Iteration 44 — Causal Dantzig M3 CLEAR INFORMATIVE · NOT REDUNDANT · M4/M5′ NEXT

In plain words: the regime-subtraction candidate survived the field test that killed its two predecessors. Turned loose on all 130 features, it did NOT go mute (the last candidate answered "nothing" for everything; this one's readings spread across the whole range — some features can get quite close to a regime-proof relationship, most cannot, a few are far off), and it is NOT an echo of an instrument we already own (its strongest resemblance to any certified reading is 76% — family resemblance, since both study steadiness, but far below our 95% "same instrument" bar). Two hurdles remain before any certificate, and they are the statistical ones: does plugging its readings into our verdict machinery actually improve verdicts on held-out data, and can it detect what it claims to detect at our sample sizes — both tested with the serial-dependence-aware designs this market demands.