Universal Natural Intelligence

This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.

Full biological parity is not a current status. It is the target world defined by these receipts.

The framing question for every lane is "What would make this lane true?" — never "why this lane fails." Existing P0..P8 definitions in CLAUDE.md are authoritative and unchanged.

LANE A — duration-only B3/B4 evidence


current receipts	B3 result `5d7a0589…` (9 models × 2 cohorts × 2 rules, 36 cells); held-out motor-equal NLPD on the frozen `sha256_mod5` split; 44/44 independent oracle; preflight 46/46; determinism byte-identical
current status	STANDS. Retained adverse result: M2 lognormal out-predicts the M3 two-timescale mixture on held-out data. All 8 motor-equal M3 contrasts `INCONCLUSIVE` at 19 holdout motors
next receipt	C02 and C10 have landed; C11 running, C01 queued. B4C02 shows the adverse M2-over-M3 pattern is generator-specific, not a generic heavy-tail artifact — so "it is just shape" is no longer an adequate account of the retained adverse result. No mechanism is established by this. Next: C11, C01
falsifier	an adversarial baseline survives or wins under corrected CI-bound tests
maps to	`P3` held-out predictive (duration-only scope)

LANE B — corrected robustness


current receipts	C03–C08 executed; D1 bootstrap repaired + tested; D3 stable seeding repaired + tested; resource costs re-measured; B4C10 LANDED at full frozen N_boot=2000 (`959a00e9…`, 8372.7 s, 1994/2000 completed, U2/U3/U4 all `OK`) on the byte-identical frozen runner; B4C02 LANDED at full frozen N=200/generator (`0633988d…`, 8.17 h, 600/600 sims, 0 failures, `ELIGIBLE_FOR_FROZEN_VERDICT`)
current status	IN PROGRESS. C10 complete; C02 complete (`GENERATOR-SPECIFIC`, frozen prediction `GENERATOR-ROBUST_ADVERSE` REFUTED at full N — the adverse M2-over-M3 result is not generic to heavy-tailed shape); C11 running; C01 queued. C11 U4 remains withdrawn pending its corrected run — B4C10's favourable result may not be transferred to C11 (M4 pooled-i.i.d. vs M7 motor-grouped; different likelihood structure)
next receipt	the two remaining corrected full-N results (C11 running, C01 queued), each hashed, with old-vs-new status tables
falsifier	corrected cells fail, or remain inconclusive at full N
maps to	`P3`, `P6` per cell

LANE C — mark process


current receipts	schema carries `nextStateN`/`direction`/`jump`; B3 never reads them; D6 quarantine policies implemented and tested
current status	QUARANTINED. Holdout mark channel burned (D5) → retrospective-only on this dataset. Closed-chain modelling blocked by D6: 2 impossible marks, 5 zero-support holdout cells, 15–17% of marks leave `{1..8}`
next receipt	raw-archive re-derivation of the 2 impossible marks and an independent prospective mark dataset with a predeclared smoothing/quarantine policy
falsifier	impossible marks persist unresolved; smoothing constant controls the outcome (it currently flips the sign); transfer fails
maps to	`P2` (mark fields), `P6`; prospective route now runs through `P4`/`P7`

LANE D — motor-stack AIF


current receipts	constrained F-side model built: 2 free params, per-motor latents integrated by quadrature, censoring-correct hazard/survival, no floor, motor-equal scoring, motor-resampling bootstrap; 385 tests pass. SCORED 2026-07-22 — `F_SIDE_MOTOR_STACK_SCORING_RESULT.json` sha256 `b3b12720…`, determinism proven byte-identical over two full executions, independent-oracle check PASS with residual exactly `0.0` against three frozen models (M3, M0, M8)
current status	BUILT AND SCORED. Verdict `NOT_ESTABLISHED` against `CONTROL_CURRENT` (M3) and against every serious adversary (M2, M8, M1); `RESOLVED_ABOVE` only against the two weakest (M0, M5). The F-side model reproduces the frozen `M7_HIERARCHICAL_MOTOR` to `2.5e-7` nats — at this resolution it is a re-derivation of the incumbent, not an advance on it. The hierarchy buys nothing measurable over its own `tau → 0` limit `M1_WEIBULL` (contrast `+0.000615`, half-width `0.0111`). G-side fenced: `DESIGN_ONLY_UNTIL_INTERVENTION_OR_TRANSFER`
retained adverse	`M2_LOGNORMAL` (`3.4093`) and `M8_EMPIRICAL_KDE` (`3.4225`) both out-predict the F-side motor stack (`3.4327`) on point estimate. Both intervals cross 0, so this is `NOT_ESTABLISHED` — not a refutation and equally not a defence. The standing lognormal adverse finding is extended, not overturned
next receipt	per-motor arrays for M4/M6/M7 DELIVERED 2026-07-22 (`751a59ef...`): all three re-derived at exact zero oracle residual; neither M4 nor M6 resolves against the candidate; M7 exposed D10 (no minimum-effect-size guard). Candidate ranks 5th of 10. Remaining: more independent motors; a model that beats its own `tau → 0` limit; censored observations to exercise the unexercised censoring branch
falsifier	an adversary's contrast interval lies entirely below 0. Not observed — so `H_FSIDE_DURATION` is not falsified, and equally not supported, against the serious competitors
maps to	`P1` on implementation (strengthened: exact-zero oracle residual + proven determinism); `P3` duration-only scope — unchanged, B3 stands. No P-level raised

The G-side requirement, stated concretely

G needs policies over actions; this dataset has none — the action set is empty, which is structural, not sample-size-limited. To make G a biological motor claim requires all three: (1) a manipulated variable with recorded onset time, (2) paired pre/post observation on the same motors, (3) enough independent motors per condition for a CI-bound verdict. Until then G is used only at ORCHESTRATE level to choose the next experiment.

LANE E — full biological parity


current receipts	`P0` H-AIF-G1 established; `P1` corrected implementations tested; `P3` duration-only B3 stands
current status	`FULL_PARITY = false`. `P8` is conjunctive; `P4` transfer is the first unsatisfied level
next receipt	`P0`–`P7` conjunctively supported
falsifier	any required level remains `NOT_ESTABLISHED`

Per-level status

level	status	missing receipt
`P0` computational integrity	holds	maintain frozen baseline; correct D7 width reporting
`P1` equation / implementation	repaired in new namespace	F-side scoring; corrected runs complete
`P2` observational	limited	raw MAT archive absent (`BLOCKED_EXTERNAL`); D6 unresolved for mark fields
`P3` held-out predictive	stands, duration-only	corrected B4 results; F-side scored comparison
`P4` transfer	`NOT_ESTABLISHED`	an independent dataset with a predeclared split — cannot be closed by modelling
`P5` intervention	`NOT_ESTABLISHED`	wet-lab perturbation with recorded response — cannot be closed by modelling
`P6` structural / mechanistic	scoped: C11 U4 withdrawn pending rerun; mark-mechanism retrospective-only; duration-only unchanged	corrected C11; a mechanism discriminator; independent mark data
`P7` independent replication	`NOT_ESTABLISHED`	a second lab/dataset — cannot be closed by modelling
`P8` full verdict	false	all of the above

P4, P5, and P7 cannot be closed by any amount of modelling, code, or analysis in this repository. They require data this project does not have. Saying so plainly is not defeatism; it is what makes the levels that are movable worth moving.

Strongest claim currently licensed

On a single 19-motor E. coli cohort (Wadhwa 2022), under a frozen sha256_mod5 split and motor-equal NLPD/CRPS scoring, a simple lognormal dwell model out-predicts the two-timescale mixture on held-out data, and every motor-equal contrast against the reference mechanism is inconclusive at this sample size. No mechanism, transfer, intervention, or parity claim is supported.

Standing frame

Nature supplies the architecture candidate. The gate supplies the status.
Full biological parity is not a current status. It is the target world defined by these receipts.

sha256 3d812461303c6dd3 — of the original file, so what was ingested stays checkable.

Biological Parity Receipt Map

How to read this page

Precise — the source document