODUCTION FLOORTREASURY MAY GO NEGATIVETHE ONE MISS NOW, AN EXPECTED ONE: APPROPRIATIONS ARE ALREADY SCALED TOTHE TREASURY; THE CLAMP IS REDUNDANTLEGIBILITY SATURATESINSTRUMENT COSTS DO NOT SCALEMISSED, THEN RETARGETED: UNSCALING MOVED TREASURY/REVENUE 1.9X TO 2.0X;THE BINDING LINK LIMITS SPENDING FIRSTRAILWAY DOES NOT EXTEND THE NETWORKPOPULATION NEVER GROWS

source: vicky game/mutants.py lines 17-78 · docs/BUILD-LOG.md lines 37, 1017-1018 (working tree over HEAD 93c9f5e; 13, 993-994 at HEAD) · current run: python -m game.mutants on a copy of the working tree at HEAD 93c9f5e plus uncommitted edits, 2026-10-07 and again 2026-10-13, 11 of 12 caught · mutantGrid.data.ts

View in context: Testing assertions by reintroducing bugs → · All figures

Andrew Fribush, software and computer engineer in Austin. I build consequential systems where software meets institutions: finance, healthcare, commerce, government data, and the mechanisms that govern AI agents. Taking new engagements: hello@andrewfribush.com.