Nishi Family › Compare › Skin Twin -- appearance derived from anatomy

Nishi Compare · capabilities, evidence and delivery

Skin Twin -- appearance derived from anatomy

Compare capabilities and delivery evidence. Source discovery, qualification, deployment and accepted outcomes are separate; an open contract names work still to verify.

A chromophore substrate for human appearance: absorbers and structure in, reflectance out, relightable and editable by construction -- vs texture-authored and scan-captured pipelines, Sept-2026 field

Overview

The purpose, declared position and evidence coverage of this domain. Source presence and completed acceptance are different measures.

Where we are. MEASURED 2026-09-03 by building the chain and running every gate, not by inheriting a claim. Before this board existed, three absence probes came back ABSENT-PROVEN with corpus_complete=1: kubelka, cat02 and bradford. The estate had CIELAB and CIEDE2000 in nx_colorsci_lib and no way to move a D65 capture into the D50 PCS, no forward model from absorber concentrations to reflectance, and no inverse from a measurement back to absorbers. It also had, already, the growth architecture this board depends on -- nx_derm computing skin as an envelope FROM the layers beneath it, nx_anat_sov carrying inside-out vasculature scoped to where it shows, nx_nxa_groom growing hair from a follicle-density ratio, nx_microscan registering follicles against the pore layer, and hr_marschner_lobe with its own referee. The gap was never growth. The gap was the step between a grown structure and a visible colour, and that step is what this board owns.

Where we need to go. A human whose appearance is COMPUTED from anatomy at every layer, such that relighting is correct by construction and a physical edit -- a scar, a tan line, a wax, a flush -- is expressible as a change to absorbers and structure rather than to pixels.

The unit. permil of CIEDE2000 agreement against a measured capture, per body region, once ST1 wires the referee. Until then the unit is per-gate teeth passed, and the board says so rather than quoting a number it cannot compute.

Research bar. SimAvatar (layered simulation-ready avatars) is measured on body, hair and garments as SEPARATE simulated layers driven from text. Theirs: the layered-and-simulation-ready bar this board's growth stack is measured against. Ours: ST3 and ST5.

Research bar. SSD-GS (scattering and shadow decomposition) is measured on decomposes a captured splat into scattering and shadow so it can be relit. Theirs: the relightability bar -- and the reason a raw splat is a capture rather than an authoring format. Ours: ST1 and ST6.

Research bar. Subsurface scattering for 3D Gaussian splatting is measured on subsurface transport added to a splat representation for translucent surfaces. Theirs: the skin-translucency bar the Kubelka-Munk core is answering. Ours: ST2 and ST6.

Research bar. Infinigen (fully procedural, no external source) is measured on every asset generated from randomized mathematical rules, with paired ground truth exported. Theirs: the procedural-sovereignty bar, and the ground-truth export schema the caption lane should adopt. Ours: ST4 and ST8.

Latest recorded release

2026-10-03 · ST6

PROMOTED nx_skinspec_gate 0f2a1cb7 (212120 B, prev ca84de7f banked as .prev) carrying the honest cross-test names, bilirubin and water in the dermis, the adding-doubling arm and the per-subject measured-side CIELAB; allow_capability_loss passed only after nx_contentdiff live-first named both lost runs (the per-transport subject-line prefix and the reworded verdict note). _offc installed and hash-identical to live; the scratch row nx_skinspec_gate_staged reaped as superseded (toolreg f566282c). The full run that justified the promote ran the SAME bytes (sha 0f2a1cb7 = the staged artifact, /tmp/skinspec_st6_full_1003b.out, 78794 B): 20/20 GREEN, so the served gate's verdict is that run by hash identity; it was not re-run (3 h 16 min of wall clock on an 8-core box at load 15 to 21).

Release history and work log

Release entries describe recorded changes; they do not establish that every capability passed evaluation.

14 of 18 capability source symbols found|0 of them authored stronger-claims|4 open|coverage 777/1000|adoption 7 full / 7 partial

Evidence profile — what the gaps on this board actually are

Measured by nx_swcompare_evidence, read back by nx_evprofile_lib. Every figure is a count with its denominator — there is deliberately no score, no grade and no percentage anywhere in this band, because a stored scalar is a field a seat can edit and a counted partition is not.

proven able to fail counts the gates that have a RECORDED RED — nx_gate_bite mutated the gate subject, rebuilt it, watched the gate go red, and that record is inside the shared TTL. never bitten is its complement over the same denominator: those gates ran and were green, and nothing has ever shown them able to detect anything, so their green is a statement about this run and not about the gate. green at 0/0 is a separate and much weaker observation — the gate printed GREEN on a zero denominator, so its own tooth counter says it examined nothing. A gate can be green, non-zero, and still never bitten; that is the common case and it is now visible instead of implied.

partition: grounded + unsupported = 11 vs present 11 · named + unnamed + withheld = 5 vs open 5 · both reconcile

liar-kill conj=GPQN · all four conjuncts held

graded document: BUILDROOT tree, 8308 bytes · gates map: PRIMARY · stamped 34d 23h ago · STALE against the shared TTL of 604800s — this is evidence about a build nobody can vouch for now · source ../knowledge/status/evstamp_skintwin.verdict

Gap class	What it is, and the work it names
STALE	This profile is older than the shared TTL, or its clock ran backwards. It is reported as unusable rather than as fresh, because being wrong in the direction of refusing is the safe direction.

This band reports the referee counts. The per-axis worklist rows — which axis is unsupported, which watch is ready to flip, which gap is unnamed — are printed by nx_swcompare_evidence skintwin itself and are not carried on the stamp, so this page names the classes and the producer names the rows. That split is stated rather than hidden: a count without a worklist is not actionable, and this band is honest about which half of that it is.

Production map

Follow the dependencies, declared acceptance criteria and recorded priorities. Inspect source binding before treating a rank as executable work.

Ranking source binding: STALE_OR_INVALID_BINDING. Recorded priorities require current acceptance evidence and resource checks before execution.

Recorded ranking withheld

Retained ranking is not bound to the current plan and matrix: STALE_OR_INVALID_BINDING. Historical data remains retained. Refresh the ranking after resolving the input differences. Latest execution outcome is unverified.

Declared roadmap — contract, acceptance, executor, effort

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Wire CIEDE2000 as the per-region render referee (ST1)	cs_de2000	THE highest-value rung on the board. The ruler is built, promoted and has ZERO adopters outside the garment lane, measured with corpus_complete=1. Until it is wired, no row here can state a rendered error and every quality claim is an assertion.	Organ	1 u
Hair absorber pair: eumelanin and pheomelanin (ST2)	km_hair_mix	Reuses the shipped Kubelka-Munk core unchanged; only the coefficient vectors differ. Unlocks natural hair and is the precondition for ST3.	Organ	1 u
Fibre classes: real, enhanced, fully synthetic (ST3) after ST2	fo_fibre_class	A wig is a material, not a colour. Covers eyelashes and extensions by the same row. Without it the wig-versus-hair difference is inexpressible.	Organ	1.5 u
Anatomical follicle fields: density, orientation, correlation (ST4)	fo_field_sample	Random placement is a direct uncanny trigger. The sibling contract is the limbal ring on the graphics board -- same law, different organ.	Organ	2 u
Layered Kubelka-Munk for sheer and lace over skin (ST5)	km_layer_over	The problem KM was invented for. Lets garments and skin share one appearance core.	Organ	2 u
Measured absorption spectra, mirrored and pinned (ST6)	km_spectra_load	Every coefficient on this board is a probe value. This rung is what turns shaped-correctly into measured.	Organ	2 u
Wetness as a surface state (ST7)	km_wet_state	Wet skin darkens and its specular sharpens; the operator named environment and body fluids as first-class states. It is a surface term, not a chromophore one, which is why it is its own rung. BUILT 2026-09-28 in runtime/nx_kubelka_lib.nx (km_fresnel, km_fresnel_hemi, km_film_internal, km_wet_lambert, km_film_specular, km_wet_state): the film-over-Lambertian series of Jensen 1999 Sec. 2.1 and 3.1 with Fresnel derived from the indices and the film's internal diffuse reflectance by Judd's identity, held by nx_kubelka_gate against the direct integral through total internal reflection, an exact energy identity (a white substrate loses exactly the film's diffuse reflectance) and a cross-identity that re-derives the published wet value from the published film reflectances. Declared deviations: Jensen's roughness exponent is not carried (the microfacet lane's), and the water-beneath-the-surface effect (Sec. 2.2) is a separate rung. Done-rule: every film value the gate publishes re-derives from two independent methods within the stop tolerance, and the wet ladder darkens strictly.	Organ	1.5 u
Bite-prove every gate on this board (ST8)	nx_gate_bite	CLOSED 2026-09-03. All five gates mutation-proven BOTH directions on the laptop build farm while the NAS build lane was storm-refused, each with a byte-identical restore verified by sha256. The kills were PRECISE rather than shotgun, which is the part that matters: kubelka RED 9 of 10 killed by roundtrip_ok=1/9 (only the ks=0 rung survives, and it should, because the mutation is invisible there); vasc RED 7 of 9 killed by BOTH depth teeth including the anti-painting one, when the depth attenuation was removed -- that is the model literally turned into a painting and the tooth firing; follicle RED 15 of 16 killed by the TYNDALL tooth ALONE, with the blonde-versus-black tooth correctly surviving because those differ by their own coefficients regardless of depth; skinfit RED 5 of 8 killed by recovery, monotonicity and the honesty tooth together when the error metric was blinded; ws_client RED 6 of 9 killed by the mask bit, the INTEROP tooth and the extended-length round trip when the masking XOR was dropped. Every gate returned to full GREEN on restore.	Organ	0.5 u
Board-level journal anchor for skintwin (ST0)	-	BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-04. This board already journals 9 log rows against the id ST0, meaning the entry is about the BOARD rather than about one rung. The .plan schema says a log row's third field is a rung id, so until this row existed every one of those entries resolved to nothing and a session inheriting this board read a ledger pointing at an address that was never declared. Declaring the anchor is the additive fix: it changes no existing row, it makes the convention READABLE instead of implicit, and nx_board_idref_gate now resolves those entries instead of counting them as dangling. Journal board-wide findings here; journal rung work against its rung.	Board	0 u

Ladder to SOTA

Two dated targets, and every rung's role toward them. Best in class is the proven, deployed leader today; the frontier is the research edge as of its month. Each target is dated by a sotabar row and judged against the current month by the bar-age ruler (FRESH, ATTESTED with the fallback month named, or refused). Each rung is substrate (what the target's own methods consume), an arm (a sovereign arm that triangulates without competing for the number), a contender (its done-rule is a number against the target) or superseded (kept as capability, retired as a claim), and carries the rungs it stands on.

best-in-class (T-BIC-SKIN), dated by SB-MHSKIN-2609 = 2026-09 ATTESTED. The shipped human skin pipeline of MetaHuman 5.8 (painted and parameterised skin, palm and sole tone, nails, the neck-seam fix, unbaked material control) judged in CIEDE2000 against ISO/DTR 26478 patches and NIST's traceable spectra; parity is a number the referee prints, never a look ref metahuman_58_notes_202609

frontier (T-FR-DERIVED), dated by SB-PHYSAPP-2609 = 2026-09 ATTESTED. Appearance COMPUTED from anatomy at every layer so that a relight and a physical edit (a tan line, a wax, a flush, a wet surface, a sheer garment over it) are changes to absorbers and structure, never to pixels; the field's nearest neighbour this month drives one avatar's appearance by physiology (Heartian); nothing derives every layer ref arxiv_co_siga2026

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Hair wetness and clumping as a state (ST9) after ST7,ST2	_ABSENT_:fo_hair_wet	Wet hair is not darker hair paint: strands clump by capillary adhesion (a clump field over the groom), the cuticle reads index-matched so the Marschner R lobe drops and TT rises, and drying is a curve. Consumes ST7's wetness field and the shipped hr_marschner_lobe; graphics GR58 renders it. No current-month rival exists (SB-WET-2512). Done-rule: a wet groom renders with a measured clump count and lobe shift published beside the dry groom, both from one wetness row.	Organ	2 u

Rung	Role	Toward	Stands on	Why
ST0	substrate	T-FR-DERIVED	-	board-level journal anchor
ST1	substrate	T-BIC-SKIN	-	the CIEDE2000 referee is what turns every claim on both targets into a number
ST2	substrate	T-FR-DERIVED	-	hair absorbers are the derived-appearance chain reaching hair
ST3	substrate	T-FR-DERIVED	ST2	fibre classes make a wig a material, part of the derived chain
ST4	substrate	T-FR-DERIVED	-	anatomical follicle fields are structure the appearance derives from
ST5	contender	T-FR-DERIVED	-	sheer fabric over skin as layered absorbers is the derivability claim on a garment
ST6	substrate	T-BIC-SKIN	-	measured spectra pinned as data are the calibration the standards tier demands
ST7	contender	T-FR-DERIVED	-	wetness as a state is a physical edit expressed in structure, the claim itself
ST8	substrate	T-BIC-SKIN	-	bite-proven gates are the evidence floor
ST9	contender	T-FR-DERIVED	ST7,ST2	wet hair as a state is the same claim on the groom

Ladder verdict. LADDERED. Targets, rung roles and bars are counted in the tables below, each partition re-derived beside its total. Bars read FRESH against the current month 2026-10.

Ladder verdict, counted (17 rows, more than the 12-row disclosure bar)

Ladder verdict, counted
Class	Count	Note
targets	2	sotatarget rows on the plan
best in class	1	targets whose class is best-in-class, by code
frontier	1	targets whose class is frontier, by code
undated	0	targets naming no bar that exists
bad class	0	targets whose class is neither word
duplicate	0	target ids an earlier row already carries
rungs	10	rung rows on the plan
placed	10	rungs carrying a rungrole row
unplaced	0	rungs with no role toward any target
bad role	0	rungrole rows naming a role the ladder does not know
bad target	0	rungrole rows naming a target no sotatarget row declares
rungs, summed	10	RECONCILED: the parts sum to the total 10
substrate	7	what the target's own methods consume
arm	0	a sovereign arm that triangulates without competing for the number
contender	3	its done-rule is a number against the target
superseded	0	kept as capability, retired as a claim
placed, summed	10	RECONCILED: the parts sum to the total 10

Bar freshness, counted
Class	Count	Note
bars	7	sotabar rows on the plan
fresh	0	the bar month is the current month
attested	7	an older bar with a this-month barscan reporting no newer entry
stale	0	a this-month scan saw a newer month the bar ignores
unattested	0	nobody looked this month
malformed	0	a sotabar row missing a field
duplicate	0	bar ids an earlier row already carries
bars, summed	7	RECONCILED: the parts sum to the total 7
current month	2026-10	the month every bar is judged against

Treatment verdict. NOT DECLARED. No rung on this board declares its three-step treatment (a treat row naming the MCP row, the API route, the primitive it composes, the workflow that composes it and the agent beat that runs it); the ranker labels every rung TREATMENT-UNDECLARED until it does.

Inspect a rung and its prerequisites

Declared nodes 10. Rank input binding: STALE_OR_INVALID_BINDING. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 897c3b1824a51be6b21bd4325501488162c847cd5f01a4c089371cd242d2fa75. Target rows 2; role rows 10. Existing risks and release worklog retain their own scope; no node completion is inferred.

Runtime observation coverage: 0 of 10 rungs have current, identity-bound execution observations. Observation may report RED. Accepted completion remains unknown; source presence, qualification and deployment do not establish it.

Use Enter or Space on a rung to inspect its contract. Prerequisite links locate another rung in this list; open its summary to inspect it. Estimates are authored effort, not forecasts.

ST1 — Wire CIEDE2000 as the per-region render referee
Prerequisites: None declared; this does not establish execution eligibility.
Contract: cs_de2000
Acceptance: THE highest-value rung on the board. The ruler is built, promoted and has ZERO adopters outside the garment lane, measured with corpus_complete=1. Until it is wired, no row here can state a rendered error and every quality claim is an assertion.
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST2 — Hair absorber pair: eumelanin and pheomelanin
Prerequisites: None declared; this does not establish execution eligibility.
Contract: km_hair_mix
Acceptance: Reuses the shipped Kubelka-Munk core unchanged; only the coefficient vectors differ. Unlocks natural hair and is the precondition for ST3.
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST3 — Fibre classes: real, enhanced, fully synthetic
Prerequisites: ST2 (acceptance unverified)
Contract: fo_fibre_class
Acceptance: A wig is a material, not a colour. Covers eyelashes and extensions by the same row. Without it the wig-versus-hair difference is inexpressible.
Authored effort: 1.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST4 — Anatomical follicle fields: density, orientation, correlation
Prerequisites: None declared; this does not establish execution eligibility.
Contract: fo_field_sample
Acceptance: Random placement is a direct uncanny trigger. The sibling contract is the limbal ring on the graphics board -- same law, different organ.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST5 — Layered Kubelka-Munk for sheer and lace over skin
Prerequisites: None declared; this does not establish execution eligibility.
Contract: km_layer_over
Acceptance: The problem KM was invented for. Lets garments and skin share one appearance core.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST6 — Measured absorption spectra, mirrored and pinned
Prerequisites: None declared; this does not establish execution eligibility.
Contract: km_spectra_load
Acceptance: Every coefficient on this board is a probe value. This rung is what turns shaped-correctly into measured.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST7 — Wetness as a surface state
Prerequisites: None declared; this does not establish execution eligibility.
Contract: km_wet_state
Acceptance: Wet skin darkens and its specular sharpens; the operator named environment and body fluids as first-class states. It is a surface term, not a chromophore one, which is why it is its own rung. BUILT 2026-09-28 in runtime/nx_kubelka_lib.nx (km_fresnel, km_fresnel_hemi, km_film_internal, km_wet_lambert, km_film_specular, km_wet_state): the film-over-Lambertian series of Jensen 1999 Sec. 2.1 and 3.1 with Fresnel derived from the indices and the film's internal diffuse reflectance by Judd's identity, held by nx_kubelka_gate against the direct integral through total internal reflection, an exact energy identity (a white substrate loses exactly the film's diffuse reflectance) and a cross-identity that re-derives the published wet value from the published film reflectances. Declared deviations: Jensen's roughness exponent is not carried (the microfacet lane's), and the water-beneath-the-surface effect (Sec. 2.2) is a separate rung. Done-rule: every film value the gate publishes re-derives from two independent methods within the stop tolerance, and the wet ladder darkens strictly.
Authored effort: 1.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST8 — Bite-prove every gate on this board
Prerequisites: None declared; this does not establish execution eligibility.
Contract: nx_gate_bite
Acceptance: CLOSED 2026-09-03. All five gates mutation-proven BOTH directions on the laptop build farm while the NAS build lane was storm-refused, each with a byte-identical restore verified by sha256. The kills were PRECISE rather than shotgun, which is the part that matters: kubelka RED 9 of 10 killed by roundtrip_ok=1/9 (only the ks=0 rung survives, and it should, because the mutation is invisible there); vasc RED 7 of 9 killed by BOTH depth teeth including the anti-painting one, when the depth attenuation was removed -- that is the model literally turned into a painting and the tooth firing; follicle RED 15 of 16 killed by the TYNDALL tooth ALONE, with the blonde-versus-black tooth correctly surviving because those differ by their own coefficients regardless of depth; skinfit RED 5 of 8 killed by recovery, monotonicity and the honesty tooth together when the error metric was blinded; ws_client RED 6 of 9 killed by the mask bit, the INTEROP tooth and the extended-length round trip when the masking XOR was dropped. Every gate returned to full GREEN on restore.
Authored effort: 0.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST0 — Board-level journal anchor for skintwin
Prerequisites: None declared; this does not establish execution eligibility.
Contract: -
Acceptance: BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-04. This board already journals 9 log rows against the id ST0, meaning the entry is about the BOARD rather than about one rung. The .plan schema says a log row's third field is a rung id, so until this row existed every one of those entries resolved to nothing and a session inheriting this board read a ledger pointing at an address that was never declared. Declaring the anchor is the additive fix: it changes no existing row, it makes the convention READABLE instead of implicit, and nx_board_idref_gate now resolves those entries instead of counting them as dangling. Journal board-wide findings here; journal rung work against its rung.
Authored effort: 0. Executor kind: Board. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
ST9 — Hair wetness and clumping as a state
Prerequisites: ST7 (acceptance unverified), ST2 (acceptance unverified)
Contract: _ABSENT_:fo_hair_wet
Acceptance: Wet hair is not darker hair paint: strands clump by capillary adhesion (a clump field over the groom), the cuticle reads index-matched so the Marschner R lobe drops and TT rises, and drying is a curve. Consumes ST7's wetness field and the shipped hr_marschner_lobe; graphics GR58 renders it. No current-month rival exists (SB-WET-2512). Done-rule: a wet groom renders with a measured clump count and lobe shift published beside the dry groom, both from one wetness row.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.

Learning and practice paths

No structured learning path is declared for this plan. Existing research, roadmap and worklog remain available above.

Rung dossiers — every record that bears on a rung, in one place

Joined by symbol, never typed. Each dossier is assembled on every publish from the rows this board already carries: the plan's rung row, the matrix row that carries its contract symbol (its status re-measured by the one symbol ruler), the gate the organ declares and the verdict the evidence referee recorded on its last run, the reference marks in the rung text, the design rows that name the rung and the work log filed against it. The capability matrix links each cell to its dossier and each dossier links back to its cell and forward to the rungs it stands on; a dependency naming no rung on this plan is written as UNRESOLVED rather than linked. Absences are named: NO-MATRIX-ROW, NO-GATE-DECLARED, UNRECORDED.

Gate verdicts: UNSTAMPED — no per-gate verdict artifact is published for this domain yet (nx_swcompare_evidence skintwin writes it beside its stamp); every gate below reads UNRECORDED until it runs.

Gate declarations: knowledge/organ_gate.conf read from this working directory (1 rows) and from the estate root (196 rows); rows in both counted once: 0; declarations read: 197. Every declaration in either copy counts, so a stub beside the publisher can never hide the root.

ST1 Wire CIEDE2000 as the per-region render referee · cell PRESENT · gate NO-GATE-DECLARED

Contract	`cs_de2000` · executor Organ · 1 u · stands on nothing
Definition of done	THE highest-value rung on the board. The ruler is built, promoted and has ZERO adopters outside the garment lane, measured with corpus_complete=1. Until it is wired, no row here can state a rendered error and every quality claim is an assertion.
Matrix cell	COLOUR: perceptual colour difference as the referee unit · `runtime/nx_colorsci_lib.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_colorsci_lib`; declare `nx_colorsci_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-10-01 measure FIRST RENDER-SIDE ADOPTION OF THE REFEREE (graphics GR59, 2026-09-30): nx_skin_ita_gate (promoted sha256 5dfb04e0, live 25/25 GREEN) reads a DERIVED tan line through cs_de2000: inside-outside DE2000 4.99 beside DE76 6.51 against the measured natural tan of Amano 2020, DE76 mean 6.03 SD 2.64 [amano_naturaltan_2020], on a spectral chain that integrates the CIE's own 1931 2-degree tables [cie_cmf_1931_2deg_202609] at 5 nm and adapts illuminant A to D65 through cs_cat_build; the zero-exposure control reads DE2000 exactly 0, and an aimed mutant that ignores the garment's coverage history is killed by the DE2000 tooth among six others. The constitutive melanin, solved from the genome's ITA on the incumbent locus, lands at 5.54 percent inside the light-skin range of [jacques_skinoptics_1998]; the facultative field is Ravnbak's measured dose-response and fade [ravnbak_tankinetics_2008]. Adopters of cs_de2000 outside the garment lane: 0 -> 1. NAMED LIMIT: the dermis is still bloodless (a* 3.0 inside where measured skin reads near 10), so the referee's a* axis waits on ST6's haemoglobin data entering through km_ks_mix k_hem

ST2 Hair absorber pair: eumelanin and pheomelanin · cell WATCHING · gate NO-GATE-DECLARED

Contract	`km_hair_mix` · executor Organ · 1 u · stands on nothing
Definition of done	Reuses the shipped Kubelka-Munk core unchanged; only the coefficient vectors differ. Unlocks natural hair and is the precondition for ST3.
Matrix cell	HAIR: the same two-absorber model with different absorbers · `runtime/nx_kubelka_lib.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_kubelka_lib`; declare `nx_kubelka_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

ST3 Fibre classes: real, enhanced, fully synthetic · cell WATCHING · gate NO-GATE-DECLARED

Contract	`fo_fibre_class` · executor Organ · 1.5 u · stands on ST2
Definition of done	A wig is a material, not a colour. Covers eyelashes and extensions by the same row. Without it the wig-versus-hair difference is inexpressible.
Matrix cell	HAIR: a wig is a MATERIAL, which is why wigs read as wigs · `runtime/nx_follicle_lib.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_follicle_lib`; declare `nx_follicle_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

ST4 Anatomical follicle fields: density, orientation, correlation · cell WATCHING · gate NO-GATE-DECLARED

Contract	`fo_field_sample` · executor Organ · 2 u · stands on nothing
Definition of done	Random placement is a direct uncanny trigger. The sibling contract is the limbal ring on the graphics board -- same law, different organ.
Matrix cell	HAIR: follicle placement follows anatomical fields, not noise · `runtime/nx_follicle_lib.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_follicle_lib`; declare `nx_follicle_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

ST5 Layered Kubelka-Munk for sheer and lace over skin · cell LANDED · gate NO-GATE-DECLARED

Contract	`km_layer_over` · executor Organ · 2 u · stands on nothing
Definition of done	The problem KM was invented for. Lets garments and skin share one appearance core.
Matrix cell	FABRIC: sheer and lace over skin is a LAYERED scattering problem · `runtime/nx_kubelka_lib.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_kubelka_lib`; declare `nx_kubelka_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-27 queue The beach-lounger reference class (2026-09-27) makes ST5 the load-bearing rung of the whole twin: the garment is sheer, so the skin term must transmit through the fabric; graphics GR57 is its render-side consumer and depends on it. nx_compare_rank reads ST5 as DARK-CONTRACT (organ exists, symbol undeclared), so km_layer_over is unbuilt, not merely unwired. 2026-09-27 land LANDED 2026-09-27: km_layer_over, km_layer_rt, km_open_strain and km_sheer_over in nx_kubelka_lib (4116 -> 9704 bytes, additive; the single-constant forward and inverse untouched); nx_kubelka_gate grew from 10 to 22 teeth (7487 -> 18165 bytes) and was built sovereign (70904 bytes, sha256 7726d1633e08cd55200b6d050cf419b60bfc3e700f0b38577f38824707261bf9), probed staged-vs-live (other_rc=0, DIFFERS only from the new teeth onward), bite-proven (a compiling comparison mutant of the lib at byte 4050 KILLED, both twins restored byte-identical), and PROMOTED with prev banked. Mechanism replicated: Kubelka's layered two-flux composition R = R_f + T_f^2 R_s / (1 - R_f R_s) over a porous layer whose open fraction follows area geometry under strain; no coefficient typed. 2026-09-27 measure nx_contentdiff live vs staged read RED lost_from_live=4: the rewritten verdict note (named, mine) and sys_openat_append, gv_cat, gv_catn; the matched-vintage control (nx_vasc_gate, source untouched since 2026-09-03, rebuilt through the SAME lane) lost exactly those three symbols and gained the same framed-append set, so the loss is the gate base class moving between 09-03 and today, not this change; promoted under allow_capability_loss=yes with every run named here. 2026-09-27 lesson A LAYERED REFLECTANCE IS NOT MONOTONE IN THE OPEN FRACTION, AND THE FIRST TOOTH ASSERTED IT WAS: a bright absorbing fabric (R 0.3, T 0.5) over darker skin (0.4) sits ABOVE the skin at rest (0.4136 fully closed), dips below it as the weave opens (0.3993 at 0.2 open, 0.3930 at 0.34) and returns to it only at full opening. The arithmetic caught the invariant before the build did; the tooth now asserts the geometry that is monotone (the open fraction under strain) and a thickness ladder asserts the two-flux property that is (reflectance rises, transmittance falls, energy conserved on every rung). 2026-09-27 queue The bite's first-kill landed in the pre-existing mix function; an AIMED bite inside the new block is owed so the layered teeth are proven non-vacuous by a mutant of their own subject, and graphics GR57 is the render-side consumer that turns this lib term into a swatch under the sky. 2026-09-27 measure AIMED BITE RESOLVED, NOT SKIPPED: nx_gate_bite aim km_layer_over 0 produced 2 numeric mutants and both FAILED TO COMPILE (invalid, excluded, verdict INCONCLUSIVE, twins restored byte-identical); the numeric operator can only mutate literals and the new arithmetic carries none (its guards are the only zeros, a consequence of the magic-number law), while the comparison operator's first-kill-wins stops at byte 4050 in the pre-existing mix function. The non-vacuity oracle for T11 to T21 is therefore the hand-derived KATs (399292, 488000, 412667, the exact halves and R_inf), which a wrong formula cannot pass; a comparison-aimed bite (aim on == sites of one function) is a bite-organ rung, filed as debt rather than left as an unnamed gap.

ST6 Measured absorption spectra, mirrored and pinned · cell LANDED · gate NO-GATE-DECLARED

Contract	`km_spectra_load` · executor Organ · 2 u · stands on nothing
Definition of done	Every coefficient on this board is a probe value. This rung is what turns shaped-correctly into measured.
Matrix cell	CALIBRATION: measured absorption spectra as pinned data · `runtime/nx_kubelka_lib.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_kubelka_lib`; declare `nx_kubelka_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-28 retract The ST6 contract was unjudgeable as written: its watch row named the DIRECTORY knowledge/fetched as the organ and st_coeff_spectra as the symbol, so no symbol ruler could ever read it, and graphics GR59 (melanin field, on the photoreal path) read its prerequisite skintwin:ST6 as UNRESOLVED(organ-unreadable) through nx_xdep_lib on 2026-09-28. Repointed: the rung and the row now name runtime/nx_kubelka_lib.nx km_spectra_load, the reader of the pinned melanin and haemoglobin absorption spectra that the lib's own header says belong in pinned reference data and not in the lib. The mirrored, pinned spectra stay the data half of the rung; the loader is the half a ruler can judge. Cell reads OPEN until km_spectra_load is built 2026-10-01 measure ACCEPT RULE PRE-DECLARED before any fit runs (2026-09-30): the bar paper 2606.27604 (Aliaga and Jarabo, mirrored knowledge/fetched/cmp_skintwin_spectral_sss_2606_27604.pdf, read through nx_pdf_text) reports its accuracy ONLY against its own Monte Carlo reference, never against measured skin: Table 1 rendering error Delta E76 on the visible band over 10 representative tones = 3.19 for its K=3 neural decoder and 3.79 for the K=3 per-tone joint optimisation (its stated fitting ceiling), and a mean reflectance error of 3.7 percent (K=3) over 25K tones; its limitations section names fitting to measured spectra as the validation axis it did NOT do. ST6 measures the estate on that missing axis: a spectral two-layer biophysical model fitted to every one of NIST's 100 measured skin spectra (JRES 122.026, data DOI 10.18434/M38597, mirrored and pinned). Rule: median Delta E76 over the 100 spectra at or below 3.19 MEETS the bar's best number, at or below 3.79 meets its fitting ceiling; the worst spectrum is held to 3.79 and published whatever it reads; median mean-relative reflectance error held to 3.7 percent. The comparison is a different test (measured vs simulated reference) and is published as such, never as like-for-like 2026-10-01 measure FIRST FULL RUN on all 100 NIST spectra (staged gate, job pid 17528, /tmp/nx_skinspec_gate_st6_1.out, 12/16 RED, every loader tooth GREEN): the two-layer model with the epidermis held at Jacques' 60 um and fitted over 400 to 900 nm reads median Delta E2000 3.457, median Delta E76 4.585 (worst 9.196, subject 72), median spectral RMS 0.0214, median mean-relative error 5.18 percent -- it does NOT meet the bar's 3.19. Pre-declared now, before the second run: the MODEL OF RECORD carries the bar's own fifth structural parameter, the epidermal thickness, fitted over the bar's own range (50 to 350 um, 2606.27604 Sec. 3), and fits the visible band 400 to 780 nm on which the bar's Delta E is measured (the near infrared carries water and the sphere's lateral loss, neither modelled); the 400 to 900 nm four-parameter run and a 400 to 780 nm four-parameter run are published beside it as the ablation, so the choice of record is not made after seeing which wins 2026-10-01 measure TRIANGULATED BAR as of 2026-10-01 for spectral skin appearance (layered chromophore models, spectral subsurface scattering, inversion from photographs): FORWARD MODEL -- the best is a multi-layer spectral transport whose layers carry chromophores and which yields a spatial scattering profile, not only hemispherical reflectance; the June 2026 bar Monte Carlo-simulates a two-layer skin (epidermis 50 to 350 micrometres, melanin-dominated, over a semi-infinite dermis with haemoglobin, bilirubin, beta-carotene and water) at 376 wavelengths from five properties (melanin and blood fraction, epidermal thickness, blood oxygenation, eumelanin to pheomelanin ratio), records total diffuse reflectance AND the radial diffusion profile, and states that hemispherical-only spectral datasets cannot characterise translucency [aliaga_spectralsss_2606] [spectral_sss_2606_27604] [arxiv_sss_202609]; the measured-capture lineage is the Donner and Jensen two-layer spectral BSSRDF with melanin blend type and epidermal hemoglobin fraction [gitlina_spectralskin_2020]. INVERSION -- complementary multispectral illumination recovers chromophore maps better than standard broadband illumination, with a network replacing lookup tables [gitlina_spectralskin_2020]; the frontier inverts one RGB albedo through a learned biophysical prior into a three-media mixture for a random-walk path tracer [aliaga_spectralsss_2606]. CALIBRATION -- traceable measured reflectance: NIST's 100 spectra [nist_skin_reflectance_202609] and, newest this month (2609.30122, submitted 2026-09-24), 98 patients' skin reflectance spectra in the public OpenOximetry Repository fitted with a melanin-corrected diffuse-optics model [arxiv_skinspecrefl_202609]. RE-CHECK of SB-SKINAPP-2606 today: the arXiv subsurface-scattering-and-skin search holds no July to October 2026 entry, so the June bar stands and September's newest on-subject entry is biomedical [arxiv_sss_202610] [arxiv_skinspecrefl_202609] ; witnesses [aliaga_spectralsss_2606] [gitlina_spectralskin_2020] [arxiv_sss_202609] [arxiv_sss_202610] [arxiv_skinspecrefl_202609] [nist_skin_reflectance_202609] -- independent groups: Aliaga and Jarabo; Ghosh's group at Imperial College London with NTNU, York, Clemson and UBC; the Cremers group (DIAMOND-SSS 2601.12020); Ha, Munoz, Gutierrez and Kim (polarimetric BSSRDF of dynamic faces, ACM doi 10.1145/3687767); the Fantini diffuse-optics group (2609.30122) -- five groups, with a peer-reviewed CGF paper and an ACM-published paper among them and NIST as the standards tier; aliaga_spectralsss_2606 and spectral_sss_2606_27604 pin the same sha256 and are ONE witness ; WHERE THEY DISAGREE: measured multispectral capture versus RGB-only learned inversion (one RGB albedo underdetermines five biophysical properties, resolved by a learned prior in the frontier and by extra spectral illuminants in the capture school); validation (the frontier validates only against its own Monte Carlo simulation and names fitting measured integrating-sphere diffusion profiles and in-vivo samples as the next step, while the capture school matches photographs); explicit chromophore layers versus purely learned neural scattering (DIAMOND-SSS reconstructs relightable translucent Gaussians, wax and jade as well as skin, from as few as ten images with no chromophore model in its abstract); and how well melanin's effect is even resolved (the oximetry fit needs a cohort roughly ten times larger to resolve its melanin dependence statistically) ; VERDICT on the running lane (updated after its third full run, log 1790826240): ST6 loads the pinned OMLC melanin and haemoglobin absorption and fits a two-layer spectral model in the board's Kubelka-Munk core to all 100 NIST measured spectra -- AT the field on calibration and AHEAD of the bar on measured validation, an axis the bar never measured (median Delta E2000 2.081, median Delta E76 2.419 beside the bar's simulated-reference 3.19, median mean-relative error 3.41 percent; a different test, published as such); BELOW the triangulated best 2026-10-01 measure THIRD FULL RUN, the optimiser corrected (staged gate pid 17884, /tmp/nx_skinspec_gate_st6_3.out, 16/17): the second run's model of record had read WORSE than the 900 nm ablation on its own window, which a superset of parameters on a subset of wavelengths cannot do unless the search fails, and it was the search: halving coordinate brackets collapsed around the wrong point on the coupled scattering and melanin pair. A compass-search second round (Torczon pattern search: steps halve only when no coordinate improves) plus the captured specular as a fitted term bounded by the epidermal Fresnel value gives, on every one of the 100 NIST spectra, model of record (400 to 780 nm, six parameters): median Delta E2000 2.081, worst 4.925; median Delta E76 2.419 (meets the bar's 3.19), worst 6.695 at subject 55 (misses the declared 3.79); median spectral RMS 0.0137, worst 0.0292; median mean-relative error 3.41 percent (meets the bar's 3.7). Ablation on the same build: 900 nm window with the thickness held at 60 um, median Delta E76 3.131, worst 7.417; visible window held, 3.170, worst 7.540. Band residuals of record (median signed model minus measured): plus 0.019 at 400 to 450 nm, plus 0.010 at 450 to 520, minus 0.004, minus 0.004, minus 0.012 at 700 to 780: the model still lacks blue absorption, the shape beta-carotene and bilirubin give, both carried by the bar's dermis. Surface witness: NIST's darkest 2500 nm reading is 0.0229 against the 1.4 Fresnel 0.0278 and the 1.55 Fresnel 0.0465, so the dry stratum corneum index is refuted and the sphere does not return the full specular of every measurement 2026-10-01 land THE CONTRACT SYMBOL LANDED IN THE CONTRACT ORGAN 2026-09-30: km_spectra_load is in runtime/nx_kubelka_lib.nx (18047 to 29088 B, strictly additive: one import of the canonical nx_sha256 and an appended block; no existing line altered). It reads a mirrored table, proves the bytes against the .refs pin by sha256, and returns one column on a declared window in a declared unit (column times num over den, Q30, integer arithmetic so a 5e5 molar extinction never meets a Q30 multiply), refusing BY NAME a wrong pin (KM_SPEC_PIN), a malformed pin (KM_SPEC_BADPIN), a truncated table (KM_SPEC_SHORT), a non-monotone wavelength column (KM_SPEC_NONMONO) and an overrun (KM_SPEC_CAP); km_spec_interp interpolates and refuses outside the span. Shared-lib discipline: the block was compiled and run first in a scratch file (nx_skinspec_load.nx, now retired to knowledge/retired), then moved verbatim. Every consumer re-proved against the grown lib: nx_kubelka_gate 37/37 GREEN, contentdiff IDENTICAL-CONTENT (82712 B both), promoted 941a87a9; nx_vasc_gate 9/9, nx_follicle_gate 16/16, nx_skinfit_gate 8/8 GREEN as staged rebuilds, NOT promoted here because their promoted binaries predate the gate base-class move of September (contentdiff loses sys_openat_append, gv_cat, gv_catn on all three, the same three the 09-27 matched-vintage control attributed to the base class) and that drift belongs to their owners' lane. The 09-03 incumbent nx_skinspectra_lib is COMPOSED, not duplicated: the model reads the melanosome law through its ss_melanin_mua; its exact-match integer ss_hb_lookup is superseded by km_spectra_load (pinned, decimal, any column, windowed, named refusals) and stays callable for its own gate. That incumbent gate, source-only since 09-03 with its build enqueued, built and ran for the first time: nx_skinspectra_gate 10/10 GREEN, promoted 3a30f194 (54541 B), registered, _offc refreshed 2026-10-01 lesson A SUPERSET OF PARAMETERS FITTED ON A SUBSET OF WAVELENGTHS CANNOT READ WORSE ON THAT SUBSET UNLESS THE SEARCH FAILED, AND THAT ONE INEQUALITY FOUND THE DEFECT: the second run's six-parameter visible fit read median Delta E76 5.44 against 4.66 for the four-parameter 900 nm fit. Halving coordinate brackets collapse around the wrong point on a coupled pair (scattering scale against melanin); a compass-search second round, whose steps halve only when no coordinate improves, took the same model to 2.42. The fix was the optimiser, not the physics, and the ablation published beside the model of record is what made the impossibility visible 2026-10-01 lesson THE INCUMBENT WAS FOUND BY A COMMENT, NOT BY A SEARCH: nx_skinspectra_lib and its gate (ST6, 2026-09-03, source-only, build enqueued) sat in buildroot/runtime while this rung was rebuilt; the catalogue was asked for nx_skinspec and its siblings but not for nx_skinspectra, and the .gates header named it only after the loader was written. Remedy taken the same hour rather than filed: the model now COMPOSES its ss_melanin_mua (one melanin law in the estate), its exact-match integer lookup is superseded by the pinned windowed loader, and its gate was built, run 10/10 and promoted. Ask nx_capsearch with the RUNG ID and the data words (ST6, spectra, melanin) before naming a new organ 2026-10-01 lesson A SPECULAR TERM IS BOUNDED BY THE DARKEST READING, NOT ARBITRATED BY THE MEDIAN: the first surface witness asked which index's 8 degree Fresnel term sat nearer the median 2500 nm reading (0.0391) and picked the dry stratum corneum's 1.55 (0.0465), ignoring that the body still returns light at 2500 nm. The sound test is that the specular can never exceed what the instrument read where the body returns least: the darkest subject reads 0.0229, below even the epidermal 1.4 term (0.0278), so 1.55 is refuted outright and the sphere's captured specular is a per-measurement fraction, now fitted under the Fresnel bound and checked against each subject's own 2500 nm reading 2026-10-01 measure THE MODEL OF RECORD MEETS THE BAR ON THE AXIS THE BAR DID NOT MEASURE (2026-09-30): nx_skinspec_gate 18/18 GREEN on the binary then promoted (sha256 5113a088, 140963 B, staged run /tmp/nx_skinspec_gate_final_1.out, the same bytes /api/promote then took with expect_sha256). Model: the bar's two-layer biophysical skin (melanosome-dominated epidermis over a semi-infinite dermis carrying oxy- and deoxy-haemoglobin and beta-carotene, Jacques baseline and Mie plus Rayleigh scattering), every absorber read from a pinned mirror through km_spectra_load, transported by Kubelka's layered two-flux solution, crossed through the skin surface by the Saunderson series with Judd's internal reflectance, rendered through the pinned CIE 1931 2 degree functions under pinned D65. Fitted to EVERY one of NIST's 100 measured forearm spectra over 400 to 780 nm: median CIEDE2000 0.769, worst 2.420; median CIE76 Delta E 1.030, worst 2.824 at subject 53; median spectral RMS 0.0128, worst 0.0259; median mean-relative error 3.12 percent. BESIDE THE BAR (2606.27604, Aliaga and Jarabo, read from the mirror): its Table 1 visible-band Delta E76 is 3.19 for the K=3 neural decoder and 3.79 for its per-tone joint optimisation, and its mean reflectance error 3.7 percent -- all against its OWN Monte Carlo reference over 10 representative tones, never against measured skin. On measured skin the estate's median Delta E76 is 3.1 times lower than the bar's best figure and its worst spectrum sits under the bar's fitting ceiling; this is a different test from the bar's and is published as such, not as like-for-like. What closed the gap from the first run's 4.585: the compass-search optimiser (to 2.419) and beta-carotene, the dermal chromophore the bar carries (to 1.030); band residuals of record plus 0.018 at 400 to 450 nm, plus 0.005, minus 0.002, minus 0.003, minus 0.010 at 700 to 780 nm. Pre-declared gates all held: median 1.030 at or below 3.19, worst 2.824 at or below 3.79, median relative error 3.12 at or below 3.7 percent 2026-10-01 measure THE PRE-DECLARED ABLATION ON THE FINAL MODEL (staged run /tmp/nx_skinspec_gate_st6_4.out, the same model before its melanosome law was rerouted through the incumbent ss_melanin_mua; its per-subject lines are digit-identical to the promoted run's, subjects 55 to 60 checked, so the composition is numerically neutral): 900 nm window with the epidermis held at Jacques' 60 um, median CIEDE2000 1.074, median Delta E76 1.589, worst 5.962, median relative error 3.48 percent; visible window with the epidermis held, 1.108, 1.376, worst 3.392, 3.96 percent; model of record (visible, thickness fitted over the bar's 50 to 350 um), 0.769, 1.030, worst 2.824, 3.12 percent. The thickness and the visible window each pay for themselves on the worst spectrum, the reason the model of record was pre-declared as it was. NAMED LIMITS read off the fitted parameters: the epidermal thickness sits at the bar's 50 um floor for most subjects (median 49 um printed, integer truncation of 50), the captured specular fits to zero for 97 of 100 subjects, and the carotene absorption reaches its declared span (16 times the baseline at its peak, 19.27 per cm) on 4 subjects (17, 44, 61, 72) -- three places where a bound, not the data, decides, each a pointer to the transport and the missing chromophores rather than a defect to tune 2026-10-01 queue NEXT, in order of what the band residuals and the bound hits say: (1) bilirubin and pheomelanin as pinned absorbers (OMLC PhotochemCAD carries bilirubin; a pheomelanin spectrum must be found and pinned before the melanin blend the bar carries can exist) -- the 400 to 450 nm residual of plus 0.018 is the blue absorption still missing; (2) the transport at the bar's fidelity class: Monte Carlo or adding-doubling over the same two layers, so the thickness and specular stop sitting on their bounds; (3) water from a pinned OMLC table so the 900 nm window joins the model of record; (4) the RGB chain switch: nx_skinfit_lib, nx_vasc and nx_follicle still carry probe coefficients and their gates (8/8, 9/9, 16/16 GREEN against the grown lib) must re-derive their moved numbers when they read km_spectra_load; (5) a mutation bite of nx_skinspec_gate over nx_kubelka_lib aimed at km_spectra_load, deferred tonight because the graphics GR59 lane imports the same shared lib concurrently and a mutated shared source must not be compiled by another lane -- run it in a quiet window (nx_gate_bite nx_skinspec_gate nx_kubelka_lib.nx 1 aim km_spectra_load); the in-gate neg-controls (wrong pin, truncated table, non-monotone table, reversed spectrum) already bite on real fixtures 2026-10-01 retract CORRECTION TO THE COUNTS IN ROW 1790827400, which were read off a partial view of the per-subject lines rather than counted: on the promoted binary's run (/tmp/nx_skinspec_gate_final_1.out) nx_txtscan counts the epidermal thickness at the bar's 50 um floor (printed 49) for 85 of 100 subjects, the captured specular fitted to zero for 94 of 100 (not 97), and the carotene absorption at its declared span (19.27 per cm) for 8 of 100 (not 4; the four named are among them). The conclusion stands -- three places where a bound decides rather than the data -- with the counts now measured by the instrument instead of recalled 2026-10-01 land SHIPPED AND CLEANED 2026-09-30: nx_skinspec_gate PROMOTED 5113a088 (140963 B), _offc installed, registered; nx_skinspectra_gate PROMOTED 3a30f194 (54541 B), _offc installed, registered; nx_kubelka_gate re-PROMOTED 941a87a9 against the grown lib (identical content, 37/37). Sources: nx_kubelka_lib.nx b0aa070d (29088 B), nx_skinspec_lib.nx 2a287f57 (27351 B), nx_skinspec_gate.nx 2feee31b (37188 B). Evidence banked out of /tmp: knowledge/status/skinspec_gate_st6_final_20260930.out (a1749fea, the promoted binary's 18/18 run) and knowledge/status/skinspec_gate_st6_ablation_20260930.out (the ablation run, 18/18). Both .gates rows added (knowledge/compare/skintwin.gates). Seven scratch registrations reaped (the _st6_staged and _final_staged rows) and their six staged files retired reversibly to knowledge/retired; the dev loader file retired the same way. Refs pinned this lane: nist_skin_reflectance_data_202609, nist_jres_122_026_paper, cie_cmf_1931_2deg_202609, cie_d65_202609, spectral_sss_2606_27604, jacques_skinoptics_1998, krishnaswamy_baranoski_2004, omlc_betacarotene_202609. nx_compare_rank skintwin now reads ST6 done (done 4, open 6) 2026-10-01 measure MUTATION BITE ATTEMPTED ON THE SAFE SUBJECT, UNPROVEN: nx_gate_bite nx_skinspec_gate over nx_skinspec_lib.nx (this lane's own file, no other importer, so no other lane could compile a mutant) aim ssp_span returned verdict=UNPROVEN: the build governor refused the mutant build (exit 6) and the baseline gate run hit the harness's 120000 ms deadline (gate_exit 124), because the full 100-spectrum measurement takes about five minutes on the loaded box. Nothing was compiled or deployed: source 2a287f57 and live 5113a088 re-hashed unchanged afterwards. Standing evidence that the measurement teeth fire: the same tooth set read RED on three earlier model states this evening (12 of 16, 11 of 16, 16 of 17) before the optimiser and carotene closed them, and the in-gate neg-controls bite on real fixtures 2026-10-01 queue A BITE MODE FOR THE MEASUREMENT GATE: nx_skinspec_gate needs an argv mode that keeps every loader tooth and fits a declared handful of named subjects (never presented as the measurement, which stays all 100) so a mutant run fits inside nx_gate_bite's deadline; then bite twice in a quiet window -- over nx_skinspec_lib.nx aim ssp_fit and over nx_kubelka_lib.nx aim km_spectra_load once no other lane is compiling the shared lib. Also owed: wire nx_skinspec_gate and nx_skinspectra_gate onto a roster beat sized for a five-minute gate 2026-10-01 measure BACKFILL-HUMAN (f) the frontier subsurface-rendering bar: full-text witnesses now THREE groups by 2026 affiliation: Meta Reality Labs Research (Aliaga and Jarabo 2606.27604, existing full-text row aliaga_spectralsss_2606; affiliation read from the arXiv HTML knowledge/fetched/cmp_skintwin_bfh_aliaga_2606_27604.html sha256 99f6b10b); TUM with LMU Munich (DIAMOND-SSS 2601.12020, refs st6_diamondsss_2601_ft, mirror knowledge/fetched/cmp_skintwin_bfh_diamondsss_2601_12020.html sha256 8418aee7, full text read: diffusion augmentation from under 7 percent of a dataset replaces up to 95 percent of captures, up to 90 percent less real data than SSS-3DGS, its own caveat that augmented samples are not strictly physically accurate); KAIST with POSTECH and Universidad de Zaragoza I3A (polarimetric BSSRDF 2501.01980, refs st6_polarbssrdf_2501_ft, mirror knowledge/fetched/cmp_skintwin_bfh_polarbssrdf_2501_01980.html sha256 7e2a094e, full text read: per-texel 3x3 Mueller-matrix BSSRDF, biophysical hemoglobin, eumelanin and pheomelanin layers, 11 individuals of diverse skin tones). Independent count 3, with a named caveat: Jarabo's former lab is Zaragoza I3A, so the lineage overlaps the polarimetric paper's Zaragoza co-authors; held strictly by lineage the count is 2. The 42 MB arXiv PDFs are mirrored too (cmp_skintwin_bfh_diamondsss_2601_12020.pdf sha256 cbd6c8ef, cmp_skintwin_bfh_polarbssrdf_2501_01980.pdf sha256 cb3d1fe8). NIST: the critic's landing-page note is STALE: the spectra DATA file is mirrored by this board's own row nist_skin_reflectance_data_202609 (knowledge/fetched/cmp_skintwin_nist_skinrefl_v3.txt, 2,119,696 B, nx_filehash ce9aac73 equals its pin) 2026-10-01 measure ADVERSARIAL VERIFIER, METHOD CLASS 3 (independent re-derivation by hand, a different method than the gate): read the pinned NIST table directly by byte offset (layout measured: 3 nm rows 250 to 2500, 2814 B per 3-digit row, 4 columns per subject R1 R2 R3 Average) and the 1 nm CIE 1931 2 deg and D65 tables, and summed S times cmf times R at a 9 nm subsample (45 points, 379 to 775 nm, every third point of the gate's own 133-point colour window), normalised to the D65 white. (a) WHITE POINT recomputed: Xn 95.045, Zn 108.928 (Y=100) against the published 1 nm D65 2 deg 95.047 and 108.883, deltas 0.002 and 0.045 (0.04 percent, consistent with the subsample and the 775 nm cutoff). (b) MEASURED-SIDE COLOUR by hand: subject 53 X 40.65 Y 39.38 Z 30.57, Lab 69.02 10.24 15.65; subject 54 X 41.15 Y 40.10 Z 32.35, Lab 69.54 9.53 14.04 (each average column checked against the mean of its three replicates at every sampled row). (c) COMPARISON: the banked gate output knowledge/status/skinspec_gate_st6_final_20260930.out does NOT print measured-side XYZ or Lab per subject (only fitted parameters, de2000 and de76 of model against measured, rms, relerr and r2500), so the measured-side Lab CANNOT be compared and is UNVERIFIED by this class, not refuted; the de2000 and de76 values need the model spectrum and are outside this class. The one measured-side per-subject value it prints, r2500_micro, was re-read from the raw 2500 nm row for subjects 1 to 54 and ALL 54 agree (gate prints the table value times 1e6 minus 1, a float floor, e.g. 0.0392 to 39199 for subject 53). VERDICT PARTIAL: loader reading and white point reproduced; per-subject measured Lab is a value the gate does not emit. OWED: nx_skinspec_gate to gv_kv the measured-side Lab per subject (or for named subjects) so a hand re-derivation like this one can disagree with it; expected values for 53 and 54 are above, within the 9 nm subsample bound 2026-10-01 land BACKFILL-HUMAN refs rows landed this lane in skintwin.refs, each mirror pinned by nx_filehash, cited here so the refs referee's inline-citation tooth sees them: [st6_diamondsss_2601_ft] [st6_polarbssrdf_2501_ft] 2026-10-01 measure HONEST NAMES, SOURCE ONLY (2026-10-01): the three T10 teeth of nx_skinspec_gate read in source (37195 B, mtime 1790829963) median-de76-measured-skin-fit-beside-paper-mc-number-3-19-cross-test, worst-de76-measured-skin-fit-beside-paper-mc-fitting-ceiling-3-79-cross-test and median-relerr-measured-skin-fit-beside-paper-mc-number-3-7-percent-cross-test, values unchanged; but the PROMOTED binary (nx_skinspec_gate.elf 140963 B, the 5113a088 of row 1790827530) still carries the old names: nx_txtscan finds median-de76-over-every-nist-spectrum-at-or-below-the-bar-paper-model-3-19 and its two siblings at offsets 138148, 138221, 138296 and counts cross-test 0 times. The rename is NOT live until the gate is rebuilt and promoted; until then every published run of the served gate still presents the cross-test as passing the paper's bar 2026-10-01 measure MIRRORED AND READ for the next rungs (2026-10-01, all through nx_research_fetch, pinned on skintwin.refs): bilirubin [st6_omlc_bilirubin_202610] is PhotochemCAD 119 measured IN CHLOROFORM, 55000 cm-1/M at 450.8 nm, 1 nm grid -- usable through km_spectra_load with a declared solvent deviation; water [st6_omlc_water_hale73_202610] is Hale and Querry 1973 at 25 nm steps (0.00058 per cm at 400 nm, 0.0060 at 700, 0.026 at 750), coarse but the field's standard; the adding-doubling known answers [st6_prahl_ad_example_202610] are Prahl's published supplement to his 1995 chapter section 5.3.5 (quadrature, redistribution matrices, R and T of dtau 0.25, 0.5 and 1.0 layers, boundary R01 and T01 summing to 1, with errata) -- a real slab table to gate an adding-doubling arm against; but the chapter PDF that states the example's albedo and anisotropy is an image scan: nx_pdf_text emitted 8829715 B in which the word doubling occurs 0 times, so those inputs are UNREAD. The second calibration set [st6_openox_physionet_202610] is the OpenOximetry Repository on PhysioNet, credentialed: registered users who sign a data use agreement, PhysioNet Restricted Health Data License 1.5.0; its v1.0.0 landing page lists oximetry, blood gas and PPG variables and no reflectance spectra 2026-10-01 queue OPEN AT HAND-BACK, no build started (the lane passed its build cutoff 1790841600 before this work began: NAS clock read 1790856331): (1) REBUILD AND PROMOTE nx_skinspec_gate so the cross-test names go live (nx_sov_build_run nx_skinspec_gate --build-only, nx_contentdiff live-first and read the LOST lines: the three old tooth names are the expected loss, then /api/promote with expect_sha256, nx_offc_install); (2) TRANSPORT AT THE PAPER'S LEVEL: an adding-doubling arm in NishiLang Q30 (Prahl 1995: Gaussian quadrature, Henyey-Greenstein redistribution with g = 0.62 + 0.00029 nm, doubling from a thin layer, adding the index-matched boundary) as nx_addoubling_lib, gated first against the four-point quadrature, redistribution and boundary values of the published supplement, which needs the section 5.3.5 albedo and anisotropy from a TEXT source (the CWeb listing ad_src.pdf or the iad repository's test inputs) before any tooth is written; then all 100 NIST spectra fitted with it beside the Kubelka-Munk fit; (3) bilirubin and water into the dermis through km_spectra_load, reporting the 400 to 450 nm residual (plus 0.018 of record) before and after; (4) SECOND CALIBRATION SET BLOCKED on an operator decision: the OpenOximetry data needs a PhysioNet account and a signed restricted-data use agreement, which a seat may not sign; even then its licence is restricted, so it can be a private calibration set and never a published mirror, and whether v1.0.0 holds reflectance spectra at all is unmeasured; (5) IDENTIFIABILITY under the Kubelka-Munk transport is already measured (row 1790827420: thickness at the 50 um floor for 85 of 100, specular at zero for 94, carotene at its span for 8); the same counts under the new transport wait on (2) 2026-10-01 lesson A RENAMED TOOTH IN SOURCE IS NOT A RENAMED TOOTH: the honest names landed in nx_skinspec_gate.nx on 2026-09-30 and were never built, so the served gate kept publishing the cross-test as at-or-below-the-bar for a day while the source read correctly; only a scan of the PROMOTED binary's strings (nx_txtscan count cross-test = 0 on the live elf) told the two apart. A naming correction is shipped only when the live artifact carries it 2026-10-03 queue OpenOximetry adjudicated against the operator licence rule (no royalty, no obligation on the product path; restricted data used only as math): the raw data IS present (openox-repo 1.0.0 on PhysioNet, spectrophotometer.csv with per-site quantitative skin colour, Monk scale, de-identified) but it is under the PhysioNet Restricted Health Data License 1.5.0 with a signed DUA, whose own text (mirrored knowledge/fetched/cmp_skintwin_physionet_restricted_license_150.html) limits use to the sole purpose of lawful scientific research and no other, forbids sharing access, and obliges contributing code associated with publications arising from the data to an open repository. Calibrating a proprietary product model is outside sole-purpose research, and publishing a number derived from it would trigger the open-code obligation against the proprietary-code policy. VERDICT: not acquired for calibration or publication; the operator may still choose a research-only private validation, which publishes nothing. Second calibration set to be sought from public-domain or permissively licensed measured skin spectra instead. 2026-10-03 land HONEST NAMES LIVE: nx_skinspec_gate rebuilt from the 2026-09-30 source and promoted ca84de7f (162696 B, prev 5113a088 banked as .prev) under allow_capability_loss after nx_contentdiff live-first named all 7 lost runs: the four old at-or-below-the-bar tooth names and the old verdict note (the rename itself) plus two sys_mmap FATAL strings the shared nx_syscalls lib reworded to its owned-arena wording; _offc installed; the served gate re-ran GREEN 15/15 (KM model of record unchanged: median de76 1.030, worst 2.824 at subject 53, median relerr 3.12 percent). The measured-side Lab per subject the verifier asked for was already printed on every subject line (meas_L/a/b_milli): subject 53 reads 69.020 10.253 15.638 against the hand re-derivation 69.02 10.24 15.65, subject 54 reads 69.545 9.544 14.040 against 69.54 9.53 14.04 -- agreement inside the 9 nm subsample bound, so method class 3 now COMPARES rather than abstains. ADDING-DOUBLING ARM LANDED: buildroot/runtime/nx_addouble_lib.nx (Prahl 1995 section 5.3 in Q30: Gauss on [0,nu_c] plus Radau on [nu_c,1] found by sign-change scan and bisection, delta-M Henyey-Greenstein over N points, the diamond start eqs 5.54-5.56, doubling, a general four-matrix adder, Fresnel boundary layers through km_fresnel; nx_kubelka_lib untouched) and its gate nx_addouble_gate PROMOTED 5db42ea2 (114502 B), registered, _offc installed, served run 20/20 GREEN. Its known answers are PARSED FROM THE PINNED SUPPLEMENT BYTES [st6_prahl_ad_example_202610] after km_pin_match, not typed: every quadrature node and weight, every entry of h-- and h+-, A and B, R and T at dtau 0.25, 0.5 and 1.0 with fluxes, the four boundary Fresnel values and the slab between two boundaries (eqs 5.73, 5.74) agree within 6 micro (5 = half the fifth printed decimal, 1 = Q30 allowance); worst observed miss 5 micro. The example's inputs are not printed and the chapter scan has no text layer, so they were DERIVED from the table and re-proven by it: n = 1.5 (Gauss nodes 0.15751, 0.58784 = nu_c (1 -+ 1/sqrt3)/2), g = 0.9 (h(1,1) = 1 + 3 chi1 + 5 chi2 + 7 chi3 = 6.84908), a = 0.9 (scaled albedo 0.7558), dtau 0.25. 2026-10-03 lesson A TRUNCATING MULTIPLY IS A BIAS THAT DOUBLING SQUARES: the first nx_addouble_gate run passed every published entry but read a non-scattering slab 4 micro off exp(-tau/nu) after 12 doublings and a 2^-10-start reference 330 micro off whatever the trial start, because fq_mul truncates its low partial and each doubling squares the carried relative error (2^(k+1)-1 units). The fix was in the arithmetic, not the bounds: adq_mulr rounds that one partial to nearest in the matrix kernels, after which the successive start-halving changes read 6, 0 and 4 micro and conservation 1 micro against a derived 15 micro bound. A deep-start reference is not a reference when the deepening itself adds arithmetic; successive differences are. 2026-10-03 measure SECOND CALIBRATION SET FOUND BY THE ESTATE'S OWN HUNT, 2026-10-03: nx_frontier_scan hunt skintwin reflectance (subject skin reflectance spectral, 10 hunt rows in knowledge/research_sources.conf: Europe PMC, Zenodo, DataCite, figshare, OSF, Dryad, PhysioNet, Wikimedia Commons, Crossref for PeerJ, data.gov; all 10 reachable on the second run) read 1102 records, admitted 198 on relevance (897 irrelevant, 7 unpaired, partition sums) and classed every admitted hit from its own record's licence: ACQUIRE 114, CALIBRATE-ONLY 1, ACQUIRE-PRIVATE-REFERENCE 17, UNKNOWN-NEEDS-LICENCE-READ 65, REFUSED 1 (worklist knowledge/status/hunt_skintwin_reflectance.worklist). Top permitted human-skin set: the International Skin Spectra Archive, ISSA (Lu, Xiao, Pointer and co-authors 2025, figshare 10.6084/m9.figshare.28228571.v4) [st6_issa_lu2025_202610], CC BY 4.0 on two independent records (the figshare article record and the DataCite rightsList) [st6_cc_by_40_legalcode_202610], mirrored at 9320520 B, equal to the size both records declare, store class cc-by, attribution owed in a NOTICE. NOT YET READ: the xlsx has not been parsed, so its population, wavelength grid and instrument are UNMEASURED and the fit against it is owed. The NIST Reference Data Set of Human Skin Reflectance read UNKNOWN (its rights text is NIST Licensing, no licence the classifier names): a licence read is owed, not an acquisition. The colour hunt (skin color colorimetry) reproduced the OpenOximetry precedent mechanically: all four openox-repo versions REFUSED, class RESTRICTED-RESEARCH-ONLY, pattern restricted health data license, read from PhysioNet's own license.name, beside ENCoDE (Credentialed licence) also REFUSED (worklist knowledge/status/hunt_skintwin_colour.worklist) 2026-10-03 queue ISSA is acquired and unread. Next: parse knowledge/fetched/cmp_skintwin_issa_v4_ISSA_17_Jan_2025_Yan_Lu.xlsx with the estate's xlsx reader (nx_xlsx), publish its population, sites, wavelength grid and instrument, and fit the spectral skin model to it as the second calibration set beside NIST's 100 spectra, with the attribution line in the release NOTICE. Read the NIST licensing page before treating the NIST set as anything but UNKNOWN 2026-10-03 retract A FITTED DERMAL WATER FRACTION, landed in source an hour earlier and never promoted, is withdrawn: the first full Kubelka refit with nine free parameters (banked knowledge/status/skinspec_gate_st6_9param_km_partial_20261003.out, all 100 subject lines) showed water displacing the higher Jacques bloodless baseline acting as a baseline-absorption reducer that ran toward its upper bound (subject 98: blood 0, water 1.000), not a water measurement; and the flat nine-parameter search LOST subject 98's earlier optimum (Delta E76 1.059 under the seven-parameter model of record, 6.281 under the nine) although that optimum lies inside the larger space -- the search, not the model, failed. Replaced by: water HELD at W = 0.65, Jacques' generic skin dermis [st6_jacques_generic_optics_2015], as the bar fixes basal water (2606.27604 text: we fix basal concentrations for bilirubin, beta-carotene and water); bilirubin kept FITTED like carotene (a declared deviation: the bar fixes a basal value it does not state); and a NESTED fit in nx_skinspec_lib ssp_fit (the model without bilirubin fitted first, then a compass polish over all parameters that accepts only improvements), so adding an absorber can never return a worse fit than the model without it. 2026-10-03 lesson OPENOXIMETRY AT THE NEW IDEA LEVEL (operator 2026-10-03: "we also have idea level that is just used for idea but doesnt require attribution"; researchreach RR26 log 1791045000): the openox-repo DATA stays REFUSED even at idea level, because a contract binds it -- PhysioNet grants access only on signing the Restricted Health Data License, whose data use agreement limits use to the sole purpose of lawful scientific research and no other, with code contributed back, so even reading the data for ideas happens under that contract. nx_licclass_lib lc_use_decide now says so mechanically (IDEA-ONLY on the licence name and on the mirrored licence text both REFUSE, reason CONTRACT-BINDS-USE, clause named; gate 64/64 GREEN, 8e70fe86). The PUBLISHED OpenOximetry paper is a different source: its findings, methods and reported numbers are idea-level ALLOWED -- read, cited in a refs row as provenance-not-attribution, and used to shape our own spectral and oximetry models, with no text, figure, code or data file copied. Applying it here: the paper's reported skin-tone bias findings may inform the ST6 bar and the error budget; the repository's measurements may not enter any fit 2026-10-03 retract log 1791045002 described the OpenOximetry paper's findings (skin-tone bias may inform the error budget) from general knowledge; the paper was not fetched or read. Claim withdrawn until the paper is mirrored, pinned as a refs row flagged provenance-not-attribution, and its findings are read from the bytes. Repository data stays REFUSED (data use agreement binds). 2026-10-03 measure THE ADDING-DOUBLING REFIT IS 35 TIMES THE TWO-FLUX COST AND WAS MADE PARALLEL RATHER THAN SAMPLED: one adding-doubling subject took 5.3 minutes serially (127 wavelengths per evaluation, each an epidermal slab of about 5 doublings plus a semi-infinite dermis of about 12, two adds and two diamond starts, over roughly 2000 nested-fit evaluations), so 100 subjects would have taken about 9 hours on one core of an 8-core host. nx_skinspec_gate sg_fit_all now forks workers by the estate's pool governor (nx_poolgov pg_width: the CPUs this process may run on, squeezed proportionally above a load equal to them), each worker fits a stripe of subjects into rows of one sys_mmap_shared region, and the parent reaps every worker and assembles and prints in subject order (the nx_parexec pattern: disjoint writes, no locks). The Kubelka arm then took about 80 s on 8 workers against about 5 minutes serially, and nx_srcdiff of the serial run's banked output (knowledge/status/skinspec_gate_st6_km_fixedwater_ad1_20261003.out) against the parallel run's matched every one of the 178 shared lines: all 100 Kubelka subject lines, every tooth and every value are byte-identical; the only differences are the new workers line and the two adding-doubling subjects the serial run had reached before it was stopped (subject 1 Delta E76 0.931, n8-vs-n4 0.248; subject 2 0.387, 0.244), which the parallel run's adding-doubling arm will be checked against. 2026-10-03 measure RESUMED 2026-10-03 (second seat on the lane): the served nx_addouble_gate (5db42ea2, 114502 B) re-ran 20/20 GREEN through nx_job_run (every published quadrature, redistribution, slab and boundary entry within 6 micro, worst boundary miss 5 micro, conservation 1 micro, exp(-tau/nu) 3 micro, start-halving 6/0/4 micro, g=0.8 neg-control bites); it was REGISTERED-UNAUTHORISED, now authorised by a minted cap (nonce 1791058700, rows nx_addouble_gate, nx_skinspec_gate, nx_skinspec_gate_staged). The staged nx_skinspec_gate (212120 B, built 1791045564, newer than every source in its closure: gate 1791045530, skinspec_lib 1791044398, addouble_lib 1791043419) against the live ca84de7f: nx_contentdiff live-first loses exactly 2 runs, both named: the subject-line prefix ' nist subject=' (now printed per transport tag, nist or ad) and the old verdict note (reworded for two transports); gains bilirubin, water, workers and n8-vs-n4 fields. A full run of the staged gate is in flight (/tmp/skinspec_st6_full_1003b.out): the Kubelka arm finished all 100 subjects in about 90 s on 8 workers; the adding-doubling refit and its 8-point check are running. 2026-10-03 land ISSA READ AND CONVERTED TO A NATIVE RECORD, 2026-10-03: the pinned workbook [st6_issa_lu2025_202610] (9320520 B, km_pin_match against the refs pin = 1) holds two sheets, Coding Scheme (sheet1, 21346 B of XML) and ISSA (sheet2, 42884831 B); the served nx_xlsx read only sheet1, so the datasheet was read through nx_xlsx_sheet_lib (any sheet by name, sized from the package, cells on the workbook's own row and column numbers, 118 shared strings, 879073 cells, 0 sanitized). New lib nx_skinspec_issa_lib parses it with the grid and the CIELAB columns READ FROM THE HEADER ROW (360 to 780 nm every 10, 43 wavelengths; L, a, b* by header text) and classifies every row: 15256 candidates = 15256 ok + 0 badnum + 0 badtext + 0 badgrid + 0 gap + 0 extra + 0 nolab. Population measured from the file: 15256 records (the archive's own figshare record declares 15,256), highest subject number 2113 (declared 2,113 subjects) but only 2107 distinct subject numbers occur; 11 origins (1501, 1805, 1326, 4112, 1159, 81, 1494, 941, 1062, 812, 963); 12 body sites (back of hand 1716, cheek 2511, cheek bone 232, chin 2072, ear lobe 480, forehead 2092, inner arm 1692, neck 1162, nose tip 646, outer arm 1632, palm 778, ring finger 243); ethnicity CA 2512, CN 2935, IQ 1326, TH 4112, SA 1405, AF 941, JP 1062, AB 963; gender F 9382, M 5856, empty 9, 'M ' (trailing space) 9 -- kept as written; age group unrecorded on 2716; instrument SP (spectrophotometer) on all 15256, SPIN SCI (specular included) on all 15256; TWO grids: 360-740 nm (5297 records) and 400-700 nm (9959). Instrument geometry beyond SP/SCI and the archive's diffuse-illumination protocol is NOT recorded per record in the file. Record written to knowledge/calibration/skintwin_issa_v4.issa (4590250 B; reflectance factor x1e6, CIELAB x1e6, attribution and pin in its header) and READ BACK identical field by field (0 mismatches); NOTICE beside it (knowledge/calibration/NOTICE: creators, citation, CC BY 4.0, original file and hash, the changes made). The workbook is a calibration input only, never a runtime dependency. Gate nx_skinspec_issa_gate 25/25 GREEN (pin bite, population, hand-read known answers from the raw sheet XML of rows 13 and 15268, round trip, five named refusals on a fixture datasheet, number-parser known answers, a short-row refusal bite). 2026-10-03 measure THE SPECTRAL MODEL OF RECORD FITTED TO EVERY ISSA RECORD (15256 of 15256, no sampling) BESIDE NIST, 2026-10-03 [st6_issa_lu2025_202610]: one binary (nx_skinspec_issa, staged e1658cd9, the nx_skinspec_lib ssp_fit nested fit on the Kubelka arm, the model built on each record's OWN grid -- 380 to 740 nm for the 5297 records measured 360 to 740, 400 to 700 nm for the 9959 measured 400 to 700; fit window 400 nm to the record's end) fitted NIST's 100 subjects and every ISSA record; fit rows in knowledge/calibration/skintwin_st6_fit_.fit. METHOD CLASS 1, REPRODUCTION: the NIST run reproduces the incumbent nx_skinspec_gate's banked Kubelka output subject by subject (subject 1 Delta E2000 0.541, Delta E76 0.611, RMS 8063 micro, rel 14294 ppm; subjects 2-4 Delta E76 0.469, 0.351, 0.113 -- identical). NIST free: Delta E76 median 0.605, worst 2.054; Delta E2000 median 0.528; mean relative error median 1.63 percent. ISSA free: Delta E76 median 1.082, worst 17.496; Delta E2000 median 0.880; relative error median 2.62 percent; partition by body site sums 15256: back of hand 0.960 (n 1716), cheek 1.171 (2511), cheek bone 0.851 (232), chin 0.805 (2072), ear lobe 2.370 (480), forehead 0.901 (2092), inner arm 1.525 (1692), neck 0.854 (1162), nose tip 1.330 (646), outer arm 1.116 (1632), palm 1.390 (778), ring finger 1.125 (243); by origin 1.683, 1.727, 1.932, 1.017, 0.922, 1.101, 1.153, 0.627, 0.555, 0.418, 0.598 (origins 1-11). Beside the bar paper's own numbers (2606.27604: Delta E76 3.19 model, 3.79 ceiling, 3.7 percent relative) as a cross-test only: the paper fits its own Monte Carlo medium, so no pass or fail is drawn. Beside what the ISSA paper itself reports [st6_issa_scidata2025_202610]: it reports NO model-fit error; it reports the measurement's own variability -- short-term repeatability within 0.40 Delta Eab, pressure differences 0.88 and 1.82, aperture differences 2.48 and 3.04 -- so the ISSA median model error (1.08) sits above repeatability and below the aperture spread, and the ear lobe (2.37) and inner-arm (1.53) sites are where the model misses most. METHOD CLASS 2, A KNOWN ANSWER RE-DERIVED: our CIELAB (D65, CIE 1931 2 degree, the pinned tables) of each measured ISSA spectrum against the archive's OWN published Lab: on the 400-700 grid Delta E76 median 0.000, worst 0.001 over 9959 records (the same arithmetic on the same band); on the 360-740 grid median 0.098, worst 0.118 over 5297 (our colour grid opens at 380 nm because the pinned haemoglobin table does, so the archive's 360 and 370 nm samples are not summed -- a declared deviation, not a disagreement). METHOD CLASS 3, HELD OUT: the three transferable parameters (scattering exponent u, captured specular, epidermal thickness) were HELD at one set's medians and only the chromophores refitted on the other. NIST medians (u -0.0840, specular 0.0278, thickness 50 um) held on ISSA: Delta E76 median 1.849; ISSA's own medians (u -0.1738, specular 0.0278, 50 um) held on ISSA: 1.902 -- NIST's medians transfer to ISSA no worse than ISSA's own. ISSA medians held on NIST: 0.836; NIST's own medians held on NIST: 0.797 -- a cross-set penalty of 0.039. The cost is freezing (ISSA 1.08 to about 1.9, NIST 0.61 to about 0.8), not the change of archive. Both sets put the median epidermal thickness on its 50 um lower bound and the median specular on the 8 degree Fresnel bound: neither is measured by these fits. Gates: nx_skinspec_issa_gate 25/25 GREEN on the NAS (promoted c55278d0, registered). 2026-10-03 queue ISSA LANE OPEN ITEMS: (1) nx_skinspec_issa is staged (e1658cd9) and registered as nx_skinspec_issa_staged; its ship (nx_organ_ship, gate nx_skinspec_issa_gate) was refused by build admission (load1 17.0 against 16.43) and enqueued as BQ-nx_skinspec_issa -- promote and register under its own name when it fires. (2) The adding-doubling arm has not been run on ISSA (about 35 times the Kubelka cost: roughly 9 core-hours for 15256 records). (3) A colour grid that reaches 360 nm needs a haemoglobin table pinned below 380 nm; until then the 360-740 records' CIELAB omits the archive's 360 and 370 nm samples. (4) The thickness and specular medians sit on their bounds on both sets: the held-out test therefore transfers bounds, not measurements, and a wider thickness span or a measured specular capture is owed before those parameters can be called calibrated. (5) nx_skinspec_issa_gate has not been mutation-bitten with nx_gate_bite (its two named bites are hand-planted). 2026-10-03 land PROMOTED nx_skinspec_gate 0f2a1cb7 (212120 B, prev ca84de7f banked as .prev) carrying the honest cross-test names, bilirubin and water in the dermis, the adding-doubling arm and the per-subject measured-side CIELAB; allow_capability_loss passed only after nx_contentdiff live-first named both lost runs (the per-transport subject-line prefix and the reworded verdict note). _offc installed and hash-identical to live; the scratch row nx_skinspec_gate_staged reaped as superseded (toolreg f566282c). The full run that justified the promote ran the SAME bytes (sha 0f2a1cb7 = the staged artifact, /tmp/skinspec_st6_full_1003b.out, 78794 B): 20/20 GREEN, so the served gate's verdict is that run by hash identity; it was not re-run (3 h 16 min of wall clock on an 8-core box at load 15 to 21). 2026-10-03 measure ALL 100 NIST SPECTRA UNDER BOTH TRANSPORTS (no sampling; water held at W 0.65, bilirubin and carotene fitted, nested fit). KUBELKA arm: median Delta E76 0.605, worst 2.054 (subject 44), median Delta E2000 0.528, worst 1.558, median relative error 1.63 percent, median RMS 0.0078 -- against the model of record before bilirubin and water (median 1.030, worst 2.824) on the same spectra. ADDING-DOUBLING arm (4-point quadrature, Prahl 1995): median Delta E76 0.752, worst 2.201 (subject 95), median Delta E2000 0.616, worst 2.054, median relative error 1.86 percent; its 8-point recompute at the fitted parameters moves the colour by median 0.253 and worst 0.320 Delta E76. PARAMETERS ON A BOUND (a value on its bound was clipped, not measured), Kubelka / adding-doubling: epidermal thickness on its 50 um floor 63 / 83; bilirubin at zero 85 / 85; carotene at zero 74 / 78; specular at zero 22 / 9 and at the Fresnel ceiling 57 / 70; melanin, blood, SO2 and scattering scale on no bound under either arm. The arms disagree on scattering magnitude: median fitted scale 0.943 under Kubelka, 2.551 under adding-doubling. Bar comparison is a CROSS-TEST only: the paper's 3.19 / 3.79 / 3.7 percent are its fit of its own Monte Carlo medium, printed beside ours, no pass drawn. Three method classes: (1) gate KATs re-run (OMLC haemoglobin, NIST, CIE, beta-carotene, bilirubin 54889 at 450 nm, water 0.00058 and 0.0060 per cm) and the reproduced run; (2) the adding-doubling code held to Prahl's published supplement by nx_addouble_gate, re-run today 20/20 (every entry within 6 micro); (3) independent re-derivation: the hand CIELAB of subjects 53 and 54 (69.02/10.24/15.65, 69.54/9.53/14.04) agrees with the gate's measured-side values (69.020/10.253/15.638, 69.545/9.544/14.040) inside the 9 nm subsample bound. 2026-10-03 lesson A TRANSPORT UPGRADE THAT FITS WORSE IS NOT A REGRESSION UNTIL THE FREE PARAMETERS ARE ASKED WHAT THEY ABSORBED: the physically stricter adding-doubling arm lands at median Delta E76 0.752 against Kubelka's 0.605 on the same 100 spectra, and its 4-point quadrature error (median 0.253) is a third of that residual, so the n4 arm is not converged relative to the number it is judged on; meanwhile Kubelka's fitted scattering scale (0.943) silently absorbs the K,S-to-mua,musp mapping that adding-doubling cannot (2.551). The bound census says what neither arm can measure from a diffuse-reflectance spectrum alone: epidermal thickness (on its floor for 63 and 83 of 100), bilirubin (zero for 85 under both) and carotene (zero for 74 and 78) are clipped, not fitted. Projection lesson: the serial 5.3 min per subject promised ~66 min on 8 workers; at load 15 to 21 the refit took 3 h 16 min -- a wall-clock budget priced at an idle box is a floor. 2026-10-03 queue OPEN after this leg, each with its recipe: (1) ADDING-DOUBLING CONVERGENCE: refit with SG_AD_N = 8 (or the extrapolated n4/n8 pair) so the transport error sits well below the fit residual before the arm's number is compared with Kubelka's; cost is roughly 8x per evaluation, so run it when the box is idle or with a coarser fit grid declared. (2) IDENTIFIABILITY: thickness, bilirubin and carotene need a second observable (a second geometry, polarisation or a spectral window where they dominate) or must be held at a cited basal value with the deviation declared; until then they are reported as clipped. (3) OPENOXIMETRY is an OPERATOR DECISION, not a seat's: the openox-repo data sits behind PhysioNet's credentialed access and a signed Restricted Health Data License 1.5.0 data use agreement (sole purpose lawful scientific research, no sharing, code contributed back for publications); a seat may not sign it, and calibrating or publishing from it conflicts with the proprietary-code rule (log 1791040900, 1791045002). The operator may choose a research-only private validation that publishes nothing. (4) ISSA second calibration set (CC BY 4.0, mirrored, log 1791043440): the estate's xlsx reader reads only sheet1 with a fixed 8 MiB part buffer, and ISSA's sheet1 is its code legend (measured today through /api/organ_run nx_xlsx read: CODES sheet, columns A..L incl. start/end wavelength and interval); the spectra sheet is unreachable until nx_xlsx (buildroot/runtime/_hdl_build, office lane) reads a named sheet through sys_read_file -- owed there, then the fit beside NIST's 100. (5) The nx_addouble_gate _build fossil (131328 B, built 1791043420) differs from the promoted 5db42ea2 (114502 B) only by source-path build metadata (nx_contentdiff GREEN, lost 0): a lane difference, nothing to promote. 2026-10-03 measure ADVERSARIAL VERIFY of the 2026-10-03 promote (receipts method class): nx_skinspec_gate.elf and _offc/nx_skinspec_gate.elf both sha 0f2a1cb7 212120 B (identical); nx_addouble_gate.elf and _offc both 5db42ea2 114502 B (identical); nx_txtscan on the promoted skinspec elf counts at-or-below 0 and the-bar 0 (the only 'below' is the base-class envelope text, present in .prev too); nx_contentdiff .prev->live loses exactly 2 runs (the ' nist subject=' prefix and the old verdict note), gains 18 incl. bil_abs_milli and de76_n8_vs_n4_milli. nx_addouble_gate re-ran through nx_job_run: 20/20 GREEN, boundaries worst 5 micro, conservation 1, exp(-tau/nu) 3, start-halving 6/0/4, g=0.8 neg-control bite-proven. The skinspec full-run artifact (/tmp/skinspec_st6_full_1003b.out, 78794 B) reads 20/20 GREEN with every published figure (KM median/worst dE76 605/2054 subj 44, AD 752/2201 subj 95, n8-vs-n4 253/320, bounds 63/83 thickness, 85/85 bilirubin) matching the report; the run prints no subject sha, so 'same bytes' rests on the new-only fields it emits, not on a hash; the skinspec gate was NOT re-run here (3 h on a loaded box). REFUTED IN THE REPORT: queue row 1791070503 item (4) and the report's ISSA open item say the ISSA spectra sheet is unreachable, but rows 1791060340/1791064760 already read all 15256 ISSA records via nx_xlsx_sheet_lib and fitted them; that queue item is stale and the live ISSA items are those of row 1791064800. 2026-10-03 measure ADVERSARIAL VERIFY, independent re-derivation method class (no organ run, hand arithmetic from the pinned mirrors). (A) PRAHL SLAB TABLE vs A DIFFERENT TRANSPORT: the two starred-form slabs of the supplement (rf_116c14da.raw, a 0.9, g 0.9, n 1.5 internal, diffuse flux incidence) re-derived by similarity scaling (a' = a(1-g)/(1-ag) = 0.4737, tau' = (1-ag) tau = 0.19 tau) and the two-stream/Kubelka-Munk diffusion closure (K = 2(1-a'), S = a' per unit tau'; a_km 3.2222, b_km 3.0631): dtau 0.5 R 0.0391 against the table's flux 0.04611 (-15 percent), T 0.8658 against 0.86135 (+0.5 percent); dtau 1.0 R 0.0684 against 0.06961 (-1.7 percent), T 0.7508 against 0.75816 (-1.0 percent); both table R lie below the semi-infinite two-stream limit 0.159 and rise with thickness. Not refuted: the thin-slab R gap is the expected failure of a diffusion closure at tau' 0.095 (grazing single scatter with the forward-peaked g 0.709 that the 0.15751 node carries, R(0.157,0.157) 0.952), and it closes to 1.7 percent at tau' 0.19; this checks the table the gate is held to, it does not re-run the gate (its 20/20 is row 1791071100's). (B) MEASURED-SIDE CIELAB, subject 1 (NIST Average column), by hand: 23 samples 385-715 nm at 15 nm on the file's 3 nm grid, CIE 1931 2 deg and D65 read from their pinned mirrors, white summed on the same samples (Xn/Yn 0.9502, Zn/Yn 1.0797): X/Xn 0.38952, Y/Yn 0.36288, Z/Zn 0.27317 -> L 66.739 a 8.524 b 12.884, against the gate's meas 66.751 / 8.457 / 12.957: Delta E76 0.10, inside the 15 nm rectangle-rule error. Not refuted. Bound stated: subject 1 only (subjects 53 and 54 were the earlier hand check), and the 15 nm subsample cannot resolve differences below about 0.1-0.2 Delta E76. 2026-10-03 measure ADVERSARIAL VERIFY (cited-method class) of nx_addouble_lib against Prahl's own iad source, mirrored this hour: ad_phase.w (knowledge/fetched/rf_9640e416.raw sha 3bf674b1), ad_start.w (rf_992d3553.raw sha d4577c1b), ad_bound.w (rf_724cd5dd.raw sha 2cc75860). NOT REFUTED. Phase: delta-M HG chi_k=(g^k-g^M)/(1-g^M) for k<M=quad points, a=a(1-g^M)/(1-ag^M), tau=(1-ag^M)tau, h=sum (2k+1)chi_k P_k P_k -- the lib matches term for term. Quadrature: Prahl's Quadrature() is gauleg on [0,mu_c] + Radau on [mu_c,1] (Radau on [0,1] at n=1) -- matches (his prose says Radau for both halves; the code is Gauss, the lib follows the code). Diamond: r_ij=a d w_j h+-_ij/(4mu_i), t_ij=delta_ij d/(2mu_i)-a d w_j h++_ij/(4mu_i), G=(I+t-C r)/2, C=r(I+t)^-1, GR=C, G(T+I)=I -- the lib's A,B,G,R=2GB(I+A)^-1,T=2G-I is the same algebra without Wiscombe's twoaw-weighted transposed solve; equivalence is shown by every starred entry. Boundary: Boundary_RT is a diagonal Fresnel R, same both sides, T=1-R, no n^4 radiance factor (Prahl removed it) -- the lib matches; its R=1 on virtual (TIR) rows of the top boundary where Prahl uses 0 touches only the virtual rows, since T01 is 0 there. Doubling start: Prahl halves until d<=mu_min (mu_min/2 for an infinite layer); the lib halves to mu_min/4 (declared deviation 2); live gate measures shift0 (Prahl's rule) vs default at 6 micro. Gate re-run live just now: nx_addouble_gate 20/20 GREEN, mirror pin h6edc82d7 matches nx_filehash, every comparison parsed from the pinned table (96 slab entries + 32 boundary-block entries), worst miss 5 micro vs 6 tol, add==double 0 units. CAVEATS: (1) the example's inputs (n 1.5, g 0.9, a 0.9, dtau 0.25 UNSCALED, start dtau 0.10237 confirmed from A_33=0.04275) are DERIVED from the table, not read from a text source -- over-determined by ~150 entries, but still a derivation; (2) T11's tooth name claims each halving changes R less, the live values are 6 -> 0 -> 4 micro (not monotone; the tooth tests only e0>e2 and e0>e4), i.e. the residual is the Q30 rounding floor and the name overstates; (3) this verifies the transport only, not its use inside the skin fit. 2026-10-04 land THE ADDING-DOUBLING ARM MADE FAST WITHOUT CHANGING A BIT, AND THE FITTED SPANS MADE DATA, 2026-10-03 (third seat on the lane). (1) nx_addouble_lib gained the entry points a cache needs: adq_slab_r and adq_slab_inf_r (the slab under the redistribution the context already holds; adq_slab and adq_slab_inf are now these plus adq_redist, so both paths run the same arithmetic), adq_redist_save / adq_redist_load (copy, never pointer swap) and adq_add_rt (only the reflection seen from above, adq_add's own first half operation for operation). nx_addouble_gate T14 holds them: a slab and a semi-infinite layer rebuilt from a saved redistribution after another anisotropy overwrote the context equal the direct forms at 0 Q30 units, adq_add_rt equals adq_add's RT at 0 units, and loading the wrong redistribution changes the slab by 59368761 units (0.055) -- the first run of that bite read VACUOUS because the 'wrong' anisotropy 0.8 was the skin slab's own, so the overwrite had changed nothing; it now loads 0.9. Gate 22/22 GREEN, promoted aabe3d69 (contentdiff lost 0), _offc installed. (2) nx_skinspec_lib's adding-doubling bundle carries a LAYER MEMO: per grid point the redistribution once, and three most-recently-used epidermal layers, dermal layers and finished reflectances keyed by the exact Q30 (albedo, optical thickness) inputs, so a fit that moves one parameter recomputes only the layer it touches (melanin and thickness touch the epidermis, blood, SO2, carotene and bilirubin the dermis, the captured specular neither). nx_skinspec_gate T16 (new) holds the memo to the recomputed spectrum at five vectors with a melanin-change bite. (3) The thickness span and the specular ceiling are context slots (ssp_set_spans, ssp_set_tepi); bar mode reproduces the model of record BYTE FOR BYTE: nx_skinspec_issa nist under the new build wrote 102 lines against the incumbent fit file's 102, nx_srcdiff surplus 1 and 1 = the header line only (it now names start and spans), all 100 Kubelka fit rows identical. ssp_polish refines a warm start by the compass round alone. (4) nx_skinspec_issa: options ad <N>, warm <fitfile>, spans bar

ST7 Wetness as a surface state · cell PRESENT · gate NO-GATE-DECLARED

Contract	`km_wet_state` · executor Organ · 1.5 u · stands on nothing
Definition of done	Wet skin darkens and its specular sharpens; the operator named environment and body fluids as first-class states. It is a surface term, not a chromophore one, which is why it is its own rung. BUILT 2026-09-28 in runtime/nx_kubelka_lib.nx (km_fresnel, km_fresnel_hemi, km_film_internal, km_wet_lambert, km_film_specular, km_wet_state): the film-over-Lambertian series of Jensen 1999 Sec. 2.1 and 3.1 with Fresnel derived from the indices and the film's internal diffuse reflectance by Judd's identity, held by nx_kubelka_gate against the direct integral through total internal reflection, an exact energy identity (a white substrate loses exactly the film's diffuse reflectance) and a cross-identity that re-derives the published wet value from the published film reflectances. Declared deviations: Jensen's roughness exponent is not carried (the microfacet lane's), and the water-beneath-the-surface effect (Sec. 2.2) is a separate rung. Done-rule: every film value the gate publishes re-derives from two independent methods within the stop tolerance, and the wet ladder darkens strictly.
Matrix cell	Wetness as a surface state · `runtime/nx_kubelka_lib.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_kubelka_lib`; declare `nx_kubelka_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-28 land WETNESS AS A SURFACE STATE BUILT AND PROMOTED 2026-09-28: nx_kubelka_lib gained km_fresnel (exact unpolarised dielectric Fresnel with total internal reflection), km_fresnel_hemi (cosine-weighted hemispherical mean by midpoint quadrature whose node count doubles until successive estimates agree within the caller's tolerance, capped at the Q30 resolution), km_film_internal (Judd's identity r_i = 1 - (1 - r_e)/n^2), km_wet_lambert (the film-over-Lambertian series (1 - r_e)(1 - r_i) R / (1 - r_i R) with the film's own diffuse reflectance beside it), km_film_specular and km_wet_state (an area state in [0,1]); 9704 -> 18047 B. Primary sources read from mirrored bytes: Jensen 1999 Sec. 2.1, 2.2, 3.1, 3.2 [jensen_wet_1999] (the Stanford copy refuses our TLS 1.2 certificate pipeline, the Eurographics digital library copy mirrored) and Pan et al. 2025 [pan_wetporous_2025], the medium-based frontier deliberately not built. nx_kubelka_gate 23 -> 33 teeth (Fresnel KATs, the two-method film identity, the strict darkening ladder, the energy identity, the state endpoints, a cross-identity, neg-controls), sovereign build 82120 B sha256 e018eb64, behaveprobe other_rc=0 on the staged binary, contentdiff lost 1 (the rewritten verdict note) named, promoted under allow_capability_loss, _offc refreshed. Rung symbol repointed from the absent fo_wetness to km_wet_state and a matrix watch row added 2026-09-28 measure nx_kubelka_gate on the promoted binary 2026-09-28, 35/35 GREEN, values published by gv_kv: Fresnel at normal incidence into glass 39999 micro (the exact identity is 40000); water film (n = 1.33) external diffuse reflectance r_e = 65932 micro reached at 256 quadrature nodes; internal diffuse reflectance r_i = 471949 micro by Judd's identity against 471948 micro by the direct integral through total internal reflection, two methods one micro apart; the film darkens a 0.2 grey to 0.109, a 0.5 grey to 0.323 and a 0.8 grey to 0.634 (strictly, every rung); a white substrate under the film returns 999999 micro of wet-plus-specular, the energy identity to one micro of rounding; the wetness state blends 0.5 dry and 0.323 wet to 411393 micro at a half state; the published wet value re-derives from the published r_e and r_i by the series inside the gate. OUTSIDE ORACLE, named not claimed: the classical diffuse-reflectance tables for n = 1.33 (Judd 1942, Egan and Hilgeman) are the referee for r_e and r_i and are owed as a pinned ref before these values are called agreed 2026-09-28 measure nx_gate_bite nx_kubelka_gate over nx_kubelka_lib.nx 2026-09-28, aim km_fresnel: valid_mutants=1 killed=1 not_reached=0, verdict GREEN (a compiling comparison mutant went RED), twins restored byte-identical (18047 B pristine, custody twins_checked=2 restored=2). NAMED LIMIT: the first-kill landed at byte 4050, the comparison inside the older single-constant forward, so this bite proves the gate is non-vacuous and does NOT reach the new film block; the aimed reach into literal-free arithmetic is debt 1790543778. The film block's own proof is the two-method identity, the exact energy identity and the cross-identity, all published above 2026-09-28 measure OUTSIDE ORACLE READ 2026-09-28: Judd 1942 Table I (NBS RP1504, mirrored and pinned as judd_1942_diffuse_fresnel [judd_1942_diffuse_fresnel]) gives for m = 1.33 an external diffuse reflectance of 0.0660 and an internal one of 0.472, computed in 1942 by summing the Fresnel formula over 0.04 radian intervals with Moon's tables, uncertainty less than 2 in the last figure. The estate's km_fresnel_hemi reads 0.065932 and Judd's identity 0.471949: both inside the table's stated uncertainty (0.00007 and 0.00005 away), a THIRD method class agreeing with the two in the gate. The water-film values are now AGREED by an outside referee, not only self-consistent; the gate gains the table as two teeth in the next promote 2026-09-28 land the outside oracle is a TOOTH now (2026-09-28): nx_kubelka_gate carries Judd 1942 Table I (m = 1.33: external 0.0660, internal 0.472, uncertainty less than 2 in the last figure) as two teeth beside the two-method identity, each holding the estate's value inside the table's stated uncertainty; promoted 82712 B sha256 3f6dda67, live run 37/37 GREEN with judd_1942_external_micro=66000 and judd_1942_internal_micro=472000 published beside water_film_external 65932 and internal 471949. Three method classes now agree on the film: the estate's quadrature, Judd's identity through total internal reflection, and a 1942 summation over Moon's tables

ST8 Bite-prove every gate on this board · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`nx_gate_bite` · executor Organ · 0.5 u · stands on nothing
Definition of done	CLOSED 2026-09-03. All five gates mutation-proven BOTH directions on the laptop build farm while the NAS build lane was storm-refused, each with a byte-identical restore verified by sha256. The kills were PRECISE rather than shotgun, which is the part that matters: kubelka RED 9 of 10 killed by roundtrip_ok=1/9 (only the ks=0 rung survives, and it should, because the mutation is invisible there); vasc RED 7 of 9 killed by BOTH depth teeth including the anti-painting one, when the depth attenuation was removed -- that is the model literally turned into a painting and the tooth firing; follicle RED 15 of 16 killed by the TYNDALL tooth ALONE, with the blonde-versus-black tooth correctly surviving because those differ by their own coefficients regardless of depth; skinfit RED 5 of 8 killed by recovery, monotonicity and the honesty tooth together when the error metric was blinded; ws_client RED 6 of 9 killed by the mask bit, the INTEROP tooth and the extended-length round trip when the masking XOR was dropped. Every gate returned to full GREEN on restore.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `nx_gate_bite`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-09-03 measure Bite proof, five gates, both directions, sources restored byte-identical: kubelka cf17b1aa 10/10 GREEN after RED 9/10; vasc d1b6409d 9/9 after RED 7/9; follicle 1b15ffd3 16/16 after RED 15/16; skinfit f4f70936 8/8 after RED 5/8; ws_client 714e6610 9/9 after RED 6/9. 2026-09-03 lesson A KILL BY THE WRONG TOOTH PROVES NOTHING ABOUT THE RIGHT ONE, so each mutation was chosen to attack ONE named claim and the surviving teeth were read as carefully as the failing ones. follicle is the clearest case: removing the per-channel attenuation killed TYNDALL and left the blonde-versus-black tooth passing, which is correct and would have looked like a weak gate to anyone counting failures instead of reading them. 2026-09-03 measure MACHINE bite records now exist for FOUR of the five declared gates, replacing the hand proofs that C4 correctly refused to accept: nx_kubelka_gate killed=1 restore 29512 B, nx_vasc_gate killed=1 restore 3967 B, nx_skinfit_gate killed=1 restore 5417 B, nx_de_referee_gate killed=1 restore 5255 B. Every run reported mutant_live=0 and every restore was byte-identical. C4 reads declared=5 bite-proven=4 never-bitten=1 sum=5. 2026-09-03 measure nx_colorsci_cat_gate is the outstanding one and its cause is MEASURED rather than guessed: SIX consecutive UNPROVEN, every one BUILD-REFUSED-BY-ADMISSION on the SUBJECT rebuild while gate_build_exit=0 said the gate itself compiled fine. nx_colorsci_lib is the largest subject in the set at 38414 bytes, so its per-mutant rebuild is the heaviest job here and the first thing marginal headroom declines. The other four went through on the same lane in the same window. It needs a quieter box, not a different method. 2026-09-03 lesson THE LAPTOP FARM CANNOT SUBSTITUTE FOR THIS ONE, AND THE REASON IS WORTH RECORDING TWICE. The laptop nx_colorsci_lib is 29200 bytes and contains ZERO occurrences of cs_cat_build -- it predates the chromatic-adaptation work entirely. Even after converging it, a laptop bite produces no record C4 can read, so it would move the seat's confidence and not the board's evidence. The last bite genuinely requires the NAS lane. 2026-09-03 measure THE LAPTOP/NAS COLORSCI DIVERGENCE IS NOW MEASURED IN BOTH DIRECTIONS, WHICH IS THE HALF THAT WAS EXPENSIVE. Local nx_colorsci_lib carries 38 functions and EVERY ONE of them also exists on the NAS; the NAS additionally carries cs_prof_lab_to_lin, cs_prof_gamut_permil, cs_prof_gamut_clip and the five cs_cat_* adaptation functions. So the NAS is a STRICT SUPERSET and the laptop is purely behind -- there is no two-way fork here and adopting the NAS copy destroys nothing. That direction was checked rather than assumed, because this estate has had files diverge BOTH ways and 'the laptop is the stale one' is a guess until someone lists the symbols. 2026-09-03 lesson AND THE DIVERGENCE FAILS LOUD, NOT SILENT, WHICH IS WHY IT IS A FRICTION AND NOT A CORRECTNESS RISK. A local build touching colour REFUSED with 19 named errors pointing at CS_P_B64; it did not miscompile quietly. That is the difference between a hazard that costs a seat ten minutes and one that ships a wrong number. The convergence is owed and is a bulk file transfer, so it is recorded here with its direction settled rather than performed at the tail of a long session where a partial copy would be the worst outcome of the three. 2026-09-03 lesson THE MACHINE IS STRICTER THAN THE SEAT WAS. nx_gate_bite reports invalid_uncompilable and void_build_lane SEPARATELY from killed, so a mutant that failed to compile and a mutant lost to the build lane are never counted as evidence. The hand method picked one mutation already known to compile. On nx_vasc_gate the symbol-targeted operator independently landed inside va_seg_visible -- the same depth-attenuation line the hand proof attacked, reached by a different route. 2026-09-03 lesson THE LAPTOP FARM CAN BITE-PROVE WHEN THE NAS CANNOT. nx_gate_bite is a NAS organ and the estate was unreachable, but the proof it automates is plant-the-known-bad, rebuild, see RED, restore, see GREEN -- and that runs entirely on the local sovereign toolchain. An unreachable build lane blocks shipping, not proving.

ST0 Board-level journal anchor for skintwin · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`-` · executor Board · 0 u · stands on nothing
Definition of done	BOARD-LEVEL JOURNAL ANCHOR, DECLARED 2026-09-04. This board already journals 9 log rows against the id ST0, meaning the entry is about the BOARD rather than about one rung. The .plan schema says a log row's third field is a rung id, so until this row existed every one of those entries resolved to nothing and a session inheriting this board read a ledger pointing at an address that was never declared. Declaring the anchor is the additive fix: it changes no existing row, it makes the convention READABLE instead of implicit, and nx_board_idref_gate now resolves those entries instead of counting them as dangling. Journal board-wide findings here; journal rung work against its rung.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `-`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-09-03 measure Three absence probes ABSENT-PROVEN with corpus_complete=1 before any build: kubelka, cat02, bradford. nx_colorsci_lib had CIELAB and CIEDE2000 and no adaptation and no scattering model. 2026-09-03 land cs_cat_build and cs_cat_xyz landed in nx_colorsci_lib (32813 to 38414 bytes, single application verified). nx_colorsci_cat_gate 9 of 9 GREEN, promoted 40520 bytes, registered. 2026-09-03 land nx_kubelka_lib and nx_kubelka_gate 10 of 10 GREEN, promoted 34560 bytes, registered. K over S of 0.25 returns exactly 0.5; round trip worst error one part per million. 2026-09-03 land nx_skinfit_lib 8 of 8 GREEN on the laptop farm, source landed on the NAS, build queued behind an I O storm. 2026-09-03 land nx_vasc_lib 9 of 9 GREEN. Red-minus-green 89090 sparse to 196295 dense; the same vessel buried deep returns 70497 against a 70497 baseline. 2026-09-03 land nx_follicle_lib 16 of 16 GREEN. Shaved ostium 3599 against waxed 0; blunt tip 17999 against tapered 4499; wax erythema moves rendered red-minus-green 272337 to 97278 over six days. 2026-09-03 lesson A SCALAR OSTIUM IS INSUFFICIENT AND THE OPERATOR CAUGHT IT. fo_ostium returns a darkness, and a blonde shaft is not a less-dark black one. Corrected the same session with fo_ostium_chroma: a dark shaft at depth measures 1577,900,404 against 3239,3239,3239 at the surface, so it reads BLUE-GREY on its own from one scattering asymmetry. Tyndall, emergent, not tinted. 2026-09-03 lesson GROWTH WAS NEVER THE GAP. nx_derm already computes skin FROM the layers beneath, nx_anat_sov already carries inside-out vasculature, nx_nxa_groom already grows hair from a density ratio, nx_hairdyn_gate already proves the hair plant is DERIVED rather than placed. The missing piece was the transfer function between structure and colour, and building a second growth model would have been a duplicate ruler. 2026-09-03 queue Five gates GREEN on the laptop build farm; NAS builds queued behind a sustained I O storm with BQ rows verified in knowledge/store/deployq-. One refused build destroyed its own _build fossil mid-flight. Nothing is lost and nothing needs re-typing. 2026-09-16 measure SCHEMA MIGRATION 2026-09-16: the eight rung rows were 6-field (video.plan grammar) and nx_compare_rank refused the whole board field-count fields=6 expected=8; migrated to rung, id, title, symbol, note, kind, cost, deps with the open and done words dropped (status is derived from the matrix) and costs DECLARED, not measured: ST1 1, ST2 1, ST3 1.5 depending on ST2, ST4 2, ST5 2, ST6 2, ST7 1.5, ST8 0.5. First September bar landed as SB-HAIRSTRAND-2609. 2026-09-27 land LADDER DECLARED 2026-09-27: two dated targets (T-BIC-SKIN on the vendor tier, T-FR-DERIVED on the physiology-driven frontier), six bars with the month each came from (skin scattering June 2026, physiology-driven appearance September 2026, fabric March 2026, wetness January 2024, standards September 2026, MetaHuman skin September 2026), ten rung roles; the board leaves NO-LADDER. The wetness, fabric and skin-scattering bars are older than this month and carry barscan attestations proving the September listings hold nothing newer. 2026-10-01 measure OCTOBER ATTESTATION 2026-10-01: nx_barfresh skintwin verdict FRESH, 7 of 7 bars ATTESTED. Five by the morning nx_frontier_scan attest, each listing re-read by this seat: hair (FurE 2609.35770 of 2026-09-28, HairCS 2609.16465), subsurface scattering (newest on-subject 2606.27604 spectral SSS from RGB, 2026-06-25), SIGGRAPH Asia 2026 venue tags (2026-09), fabric (newest on-subject FabricGen 2603.07240, the 2026-06 entry is cond-mat and off-subject), wetness (2025-04; the listing is dominated by off-subject hits, a hep-th Wet Hair paper on top, so its query is due a narrower rewrite). Two by the seat: SB-SKINSTD-2609 (NIST skin reflectance page unchanged, last modified 2026-05-07) and SB-MHSKIN-2609 (MetaHuman notes still 5.8). RANK by nx_compare_rank (artefact sha256 cd89aa52; no ver ladder, so value alone, LABELED): #1 ST2 the eumelanin and pheomelanin hair absorber pair (km_hair_mix, priority 1000 pulled above ST3), #2 ST6 measured absorption spectra mirrored and pinned (km_spectra_load, 800), #3 ST3 fibre classes (333), #4 ST4 follicle fields (300), #5 ST9 hair wetness (100); PM filed 0, the board needs ver rows to stage a critical path

ST9 Hair wetness and clumping as a state · cell WATCHING · gate NO-GATE-DECLARED

Contract	`_ABSENT_:fo_hair_wet` · executor Organ · 2 u · stands on ST7, ST2
Definition of done	Wet hair is not darker hair paint: strands clump by capillary adhesion (a clump field over the groom), the cuticle reads index-matched so the Marschner R lobe drops and TT rises, and drying is a curve. Consumes ST7's wetness field and the shipped hr_marschner_lobe; graphics GR58 renders it. No current-month rival exists (SB-WET-2512). Done-rule: a wet groom renders with a measured clump count and lobe shift published beside the dry groom, both from one wetness row.
Matrix cell	HAIR: wet hair clumps and its cuticle reads index-matched, a state not a paint · `runtime/nx_follicle_lib.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_follicle_lib`; declare `nx_follicle_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

dossiers=10 cells_joined=8 no_matrix_row=2 (sum 10) gates_recorded=0 gates_unrecorded=0 no_gate_declared=8 (sum 8 of cells_joined) evgates=UNSTAMPED

Capability comparisons

Compare the field, search individual capabilities and open their source and adoption evidence. Documented presence does not establish comparative quality.

Position map — centrality and distinctiveness

The four-quadrant map the field uses for brand strategy (Dawar and Bagga, HBR June 2015), re-derived from this matrix on every publish. Centrality is the share of the category's feature mass a player covers, each feature weighted by how many hold it; distinctiveness is the average lead over each rival on the rows the player holds; breadth is the depth-weighted share of the whole matrix (the bubble); depth is how deeply the rows held are held; momentum is the day-over-day move off the spine (green rising, red falling, grey until day two); the dashed path runs first day → previous day → today. Dividers are the category means. Axes are fitted to the field of play, so read the tick numerals, not the frame. Rival marks are documented presence, so a rival's position reads the record, never its quality. The picture grades its own readability below; the table beside it is the same data for a screen reader or a second method.

Rows. 13 feature rows mapped of 18 matrix rows; 5 substrate rows (kernels, bounds and instruments the plan's rungrole marks substrate) are the estate's scaffolding, not category features, and stay off the map

Views. 2D cut: centrality, distinctiveness · 3D cube: centrality, distinctiveness, breadth · axes registered: centrality, distinctiveness, breadth, depth, momentum · a board picks its own in knowledge/compare/skintwin.cdmap (cut|x|y, cube|x|y|z); absent = the HBR defaults

bubble area = breadth · ring = momentum (green rising, red falling, grey until day two) · dashed = category means · axes fitted to the field of play: centrality 300–1000, distinctiveness 0–700 of 0–1000 permil (the full range put every player in one corner)
readability of the 2D cut, self-graded by the layout ruler: label overlaps 3 · labels over marks 0 · off-canvas 0 · unresolved labels 2 · mark overlaps 3 (a fact of the data: two players that close are that close) · data spread 765 permil of the plot · quadrant words unseated 0
text contrast, measured with wcag2-ratio (floors from contrast.conf), light theme: labels 16.24 (floor 4.50) · notes 16.24 (floor 4.50) · quadrant words 5.89 (floor 4.50) · tick numerals 5.89 (floor 4.50) · axis titles 5.59 (floor 4.50) · dark theme: labels 13.78 (floor 4.50) · quadrant words 5.55 (floor 4.50) · axis titles 5.97 (floor 4.50) · dark classes under their floor 0 (one figure serves both themes: a dark shortfall is a token to fix, never a class to hide) · classes refused under their floor 0 (a refused class is not drawn; the scale classes are measured, never hidden) · export: SVG PNG (receipt, rendered by the estate's own rasteriser from this page)

axes fitted: centrality 300–1000 · distinctiveness 0–700 · breadth 0–700 permil · farther bubbles are painted first, nearer ones over them · the floor shadow is each bubble's (x, y) at height 0
readability of the 3D cube, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 2 · unresolved labels 0 · mark overlaps 6 (a fact of the data: two players that close are that close) · data spread 141 permil of the plot

Nishi MetaHuman Daz and Character Creator Scan plus 3DGS avatar pipelines Offline VFX skin (Weta and ILM class)
10 panels over 5 registered axes, every pair once (the lower axis on x, the higher on y) · each panel fitted to its own field of play, first and last tick numerals shown · no labels in a small cut, the legend names the colours; the grade below reports the mark terms only (MM summed over the panels, spread averaged)
readability of the small multiples, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 33 (a fact of the data: two players that close are that close) · data spread 588 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Aspirational	1000	527	615	666	0 since 2026-10-03	12	13	2026-09-15
MetaHuman	Peripheral	454	0	102	666	0 since 2026-10-03	2	13	2026-09-15
Daz and Character Creator	Peripheral	454	0	102	666	0 since 2026-10-03	2	13	2026-09-15
Scan plus 3DGS avatar pipelines	Peripheral	454	0	102	666	0 since 2026-10-03	2	13	2026-09-15
Offline VFX skin (Weta and ILM class)	Aspirational	636	416	256	833	0 since 2026-10-03	4	13	2026-09-15

Day	Matrix rows reviewed	Players recorded
2026-09-15	16	5
2026-09-16	16	5
2026-09-17	16	5
2026-09-18	16	5
2026-09-19	16	5
2026-09-20	16	5
2026-09-21	16	5
2026-09-22	16	5
2026-09-23	16	5
2026-09-27	12	5
2026-09-28	13	5
2026-10-03	13	5
2026-10-08	13	5

How to read this evidence. A Nishi source mark records whether the named symbol is declared in the selected source file. It does not establish working behavior or accepted delivery. A watching tag names an expected source entry; discovering it updates source coverage only. A dark tag means the file exists but the named symbol was not found: inspect the source and contract before assuming the capability is missing or renaming code. Peer codes are authored assessments; unassessed means no comparison is established, not absence. Qualification, deployment identity, runtime observations and acceptance must carry their own evidence. Adoption records the catalog's observed path through source, build, promotion, registration and invocation; libraries use importer evidence and gates use declared execution surfaces. A full catalog path does not establish quality, freshness or a successful user outcome. Partial and unknown states retain their stated next action. Census stamps: importers asof 1790637268, gate census asof 1790104131 (unix seconds; -1 = census absent).

Capability matrix — measured against source

◉ authored stronger claim; source present● present◐ partial○ absent? unassessed; excluded from relative comparison · click any capability for its evidence

Peer cell coverage: authored assessed 72, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	MetaHuman	Daz and Character Creator	Scan plus 3DGS avatar pipelines	Offline VFX skin (Weta and ILM class)
COLOUR
chromatic adaptation between capture and working white Measured: `cs_cat_build` exists in `runtime/nx_colorsci_lib.nx`, verified at emit. The ICC PCS is D50 and captures are D65, so the profile path alone leaves a white-point error no gamut work downstream can undo. Bradford and CAT02, KAT against the published D65-to-D50 matrix on all nine coefficients. Gate 9 of 9 GREEN, promoted rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=6 nonval=3` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Offline VFX skin (Weta and ILM class) (leads)	●	●	●	●	◉
perceptual colour difference as the referee unit Measured: `cs_de2000` exists in `runtime/nx_colorsci_lib.nx`, verified at emit. CIEDE2000 with the Sharma-Wu-Dalal table as data. BUILT AND UNWIRED to the render lane, which is why no row here can yet state a rendered error. That gap is rung ST1 and it is the highest-value one on the board rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=6 nonval=3` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	●
SKIN
two-flux reflectance from absorber concentrations Measured: `km_r_from_ks` exists in `runtime/nx_kubelka_lib.nx`, verified at emit. Kubelka-Munk single-constant, the form the field uses for an optically thick layer. Exact where the algebra is exact: K over S of one quarter returns a reflectance of one half. Gate 10 of 10 GREEN, promoted rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=4 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	●
the closed-form inverse that turns a measurement into chromophores Measured: `km_ks_from_r` exists in `runtime/nx_kubelka_lib.nx`, verified at emit. The inverse is what lets a captured photograph become chromophore ground truth instead of a baked texture. Round trip proven on every rung of the ladder with a worst error of one part per million Adoption: `LIB-WIRED importers=4 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
melanin and haemoglobin recovered from three channels Measured: `sf_fit` exists in `runtime/nx_skinfit_lib.nx`, verified at emit. Two parameters against three channels, inverted by deterministic coarse-to-fine search: no gradients, no autograd, no float. Identifiable only because the two absorbers have different spectral shapes, and the gate carries the PROPORTIONAL-coefficient case as a neg-control that must miss. Worst recovery error 2.2e-4, residual zero. Gate 8 of 8 GREEN Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
the residual travels with the fit Measured: `sf_sse` exists in `runtime/nx_skinfit_lib.nx`, verified at emit. A fitter that returns its best guess without saying how badly it missed will report chromophores for a measurement no skin could produce. An unreachable measurement comes back at 0.5 against a floor of 0.002 Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
ANATOMY
grown vasculature becomes the haemoglobin field Measured: `va_field` exists in `runtime/nx_vasc_lib.nx`, verified at emit. Double-pass Beer-Lambert: a vessel one millimetre down contributes far less than the same vessel just under the surface, which is why vasculature reads on the wrist and not the palm. Denser vasculature moves rendered red-minus-green from 89090 to 196295. Gate 9 of 9 GREEN Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
appearance is a function of where the structure is Measured: `va_seg_visible` exists in `runtime/nx_vasc_lib.nx`, verified at emit. THE ANTI-PAINTING PROOF. The same vessel buried deep returns the surface to 70497 against a vessel-free baseline of 70497 -- identical. A painted texture cannot do that, because its appearance is not a function of depth Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
a pore that grows -- waxed, shaved, regrowing Measured: `fo_ostium` exists in `runtime/nx_follicle_lib.nx`, verified at emit. nx_microscan measured the pore layer as scoring AT CHANCE, which is the correct verdict on a pore carrying only a position. Shaving cuts the shaft and leaves the canal FULL at 3599; waxing pulls it from the root and leaves it EMPTY at 0. The five-o-clock shadow is occupancy Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
the ostium carries a colour, not a darkness Measured: `fo_ostium_chroma` exists in `runtime/nx_follicle_lib.nx`, verified at emit. A blonde shaft does not read as a less-dark black one. And a dark shaft at depth reads BLUE-GREY, because tissue scatters short wavelengths back before they reach it while long ones penetrate and are absorbed: measured 3239,3239,3239 at the surface against 1577,900,404 at depth. Tyndall, emergent from one asymmetry, not tinted rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelinesCons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	●
a removal event reaches rendered colour and decays Measured: `fo_erythema` exists in `runtime/nx_follicle_lib.nx`, verified at emit. Freshly waxed renders red-minus-green at 272337 and the same skin six days later at 97278. A follicle EVENT propagates through perfusion to Kubelka-Munk and fades on a clock. Gate 16 of 16 GREEN Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class)Cons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○
HAIR
the same two-absorber model with different absorbers DARK — `runtime/nx_kubelka_lib.nx` EXISTS but does not declare `km_hair_mix`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Hair colour is eumelanin and pheomelanin -- the identical structure to skin with different coefficients, so the shipped core applies unchanged. Contract = the hair absorber pair, which unlocks natural hair and is the precondition for the wig row below rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching km_hair_mix Pros none measured yetCons behind MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class) (leads); open contract; expected source declaration not established	○	●	●	●	◉
a wig is a MATERIAL, which is why wigs read as wigs DARK — `runtime/nx_follicle_lib.nx` EXISTS but does not declare `fo_fibre_class`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Synthetic fibre has no medulla, no melanin, a uniform dye and a sharper specular. Modelled as colour-plus-strands the difference is inexpressible and cosplay wigs will always look like hair that is slightly wrong. Contract = real, enhanced and fully synthetic as fibre classes, covering eyelashes and extensions by the same row rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching fo_fibre_class Pros none measured yetCons behind MetaHuman, Daz and Character Creator, Offline VFX skin (Weta and ILM class) (leads); open contract; expected source declaration not established	○	●	●	○	◉
follicle placement follows anatomical fields, not noise DARK — `runtime/nx_follicle_lib.nx` EXISTS but does not declare `fo_field_sample`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Random placement is a direct uncanny trigger. Real follicles carry a density gradient by body site, an orientation field with streams and whorls, and neighbour correlation. Contract = sample placement from those fields. The same law governs limbal rings in the eye, which is the sibling contract on the graphics board rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching fo_field_sample Pros none measured yetCons behind MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class) (leads); open contract; expected source declaration not established	○	●	●	●	◉
FABRIC
sheer and lace over skin is a LAYERED scattering problem Measured: `km_layer_over` exists in `runtime/nx_kubelka_lib.nx`, verified at emit. Lace has holes and sheer transmits, so light crosses the fabric, scatters in skin and returns through it. That is the problem Kubelka-Munk was invented for, so garments and skin can share one appearance core rather than forking into two renderers rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=4 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Offline VFX skin (Weta and ILM class) (leads)	●	●	●	●	◉
CALIBRATION
measured absorption spectra as pinned data Measured: `km_spectra_load` exists in `runtime/nx_kubelka_lib.nx`, verified at emit. Every coefficient on this board is a PROBE VALUE and both the libs and the gates say so in their own text. Contract = mirrored, pinned melanin and haemoglobin absorption spectra so the numbers stop being shaped-correctly and start being measured. The three mirrors this row is grounded on are already fetched and pinned and are cited here so the declaration and the evidence cannot drift apart: haemoglobin [omlc_hb] and melanin [omlc_mel] absorption spectra as pinned datasets, and the procedural-generation release [infinigen_rel] the coefficient shapes were compared against. One bracket per key on purpose -- a compound mark reads as a single unknown key to the referee. Adoption: `LIB-WIRED importers=4 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind MetaHuman (leads), Daz and Character Creator (leads), Scan plus 3DGS avatar pipelines (leads), Offline VFX skin (Weta and ILM class) (leads)	●	◉	◉	◉	◉
HAIR
wet hair clumps and its cuticle reads index-matched, a state not a paint DARK — `runtime/nx_follicle_lib.nx` EXISTS but does not declare `fo_hair_wet`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. ST9 (2026-09-27). Strands clump by capillary adhesion (a clump field over the groom), the Marschner R lobe drops and TT rises as the cuticle reads index-matched, and drying is a curve; consumes ST7's wetness field and the shipped hr_marschner_lobe, rendered by graphics GR58. The estate's wet-appearance scan holds no 2026 paper on the subject [arxiv_wet_202609]. Watch contract: a wet groom renders with a measured clump count and lobe shift published beside the dry groom from one wetness row. Rival cells coded 0: not witnessed by a mirrored source. dossierwatching fo_hair_wet Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○
Wetness as a surface state Measured: `km_wet_state` exists in `runtime/nx_kubelka_lib.nx`, verified at emit. ST7 (2026-09-28). A smooth water film over a Lambertian substrate: Fresnel derived from the indices, the film's internal diffuse reflectance by Judd's identity held against the direct integral through total internal reflection, the darkening series (1 - r_e)(1 - r_i) R / (1 - r_i R) with the film's mirror reflectance beside it, wetness as an area state in [0,1]. Replicates Jensen 1999 Sec. 2.1 and 3.1 [jensen_wet_1999]; the medium-based frontier is WetSpongeCake [pan_wetporous_2025], deliberately not this rung. Rival cells coded 0: not witnessed by a mirrored source. Adoption: `LIB-WIRED importers=4 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of MetaHuman, Daz and Character Creator, Scan plus 3DGS avatar pipelines, Offline VFX skin (Weta and ILM class); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○

Rival-claim provenance. Every Yes, Best or Part mark in a rival column is a claim about someone else's product. A cited row carries a reference mark that resolves to a pinned mirror (the refs gate measures that); an uncited row is an observation read and is badged in its evidence until a reference lands. The badge is the receipt, never the proof: a mark proves a mirror exists, not that the mirror supports the code. The rows are counted below.

Rival claims, counted
Class	Count	Note
rows with rival marks	9	rows carrying a Yes, Best or Part in a rival column
cited	1	a reference mark resolves to a pinned mirror
uncited	8	an observation read, badged until a reference lands
rows, summed	9	RECONCILED: the parts sum to the total 9

Delivery and evidence

Inspect risks, technical debt, rendered observations, experiments and references. Read scope and limitations alongside every result.

Release history and work log

The worked plan, on the board. Every leg appends what it measured, landed, retracted, learned and left queued, per rung, so a crash leaves the next seat a ledger here rather than a transcript to mine. Kinds: measure, land, retract, lesson, queue. Newest last.

When	Rung	Kind	Entry
2026-09-03	ST8	measure	Bite proof, five gates, both directions, sources restored byte-identical: kubelka cf17b1aa 10/10 GREEN after RED 9/10; vasc d1b6409d 9/9 after RED 7/9; follicle 1b15ffd3 16/16 after RED 15/16; skinfit f4f70936 8/8 after RED 5/8; ws_client 714e6610 9/9 after RED 6/9.
2026-09-03	ST8	lesson	A KILL BY THE WRONG TOOTH PROVES NOTHING ABOUT THE RIGHT ONE, so each mutation was chosen to attack ONE named claim and the surviving teeth were read as carefully as the failing ones. follicle is the clearest case: removing the per-channel attenuation killed TYNDALL and left the blonde-versus-black tooth passing, which is correct and would have looked like a weak gate to anyone counting failures instead of reading them.
2026-09-03	ST8	measure	MACHINE bite records now exist for FOUR of the five declared gates, replacing the hand proofs that C4 correctly refused to accept: nx_kubelka_gate killed=1 restore 29512 B, nx_vasc_gate killed=1 restore 3967 B, nx_skinfit_gate killed=1 restore 5417 B, nx_de_referee_gate killed=1 restore 5255 B. Every run reported mutant_live=0 and every restore was byte-identical. C4 reads declared=5 bite-proven=4 never-bitten=1 sum=5.
2026-09-03	ST8	measure	nx_colorsci_cat_gate is the outstanding one and its cause is MEASURED rather than guessed: SIX consecutive UNPROVEN, every one BUILD-REFUSED-BY-ADMISSION on the SUBJECT rebuild while gate_build_exit=0 said the gate itself compiled fine. nx_colorsci_lib is the largest subject in the set at 38414 bytes, so its per-mutant rebuild is the heaviest job here and the first thing marginal headroom declines. The other four went through on the same lane in the same window. It needs a quieter box, not a different method.
2026-09-03	ST8	lesson	THE LAPTOP FARM CANNOT SUBSTITUTE FOR THIS ONE, AND THE REASON IS WORTH RECORDING TWICE. The laptop nx_colorsci_lib is 29200 bytes and contains ZERO occurrences of cs_cat_build -- it predates the chromatic-adaptation work entirely. Even after converging it, a laptop bite produces no record C4 can read, so it would move the seat's confidence and not the board's evidence. The last bite genuinely requires the NAS lane.
2026-09-03	ST8	measure	THE LAPTOP/NAS COLORSCI DIVERGENCE IS NOW MEASURED IN BOTH DIRECTIONS, WHICH IS THE HALF THAT WAS EXPENSIVE. Local nx_colorsci_lib carries 38 functions and EVERY ONE of them also exists on the NAS; the NAS additionally carries cs_prof_lab_to_lin, cs_prof_gamut_permil, cs_prof_gamut_clip and the five cs_cat_* adaptation functions. So the NAS is a STRICT SUPERSET and the laptop is purely behind -- there is no two-way fork here and adopting the NAS copy destroys nothing. That direction was checked rather than assumed, because this estate has had files diverge BOTH ways and 'the laptop is the stale one' is a guess until someone lists the symbols.
2026-09-03	ST8	lesson	AND THE DIVERGENCE FAILS LOUD, NOT SILENT, WHICH IS WHY IT IS A FRICTION AND NOT A CORRECTNESS RISK. A local build touching colour REFUSED with 19 named errors pointing at CS_P_B64; it did not miscompile quietly. That is the difference between a hazard that costs a seat ten minutes and one that ships a wrong number. The convergence is owed and is a bulk file transfer, so it is recorded here with its direction settled rather than performed at the tail of a long session where a partial copy would be the worst outcome of the three.
2026-09-03	ST8	lesson	THE MACHINE IS STRICTER THAN THE SEAT WAS. nx_gate_bite reports invalid_uncompilable and void_build_lane SEPARATELY from killed, so a mutant that failed to compile and a mutant lost to the build lane are never counted as evidence. The hand method picked one mutation already known to compile. On nx_vasc_gate the symbol-targeted operator independently landed inside va_seg_visible -- the same depth-attenuation line the hand proof attacked, reached by a different route.
2026-09-03	ST8	lesson	THE LAPTOP FARM CAN BITE-PROVE WHEN THE NAS CANNOT. nx_gate_bite is a NAS organ and the estate was unreachable, but the proof it automates is plant-the-known-bad, rebuild, see RED, restore, see GREEN -- and that runs entirely on the local sovereign toolchain. An unreachable build lane blocks shipping, not proving.
2026-09-03	ST0	measure	Three absence probes ABSENT-PROVEN with corpus_complete=1 before any build: kubelka, cat02, bradford. nx_colorsci_lib had CIELAB and CIEDE2000 and no adaptation and no scattering model.
2026-09-03	ST0	land	cs_cat_build and cs_cat_xyz landed in nx_colorsci_lib (32813 to 38414 bytes, single application verified). nx_colorsci_cat_gate 9 of 9 GREEN, promoted 40520 bytes, registered.
2026-09-03	ST0	land	nx_kubelka_lib and nx_kubelka_gate 10 of 10 GREEN, promoted 34560 bytes, registered. K over S of 0.25 returns exactly 0.5; round trip worst error one part per million.
2026-09-03	ST0	land	nx_skinfit_lib 8 of 8 GREEN on the laptop farm, source landed on the NAS, build queued behind an I O storm.
2026-09-03	ST0	land	nx_vasc_lib 9 of 9 GREEN. Red-minus-green 89090 sparse to 196295 dense; the same vessel buried deep returns 70497 against a 70497 baseline.
2026-09-03	ST0	land	nx_follicle_lib 16 of 16 GREEN. Shaved ostium 3599 against waxed 0; blunt tip 17999 against tapered 4499; wax erythema moves rendered red-minus-green 272337 to 97278 over six days.
2026-09-03	ST0	lesson	A SCALAR OSTIUM IS INSUFFICIENT AND THE OPERATOR CAUGHT IT. fo_ostium returns a darkness, and a blonde shaft is not a less-dark black one. Corrected the same session with fo_ostium_chroma: a dark shaft at depth measures 1577,900,404 against 3239,3239,3239 at the surface, so it reads BLUE-GREY on its own from one scattering asymmetry. Tyndall, emergent, not tinted.
2026-09-03	ST0	lesson	GROWTH WAS NEVER THE GAP. nx_derm already computes skin FROM the layers beneath, nx_anat_sov already carries inside-out vasculature, nx_nxa_groom already grows hair from a density ratio, nx_hairdyn_gate already proves the hair plant is DERIVED rather than placed. The missing piece was the transfer function between structure and colour, and building a second growth model would have been a duplicate ruler.
2026-09-03	ST0	queue	Five gates GREEN on the laptop build farm; NAS builds queued behind a sustained I O storm with BQ rows verified in knowledge/store/deployq-. One refused build destroyed its own _build fossil mid-flight. Nothing is lost and nothing needs re-typing.
2026-09-16	ST0	measure	SCHEMA MIGRATION 2026-09-16: the eight rung rows were 6-field (video.plan grammar) and nx_compare_rank refused the whole board field-count fields=6 expected=8; migrated to rung, id, title, symbol, note, kind, cost, deps with the open and done words dropped (status is derived from the matrix) and costs DECLARED, not measured: ST1 1, ST2 1, ST3 1.5 depending on ST2, ST4 2, ST5 2, ST6 2, ST7 1.5, ST8 0.5. First September bar landed as SB-HAIRSTRAND-2609.
2026-09-27	ST0	land	LADDER DECLARED 2026-09-27: two dated targets (T-BIC-SKIN on the vendor tier, T-FR-DERIVED on the physiology-driven frontier), six bars with the month each came from (skin scattering June 2026, physiology-driven appearance September 2026, fabric March 2026, wetness January 2024, standards September 2026, MetaHuman skin September 2026), ten rung roles; the board leaves NO-LADDER. The wetness, fabric and skin-scattering bars are older than this month and carry barscan attestations proving the September listings hold nothing newer.
2026-09-27	ST5	queue	The beach-lounger reference class (2026-09-27) makes ST5 the load-bearing rung of the whole twin: the garment is sheer, so the skin term must transmit through the fabric; graphics GR57 is its render-side consumer and depends on it. nx_compare_rank reads ST5 as DARK-CONTRACT (organ exists, symbol undeclared), so km_layer_over is unbuilt, not merely unwired.
2026-09-27	ST5	land	LANDED 2026-09-27: km_layer_over, km_layer_rt, km_open_strain and km_sheer_over in nx_kubelka_lib (4116 -> 9704 bytes, additive; the single-constant forward and inverse untouched); nx_kubelka_gate grew from 10 to 22 teeth (7487 -> 18165 bytes) and was built sovereign (70904 bytes, sha256 7726d1633e08cd55200b6d050cf419b60bfc3e700f0b38577f38824707261bf9), probed staged-vs-live (other_rc=0, DIFFERS only from the new teeth onward), bite-proven (a compiling comparison mutant of the lib at byte 4050 KILLED, both twins restored byte-identical), and PROMOTED with prev banked. Mechanism replicated: Kubelka's layered two-flux composition R = R_f + T_f^2 R_s / (1 - R_f R_s) over a porous layer whose open fraction follows area geometry under strain; no coefficient typed.
2026-09-27	ST5	measure	nx_contentdiff live vs staged read RED lost_from_live=4: the rewritten verdict note (named, mine) and sys_openat_append, gv_cat, gv_catn; the matched-vintage control (nx_vasc_gate, source untouched since 2026-09-03, rebuilt through the SAME lane) lost exactly those three symbols and gained the same framed-append set, so the loss is the gate base class moving between 09-03 and today, not this change; promoted under allow_capability_loss=yes with every run named here.
2026-09-27	ST5	lesson	A LAYERED REFLECTANCE IS NOT MONOTONE IN THE OPEN FRACTION, AND THE FIRST TOOTH ASSERTED IT WAS: a bright absorbing fabric (R 0.3, T 0.5) over darker skin (0.4) sits ABOVE the skin at rest (0.4136 fully closed), dips below it as the weave opens (0.3993 at 0.2 open, 0.3930 at 0.34) and returns to it only at full opening. The arithmetic caught the invariant before the build did; the tooth now asserts the geometry that is monotone (the open fraction under strain) and a thickness ladder asserts the two-flux property that is (reflectance rises, transmittance falls, energy conserved on every rung).
2026-09-27	ST5	queue	The bite's first-kill landed in the pre-existing mix function; an AIMED bite inside the new block is owed so the layered teeth are proven non-vacuous by a mutant of their own subject, and graphics GR57 is the render-side consumer that turns this lib term into a swatch under the sky.
2026-09-27	ST5	measure	AIMED BITE RESOLVED, NOT SKIPPED: nx_gate_bite aim km_layer_over 0 produced 2 numeric mutants and both FAILED TO COMPILE (invalid, excluded, verdict INCONCLUSIVE, twins restored byte-identical); the numeric operator can only mutate literals and the new arithmetic carries none (its guards are the only zeros, a consequence of the magic-number law), while the comparison operator's first-kill-wins stops at byte 4050 in the pre-existing mix function. The non-vacuity oracle for T11 to T21 is therefore the hand-derived KATs (399292, 488000, 412667, the exact halves and R_inf), which a wrong formula cannot pass; a comparison-aimed bite (aim on == sites of one function) is a bite-organ rung, filed as debt rather than left as an unnamed gap.
2026-09-28	ST7	land	WETNESS AS A SURFACE STATE BUILT AND PROMOTED 2026-09-28: nx_kubelka_lib gained km_fresnel (exact unpolarised dielectric Fresnel with total internal reflection), km_fresnel_hemi (cosine-weighted hemispherical mean by midpoint quadrature whose node count doubles until successive estimates agree within the caller's tolerance, capped at the Q30 resolution), km_film_internal (Judd's identity r_i = 1 - (1 - r_e)/n^2), km_wet_lambert (the film-over-Lambertian series (1 - r_e)(1 - r_i) R / (1 - r_i R) with the film's own diffuse reflectance beside it), km_film_specular and km_wet_state (an area state in [0,1]); 9704 -> 18047 B. Primary sources read from mirrored bytes: Jensen 1999 Sec. 2.1, 2.2, 3.1, 3.2 [jensen_wet_1999] (the Stanford copy refuses our TLS 1.2 certificate pipeline, the Eurographics digital library copy mirrored) and Pan et al. 2025 [pan_wetporous_2025], the medium-based frontier deliberately not built. nx_kubelka_gate 23 -> 33 teeth (Fresnel KATs, the two-method film identity, the strict darkening ladder, the energy identity, the state endpoints, a cross-identity, neg-controls), sovereign build 82120 B sha256 e018eb64, behaveprobe other_rc=0 on the staged binary, contentdiff lost 1 (the rewritten verdict note) named, promoted under allow_capability_loss, _offc refreshed. Rung symbol repointed from the absent fo_wetness to km_wet_state and a matrix watch row added
2026-09-28	ST7	measure	nx_kubelka_gate on the promoted binary 2026-09-28, 35/35 GREEN, values published by gv_kv: Fresnel at normal incidence into glass 39999 micro (the exact identity is 40000); water film (n = 1.33) external diffuse reflectance r_e = 65932 micro reached at 256 quadrature nodes; internal diffuse reflectance r_i = 471949 micro by Judd's identity against 471948 micro by the direct integral through total internal reflection, two methods one micro apart; the film darkens a 0.2 grey to 0.109, a 0.5 grey to 0.323 and a 0.8 grey to 0.634 (strictly, every rung); a white substrate under the film returns 999999 micro of wet-plus-specular, the energy identity to one micro of rounding; the wetness state blends 0.5 dry and 0.323 wet to 411393 micro at a half state; the published wet value re-derives from the published r_e and r_i by the series inside the gate. OUTSIDE ORACLE, named not claimed: the classical diffuse-reflectance tables for n = 1.33 (Judd 1942, Egan and Hilgeman) are the referee for r_e and r_i and are owed as a pinned ref before these values are called agreed
2026-09-28	ST7	measure	nx_gate_bite nx_kubelka_gate over nx_kubelka_lib.nx 2026-09-28, aim km_fresnel: valid_mutants=1 killed=1 not_reached=0, verdict GREEN (a compiling comparison mutant went RED), twins restored byte-identical (18047 B pristine, custody twins_checked=2 restored=2). NAMED LIMIT: the first-kill landed at byte 4050, the comparison inside the older single-constant forward, so this bite proves the gate is non-vacuous and does NOT reach the new film block; the aimed reach into literal-free arithmetic is debt 1790543778. The film block's own proof is the two-method identity, the exact energy identity and the cross-identity, all published above
2026-09-28	ST7	measure	OUTSIDE ORACLE READ 2026-09-28: Judd 1942 Table I (NBS RP1504, mirrored and pinned as judd_1942_diffuse_fresnel [judd_1942_diffuse_fresnel]) gives for m = 1.33 an external diffuse reflectance of 0.0660 and an internal one of 0.472, computed in 1942 by summing the Fresnel formula over 0.04 radian intervals with Moon's tables, uncertainty less than 2 in the last figure. The estate's km_fresnel_hemi reads 0.065932 and Judd's identity 0.471949: both inside the table's stated uncertainty (0.00007 and 0.00005 away), a THIRD method class agreeing with the two in the gate. The water-film values are now AGREED by an outside referee, not only self-consistent; the gate gains the table as two teeth in the next promote
2026-09-28	ST7	land	the outside oracle is a TOOTH now (2026-09-28): nx_kubelka_gate carries Judd 1942 Table I (m = 1.33: external 0.0660, internal 0.472, uncertainty less than 2 in the last figure) as two teeth beside the two-method identity, each holding the estate's value inside the table's stated uncertainty; promoted 82712 B sha256 3f6dda67, live run 37/37 GREEN with judd_1942_external_micro=66000 and judd_1942_internal_micro=472000 published beside water_film_external 65932 and internal 471949. Three method classes now agree on the film: the estate's quadrature, Judd's identity through total internal reflection, and a 1942 summation over Moon's tables
2026-09-28	ST6	retract	The ST6 contract was unjudgeable as written: its watch row named the DIRECTORY knowledge/fetched as the organ and st_coeff_spectra as the symbol, so no symbol ruler could ever read it, and graphics GR59 (melanin field, on the photoreal path) read its prerequisite skintwin:ST6 as UNRESOLVED(organ-unreadable) through nx_xdep_lib on 2026-09-28. Repointed: the rung and the row now name runtime/nx_kubelka_lib.nx km_spectra_load, the reader of the pinned melanin and haemoglobin absorption spectra that the lib's own header says belong in pinned reference data and not in the lib. The mirrored, pinned spectra stay the data half of the rung; the loader is the half a ruler can judge. Cell reads OPEN until km_spectra_load is built
2026-10-01	ST6	measure	ACCEPT RULE PRE-DECLARED before any fit runs (2026-09-30): the bar paper 2606.27604 (Aliaga and Jarabo, mirrored knowledge/fetched/cmp_skintwin_spectral_sss_2606_27604.pdf, read through nx_pdf_text) reports its accuracy ONLY against its own Monte Carlo reference, never against measured skin: Table 1 rendering error Delta E76 on the visible band over 10 representative tones = 3.19 for its K=3 neural decoder and 3.79 for the K=3 per-tone joint optimisation (its stated fitting ceiling), and a mean reflectance error of 3.7 percent (K=3) over 25K tones; its limitations section names fitting to measured spectra as the validation axis it did NOT do. ST6 measures the estate on that missing axis: a spectral two-layer biophysical model fitted to every one of NIST's 100 measured skin spectra (JRES 122.026, data DOI 10.18434/M38597, mirrored and pinned). Rule: median Delta E76 over the 100 spectra at or below 3.19 MEETS the bar's best number, at or below 3.79 meets its fitting ceiling; the worst spectrum is held to 3.79 and published whatever it reads; median mean-relative reflectance error held to 3.7 percent. The comparison is a different test (measured vs simulated reference) and is published as such, never as like-for-like
2026-10-01	ST6	measure	FIRST FULL RUN on all 100 NIST spectra (staged gate, job pid 17528, /tmp/nx_skinspec_gate_st6_1.out, 12/16 RED, every loader tooth GREEN): the two-layer model with the epidermis held at Jacques' 60 um and fitted over 400 to 900 nm reads median Delta E2000 3.457, median Delta E76 4.585 (worst 9.196, subject 72), median spectral RMS 0.0214, median mean-relative error 5.18 percent -- it does NOT meet the bar's 3.19. Pre-declared now, before the second run: the MODEL OF RECORD carries the bar's own fifth structural parameter, the epidermal thickness, fitted over the bar's own range (50 to 350 um, 2606.27604 Sec. 3), and fits the visible band 400 to 780 nm on which the bar's Delta E is measured (the near infrared carries water and the sphere's lateral loss, neither modelled); the 400 to 900 nm four-parameter run and a 400 to 780 nm four-parameter run are published beside it as the ablation, so the choice of record is not made after seeing which wins
2026-10-01	ST0	measure	OCTOBER ATTESTATION 2026-10-01: nx_barfresh skintwin verdict FRESH, 7 of 7 bars ATTESTED. Five by the morning nx_frontier_scan attest, each listing re-read by this seat: hair (FurE 2609.35770 of 2026-09-28, HairCS 2609.16465), subsurface scattering (newest on-subject 2606.27604 spectral SSS from RGB, 2026-06-25), SIGGRAPH Asia 2026 venue tags (2026-09), fabric (newest on-subject FabricGen 2603.07240, the 2026-06 entry is cond-mat and off-subject), wetness (2025-04; the listing is dominated by off-subject hits, a hep-th Wet Hair paper on top, so its query is due a narrower rewrite). Two by the seat: SB-SKINSTD-2609 (NIST skin reflectance page unchanged, last modified 2026-05-07) and SB-MHSKIN-2609 (MetaHuman notes still 5.8). RANK by nx_compare_rank (artefact sha256 cd89aa52; no ver ladder, so value alone, LABELED): #1 ST2 the eumelanin and pheomelanin hair absorber pair (km_hair_mix, priority 1000 pulled above ST3), #2 ST6 measured absorption spectra mirrored and pinned (km_spectra_load, 800), #3 ST3 fibre classes (333), #4 ST4 follicle fields (300), #5 ST9 hair wetness (100); PM filed 0, the board needs ver rows to stage a critical path
2026-10-01	ST6	measure	TRIANGULATED BAR as of 2026-10-01 for spectral skin appearance (layered chromophore models, spectral subsurface scattering, inversion from photographs): FORWARD MODEL -- the best is a multi-layer spectral transport whose layers carry chromophores and which yields a spatial scattering profile, not only hemispherical reflectance; the June 2026 bar Monte Carlo-simulates a two-layer skin (epidermis 50 to 350 micrometres, melanin-dominated, over a semi-infinite dermis with haemoglobin, bilirubin, beta-carotene and water) at 376 wavelengths from five properties (melanin and blood fraction, epidermal thickness, blood oxygenation, eumelanin to pheomelanin ratio), records total diffuse reflectance AND the radial diffusion profile, and states that hemispherical-only spectral datasets cannot characterise translucency [aliaga_spectralsss_2606] [spectral_sss_2606_27604] [arxiv_sss_202609]; the measured-capture lineage is the Donner and Jensen two-layer spectral BSSRDF with melanin blend type and epidermal hemoglobin fraction [gitlina_spectralskin_2020]. INVERSION -- complementary multispectral illumination recovers chromophore maps better than standard broadband illumination, with a network replacing lookup tables [gitlina_spectralskin_2020]; the frontier inverts one RGB albedo through a learned biophysical prior into a three-media mixture for a random-walk path tracer [aliaga_spectralsss_2606]. CALIBRATION -- traceable measured reflectance: NIST's 100 spectra [nist_skin_reflectance_202609] and, newest this month (2609.30122, submitted 2026-09-24), 98 patients' skin reflectance spectra in the public OpenOximetry Repository fitted with a melanin-corrected diffuse-optics model [arxiv_skinspecrefl_202609]. RE-CHECK of SB-SKINAPP-2606 today: the arXiv subsurface-scattering-and-skin search holds no July to October 2026 entry, so the June bar stands and September's newest on-subject entry is biomedical [arxiv_sss_202610] [arxiv_skinspecrefl_202609] ; witnesses [aliaga_spectralsss_2606] [gitlina_spectralskin_2020] [arxiv_sss_202609] [arxiv_sss_202610] [arxiv_skinspecrefl_202609] [nist_skin_reflectance_202609] -- independent groups: Aliaga and Jarabo; Ghosh's group at Imperial College London with NTNU, York, Clemson and UBC; the Cremers group (DIAMOND-SSS 2601.12020); Ha, Munoz, Gutierrez and Kim (polarimetric BSSRDF of dynamic faces, ACM doi 10.1145/3687767); the Fantini diffuse-optics group (2609.30122) -- five groups, with a peer-reviewed CGF paper and an ACM-published paper among them and NIST as the standards tier; aliaga_spectralsss_2606 and spectral_sss_2606_27604 pin the same sha256 and are ONE witness ; WHERE THEY DISAGREE: measured multispectral capture versus RGB-only learned inversion (one RGB albedo underdetermines five biophysical properties, resolved by a learned prior in the frontier and by extra spectral illuminants in the capture school); validation (the frontier validates only against its own Monte Carlo simulation and names fitting measured integrating-sphere diffusion profiles and in-vivo samples as the next step, while the capture school matches photographs); explicit chromophore layers versus purely learned neural scattering (DIAMOND-SSS reconstructs relightable translucent Gaussians, wax and jade as well as skin, from as few as ten images with no chromophore model in its abstract); and how well melanin's effect is even resolved (the oximetry fit needs a cohort roughly ten times larger to resolve its melanin dependence statistically) ; VERDICT on the running lane (updated after its third full run, log 1790826240): ST6 loads the pinned OMLC melanin and haemoglobin absorption and fits a two-layer spectral model in the board's Kubelka-Munk core to all 100 NIST measured spectra -- AT the field on calibration and AHEAD of the bar on measured validation, an axis the bar never measured (median Delta E2000 2.081, median Delta E76 2.419 beside the bar's simulated-reference 3.19, median mean-relative error 3.41 percent; a different test, published as such); BELOW the triangulated best on three named axes: transport (a two-flux fit to directional-hemispherical reflectance has no lateral transport, so it cannot produce the radial profile, scattering radius or anisotropy the bar calls necessary for translucency), dermal chromophores (its own blue-band residual names the beta-carotene and bilirubin the bar's dermis carries), and the tail (worst Delta E76 6.695 at subject 55 misses the declared 3.79); fit a Monte Carlo two-layer reference to the same 100 spectra beside it
2026-10-01	ST1	measure	FIRST RENDER-SIDE ADOPTION OF THE REFEREE (graphics GR59, 2026-09-30): nx_skin_ita_gate (promoted sha256 5dfb04e0, live 25/25 GREEN) reads a DERIVED tan line through cs_de2000: inside-outside DE2000 4.99 beside DE76 6.51 against the measured natural tan of Amano 2020, DE76 mean 6.03 SD 2.64 [amano_naturaltan_2020], on a spectral chain that integrates the CIE's own 1931 2-degree tables [cie_cmf_1931_2deg_202609] at 5 nm and adapts illuminant A to D65 through cs_cat_build; the zero-exposure control reads DE2000 exactly 0, and an aimed mutant that ignores the garment's coverage history is killed by the DE2000 tooth among six others. The constitutive melanin, solved from the genome's ITA on the incumbent locus, lands at 5.54 percent inside the light-skin range of [jacques_skinoptics_1998]; the facultative field is Ravnbak's measured dose-response and fade [ravnbak_tankinetics_2008]. Adopters of cs_de2000 outside the garment lane: 0 -> 1. NAMED LIMIT: the dermis is still bloodless (a* 3.0 inside where measured skin reads near 10), so the referee's a* axis waits on ST6's haemoglobin data entering through km_ks_mix k_hem
2026-10-01	ST6	measure	THIRD FULL RUN, the optimiser corrected (staged gate pid 17884, /tmp/nx_skinspec_gate_st6_3.out, 16/17): the second run's model of record had read WORSE than the 900 nm ablation on its own window, which a superset of parameters on a subset of wavelengths cannot do unless the search fails, and it was the search: halving coordinate brackets collapsed around the wrong point on the coupled scattering and melanin pair. A compass-search second round (Torczon pattern search: steps halve only when no coordinate improves) plus the captured specular as a fitted term bounded by the epidermal Fresnel value gives, on every one of the 100 NIST spectra, model of record (400 to 780 nm, six parameters): median Delta E2000 2.081, worst 4.925; median Delta E76 2.419 (meets the bar's 3.19), worst 6.695 at subject 55 (misses the declared 3.79); median spectral RMS 0.0137, worst 0.0292; median mean-relative error 3.41 percent (meets the bar's 3.7). Ablation on the same build: 900 nm window with the thickness held at 60 um, median Delta E76 3.131, worst 7.417; visible window held, 3.170, worst 7.540. Band residuals of record (median signed model minus measured): plus 0.019 at 400 to 450 nm, plus 0.010 at 450 to 520, minus 0.004, minus 0.004, minus 0.012 at 700 to 780: the model still lacks blue absorption, the shape beta-carotene and bilirubin give, both carried by the bar's dermis. Surface witness: NIST's darkest 2500 nm reading is 0.0229 against the 1.4 Fresnel 0.0278 and the 1.55 Fresnel 0.0465, so the dry stratum corneum index is refuted and the sphere does not return the full specular of every measurement
2026-10-01	ST6	land	THE CONTRACT SYMBOL LANDED IN THE CONTRACT ORGAN 2026-09-30: km_spectra_load is in runtime/nx_kubelka_lib.nx (18047 to 29088 B, strictly additive: one import of the canonical nx_sha256 and an appended block; no existing line altered). It reads a mirrored table, proves the bytes against the .refs pin by sha256, and returns one column on a declared window in a declared unit (column times num over den, Q30, integer arithmetic so a 5e5 molar extinction never meets a Q30 multiply), refusing BY NAME a wrong pin (KM_SPEC_PIN), a malformed pin (KM_SPEC_BADPIN), a truncated table (KM_SPEC_SHORT), a non-monotone wavelength column (KM_SPEC_NONMONO) and an overrun (KM_SPEC_CAP); km_spec_interp interpolates and refuses outside the span. Shared-lib discipline: the block was compiled and run first in a scratch file (nx_skinspec_load.nx, now retired to knowledge/retired), then moved verbatim. Every consumer re-proved against the grown lib: nx_kubelka_gate 37/37 GREEN, contentdiff IDENTICAL-CONTENT (82712 B both), promoted 941a87a9; nx_vasc_gate 9/9, nx_follicle_gate 16/16, nx_skinfit_gate 8/8 GREEN as staged rebuilds, NOT promoted here because their promoted binaries predate the gate base-class move of September (contentdiff loses sys_openat_append, gv_cat, gv_catn on all three, the same three the 09-27 matched-vintage control attributed to the base class) and that drift belongs to their owners' lane. The 09-03 incumbent nx_skinspectra_lib is COMPOSED, not duplicated: the model reads the melanosome law through its ss_melanin_mua; its exact-match integer ss_hb_lookup is superseded by km_spectra_load (pinned, decimal, any column, windowed, named refusals) and stays callable for its own gate. That incumbent gate, source-only since 09-03 with its build enqueued, built and ran for the first time: nx_skinspectra_gate 10/10 GREEN, promoted 3a30f194 (54541 B), registered, _offc refreshed
2026-10-01	ST6	lesson	A SUPERSET OF PARAMETERS FITTED ON A SUBSET OF WAVELENGTHS CANNOT READ WORSE ON THAT SUBSET UNLESS THE SEARCH FAILED, AND THAT ONE INEQUALITY FOUND THE DEFECT: the second run's six-parameter visible fit read median Delta E76 5.44 against 4.66 for the four-parameter 900 nm fit. Halving coordinate brackets collapse around the wrong point on a coupled pair (scattering scale against melanin); a compass-search second round, whose steps halve only when no coordinate improves, took the same model to 2.42. The fix was the optimiser, not the physics, and the ablation published beside the model of record is what made the impossibility visible
2026-10-01	ST6	lesson	THE INCUMBENT WAS FOUND BY A COMMENT, NOT BY A SEARCH: nx_skinspectra_lib and its gate (ST6, 2026-09-03, source-only, build enqueued) sat in buildroot/runtime while this rung was rebuilt; the catalogue was asked for nx_skinspec and its siblings but not for nx_skinspectra, and the .gates header named it only after the loader was written. Remedy taken the same hour rather than filed: the model now COMPOSES its ss_melanin_mua (one melanin law in the estate), its exact-match integer lookup is superseded by the pinned windowed loader, and its gate was built, run 10/10 and promoted. Ask nx_capsearch with the RUNG ID and the data words (ST6, spectra, melanin) before naming a new organ
2026-10-01	ST6	lesson	A SPECULAR TERM IS BOUNDED BY THE DARKEST READING, NOT ARBITRATED BY THE MEDIAN: the first surface witness asked which index's 8 degree Fresnel term sat nearer the median 2500 nm reading (0.0391) and picked the dry stratum corneum's 1.55 (0.0465), ignoring that the body still returns light at 2500 nm. The sound test is that the specular can never exceed what the instrument read where the body returns least: the darkest subject reads 0.0229, below even the epidermal 1.4 term (0.0278), so 1.55 is refuted outright and the sphere's captured specular is a per-measurement fraction, now fitted under the Fresnel bound and checked against each subject's own 2500 nm reading
2026-10-01	ST6	measure	THE MODEL OF RECORD MEETS THE BAR ON THE AXIS THE BAR DID NOT MEASURE (2026-09-30): nx_skinspec_gate 18/18 GREEN on the binary then promoted (sha256 5113a088, 140963 B, staged run /tmp/nx_skinspec_gate_final_1.out, the same bytes /api/promote then took with expect_sha256). Model: the bar's two-layer biophysical skin (melanosome-dominated epidermis over a semi-infinite dermis carrying oxy- and deoxy-haemoglobin and beta-carotene, Jacques baseline and Mie plus Rayleigh scattering), every absorber read from a pinned mirror through km_spectra_load, transported by Kubelka's layered two-flux solution, crossed through the skin surface by the Saunderson series with Judd's internal reflectance, rendered through the pinned CIE 1931 2 degree functions under pinned D65. Fitted to EVERY one of NIST's 100 measured forearm spectra over 400 to 780 nm: median CIEDE2000 0.769, worst 2.420; median CIE76 Delta E 1.030, worst 2.824 at subject 53; median spectral RMS 0.0128, worst 0.0259; median mean-relative error 3.12 percent. BESIDE THE BAR (2606.27604, Aliaga and Jarabo, read from the mirror): its Table 1 visible-band Delta E76 is 3.19 for the K=3 neural decoder and 3.79 for its per-tone joint optimisation, and its mean reflectance error 3.7 percent -- all against its OWN Monte Carlo reference over 10 representative tones, never against measured skin. On measured skin the estate's median Delta E76 is 3.1 times lower than the bar's best figure and its worst spectrum sits under the bar's fitting ceiling; this is a different test from the bar's and is published as such, not as like-for-like. What closed the gap from the first run's 4.585: the compass-search optimiser (to 2.419) and beta-carotene, the dermal chromophore the bar carries (to 1.030); band residuals of record plus 0.018 at 400 to 450 nm, plus 0.005, minus 0.002, minus 0.003, minus 0.010 at 700 to 780 nm. Pre-declared gates all held: median 1.030 at or below 3.19, worst 2.824 at or below 3.79, median relative error 3.12 at or below 3.7 percent
2026-10-01	ST6	measure	THE PRE-DECLARED ABLATION ON THE FINAL MODEL (staged run /tmp/nx_skinspec_gate_st6_4.out, the same model before its melanosome law was rerouted through the incumbent ss_melanin_mua; its per-subject lines are digit-identical to the promoted run's, subjects 55 to 60 checked, so the composition is numerically neutral): 900 nm window with the epidermis held at Jacques' 60 um, median CIEDE2000 1.074, median Delta E76 1.589, worst 5.962, median relative error 3.48 percent; visible window with the epidermis held, 1.108, 1.376, worst 3.392, 3.96 percent; model of record (visible, thickness fitted over the bar's 50 to 350 um), 0.769, 1.030, worst 2.824, 3.12 percent. The thickness and the visible window each pay for themselves on the worst spectrum, the reason the model of record was pre-declared as it was. NAMED LIMITS read off the fitted parameters: the epidermal thickness sits at the bar's 50 um floor for most subjects (median 49 um printed, integer truncation of 50), the captured specular fits to zero for 97 of 100 subjects, and the carotene absorption reaches its declared span (16 times the baseline at its peak, 19.27 per cm) on 4 subjects (17, 44, 61, 72) -- three places where a bound, not the data, decides, each a pointer to the transport and the missing chromophores rather than a defect to tune
2026-10-01	ST6	queue	NEXT, in order of what the band residuals and the bound hits say: (1) bilirubin and pheomelanin as pinned absorbers (OMLC PhotochemCAD carries bilirubin; a pheomelanin spectrum must be found and pinned before the melanin blend the bar carries can exist) -- the 400 to 450 nm residual of plus 0.018 is the blue absorption still missing; (2) the transport at the bar's fidelity class: Monte Carlo or adding-doubling over the same two layers, so the thickness and specular stop sitting on their bounds; (3) water from a pinned OMLC table so the 900 nm window joins the model of record; (4) the RGB chain switch: nx_skinfit_lib, nx_vasc and nx_follicle still carry probe coefficients and their gates (8/8, 9/9, 16/16 GREEN against the grown lib) must re-derive their moved numbers when they read km_spectra_load; (5) a mutation bite of nx_skinspec_gate over nx_kubelka_lib aimed at km_spectra_load, deferred tonight because the graphics GR59 lane imports the same shared lib concurrently and a mutated shared source must not be compiled by another lane -- run it in a quiet window (nx_gate_bite nx_skinspec_gate nx_kubelka_lib.nx 1 aim km_spectra_load); the in-gate neg-controls (wrong pin, truncated table, non-monotone table, reversed spectrum) already bite on real fixtures
2026-10-01	ST6	retract	CORRECTION TO THE COUNTS IN ROW 1790827400, which were read off a partial view of the per-subject lines rather than counted: on the promoted binary's run (/tmp/nx_skinspec_gate_final_1.out) nx_txtscan counts the epidermal thickness at the bar's 50 um floor (printed 49) for 85 of 100 subjects, the captured specular fitted to zero for 94 of 100 (not 97), and the carotene absorption at its declared span (19.27 per cm) for 8 of 100 (not 4; the four named are among them). The conclusion stands -- three places where a bound decides rather than the data -- with the counts now measured by the instrument instead of recalled
2026-10-01	ST6	land	SHIPPED AND CLEANED 2026-09-30: nx_skinspec_gate PROMOTED 5113a088 (140963 B), _offc installed, registered; nx_skinspectra_gate PROMOTED 3a30f194 (54541 B), _offc installed, registered; nx_kubelka_gate re-PROMOTED 941a87a9 against the grown lib (identical content, 37/37). Sources: nx_kubelka_lib.nx b0aa070d (29088 B), nx_skinspec_lib.nx 2a287f57 (27351 B), nx_skinspec_gate.nx 2feee31b (37188 B). Evidence banked out of /tmp: knowledge/status/skinspec_gate_st6_final_20260930.out (a1749fea, the promoted binary's 18/18 run) and knowledge/status/skinspec_gate_st6_ablation_20260930.out (the ablation run, 18/18). Both .gates rows added (knowledge/compare/skintwin.gates). Seven scratch registrations reaped (the _st6_staged and _final_staged rows) and their six staged files retired reversibly to knowledge/retired; the dev loader file retired the same way. Refs pinned this lane: nist_skin_reflectance_data_202609, nist_jres_122_026_paper, cie_cmf_1931_2deg_202609, cie_d65_202609, spectral_sss_2606_27604, jacques_skinoptics_1998, krishnaswamy_baranoski_2004, omlc_betacarotene_202609. nx_compare_rank skintwin now reads ST6 done (done 4, open 6)
2026-10-01	ST6	measure	MUTATION BITE ATTEMPTED ON THE SAFE SUBJECT, UNPROVEN: nx_gate_bite nx_skinspec_gate over nx_skinspec_lib.nx (this lane's own file, no other importer, so no other lane could compile a mutant) aim ssp_span returned verdict=UNPROVEN: the build governor refused the mutant build (exit 6) and the baseline gate run hit the harness's 120000 ms deadline (gate_exit 124), because the full 100-spectrum measurement takes about five minutes on the loaded box. Nothing was compiled or deployed: source 2a287f57 and live 5113a088 re-hashed unchanged afterwards. Standing evidence that the measurement teeth fire: the same tooth set read RED on three earlier model states this evening (12 of 16, 11 of 16, 16 of 17) before the optimiser and carotene closed them, and the in-gate neg-controls bite on real fixtures
2026-10-01	ST6	queue	A BITE MODE FOR THE MEASUREMENT GATE: nx_skinspec_gate needs an argv mode that keeps every loader tooth and fits a declared handful of named subjects (never presented as the measurement, which stays all 100) so a mutant run fits inside nx_gate_bite's deadline; then bite twice in a quiet window -- over nx_skinspec_lib.nx aim ssp_fit and over nx_kubelka_lib.nx aim km_spectra_load once no other lane is compiling the shared lib. Also owed: wire nx_skinspec_gate and nx_skinspectra_gate onto a roster beat sized for a five-minute gate
2026-10-01	ST6	measure	BACKFILL-HUMAN (f) the frontier subsurface-rendering bar: full-text witnesses now THREE groups by 2026 affiliation: Meta Reality Labs Research (Aliaga and Jarabo 2606.27604, existing full-text row aliaga_spectralsss_2606; affiliation read from the arXiv HTML knowledge/fetched/cmp_skintwin_bfh_aliaga_2606_27604.html sha256 99f6b10b); TUM with LMU Munich (DIAMOND-SSS 2601.12020, refs st6_diamondsss_2601_ft, mirror knowledge/fetched/cmp_skintwin_bfh_diamondsss_2601_12020.html sha256 8418aee7, full text read: diffusion augmentation from under 7 percent of a dataset replaces up to 95 percent of captures, up to 90 percent less real data than SSS-3DGS, its own caveat that augmented samples are not strictly physically accurate); KAIST with POSTECH and Universidad de Zaragoza I3A (polarimetric BSSRDF 2501.01980, refs st6_polarbssrdf_2501_ft, mirror knowledge/fetched/cmp_skintwin_bfh_polarbssrdf_2501_01980.html sha256 7e2a094e, full text read: per-texel 3x3 Mueller-matrix BSSRDF, biophysical hemoglobin, eumelanin and pheomelanin layers, 11 individuals of diverse skin tones). Independent count 3, with a named caveat: Jarabo's former lab is Zaragoza I3A, so the lineage overlaps the polarimetric paper's Zaragoza co-authors; held strictly by lineage the count is 2. The 42 MB arXiv PDFs are mirrored too (cmp_skintwin_bfh_diamondsss_2601_12020.pdf sha256 cbd6c8ef, cmp_skintwin_bfh_polarbssrdf_2501_01980.pdf sha256 cb3d1fe8). NIST: the critic's landing-page note is STALE: the spectra DATA file is mirrored by this board's own row nist_skin_reflectance_data_202609 (knowledge/fetched/cmp_skintwin_nist_skinrefl_v3.txt, 2,119,696 B, nx_filehash ce9aac73 equals its pin)
2026-10-01	ST6	measure	ADVERSARIAL VERIFIER, METHOD CLASS 3 (independent re-derivation by hand, a different method than the gate): read the pinned NIST table directly by byte offset (layout measured: 3 nm rows 250 to 2500, 2814 B per 3-digit row, 4 columns per subject R1 R2 R3 Average) and the 1 nm CIE 1931 2 deg and D65 tables, and summed S times cmf times R at a 9 nm subsample (45 points, 379 to 775 nm, every third point of the gate's own 133-point colour window), normalised to the D65 white. (a) WHITE POINT recomputed: Xn 95.045, Zn 108.928 (Y=100) against the published 1 nm D65 2 deg 95.047 and 108.883, deltas 0.002 and 0.045 (0.04 percent, consistent with the subsample and the 775 nm cutoff). (b) MEASURED-SIDE COLOUR by hand: subject 53 X 40.65 Y 39.38 Z 30.57, Lab 69.02 10.24 15.65; subject 54 X 41.15 Y 40.10 Z 32.35, Lab 69.54 9.53 14.04 (each average column checked against the mean of its three replicates at every sampled row). (c) COMPARISON: the banked gate output knowledge/status/skinspec_gate_st6_final_20260930.out does NOT print measured-side XYZ or Lab per subject (only fitted parameters, de2000 and de76 of model against measured, rms, relerr and r2500), so the measured-side Lab CANNOT be compared and is UNVERIFIED by this class, not refuted; the de2000 and de76 values need the model spectrum and are outside this class. The one measured-side per-subject value it prints, r2500_micro, was re-read from the raw 2500 nm row for subjects 1 to 54 and ALL 54 agree (gate prints the table value times 1e6 minus 1, a float floor, e.g. 0.0392 to 39199 for subject 53). VERDICT PARTIAL: loader reading and white point reproduced; per-subject measured Lab is a value the gate does not emit. OWED: nx_skinspec_gate to gv_kv the measured-side Lab per subject (or for named subjects) so a hand re-derivation like this one can disagree with it; expected values for 53 and 54 are above, within the 9 nm subsample bound
2026-10-01	ST6	land	BACKFILL-HUMAN refs rows landed this lane in skintwin.refs, each mirror pinned by nx_filehash, cited here so the refs referee's inline-citation tooth sees them: [st6_diamondsss_2601_ft] [st6_polarbssrdf_2501_ft]
2026-10-01	ST6	measure	HONEST NAMES, SOURCE ONLY (2026-10-01): the three T10 teeth of nx_skinspec_gate read in source (37195 B, mtime 1790829963) median-de76-measured-skin-fit-beside-paper-mc-number-3-19-cross-test, worst-de76-measured-skin-fit-beside-paper-mc-fitting-ceiling-3-79-cross-test and median-relerr-measured-skin-fit-beside-paper-mc-number-3-7-percent-cross-test, values unchanged; but the PROMOTED binary (nx_skinspec_gate.elf 140963 B, the 5113a088 of row 1790827530) still carries the old names: nx_txtscan finds median-de76-over-every-nist-spectrum-at-or-below-the-bar-paper-model-3-19 and its two siblings at offsets 138148, 138221, 138296 and counts cross-test 0 times. The rename is NOT live until the gate is rebuilt and promoted; until then every published run of the served gate still presents the cross-test as passing the paper's bar
2026-10-01	ST6	measure	MIRRORED AND READ for the next rungs (2026-10-01, all through nx_research_fetch, pinned on skintwin.refs): bilirubin [st6_omlc_bilirubin_202610] is PhotochemCAD 119 measured IN CHLOROFORM, 55000 cm-1/M at 450.8 nm, 1 nm grid -- usable through km_spectra_load with a declared solvent deviation; water [st6_omlc_water_hale73_202610] is Hale and Querry 1973 at 25 nm steps (0.00058 per cm at 400 nm, 0.0060 at 700, 0.026 at 750), coarse but the field's standard; the adding-doubling known answers [st6_prahl_ad_example_202610] are Prahl's published supplement to his 1995 chapter section 5.3.5 (quadrature, redistribution matrices, R and T of dtau 0.25, 0.5 and 1.0 layers, boundary R01 and T01 summing to 1, with errata) -- a real slab table to gate an adding-doubling arm against; but the chapter PDF that states the example's albedo and anisotropy is an image scan: nx_pdf_text emitted 8829715 B in which the word doubling occurs 0 times, so those inputs are UNREAD. The second calibration set [st6_openox_physionet_202610] is the OpenOximetry Repository on PhysioNet, credentialed: registered users who sign a data use agreement, PhysioNet Restricted Health Data License 1.5.0; its v1.0.0 landing page lists oximetry, blood gas and PPG variables and no reflectance spectra
2026-10-01	ST6	queue	OPEN AT HAND-BACK, no build started (the lane passed its build cutoff 1790841600 before this work began: NAS clock read 1790856331): (1) REBUILD AND PROMOTE nx_skinspec_gate so the cross-test names go live (nx_sov_build_run nx_skinspec_gate --build-only, nx_contentdiff live-first and read the LOST lines: the three old tooth names are the expected loss, then /api/promote with expect_sha256, nx_offc_install); (2) TRANSPORT AT THE PAPER'S LEVEL: an adding-doubling arm in NishiLang Q30 (Prahl 1995: Gaussian quadrature, Henyey-Greenstein redistribution with g = 0.62 + 0.00029 nm, doubling from a thin layer, adding the index-matched boundary) as nx_addoubling_lib, gated first against the four-point quadrature, redistribution and boundary values of the published supplement, which needs the section 5.3.5 albedo and anisotropy from a TEXT source (the CWeb listing ad_src.pdf or the iad repository's test inputs) before any tooth is written; then all 100 NIST spectra fitted with it beside the Kubelka-Munk fit; (3) bilirubin and water into the dermis through km_spectra_load, reporting the 400 to 450 nm residual (plus 0.018 of record) before and after; (4) SECOND CALIBRATION SET BLOCKED on an operator decision: the OpenOximetry data needs a PhysioNet account and a signed restricted-data use agreement, which a seat may not sign; even then its licence is restricted, so it can be a private calibration set and never a published mirror, and whether v1.0.0 holds reflectance spectra at all is unmeasured; (5) IDENTIFIABILITY under the Kubelka-Munk transport is already measured (row 1790827420: thickness at the 50 um floor for 85 of 100, specular at zero for 94, carotene at its span for 8); the same counts under the new transport wait on (2)
2026-10-01	ST6	lesson	A RENAMED TOOTH IN SOURCE IS NOT A RENAMED TOOTH: the honest names landed in nx_skinspec_gate.nx on 2026-09-30 and were never built, so the served gate kept publishing the cross-test as at-or-below-the-bar for a day while the source read correctly; only a scan of the PROMOTED binary's strings (nx_txtscan count cross-test = 0 on the live elf) told the two apart. A naming correction is shipped only when the live artifact carries it
2026-10-03	ST6	queue	OpenOximetry adjudicated against the operator licence rule (no royalty, no obligation on the product path; restricted data used only as math): the raw data IS present (openox-repo 1.0.0 on PhysioNet, spectrophotometer.csv with per-site quantitative skin colour, Monk scale, de-identified) but it is under the PhysioNet Restricted Health Data License 1.5.0 with a signed DUA, whose own text (mirrored knowledge/fetched/cmp_skintwin_physionet_restricted_license_150.html) limits use to the sole purpose of lawful scientific research and no other, forbids sharing access, and obliges contributing code associated with publications arising from the data to an open repository. Calibrating a proprietary product model is outside sole-purpose research, and publishing a number derived from it would trigger the open-code obligation against the proprietary-code policy. VERDICT: not acquired for calibration or publication; the operator may still choose a research-only private validation, which publishes nothing. Second calibration set to be sought from public-domain or permissively licensed measured skin spectra instead.
2026-10-03	ST6	land	HONEST NAMES LIVE: nx_skinspec_gate rebuilt from the 2026-09-30 source and promoted ca84de7f (162696 B, prev 5113a088 banked as .prev) under allow_capability_loss after nx_contentdiff live-first named all 7 lost runs: the four old at-or-below-the-bar tooth names and the old verdict note (the rename itself) plus two sys_mmap FATAL strings the shared nx_syscalls lib reworded to its owned-arena wording; _offc installed; the served gate re-ran GREEN 15/15 (KM model of record unchanged: median de76 1.030, worst 2.824 at subject 53, median relerr 3.12 percent). The measured-side Lab per subject the verifier asked for was already printed on every subject line (meas_L/a/b_milli): subject 53 reads 69.020 10.253 15.638 against the hand re-derivation 69.02 10.24 15.65, subject 54 reads 69.545 9.544 14.040 against 69.54 9.53 14.04 -- agreement inside the 9 nm subsample bound, so method class 3 now COMPARES rather than abstains. ADDING-DOUBLING ARM LANDED: buildroot/runtime/nx_addouble_lib.nx (Prahl 1995 section 5.3 in Q30: Gauss on [0,nu_c] plus Radau on [nu_c,1] found by sign-change scan and bisection, delta-M Henyey-Greenstein over N points, the diamond start eqs 5.54-5.56, doubling, a general four-matrix adder, Fresnel boundary layers through km_fresnel; nx_kubelka_lib untouched) and its gate nx_addouble_gate PROMOTED 5db42ea2 (114502 B), registered, _offc installed, served run 20/20 GREEN. Its known answers are PARSED FROM THE PINNED SUPPLEMENT BYTES [st6_prahl_ad_example_202610] after km_pin_match, not typed: every quadrature node and weight, every entry of h-- and h+-, A and B, R and T at dtau 0.25, 0.5 and 1.0 with fluxes, the four boundary Fresnel values and the slab between two boundaries (eqs 5.73, 5.74) agree within 6 micro (5 = half the fifth printed decimal, 1 = Q30 allowance); worst observed miss 5 micro. The example's inputs are not printed and the chapter scan has no text layer, so they were DERIVED from the table and re-proven by it: n = 1.5 (Gauss nodes 0.15751, 0.58784 = nu_c (1 -+ 1/sqrt3)/2), g = 0.9 (h(1,1) = 1 + 3 chi1 + 5 chi2 + 7 chi3 = 6.84908), a = 0.9 (scaled albedo 0.7558), dtau 0.25.
2026-10-03	ST6	lesson	A TRUNCATING MULTIPLY IS A BIAS THAT DOUBLING SQUARES: the first nx_addouble_gate run passed every published entry but read a non-scattering slab 4 micro off exp(-tau/nu) after 12 doublings and a 2^-10-start reference 330 micro off whatever the trial start, because fq_mul truncates its low partial and each doubling squares the carried relative error (2^(k+1)-1 units). The fix was in the arithmetic, not the bounds: adq_mulr rounds that one partial to nearest in the matrix kernels, after which the successive start-halving changes read 6, 0 and 4 micro and conservation 1 micro against a derived 15 micro bound. A deep-start reference is not a reference when the deepening itself adds arithmetic; successive differences are.
2026-10-03	ST6	measure	SECOND CALIBRATION SET FOUND BY THE ESTATE'S OWN HUNT, 2026-10-03: nx_frontier_scan hunt skintwin reflectance (subject skin reflectance spectral, 10 hunt rows in knowledge/research_sources.conf: Europe PMC, Zenodo, DataCite, figshare, OSF, Dryad, PhysioNet, Wikimedia Commons, Crossref for PeerJ, data.gov; all 10 reachable on the second run) read 1102 records, admitted 198 on relevance (897 irrelevant, 7 unpaired, partition sums) and classed every admitted hit from its own record's licence: ACQUIRE 114, CALIBRATE-ONLY 1, ACQUIRE-PRIVATE-REFERENCE 17, UNKNOWN-NEEDS-LICENCE-READ 65, REFUSED 1 (worklist knowledge/status/hunt_skintwin_reflectance.worklist). Top permitted human-skin set: the International Skin Spectra Archive, ISSA (Lu, Xiao, Pointer and co-authors 2025, figshare 10.6084/m9.figshare.28228571.v4) [st6_issa_lu2025_202610], CC BY 4.0 on two independent records (the figshare article record and the DataCite rightsList) [st6_cc_by_40_legalcode_202610], mirrored at 9320520 B, equal to the size both records declare, store class cc-by, attribution owed in a NOTICE. NOT YET READ: the xlsx has not been parsed, so its population, wavelength grid and instrument are UNMEASURED and the fit against it is owed. The NIST Reference Data Set of Human Skin Reflectance read UNKNOWN (its rights text is NIST Licensing, no licence the classifier names): a licence read is owed, not an acquisition. The colour hunt (skin color colorimetry) reproduced the OpenOximetry precedent mechanically: all four openox-repo versions REFUSED, class RESTRICTED-RESEARCH-ONLY, pattern restricted health data license, read from PhysioNet's own license.name, beside ENCoDE (Credentialed licence) also REFUSED (worklist knowledge/status/hunt_skintwin_colour.worklist)
2026-10-03	ST6	queue	ISSA is acquired and unread. Next: parse knowledge/fetched/cmp_skintwin_issa_v4_ISSA_17_Jan_2025_Yan_Lu.xlsx with the estate's xlsx reader (nx_xlsx), publish its population, sites, wavelength grid and instrument, and fit the spectral skin model to it as the second calibration set beside NIST's 100 spectra, with the attribution line in the release NOTICE. Read the NIST licensing page before treating the NIST set as anything but UNKNOWN
2026-10-03	ST6	retract	A FITTED DERMAL WATER FRACTION, landed in source an hour earlier and never promoted, is withdrawn: the first full Kubelka refit with nine free parameters (banked knowledge/status/skinspec_gate_st6_9param_km_partial_20261003.out, all 100 subject lines) showed water displacing the higher Jacques bloodless baseline acting as a baseline-absorption reducer that ran toward its upper bound (subject 98: blood 0, water 1.000), not a water measurement; and the flat nine-parameter search LOST subject 98's earlier optimum (Delta E76 1.059 under the seven-parameter model of record, 6.281 under the nine) although that optimum lies inside the larger space -- the search, not the model, failed. Replaced by: water HELD at W = 0.65, Jacques' generic skin dermis [st6_jacques_generic_optics_2015], as the bar fixes basal water (2606.27604 text: we fix basal concentrations for bilirubin, beta-carotene and water); bilirubin kept FITTED like carotene (a declared deviation: the bar fixes a basal value it does not state); and a NESTED fit in nx_skinspec_lib ssp_fit (the model without bilirubin fitted first, then a compass polish over all parameters that accepts only improvements), so adding an absorber can never return a worse fit than the model without it.
2026-10-03	ST6	lesson	OPENOXIMETRY AT THE NEW IDEA LEVEL (operator 2026-10-03: "we also have idea level that is just used for idea but doesnt require attribution"; researchreach RR26 log 1791045000): the openox-repo DATA stays REFUSED even at idea level, because a contract binds it -- PhysioNet grants access only on signing the Restricted Health Data License, whose data use agreement limits use to the sole purpose of lawful scientific research and no other, with code contributed back, so even reading the data for ideas happens under that contract. nx_licclass_lib lc_use_decide now says so mechanically (IDEA-ONLY on the licence name and on the mirrored licence text both REFUSE, reason CONTRACT-BINDS-USE, clause named; gate 64/64 GREEN, 8e70fe86). The PUBLISHED OpenOximetry paper is a different source: its findings, methods and reported numbers are idea-level ALLOWED -- read, cited in a refs row as provenance-not-attribution, and used to shape our own spectral and oximetry models, with no text, figure, code or data file copied. Applying it here: the paper's reported skin-tone bias findings may inform the ST6 bar and the error budget; the repository's measurements may not enter any fit
2026-10-03	ST6	retract	log 1791045002 described the OpenOximetry paper's findings (skin-tone bias may inform the error budget) from general knowledge; the paper was not fetched or read. Claim withdrawn until the paper is mirrored, pinned as a refs row flagged provenance-not-attribution, and its findings are read from the bytes. Repository data stays REFUSED (data use agreement binds).
2026-10-03	ST6	measure	THE ADDING-DOUBLING REFIT IS 35 TIMES THE TWO-FLUX COST AND WAS MADE PARALLEL RATHER THAN SAMPLED: one adding-doubling subject took 5.3 minutes serially (127 wavelengths per evaluation, each an epidermal slab of about 5 doublings plus a semi-infinite dermis of about 12, two adds and two diamond starts, over roughly 2000 nested-fit evaluations), so 100 subjects would have taken about 9 hours on one core of an 8-core host. nx_skinspec_gate sg_fit_all now forks workers by the estate's pool governor (nx_poolgov pg_width: the CPUs this process may run on, squeezed proportionally above a load equal to them), each worker fits a stripe of subjects into rows of one sys_mmap_shared region, and the parent reaps every worker and assembles and prints in subject order (the nx_parexec pattern: disjoint writes, no locks). The Kubelka arm then took about 80 s on 8 workers against about 5 minutes serially, and nx_srcdiff of the serial run's banked output (knowledge/status/skinspec_gate_st6_km_fixedwater_ad1_20261003.out) against the parallel run's matched every one of the 178 shared lines: all 100 Kubelka subject lines, every tooth and every value are byte-identical; the only differences are the new workers line and the two adding-doubling subjects the serial run had reached before it was stopped (subject 1 Delta E76 0.931, n8-vs-n4 0.248; subject 2 0.387, 0.244), which the parallel run's adding-doubling arm will be checked against.
2026-10-03	ST6	measure	RESUMED 2026-10-03 (second seat on the lane): the served nx_addouble_gate (5db42ea2, 114502 B) re-ran 20/20 GREEN through nx_job_run (every published quadrature, redistribution, slab and boundary entry within 6 micro, worst boundary miss 5 micro, conservation 1 micro, exp(-tau/nu) 3 micro, start-halving 6/0/4 micro, g=0.8 neg-control bites); it was REGISTERED-UNAUTHORISED, now authorised by a minted cap (nonce 1791058700, rows nx_addouble_gate, nx_skinspec_gate, nx_skinspec_gate_staged). The staged nx_skinspec_gate (212120 B, built 1791045564, newer than every source in its closure: gate 1791045530, skinspec_lib 1791044398, addouble_lib 1791043419) against the live ca84de7f: nx_contentdiff live-first loses exactly 2 runs, both named: the subject-line prefix ' nist subject=' (now printed per transport tag, nist or ad) and the old verdict note (reworded for two transports); gains bilirubin, water, workers and n8-vs-n4 fields. A full run of the staged gate is in flight (/tmp/skinspec_st6_full_1003b.out): the Kubelka arm finished all 100 subjects in about 90 s on 8 workers; the adding-doubling refit and its 8-point check are running.
2026-10-03	ST6	land	ISSA READ AND CONVERTED TO A NATIVE RECORD, 2026-10-03: the pinned workbook [st6_issa_lu2025_202610] (9320520 B, km_pin_match against the refs pin = 1) holds two sheets, Coding Scheme (sheet1, 21346 B of XML) and ISSA (sheet2, 42884831 B); the served nx_xlsx read only sheet1, so the datasheet was read through nx_xlsx_sheet_lib (any sheet by name, sized from the package, cells on the workbook's own row and column numbers, 118 shared strings, 879073 cells, 0 sanitized). New lib nx_skinspec_issa_lib parses it with the grid and the CIELAB columns READ FROM THE HEADER ROW (360 to 780 nm every 10, 43 wavelengths; L, a, b* by header text) and classifies every row: 15256 candidates = 15256 ok + 0 badnum + 0 badtext + 0 badgrid + 0 gap + 0 extra + 0 nolab. Population measured from the file: 15256 records (the archive's own figshare record declares 15,256), highest subject number 2113 (declared 2,113 subjects) but only 2107 distinct subject numbers occur; 11 origins (1501, 1805, 1326, 4112, 1159, 81, 1494, 941, 1062, 812, 963); 12 body sites (back of hand 1716, cheek 2511, cheek bone 232, chin 2072, ear lobe 480, forehead 2092, inner arm 1692, neck 1162, nose tip 646, outer arm 1632, palm 778, ring finger 243); ethnicity CA 2512, CN 2935, IQ 1326, TH 4112, SA 1405, AF 941, JP 1062, AB 963; gender F 9382, M 5856, empty 9, 'M ' (trailing space) 9 -- kept as written; age group unrecorded on 2716; instrument SP (spectrophotometer) on all 15256, SPIN SCI (specular included) on all 15256; TWO grids: 360-740 nm (5297 records) and 400-700 nm (9959). Instrument geometry beyond SP/SCI and the archive's diffuse-illumination protocol is NOT recorded per record in the file. Record written to knowledge/calibration/skintwin_issa_v4.issa (4590250 B; reflectance factor x1e6, CIELAB x1e6, attribution and pin in its header) and READ BACK identical field by field (0 mismatches); NOTICE beside it (knowledge/calibration/NOTICE: creators, citation, CC BY 4.0, original file and hash, the changes made). The workbook is a calibration input only, never a runtime dependency. Gate nx_skinspec_issa_gate 25/25 GREEN (pin bite, population, hand-read known answers from the raw sheet XML of rows 13 and 15268, round trip, five named refusals on a fixture datasheet, number-parser known answers, a short-row refusal bite).
2026-10-03	ST6	measure	THE SPECTRAL MODEL OF RECORD FITTED TO EVERY ISSA RECORD (15256 of 15256, no sampling) BESIDE NIST, 2026-10-03 [st6_issa_lu2025_202610]: one binary (nx_skinspec_issa, staged e1658cd9, the nx_skinspec_lib ssp_fit nested fit on the Kubelka arm, the model built on each record's OWN grid -- 380 to 740 nm for the 5297 records measured 360 to 740, 400 to 700 nm for the 9959 measured 400 to 700; fit window 400 nm to the record's end) fitted NIST's 100 subjects and every ISSA record; fit rows in knowledge/calibration/skintwin_st6_fit_.fit. METHOD CLASS 1, REPRODUCTION: the NIST run reproduces the incumbent nx_skinspec_gate's banked Kubelka output subject by subject (subject 1 Delta E2000 0.541, Delta E76 0.611, RMS 8063 micro, rel 14294 ppm; subjects 2-4 Delta E76 0.469, 0.351, 0.113 -- identical). NIST free: Delta E76 median 0.605, worst 2.054; Delta E2000 median 0.528; mean relative error median 1.63 percent. ISSA free: Delta E76 median 1.082, worst 17.496; Delta E2000 median 0.880; relative error median 2.62 percent; partition by body site sums 15256: back of hand 0.960 (n 1716), cheek 1.171 (2511), cheek bone 0.851 (232), chin 0.805 (2072), ear lobe 2.370 (480), forehead 0.901 (2092), inner arm 1.525 (1692), neck 0.854 (1162), nose tip 1.330 (646), outer arm 1.116 (1632), palm 1.390 (778), ring finger 1.125 (243); by origin 1.683, 1.727, 1.932, 1.017, 0.922, 1.101, 1.153, 0.627, 0.555, 0.418, 0.598 (origins 1-11). Beside the bar paper's own numbers (2606.27604: Delta E76 3.19 model, 3.79 ceiling, 3.7 percent relative) as a cross-test only: the paper fits its own Monte Carlo medium, so no pass or fail is drawn. Beside what the ISSA paper itself reports [st6_issa_scidata2025_202610]: it reports NO model-fit error; it reports the measurement's own variability -- short-term repeatability within 0.40 Delta Eab, pressure differences 0.88 and 1.82, aperture differences 2.48 and 3.04 -- so the ISSA median model error (1.08) sits above repeatability and below the aperture spread, and the ear lobe (2.37) and inner-arm (1.53) sites are where the model misses most. METHOD CLASS 2, A KNOWN ANSWER RE-DERIVED: our CIELAB (D65, CIE 1931 2 degree, the pinned tables) of each measured ISSA spectrum against the archive's OWN published Lab*: on the 400-700 grid Delta E76 median 0.000, worst 0.001 over 9959 records (the same arithmetic on the same band); on the 360-740 grid median 0.098, worst 0.118 over 5297 (our colour grid opens at 380 nm because the pinned haemoglobin table does, so the archive's 360 and 370 nm samples are not summed -- a declared deviation, not a disagreement). METHOD CLASS 3, HELD OUT: the three transferable parameters (scattering exponent u, captured specular, epidermal thickness) were HELD at one set's medians and only the chromophores refitted on the other. NIST medians (u -0.0840, specular 0.0278, thickness 50 um) held on ISSA: Delta E76 median 1.849; ISSA's own medians (u -0.1738, specular 0.0278, 50 um) held on ISSA: 1.902 -- NIST's medians transfer to ISSA no worse than ISSA's own. ISSA medians held on NIST: 0.836; NIST's own medians held on NIST: 0.797 -- a cross-set penalty of 0.039. The cost is freezing (ISSA 1.08 to about 1.9, NIST 0.61 to about 0.8), not the change of archive. Both sets put the median epidermal thickness on its 50 um lower bound and the median specular on the 8 degree Fresnel bound: neither is measured by these fits. Gates: nx_skinspec_issa_gate 25/25 GREEN on the NAS (promoted c55278d0, registered).
2026-10-03	ST6	queue	ISSA LANE OPEN ITEMS: (1) nx_skinspec_issa is staged (e1658cd9) and registered as nx_skinspec_issa_staged; its ship (nx_organ_ship, gate nx_skinspec_issa_gate) was refused by build admission (load1 17.0 against 16.43) and enqueued as BQ-nx_skinspec_issa -- promote and register under its own name when it fires. (2) The adding-doubling arm has not been run on ISSA (about 35 times the Kubelka cost: roughly 9 core-hours for 15256 records). (3) A colour grid that reaches 360 nm needs a haemoglobin table pinned below 380 nm; until then the 360-740 records' CIELAB omits the archive's 360 and 370 nm samples. (4) The thickness and specular medians sit on their bounds on both sets: the held-out test therefore transfers bounds, not measurements, and a wider thickness span or a measured specular capture is owed before those parameters can be called calibrated. (5) nx_skinspec_issa_gate has not been mutation-bitten with nx_gate_bite (its two named bites are hand-planted).
2026-10-03	ST6	land	PROMOTED nx_skinspec_gate 0f2a1cb7 (212120 B, prev ca84de7f banked as .prev) carrying the honest cross-test names, bilirubin and water in the dermis, the adding-doubling arm and the per-subject measured-side CIELAB; allow_capability_loss passed only after nx_contentdiff live-first named both lost runs (the per-transport subject-line prefix and the reworded verdict note). _offc installed and hash-identical to live; the scratch row nx_skinspec_gate_staged reaped as superseded (toolreg f566282c). The full run that justified the promote ran the SAME bytes (sha 0f2a1cb7 = the staged artifact, /tmp/skinspec_st6_full_1003b.out, 78794 B): 20/20 GREEN, so the served gate's verdict is that run by hash identity; it was not re-run (3 h 16 min of wall clock on an 8-core box at load 15 to 21).
2026-10-03	ST6	measure	ALL 100 NIST SPECTRA UNDER BOTH TRANSPORTS (no sampling; water held at W 0.65, bilirubin and carotene fitted, nested fit). KUBELKA arm: median Delta E76 0.605, worst 2.054 (subject 44), median Delta E2000 0.528, worst 1.558, median relative error 1.63 percent, median RMS 0.0078 -- against the model of record before bilirubin and water (median 1.030, worst 2.824) on the same spectra. ADDING-DOUBLING arm (4-point quadrature, Prahl 1995): median Delta E76 0.752, worst 2.201 (subject 95), median Delta E2000 0.616, worst 2.054, median relative error 1.86 percent; its 8-point recompute at the fitted parameters moves the colour by median 0.253 and worst 0.320 Delta E76. PARAMETERS ON A BOUND (a value on its bound was clipped, not measured), Kubelka / adding-doubling: epidermal thickness on its 50 um floor 63 / 83; bilirubin at zero 85 / 85; carotene at zero 74 / 78; specular at zero 22 / 9 and at the Fresnel ceiling 57 / 70; melanin, blood, SO2 and scattering scale on no bound under either arm. The arms disagree on scattering magnitude: median fitted scale 0.943 under Kubelka, 2.551 under adding-doubling. Bar comparison is a CROSS-TEST only: the paper's 3.19 / 3.79 / 3.7 percent are its fit of its own Monte Carlo medium, printed beside ours, no pass drawn. Three method classes: (1) gate KATs re-run (OMLC haemoglobin, NIST, CIE, beta-carotene, bilirubin 54889 at 450 nm, water 0.00058 and 0.0060 per cm) and the reproduced run; (2) the adding-doubling code held to Prahl's published supplement by nx_addouble_gate, re-run today 20/20 (every entry within 6 micro); (3) independent re-derivation: the hand CIELAB of subjects 53 and 54 (69.02/10.24/15.65, 69.54/9.53/14.04) agrees with the gate's measured-side values (69.020/10.253/15.638, 69.545/9.544/14.040) inside the 9 nm subsample bound.
2026-10-03	ST6	lesson	A TRANSPORT UPGRADE THAT FITS WORSE IS NOT A REGRESSION UNTIL THE FREE PARAMETERS ARE ASKED WHAT THEY ABSORBED: the physically stricter adding-doubling arm lands at median Delta E76 0.752 against Kubelka's 0.605 on the same 100 spectra, and its 4-point quadrature error (median 0.253) is a third of that residual, so the n4 arm is not converged relative to the number it is judged on; meanwhile Kubelka's fitted scattering scale (0.943) silently absorbs the K,S-to-mua,musp mapping that adding-doubling cannot (2.551). The bound census says what neither arm can measure from a diffuse-reflectance spectrum alone: epidermal thickness (on its floor for 63 and 83 of 100), bilirubin (zero for 85 under both) and carotene (zero for 74 and 78) are clipped, not fitted. Projection lesson: the serial 5.3 min per subject promised ~66 min on 8 workers; at load 15 to 21 the refit took 3 h 16 min -- a wall-clock budget priced at an idle box is a floor.
2026-10-03	ST6	queue	OPEN after this leg, each with its recipe: (1) ADDING-DOUBLING CONVERGENCE: refit with SG_AD_N = 8 (or the extrapolated n4/n8 pair) so the transport error sits well below the fit residual before the arm's number is compared with Kubelka's; cost is roughly 8x per evaluation, so run it when the box is idle or with a coarser fit grid declared. (2) IDENTIFIABILITY: thickness, bilirubin and carotene need a second observable (a second geometry, polarisation or a spectral window where they dominate) or must be held at a cited basal value with the deviation declared; until then they are reported as clipped. (3) OPENOXIMETRY is an OPERATOR DECISION, not a seat's: the openox-repo data sits behind PhysioNet's credentialed access and a signed Restricted Health Data License 1.5.0 data use agreement (sole purpose lawful scientific research, no sharing, code contributed back for publications); a seat may not sign it, and calibrating or publishing from it conflicts with the proprietary-code rule (log 1791040900, 1791045002). The operator may choose a research-only private validation that publishes nothing. (4) ISSA second calibration set (CC BY 4.0, mirrored, log 1791043440): the estate's xlsx reader reads only sheet1 with a fixed 8 MiB part buffer, and ISSA's sheet1 is its code legend (measured today through /api/organ_run nx_xlsx read: CODES sheet, columns A..L incl. start/end wavelength and interval); the spectra sheet is unreachable until nx_xlsx (buildroot/runtime/_hdl_build, office lane) reads a named sheet through sys_read_file -- owed there, then the fit beside NIST's 100. (5) The nx_addouble_gate _build fossil (131328 B, built 1791043420) differs from the promoted 5db42ea2 (114502 B) only by source-path build metadata (nx_contentdiff GREEN, lost 0): a lane difference, nothing to promote.
2026-10-03	ST6	measure	ADVERSARIAL VERIFY of the 2026-10-03 promote (receipts method class): nx_skinspec_gate.elf and _offc/nx_skinspec_gate.elf both sha 0f2a1cb7 212120 B (identical); nx_addouble_gate.elf and _offc both 5db42ea2 114502 B (identical); nx_txtscan on the promoted skinspec elf counts at-or-below 0 and the-bar 0 (the only 'below' is the base-class envelope text, present in .prev too); nx_contentdiff .prev->live loses exactly 2 runs (the ' nist subject=' prefix and the old verdict note), gains 18 incl. bil_abs_milli and de76_n8_vs_n4_milli. nx_addouble_gate re-ran through nx_job_run: 20/20 GREEN, boundaries worst 5 micro, conservation 1, exp(-tau/nu) 3, start-halving 6/0/4, g=0.8 neg-control bite-proven. The skinspec full-run artifact (/tmp/skinspec_st6_full_1003b.out, 78794 B) reads 20/20 GREEN with every published figure (KM median/worst dE76 605/2054 subj 44, AD 752/2201 subj 95, n8-vs-n4 253/320, bounds 63/83 thickness, 85/85 bilirubin) matching the report; the run prints no subject sha, so 'same bytes' rests on the new-only fields it emits, not on a hash; the skinspec gate was NOT re-run here (3 h on a loaded box). REFUTED IN THE REPORT: queue row 1791070503 item (4) and the report's ISSA open item say the ISSA spectra sheet is unreachable, but rows 1791060340/1791064760 already read all 15256 ISSA records via nx_xlsx_sheet_lib and fitted them; that queue item is stale and the live ISSA items are those of row 1791064800.
2026-10-03	ST6	measure	ADVERSARIAL VERIFY, independent re-derivation method class (no organ run, hand arithmetic from the pinned mirrors). (A) PRAHL SLAB TABLE vs A DIFFERENT TRANSPORT: the two starred-form slabs of the supplement (rf_116c14da.raw, a 0.9, g 0.9, n 1.5 internal, diffuse flux incidence) re-derived by similarity scaling (a' = a(1-g)/(1-ag) = 0.4737, tau' = (1-ag) tau = 0.19 tau) and the two-stream/Kubelka-Munk diffusion closure (K = 2(1-a'), S = a' per unit tau'; a_km 3.2222, b_km 3.0631): dtau 0.5 R 0.0391 against the table's flux 0.04611 (-15 percent), T 0.8658 against 0.86135 (+0.5 percent); dtau 1.0 R 0.0684 against 0.06961 (-1.7 percent), T 0.7508 against 0.75816 (-1.0 percent); both table R lie below the semi-infinite two-stream limit 0.159 and rise with thickness. Not refuted: the thin-slab R gap is the expected failure of a diffusion closure at tau' 0.095 (grazing single scatter with the forward-peaked g* 0.709 that the 0.15751 node carries, R(0.157,0.157) 0.952), and it closes to 1.7 percent at tau' 0.19; this checks the table the gate is held to, it does not re-run the gate (its 20/20 is row 1791071100's). (B) MEASURED-SIDE CIELAB, subject 1 (NIST Average column), by hand: 23 samples 385-715 nm at 15 nm on the file's 3 nm grid, CIE 1931 2 deg and D65 read from their pinned mirrors, white summed on the same samples (Xn/Yn 0.9502, Zn/Yn 1.0797): X/Xn 0.38952, Y/Yn 0.36288, Z/Zn 0.27317 -> L 66.739 a 8.524 b 12.884, against the gate's meas 66.751 / 8.457 / 12.957: Delta E76 0.10, inside the 15 nm rectangle-rule error. Not refuted. Bound stated: subject 1 only (subjects 53 and 54 were the earlier hand check), and the 15 nm subsample cannot resolve differences below about 0.1-0.2 Delta E76.
2026-10-03	ST6	measure	ADVERSARIAL VERIFY (cited-method class) of nx_addouble_lib against Prahl's own iad source, mirrored this hour: ad_phase.w (knowledge/fetched/rf_9640e416.raw sha 3bf674b1), ad_start.w (rf_992d3553.raw sha d4577c1b), ad_bound.w (rf_724cd5dd.raw sha 2cc75860). NOT REFUTED. Phase: delta-M HG chi_k=(g^k-g^M)/(1-g^M) for k<M=quad points, a=a(1-g^M)/(1-ag^M), tau=(1-ag^M)tau, h=sum (2k+1)chi_k P_k P_k -- the lib matches term for term. Quadrature: Prahl's Quadrature() is gauleg on [0,mu_c] + Radau on [mu_c,1] (Radau on [0,1] at n=1) -- matches (his prose says Radau for both halves; the code is Gauss, the lib follows the code). Diamond: r_ij=a d w_j h+-_ij/(4mu_i), t_ij=delta_ij d/(2mu_i)-a d w_j h++_ij/(4mu_i), G=(I+t-C r)/2, C=r(I+t)^-1, GR=C, G(T+I)=I -- the lib's A,B,G,R=2GB(I+A)^-1,T=2G-I is the same algebra without Wiscombe's twoaw-weighted transposed solve; equivalence is shown by every starred entry. Boundary: Boundary_RT is a diagonal Fresnel R, same both sides, T=1-R, no n^4 radiance factor (Prahl removed it) -- the lib matches; its R=1 on virtual (TIR) rows of the top boundary where Prahl uses 0 touches only the virtual rows, since T01 is 0 there. Doubling start: Prahl halves until d<=mu_min (mu_min/2 for an infinite layer); the lib halves to mu_min/4 (declared deviation 2); live gate measures shift0 (Prahl's rule) vs default at 6 micro. Gate re-run live just now: nx_addouble_gate 20/20 GREEN, mirror pin h6edc82d7 matches nx_filehash, every comparison parsed from the pinned table (96 slab entries + 32 boundary-block entries), worst miss 5 micro vs 6 tol, add==double 0 units. CAVEATS: (1) the example's inputs (n 1.5, g 0.9, a 0.9, dtau 0.25 UNSCALED, start dtau 0.10237 confirmed from A_33=0.04275) are DERIVED from the table, not read from a text source -- over-determined by ~150 entries, but still a derivation; (2) T11's tooth name claims each halving changes R less, the live values are 6 -> 0 -> 4 micro (not monotone; the tooth tests only e0>e2 and e0>e4), i.e. the residual is the Q30 rounding floor and the name overstates; (3) this verifies the transport only, not its use inside the skin fit.
2026-10-04	ST6	land	THE ADDING-DOUBLING ARM MADE FAST WITHOUT CHANGING A BIT, AND THE FITTED SPANS MADE DATA, 2026-10-03 (third seat on the lane). (1) nx_addouble_lib gained the entry points a cache needs: adq_slab_r and adq_slab_inf_r (the slab under the redistribution the context already holds; adq_slab and adq_slab_inf are now these plus adq_redist, so both paths run the same arithmetic), adq_redist_save / adq_redist_load (copy, never pointer swap) and adq_add_rt (only the reflection seen from above, adq_add's own first half operation for operation). nx_addouble_gate T14 holds them: a slab and a semi-infinite layer rebuilt from a saved redistribution after another anisotropy overwrote the context equal the direct forms at 0 Q30 units, adq_add_rt equals adq_add's RT at 0 units, and loading the wrong redistribution changes the slab by 59368761 units (0.055) -- the first run of that bite read VACUOUS because the 'wrong' anisotropy 0.8 was the skin slab's own, so the overwrite had changed nothing; it now loads 0.9. Gate 22/22 GREEN, promoted aabe3d69 (contentdiff lost 0), _offc installed. (2) nx_skinspec_lib's adding-doubling bundle carries a LAYER MEMO: per grid point the redistribution once, and three most-recently-used epidermal layers, dermal layers and finished reflectances keyed by the exact Q30 (albedo, optical thickness) inputs, so a fit that moves one parameter recomputes only the layer it touches (melanin and thickness touch the epidermis, blood, SO2, carotene and bilirubin the dermis, the captured specular neither). nx_skinspec_gate T16 (new) holds the memo to the recomputed spectrum at five vectors with a melanin-change bite. (3) The thickness span and the specular ceiling are context slots (ssp_set_spans, ssp_set_tepi); bar mode reproduces the model of record BYTE FOR BYTE: nx_skinspec_issa nist under the new build wrote 102 lines against the incumbent fit file's 102, nx_srcdiff surplus 1 and 1 = the header line only (it now names start and spans), all 100 Kubelka fit rows identical. ssp_polish refines a warm start by the compass round alone. (4) nx_skinspec_issa: options ad <N>, warm <fitfile>, spans bar

On these two registers. Rows are declared in the domain's plan file and carry the debt id, which is the join key back to the sovereign debt plane — that plane, not this page, is the authority on state. Reconciling them automatically (the regen reading the plane and refreshing these rows) is a named, owed rung; until it lands, treat an id here as a pointer to look up, not a status to trust.

Honest verdict. NO ROW ON THIS BOARD CLAIMS AN EXCEED, and the reason is stated rather than softened: the rivals are mature, and three of the four produce better-looking humans today than anything here renders. What this board asserts is narrower and checkable -- that appearance is computed FROM anatomy. A texture pipeline stores what skin looked like; a chromophore pipeline stores WHY it looked that way, which is the only form that survives a relight or an edit. The measured consequence, from the gates below: burying the same vessel returns the surface to its vessel-free baseline byte-identically, a dark shaft at depth turns blue-grey on its own from one scattering asymmetry, and a wax event propagates through erythema to rendered colour and decays on a clock. None of those three behaviours is available to a painted layer at any resolution, and none of them was tinted in by hand. What is NOT here is equally plain: the absorption coefficients are probe values awaiting measured spectra, the CIEDE2000 referee is built and not yet wired to the render lane, and hair and fabric have their transfer functions specified but not shipped. The honest summary is a proven substrate with an unproven calibration.

Person · product · place — not yet measured for this domain

Every compare carries this layer. Declare knowledge/compare/skintwin.ppp (rows surface|nishi or c1..c4|label|url|connect naming OUR live surface and each rival's front door), run nx_ppp_probe domain skintwin, and this section fills itself on the next beat: the same ruler on both sides — privacy and CX (third-party hosts, tracker classes, cookies, security headers), design and longevity (design hygiene, computed WCAG contrast, render-blocking resources, unsized media, script weight, theme and motion queries), findability (landmarks, skip link, on-site search, breadcrumb, headings, internal links).

Exclusive and shared capabilities

One capability is graded on one board; every other board references it. The fleet duplicate census (nx_rivalcensus dupes, on the fleetdupes clock row) reads every board's matrix and names each capability graded as a row on two or more boards, with a home derived from evidence: rung when exactly one board's plan contracts the symbol (that board is starred), gates when no plan contracts it and exactly one board's executed gate map runs the organ's gate (that board is plussed), ambiguous when several boards carry the deciding evidence (a decision for their seats), and none when no evidence decides it, with the reason beside it: unmapped (no board of the duplicate declares a gates map; declare the maps) or unproven (a map exists and runs none of the organ's gates; gate the organ on its proving board, or declare it a shared library by policy). A home elsewhere means this board should carry an @xrow reference to the home board's row rather than a graded row of its own; a home here means the other boards should. The table is this board's slice of one estate-wide census: the shape is shared, the rows are the board's.

This board grades 18 capabilities: 17 exclusive to it and 1 shared with other boards. Shared rows on this page: 1 = home here 0 + home elsewhere 1 + ambiguous 0 + undecided 0 + refused 0 (partition 1); undecided = unmapped 0 + unproven 0 + unevidenced 0; homonyms 0; since the previous census: new 0, resolved 0. Census FRESH age_s=296.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
cs_de2000	`nx_colorsci_lib.nx`	phototwin+, skintwin	phototwin	rung	plan+gates	home elsewhere adopt: `@xrow phototwin\|cs_de2000\|home by rung (fleet duplicate census)`

Gaps from the record — what the estate does that no board carries

The record census (nx_goalmap record) reads the invoked-tool population and every plan queue row and files each organ or directive that NO matrix, plan or gates row names. A row here is a callout the boards missed: adjudicate it onto a board or declare it infrastructure. Census state STALE (age 1396743 s), sources read 4 of 7 declared — a BLIND census is a FLOOR: unread sources can only add rows.

kind	name	board	source	evidence

rows shown 0|this board's directives 0|estate-wide un-boarded organs 452 (listed in full on /compare/ecosystem)|census rows 639|malformed 0 (counted, never rendered)

References

Beyond a link list. Every reference below resolves twice — the publisher's copy and, where banked, the estate's own non-rottable library mirror with a content pin — and carries its evidence class plus the exact claim on this page it grounds. Keyed marks like [key] in the matrix notes jump here. A dash means honestly absent, never assumed.

[omlc_hb] Prahl S, Tabulated Molar Extinction Coefficient for Hemoglobin in Water, compiled from W B Gratzer (Medical Research Council Labs, Holly Hill, London) and N Kollias (Wellman Laboratories, Harvard Medical School, Boston) publisher · read in our library knowledge/fetched/cmp_skintwin_hemoglobin_omlc.html · pin hc5b3a312f3c8e135bf69845ba11afa94d92db0a3f784d344d92afcf9dc290c66 · accessed 2026-09-03 · datasetGrounds: CALIBRATION: measured absorption spectra as pinned data
[omlc_mel] Jacques S L and McAuliffe D J 1987, and Jacques, Glickman and Schwartz 1996, Melanosome Absorption Coefficient, Oregon Medical Laser Center. Melanosome absorption measured by the threshold for explosive vaporization under pulsed lasers, with a typical value of 296 per cm at 694 nm publisher · read in our library knowledge/fetched/cmp_skintwin_melanin_omlc.html · pin h0189e5bfd62a258bace15270d5e65615c1a6ee7093583766b308d8b40cd96d56 · accessed 2026-09-03 · datasetGrounds: CALIBRATION: measured absorption spectra as pinned data
[infinigen_rel] Infinigen release feed, princeton-vl/infinigen, read live from the GitHub releases API unauthenticated. tag_name v2.0.0a1, which also settles that Infinigen 2 exists as an alpha publisher · read in our library knowledge/fetched/cmp_skintwin_bar_infinigen.json · pin ha19c3b5689874ed550a7ddf0ecbc05588edca8c5481295bf02bde76952a5edba · accessed 2026-09-03 · vendor-docGrounds: CALIBRATION: measured absorption spectra as pinned data
[arxiv_hair_202609] arXiv API listing, hair AND strand, newest first, 20 entries, read 2026-09-16: HairCS 2609.16465 with a public strand dataset, strand hairstyle generation 2608.13679 publisher · read in our library knowledge/fetched/cmp_charsim_arxiv_hair_202609.xml · pin h464a17f280ce33ac80d731b9f5b3013710c0f20e9ae1959a95befe2f7f7c6cf5 · accessed 2026-09-16 · mirrored-indexGrounds: SB-HAIRSTRAND-2609
[arxiv_fabric_202609] arXiv API listing, sheer or translucent or thin fabric or woven or yarn AND cloth or fabric or garment AND rendering or appearance or BSDF, 17 entries, newest 2026-06-02, read 2026-09-27: FabricGen microstructure-aware woven fabric generation 2603.07240 (2026-03), fiber-level woven fabric capture from a single photo 2409.06368, reflection-transmission woven capture 2406.19398, real-time neural woven fabric rendering 2406.17782, a multi-scale yarn appearance model with fiber details 2401.12724, a neural cloth appearance model 2311.04061, mechanics-aware cloth appearance 1904.11116 publisher · read in our library knowledge/fetched/cmp_skintwin_arxiv_fabric_202609.xml · pin h503eb7f8735c9bc97d3eed6872b5a6297d36a118644e935c1ca502906e99e106 · accessed 2026-09-27 · mirrored-indexGrounds: SB-FABRIC-2603
[arxiv_wet_202609] arXiv API listing, wet AND skin or hair or sand AND rendering or appearance or reflectance, 11 entries, read 2026-09-27: the only rendering entry is the micro-ellipsoid model for wet porous materials 2401.15628; the newest entry (2512.11025) is a quantum-gravity homonym on the phrase wet hair; no 2026 wet-appearance paper exists in the listing publisher · read in our library knowledge/fetched/cmp_skintwin_arxiv_wet_202609.xml · pin h604dd11ddb72f0f26c2bf4c64c3323050c5d3ca9a19829978c861244143931c3 · accessed 2026-09-27 · mirrored-indexGrounds: SB-WET-2401
[arxiv_sss_202609] arXiv API listing, subsurface scattering AND skin or real-time or face, 14 entries, newest 2026-06-25, read 2026-09-27: Spectral Subsurface Scattering from RGB via Biophysical Skin Inversion 2606.27604, DIAMOND-SSS 2601.12020, 3DPR single-image portrait relighting 2510.15846, HumanMaterial 2507.18385, polarimetric BSSRDF acquisition of dynamic faces 2501.01980, subsurface scattering for 3D Gaussian splatting 2408.12282 publisher · read in our library knowledge/fetched/cmp_skintwin_arxiv_sss_202609.xml · pin h566ce40d54b910bd269c422d2f37e8ef27dfab76376eb260903a9603dd96a9e7 · accessed 2026-09-27 · mirrored-indexGrounds: SB-SKINAPP-2606
[arxiv_co_siga2026] arXiv API venue-tag scan, comment phrase SIGGRAPH Asia 2026, 55 entries, newest 2026-09-24, read 2026-09-27: Heartian physiology-aware relightable Gaussian head avatar 2609.28539 (2026-09-22) models intrinsic facial appearance as time-varying physiology rather than a static texture publisher · read in our library knowledge/fetched/cmp_graphics_arxiv_co_siga2026.xml · pin h51d49183aafe7aa1b0dbf264bc0c029b4716c07ea617c03a35fa2b58286599df · accessed 2026-09-27 · mirrored-indexGrounds: SB-PHYSAPP-2609
[nist_skin_reflectance_202609] NIST, Reference Data Set of Human Skin Reflectance: 100 reflectance spectra of human skin from 250 nm to 2500 nm acquired with a commercial spectrophotometer and traceable to the national scale for directional-hemispherical reflectance factor; the publication page mirrored 2026-09-27 publisher · read in our library knowledge/fetched/cmp_skintwin_nist_skin_reflectance_202609.html · pin h5630d69eaa5bd2ba7c801de22f6d27e3dbaa27aa5ffff60d191d8de120a7af6a · accessed 2026-09-27 · datasetGrounds: SB-SKINSTD-2609
[iso_dtr_26478_202609] ISO/DTR 26478, Digital imaging, skin tone representation for use in photographic testing including test charts and test spectra: selection of colour patches and spectral reflectance data representing skin tones; the draft technical report's vote runs 2026-08-05 to 2026-09-30; catalogue page mirrored 2026-09-27 publisher · read in our library knowledge/fetched/cmp_skintwin_iso_dtr_26478_202609.html · pin h24b1d7d840b764793c8b952185d6983fc08bf23f7037700d671319ea656ff035 · accessed 2026-09-27 · published-standardGrounds: SB-SKINSTD-2609
[metahuman_58_notes_202609] MetaHuman 5.8 release notes on the Epic developer documentation site, mirrored 2026-09-27 (259454 bytes): OpenRigLogic MIT release of RigLogic and DNA, a Devkit outside Unreal, unbaked texture and material control, automatic palm and sole skin colour, nail parameters, a neck-seam skin-tone fix publisher · read in our library knowledge/fetched/cmp_anatomy_metahuman_58_release_notes_202609.html · pin h19ff63681ba925d4a06cfcdc51cd16104b92160a846c5cc05aabe6b412d33953 · accessed 2026-09-27 · vendor-docGrounds: SB-MHSKIN-2609
[jensen_wet_1999] Jensen, Legakis and Dorsey 1999, Rendering of Wet Materials, Proceedings of the 10th Eurographics Workshop on Rendering, mirrored from the Eurographics digital library and read 2026-09-28 through nx_pdf_text: Sec. 2.1 a thin water film on a Lambertian surface darkens it because total internal reflection at the water-air boundary sends returned light back for further absorption (Lekner and Dorf 1988); Sec. 2.2 water beneath the surface lowers the relative index, narrows scattering and raises absorption (Twomey et al. 1986); Sec. 3.1 a two-layer surface model with Fresnel at the air-liquid and liquid-material interfaces, Eq. 1 to 4, a roughness exponent k on the second interface; Sec. 3.2 a two-term Henyey-Greenstein phase function whose forward lobe rises with wetness (paper 0.05 to 0.8, sand 0.2 to 0.7) publisher · read in our library knowledge/fetched/cmp_skintwin_jensen_wet_1999.pdf · pin h9eff4a2f34eb0255f1518cd3f3e037b748a25efcc530ef5ca5a9073f32886a7d · accessed 2026-09-28 · published-paperGrounds: ST7
[pan_wetporous_2025] Pan et al. 2025, A Micro-Ellipsoid Model for Wet Porous Materials Rendering, ACM Transactions on Graphics (arXiv 2401.15628v3), mirrored and read 2026-09-28: a generalized anisotropic radiative transfer equation with porosity and saturation, a transmittable micro-ellipsoid particle model and the WetSpongeCake position-free BSDF; it places Lekner and Dorf 1988 and Bajo et al. 2021 as the surface-based models and Jensen 1999 as the combined surface and subsurface model lacking physically meaningful parameters; a higher liquid index and saturation darken by forward scattering and absorption publisher · read in our library knowledge/fetched/cmp_skintwin_wetporous_2401_15628.pdf · pin h486a97270604ad896853838345e02e7b3a83260aa691ab3af41c0ca1437f985e · accessed 2026-09-28 · published-paperGrounds: ST7
[judd_1942_diffuse_fresnel] Judd 1942, Fresnel reflection of diffusely incident light, Journal of Research of the National Bureau of Standards 29, 329 (RP1504), mirrored from NIST and read 2026-09-28 through nx_pdf_text: Table I tabulates, per relative index m, the reflectance at perpendicular incidence and the reflectance for completely diffuse incidence from the rarer (external) and the denser (internal) side, computed by summing the Fresnel formula over 0.04 radian intervals (0.0005 radian across the critical angle) with Moon's tables, uncertainty stated as less than 2 in the last figure; the m = 1.33 row reads 0.02006, 0.0660 and 0.472 publisher · read in our library knowledge/fetched/cmp_skintwin_judd_1942_diffuse_fresnel.pdf · pin h5cb45d3699f518d83c073d5cffb597bb4c43a68c44aa10e636f039f40a785631 · accessed 2026-09-28 · published-paperGrounds: ST7
[nist_skin_reflectance_data_202609] Cooksey C C, Allen D W and Tsai B K 2017, Reference Data Set of Human Skin Reflectance, J. Res. NIST 122:26, the DATA FILE behind data DOI 10.18434/M38597 (resolved 2026-09-30 to the NIST MIDAS object below): column 0 the wavelength, then per subject R1, R2, R3 and their Average, 100 subjects, the 8/h reflectance factor of the bare right forearm from 250 to 2500 nm at 3 nm publisher · read in our library knowledge/fetched/cmp_skintwin_nist_skinrefl_v3.txt · pin hce9aac735d9d83b4b874b98d45f1ebda682cdc0e652f804efe00ab2fce0c25d9 · accessed 2026-09-30 · datasetGrounds: ST6
[nist_jres_122_026_paper] Cooksey, Allen and Tsai 2017, Reference Data Set of Human Skin Reflectance, J. Res. NIST 122:26, the paper, read 2026-09-30 through nx_pdf_text: the subject places the bare right forearm at the sample port of the integrating sphere; the 8/h spectral reflectance factor over 250 to 2500 nm at a 3 nm interval, 3 scans per subject, a 3 nm bandwidth to 860 nm, traceable to the NIST STARR scale through sintered PTFE publisher · read in our library knowledge/fetched/cmp_skintwin_nist_jres_122_026.pdf · pin h582a1d5a2645e5ca365cdcd2f2e472ad6407d957280fd35cc2565a9f0949a180 · accessed 2026-09-30 · published-paperGrounds: ST6
[cie_cmf_1931_2deg_202609] CIE data table, CIE 1931 colour-matching functions, 2 degree observer, 360 to 830 nm at 1 nm, the CSV the CIE datatable page links (page mirrored as knowledge/fetched/cmp_skintwin_cie_1931_2deg_page.html) publisher · read in our library knowledge/fetched/cmp_skintwin_cie_xyz_1931_2deg.csv · pin hfa663e3535a7e0763a745993a1f0a192eb0275ac46ad2d1befd7626841e713c1 · accessed 2026-09-30 · published-standardGrounds: ST6
[cie_d65_202609] CIE data table, CIE standard illuminant D65 relative spectral power distribution, 300 to 830 nm at 1 nm publisher · read in our library knowledge/fetched/cmp_skintwin_cie_std_illum_d65.csv · pin he76f210bffff3d552ef7113025da5f325d5dfec200dd4b878b1a2f3a507032cb · accessed 2026-09-30 · published-standardGrounds: ST6
[spectral_sss_2606_27604] Aliaga C and Jarabo A 2026, Spectral Subsurface Scattering from RGB via Biophysical Skin Inversion, arXiv 2606.27604 (Meta Reality Labs Research), read 2026-09-30 through nx_pdf_text: a two-layer skin (an epidermis 50 to 350 um thick, melanin-dominated, Jacques scattering with g = 0.62 + 0.00029 nm, over a semi-infinite dermis carrying haemoglobin, bilirubin, beta-carotene and water) with five biophysical parameters and GPU Monte Carlo ground truth at 376 wavelengths; Table 1 rendering error Delta E76 on the visible band over 10 representative tones 3.19 (K=3 neural decoder) and 3.79 (K=3 joint optimisation); mean reflectance error 3.7 percent for K=3; its validation relies on Monte Carlo simulation and fitting measured data is named as future work publisher · read in our library knowledge/fetched/cmp_skintwin_spectral_sss_2606_27604.pdf · pin h4f662f44e01fa0f8470b88e4868786a0dbb3ebf845e40fe116475225144215c6 · accessed 2026-09-30 · published-paperGrounds: SB-SKINAPP-2606
[jacques_skinoptics_1998] Jacques S L 1998, Skin Optics, Oregon Medical Laser Center News, January 1998: baseline absorption 0.244 + 85.3 exp(-(nm - 154)/66.2) per cm (Sec. 1.1), melanosome volume fractions under a 60 um epidermal thickness (Sec. 1.3), epidermal and dermal absorption as volume-fraction mixes (Sec. 1.4 and 2.3), reduced scattering 2e5 nm^-1.5 + 2e12 nm^-4 per cm (Sec. 2.4) publisher · read in our library knowledge/fetched/cmp_skintwin_jacques_skinoptics_1998.html · pin h551d1255c089c85b7c780f258e9ad44c475677d816da2f529e3d6c0e587bb083 · accessed 2026-09-30 · published-paperGrounds: ST6
[krishnaswamy_baranoski_2004] Krishnaswamy A and Baranoski G V G 2004, A Study on Skin Optics, University of Waterloo technical report CS-2004-01, read 2026-09-30 through nx_pdf_text: its parameter table gives refractive indices stratum corneum 1.55 [Anderson and Parrish 1981], epidermis 1.4 [Tuchin 2000], papillary dermis 1.36 and reticular dermis 1.38 [Jacques et al. 1987] publisher · read in our library knowledge/fetched/cmp_skintwin_krishnaswamy_baranoski_skinoptics_2004.pdf · pin hc20831bcea1846c0e58e32a0aa3baf8683f9086945f63503855cee81886a3177 · accessed 2026-09-30 · published-paperGrounds: ST6
[arxiv_skinspecrefl_202609] arXiv search listing, abstract terms skin spectral reflectance, ordered newest submission first, 28 results, read 2026-10-01 (UTC): the newest on-subject entry is 2609.30122 Melanin- and linewidth-corrected pulse-oximetry (Blaney, Frias, Durbha, Fantini, Koomson, submitted 2026-09-24), which fits the empirical, the diffuse-optics and a melanin-corrected diffuse-optics calibration to 98 patients with paired arterial draws, photoplethysmograms and skin reflectance spectra in the public OpenOximetry Repository and reports the saturation error's melanin dependence falling from +2.8 to +1.2 to +0.1 percentage points per unit melanin volume fraction, none statistically resolved at that cohort size; the newest rendering entry is still Spectral Subsurface Scattering from RGB via Biophysical Skin Inversion 2606.27604 (2026-06-25); no October 2026 entry; the remaining hits are homonyms (non-Hermitian skin effect, accretion-disc skin, EM skins) publisher · read in our library knowledge/fetched/cmp_skintwin_arxiv_skinspecrefl_202609.html · pin h31e754f7d45dd138c0f133e5b00689fbca93dbae34b1e5ac92671e3e64d20279 · accessed 2026-10-01 · mirrored-indexGrounds: SB-SKINAPP-2606
[gitlina_spectralskin_2020] Gitlina, Guarnera, Dhillon, Hansen, Lattas, Pai and Ghosh 2020, Practical Measurement and Reconstruction of Spectral Skin Reflectance, Computer Graphics Forum 39(4), Eurographics Symposium on Rendering 2020 (Imperial College London with NTNU, University of York, Clemson University and UBC), mirrored from the Imperial group page and read 2026-10-01 through nx_pdf_text: two practical spectral measurement methods for live subjects drive a spectral BSSRDF of appropriate complexity to match skin in photographs; a multispectral LED sphere under two complementary uniform spectral illuminations estimates spatially varying melanin and hemoglobin concentration, melanin blend-type fraction and epidermal hemoglobin fraction better than standard broadband illumination; a neural network replaces lookup-table search for the inversion and avoids parameter quantization; its related work names the lineage: Donner and Jensen multipole multi-layer diffusion, their two-layer spectral BSSRDF with epidermis and dermis parameterised by chromophore concentrations, the Donner et al. 2008 extension with epidermal hemoglobin fraction, inter-layer absorption and path-traced heterogeneous scattering, and Tsumura et al. 2003 ICA melanin and hemoglobin maps from ordinary photographs publisher · read in our library knowledge/fetched/cmp_skintwin_gitlina_spectralskin_2020.pdf · pin h89cc4c7f8d606381673755ed5351649b0d851f6211fb045da92ff0841baebbb7 · accessed 2026-10-01 · published-paperGrounds: SB-SKINAPP-2606
[arxiv_sss_202610] arXiv search listing, abstract phrase subsurface scattering with skin, ordered newest submission first, 8 results, read 2026-10-01 (UTC) as a re-check of SB-SKINAPP-2606: the newest entry is still Spectral Subsurface Scattering from RGB via Biophysical Skin Inversion 2606.27604 (Aliaga and Jarabo, submitted 2026-06-25), then DIAMOND-SSS 2601.12020 (Figueroa-Araneda, Jimenez, Hofherr, Ko, Andrade-Loarca, Cremers, 2026-01-17), HumanMaterial 2507.18385, SFDM, Polarimetric BSSRDF Acquisition of Dynamic Faces 2501.01980, physics-based blemish retouching in a melanin and haemoglobin space, a light-sampling-field BRDF paper and AvatarMe++; no July, August, September or October 2026 skin-scattering entry exists in it publisher · read in our library knowledge/fetched/cmp_skintwin_arxiv_sss_202610.html · pin h289848688d9e47a9256d59e64f0b44a1c1c7c08f000f7286638e67b0fcff518d · accessed 2026-10-01 · mirrored-indexGrounds: SB-SKINAPP-2606
[amano_naturaltan_2020] Amano K, Xiao K, Wuerger S, Meyer G 2020, A colorimetric comparison of sunless with natural skin tan, PLoS ONE 15(12):e0233816, CC BY 4.0, JATS XML read 2026-09-30: control group n=60, inner forearm (infrequently sun-exposed) against outer forearm (frequently sun-exposed) defined as the naturally occurring tan; Delta E out-inn mean 6.03 SD 2.64; outer-forearm ITA mean 38.8; a natural tan is a decrease in lightness without an increase in yellowness publisher · read in our library knowledge/fetched/rf_803ff2ff.raw · pin hd24c54fcc23d79b3cf224bf228b222c32e1669bb1a5064a2622730b53e5a9515 · accessed 2026-09-30 · published-paperGrounds: ST1
[ravnbak_tankinetics_2008] Ravnbak MH, Philipsen PA, Wiegell SR, Wulf HC 2008, Skin pigmentation kinetics after UVB exposure, Acta Derm Venereol 88:223-228, doi 10.2340/00015555-0431, open access, read 2026-09-30 through nx_pdf_text: Table I dose to minimal pigmentation per Fitzpatrick type and pre-exposure pigmentation; Table II absolute steady-state pigmentation increase at 0.4, 0.6 and 0.8 MMD after 6 and 12 exposures, independent of skin type; the dose fraction to minimal pigmentation lowered by a factor 2 after 6 and 3 after 12 exposures; mean days to total fading 132 to 175 by linear regression; peak one week after the last exposure publisher · read in our library knowledge/fetched/rf_e2ec0d52.raw · pin hacd1bc8a51fcaf8d2de21ea7950bfb0c7a08abb75a098aebfd0bd1d4d53f3450 · accessed 2026-09-30 · published-paperGrounds: ST1
[aliaga_spectralsss_2606] Aliaga and Jarabo 2026, Spectral Subsurface Scattering from RGB via Biophysical Skin Inversion, arXiv 2606.27604v1 (submitted 2026-06-25, cs.GR), full PDF mirrored and read 2026-10-01 through nx_pdf_text: skin is a two-layer medium, an epidermis 50 to 350 micrometres thick with melanin-dominated absorption and Jacques scattering with g = 0.62 + 0.00029 lambda over a semi-infinite dermis carrying haemoglobin, bilirubin, beta-carotene and water, each tone defined by five properties (melanin and blood fraction, epidermal thickness, blood oxygenation, eumelanin to pheomelanin ratio); a GPU Monte Carlo photon simulator propagates 376 wavelengths from 250 to 1000 nm and records both total diffuse reflectance and the radial diffusion profile in 512 log-spaced bins to 5 cm, the authors stating that prior spectral skin datasets storing only hemispherical reflectance cannot characterise translucency; 25000 training tones at one million photons per wavelength feed a chained neural decoder from RGB albedo to a mixture of three media for a random-walk path tracer; it positions itself after Gitlina et al. 2020 and Aliaga et al. 2023 as moving single-image biophysical inversion into production path tracing publisher · read in our library knowledge/fetched/cmp_skintwin_aliaga_spectralsss_2606_27604.pdf · pin h4f662f44e01fa0f8470b88e4868786a0dbb3ebf845e40fe116475225144215c6 · accessed 2026-10-01 · published-paperGrounds: SB-SKINAPP-2606
[omlc_betacarotene_202609] Prahl S (OMLC), Beta-carotene absorption from PhotochemCAD 2.1a (Du et al. 1998, Dixon et al. 2005): measured by R.-C. A. Fuh in 1995 on a Cary 3 in hexane at a 0.25 nm data interval and scaled so the molar extinction is 139500 cm-1/M at 451 nm (Zechmeister and Polgar 1943); the tabulation ends at 700 nm; one of the dermal chromophores the bar's model carries (page mirrored as knowledge/fetched/cmp_skintwin_betacarotene_omlc_page.html) publisher · read in our library knowledge/fetched/cmp_skintwin_betacarotene_omlc_abs.txt · pin h8de6dc03df3c8a88b906a8dcbad024ffd4f2b84903079e4e589f16fa0d322f3d · accessed 2026-09-30 · datasetGrounds: ST6
[st6_diamondsss_2601_ft] Figueroa-Araneda, Jimenez, Hofherr, Ko, Andrade-Loarca and Cremers (Technical University of Munich; LMU Munich; an independent researcher), DIAMOND-SSS: Diffusion-Augmented Multi-View Optimization for Data-efficient SubSurface Scattering, arXiv 2601.12020, FULL TEXT read 2026-10-01 from the arXiv HTML (the 42.9 MB PDF is also mirrored at knowledge/fetched/cmp_skintwin_bfh_diamondsss_2601_12020.pdf): builds on SSS-3DGS; diffusion models fine-tuned on under 7 percent of the dataset synthesise novel views and relighting that replace up to 95 percent of missing captures, with illumination-independent multi-view silhouette and depth consistency losses; reconstruction from as few as ten real captures, up to 90 percent less real data than prior SSS-3DGS; its own stated limit: diffusion-augmented samples are not strictly physically accurate and may carry slight colour or scattering biases publisher · read in our library knowledge/fetched/cmp_skintwin_bfh_diamondsss_2601_12020.html · pin h8418aee73b0549b0c99d931450e081da255f51444a31e182a347c4c72c7f9498 · accessed 2026-10-01 · published-paperGrounds: ST6
[st6_polarbssrdf_2501_ft] Ha H, Hwang I, Monzon N, Cho J, Kim D, Baek S-H, Munoz A, Gutierrez D and Kim MH (KAIST; POSTECH; Universidad de Zaragoza I3A), Polarimetric BSSRDF Acquisition of Dynamic Faces, arXiv 2501.01980, FULL TEXT read 2026-10-01 from the arXiv HTML (PDF also mirrored at knowledge/fetched/cmp_skintwin_bfh_polarbssrdf_2501_01980.pdf): a per-texel polarimetric BSSRDF stored as a 3x3 Mueller matrix of linear polarization, nine times the size of a conventional BSSRDF, with index of refraction, specular and single-scattering roughness and intensity, heterogeneous subsurface scattering, and biophysical inner- and outer-layer hemoglobin, eumelanin and pheomelanin recovered from multispectral absorption; inner-layer reduced scattering fixed at 50 percent of the outer; 11 individuals of diverse skin tones captured dynamically; code at github KAIST-VCLAB; venue not stated in the mirrored HTML header publisher · read in our library knowledge/fetched/cmp_skintwin_bfh_polarbssrdf_2501_01980.html · pin h7e2a094efdc562706f0987a95d63c9c314908c120b7a4cbb0de5e786e7489414 · accessed 2026-10-01 · published-paperGrounds: ST6
[st6_omlc_bilirubin_202610] Prahl S (OMLC), Bilirubin absorption from PhotochemCAD 2.1a (Du et al. 1998, Dixon et al. 2005): measured by J. Li on 1997-12-11 on a Cary 3 IN CHLOROFORM at 0.25 nm data interval, scaled so the molar extinction is 55000 cm-1/M at 450.8 nm (Agati and Fusi 1990), interpolated to 1 nm; the dermal bilirubin the bar's model carries (solvent is chloroform, not plasma: a declared deviation when it enters the dermis) publisher · read in our library knowledge/fetched/rf_ddf2ca34.raw · pin h6efad57b3e5472011a83e86605e95e0c6fe4ea389f3767df38e9570993016604 · accessed 2026-10-01 · datasetGrounds: ST6
[st6_omlc_water_hale73_202610] Hale G M and Querry M R 1973, Optical constants of water in the 200 nm to 200 micron wavelength region, Appl. Opt. 12, 555-563, as tabulated by OMLC: absorption coefficient in 1/cm at 25 nm steps in the visible (400 nm 0.00058, 700 nm 0.0060, 750 nm 0.026); the water absorber for the dermis so the 900 nm window can join the model of record publisher · read in our library knowledge/fetched/rf_73ea69ee.raw · pin hc636c60fc0f44c6630de1dd5b9291139837fd2ce3aeea943da07fcd06efdeb6b · accessed 2026-10-01 · datasetGrounds: ST6
[st6_prahl_ad_example_202610] Prahl S A, numerical results supplementing The Adding-Doubling Method (Optical-Thermal Response of Laser Irradiated Tissue, Welch and van Gemert eds, Plenum 1995, pp 101-129) section 5.3.5: four-point quadrature angles and weights, the redistribution matrices, R and T of a dtau 0.25, 0.5 and 1.0 layer before and after the boundaries (eq 5.65 to 5.76), with the author's errata for 5.64 and 5.72; the published known-answer table an estate adding-doubling transport arm must reproduce. The chapter PDF (knowledge/fetched/rf_243672e6.raw, 17151649 B, sha256 fe1bea40) is an image scan: nx_pdf_text extracted no text from it, so the example's albedo and anisotropy are UNREAD and must come from a text source before the table can be gated publisher · read in our library knowledge/fetched/rf_116c14da.raw · pin h6edc82d7350e879757fb1506c1c8b11ad31de682f71ce98d12efebf093d88418 · accessed 2026-10-01 · published-courseGrounds: ST6
[st6_openox_physionet_202610] OpenOximetry Repository v1.0.0 on PhysioNet (doi 10.13026/cc78-ad74, published 2024-02-27), the repository 2609.30122 names for its 98 patients' skin reflectance spectra: ACCESS POLICY only registered users who sign the data use agreement, LICENSE PhysioNet Restricted Health Data License 1.5.0; the landing page lists oximetry, blood gas and PPG variables and does not mention reflectance spectra, so whether v1.0.0 carries them is UNMEASURED; landing page mirrored, data NOT mirrored publisher · read in our library knowledge/fetched/rf_418b7fa9.raw · pin h204f417ac3443e5d6aae3d4c201f4d128fc0227d133b8dc4277369232b6ebc2b · accessed 2026-10-01 · datasetGrounds: ST6
[st6_issa_lu2025_202610] Lu Y, Xiao K, Pointer M, He R, Zhou S and co-authors 2025, The International Skin Spectra Archive (ISSA): a multicultural human skin phenotype and colour spectra collection, figshare dataset doi 10.6084/m9.figshare.28228571.v4 (file ISSA_17_Jan_2025_Yan_Lu.xlsx, 9320520 B, figshare computed_md5 67e15398f96dd9af08b5d209d2522e19), EPSRC EP/L001012/1 and EP/K040057/1; licence CC BY 4.0 read from TWO independent records (the figshare article record license.name and the DataCite rightsList rightsIdentifier cc-by-4.0); found by nx_frontier_scan hunt skintwin reflectance and classed ACQUIRE / store cc-by by nx_licclass_lib; mirrored byte count equals the size both records declare; the second calibration set for the spectral skin model, attribution owed in a NOTICE publisher · read in our library knowledge/fetched/cmp_skintwin_issa_v4_ISSA_17_Jan_2025_Yan_Lu.xlsx · pin h7396aa7608f1a6f91ec70bf8f8421252a9d81f71a2bca066eb7e98155b9d1161 · accessed 2026-10-03 · datasetGrounds: ST6
[st6_cc_by_40_legalcode_202610] Creative Commons Attribution 4.0 International Public License, legal code (plain text), the licence the ISSA dataset carries; read in full: Section 3(a) attribution conditions apply when You Share (identify the creator, keep the notices, indicate changes, link the licence), Section 2(b)(3) waives royalties for the licensed rights, there is no share-alike term, and the phrase public domain (with CC0) occurs once, in the closing Creative Commons notice that dedicates the licence TEXT to the public domain -- which is why the licence classifier runs its attribution rows before its public-domain rows publisher · read in our library knowledge/fetched/cmp_skintwin_cc_by_4_0_legalcode.txt · pin h9ba9550ad48438d0836ddab3da480b3b69ffa0aac7b7878b5a0039e7ab429411 · accessed 2026-10-03 · published-standardGrounds: ST6
[st6_jacques_generic_optics_2015] Jacques S L 2015, Generic tissue optical properties, Oregon Medical Laser Center news, page dated Feb. 24, 2015, read 2026-10-03: tissue absorption as volume fractions of blood (B, oxygen saturation S), water (W, Hale and Querry 1973), melanosomes (M) and fat (F); its worked skin-dermis example B = 0.002, S = 0.70, W = 0.65, M = 0, F = 0, reduced scattering 43 per cm at 500 nm with Rayleigh fraction 0.64 and Mie power 0.91; the dermal water volume fraction the ST6 model now holds fixed, as the bar (2606.27604) fixes basal water rather than fitting it publisher · read in our library knowledge/fetched/rf_422ae4a6.raw · pin h3f427240302c86ab9c98eb25b97ea41836e572d76107499047e15466f91945f2 · accessed 2026-10-03 · published-courseGrounds: ST6
[st6_issa_scidata2025_202610] Lu Y, Xiao K, Pointer M, et al. 2025, The International Skin Spectra Archive (ISSA): a multicultural human skin phenotype and colour spectra collection, Scientific Data 12 (2025), doi 10.1038/s41597-025-04857-5 (found by a Crossref bibliographic query for the dataset title, mirrored as knowledge/fetched/cmp_skintwin_issa_crossref_query_20261003.json; the figshare record links no paper); read 2026-10-03 from the mirrored article HTML: datasets 1, 2, 3, 6 and 10 measured with a Konica Minolta CM-2600d, datasets 5, 7, 8, 9 and 11 with a CM-700d, dataset 4 with an X-Rite SP62, all di:8 degree geometry (diffuse illumination, 8 degree detection) with the specular component included; Technical Validation reports short-term repeatability of the spectrophotometer within 0.40 Delta E*ab, mean differences between measurement pressures of 0.88 (medium aperture) and 1.82 (small aperture), and between the two apertures of 2.48 and 3.04 -- the measurement noise floor a fitted model is judged against; the paper reports no model-fit error (no custom code was used) publisher · read in our library knowledge/fetched/cmp_skintwin_issa_scidata_2025.html · pin hd4fe755d543198f6faa38a5311fe229b504ab6e536f654c56b5a6962aacd30df · accessed 2026-10-03 · published-paperGrounds: ST6

generated by nx_swcompare_matrix (sovereign NishiLang organ) from knowledge/compare/skintwin.matrix · source checks show implementation presence; runtime and user-outcome evidence are reported separately · JavaScript supports page controls