nx_bodydensity_gate.nx

nx_bodydensity_gate.nx -- THE TESSELLATION RATCHET: adding triangles to a surface that has no new detail in it must NOT be scorable as a form improvement. WHY THIS GATE EXISTS. A density ladder measured 2026-08-23 (knowledge/bodydensity.conf) found that across a 58x TRIANGLE RANGE (23,472 -> 1,365,120) on one fixed canon, silhouette IoU moved +5 permil, shape moved ZERO, and the DETAIL judge PEAKED at 88,992 and then DECLINED. The generator's analytic loft has no high-frequency surface mechanism, so there is nothing for extra triangles to carry -- proven separately by pushing relief to 1000 at 1,365,120 tris and getting detail 327, IDENTICAL to relief 0 at the same density. ★★★★★★ THIS GATE DOES NOT FORBID RAISING DENSITY, AND MUST NEVER BE READ THAT WAY. Density and detail CO-SCALE. The order is: build surface detail, THEN raise density to carry it, THEN re-measure. A triangle budget is a CONSEQUENCE of the surface, never a target chosen ahead of it. 88,992 is where THE CURRENT GENERATOR SATURATES -- it is not a ship target and not a cap. When a high-frequency surface mechanism ships, THIS GATE IS RE-DERIVED against the new generator and its saturation point is EXPECTED TO MOVE UP. A tooth that pinned 88,992 forever would be a capability cap wearing a gate's clothes. ⚠ RULER BLINDNESS, DECLARED: nx_bodybench sees HIGH-frequency local normal variance. It is BLIND to LOW-frequency radius perturbation (relief moves geometry -- different sha256, identical byte count -- with zero score change at any density). NOBODY SHOULD TUNE RELIEF AGAINST IT. And it scores agreement with a CADAVER oracle, so it is a DIAGNOSTIC for the beauty climb, never the target: past a point, closing that gap moves AWAY from beauty. Subject: ./nx_bodybench.elf over the shipped artifacts. e2e fork via nx_gatekit_lib, never a hand-rolled pipe (the banked deadlock: an undrained capture pipe wedges forever). license_tier: ORIGINAL expect_exit: 0 No hw writes (Rule 26).