nx_skinspectra_gate.nx

nx_skinspectra_gate.nx -- THE GATE FOR THE MEASURED SPECTRA, 2026-09-03 (rung ST6). SUBJECT: ss_melanin_mua / ss_melanin_mua_at / ss_hb_mua / ss_hb_lookup in-process, against the SHA-PINNED mirrors in knowledge/fetched. THE STRONGEST TOOTH IS T1 AND IT IS PREFACTOR-FREE. The published melanin law has a prefactor the source itself calls tenfold-variable, so any tooth that depends on it is testing a number nobody can pin. The RATIO of two wavelengths cancels the prefactor entirely and leaves only the exponent: mua(450)/mua(600) must equal (600/450)^3.48 whatever the concentration. That is the part of the law that IS pinned, and it is the part this gate checks. MEASURED: 2721365 against 2721365 micro, exact. T5 AND T7 ARE THE ANTI-STUB PAIR. A lookup that returned a fixed row would pass "the value is non-zero" forever. T5 demands two DIFFERENT wavelengths give DIFFERENT values, and T7 demands a wavelength the table does not contain be REFUSED rather than answered with a neighbour -- handing back the nearest row would be inventing a measurement and calling it data. T7 CARRIES A CORRECTION WORTH READING. On its first run, on the laptop where the mirror is absent, T7 PASSED -- and passed for the wrong reason entirely: the lookup returned REFUSED because the FILE was missing, not because 451 nm is untabulated. A green earned by an unrelated failure is worse than a red, so the tooth now requires T4 to have genuinely parsed the table first and cannot pass anywhere the mirror cannot be read. Verified by the fix dropping the local run from 7/10 to 6/10. Teeth, in order: T1 PREFACTOR-FREE KAT: the 450-to-600 absorption ratio equals (600/450)^3.48. T2 monotone: a shorter wavelength absorbs more. T3 concentration scales linearly -- doubling it doubles the absorption. T4 the mirrored haemoglobin table PARSES: a tabulated wavelength returns two non-zero columns. T5 ANTI-STUB: two different tabulated wavelengths return different values. T6 the two columns DIFFER at one wavelength -- HbO2 and Hb are different molecules, so a lookup returning one column twice cannot hide behind T4 and T5. T7 NEG-CONTROL: an untabulated wavelength REFUSES, with the table PROVEN readable first. T8 NEG-CONTROL: a missing mirror REFUSES rather than returning zero. T9 NEG-CONTROL: a wavelength outside the fitted band REFUSES rather than extrapolating the formula. T10 the source's own conversion arithmetic: mua = 2.303 e x / 64500. license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0