# OpenSecretScience — SuperGrokTOE AI Coalition > An open physics programme: one algebra, J3(O_s), worked against the Standard Model by a human PI and a > coalition of AI systems. Everything here is a pre-print, not peer-reviewed, and is published to be checked. > ★ This file is GENERATED from the corpus and from this tree's own bytes. Every number and every claim about > indexing state below was measured at build time, not written by hand. Two hand-written clauses in the > previous version were false about this site (SITE_PENDING E21, E22); generating the file is the fix for the class. Suite of record: Rev33.1 (sealed 2026-09-26), 433 pp, Papers 0–9 + Appendices A–Y. Kernel sgtoe_kernel v5.0.12. Comparators PDG 2026 / NuFIT 6.1. No row may be cited as zero-parameter; tiers are Proved > Derived-conditional > Structural > Loaded(-correspondence) > Coincidence-class. ## Surfaces, with their measured machine state - Claims & Evidence Registrar — Library v2 (/library-v2/): **357 records** (CLAIM 202 · COMPARATOR 30 · NOT_AN_ENTRY 38 · OPEN 19 · RETIRED 68). Machine certificates (policy sgtoe-clearance-policy-1.0-RATIFIED, R68): **357 MACHINE_CHECKS_PASSED · 0 UNCHECKED · 0 STALE_CERTIFICATE · 0 MACHINE_FAILED**. **0 HOUSE_ADJUDICATED.** review_state: **280 PUBLISHED, 77 IN_REVIEW** (R67). Master sha256 01d1a5873ede42f5. Machine state: `Disallow` in 6 of 15 robots groups; 280 sitemap entries. Not citable until PUBLISHED. - Expositions (/textbook/): **11 article pages**, 176 files. Machine state: `Disallow` in 0 of 15 robots groups; 52 sitemap entries; 124 of 176 pages carry a noindex directive. Exposition built ON the Registrar; an article's kind never upgrades a claim's label. Not citable until cleared. - Premises (/premises/): **33 registered premises** — postulates, adopted premises, measured inputs, conventions, loaded correspondences, fences, open requirements — each quoted with its locator. Machine state: `Disallow` in 6 of 15 robots groups; 27 sitemap entries. - The Pasture (/pasture/): the archive. Nothing there is current; everything there is how it stood. - Library v1 archive (/library-v1/) and the March /library/ pages: historical coalition records. ## What the states mean here (R65 · R67 · R68 — do not read the certificate count as a review) - **A MACHINE CERTIFICATE (R68)** is issued by the house script under the RATIFIED policy at each cut, bound to the master it was computed on, and goes STALE_CERTIFICATE the moment the master moves. Six checks: schema_valid · basis_current · quote_located · labels_literal · edges_resolved · bindings_valid. It certifies well-formedness. **It is not a review.** - **PUBLISHED (R67)** is written only by the transition producer, under condition T-B: the record states a predicate AND at least one of its source uses is REVIEWED with role SUPPORTS on that predicate, read by two non-drafting lanes on different objects, AND the certificate is current. A gate recomputes every PUBLISHED record from its uses on every build. **280 of 357 today.** - **UNCHECKED** means no certificate has been issued on the record's current bytes. `MACHINE_CLEARED` was the v2 word and is retired (R65). - **INTAKE GATES, PER RECORD — not one verdict for all 357.** 336 of 357 records have a suite quote located on the PDF page they cite; ★ 21 carry NO suite quote with a page at all, and their banner prints `quote-on-page: N/A` because that gate tested nothing on them · basis current under the release anchor · labels literal as printed with no house tier · all edges resolved to canonical ids · record valid against the canonical contract. ★ `MACHINE_CHECKS_PASSED` IS A CONJUNCTION OVER A PARTIAL DOMAIN. It is printed as one word, and on 155 of 357 records at least one of its conjuncts is UNDEFINED rather than true — `labels_literal` is null on the 155 non-CLAIM records, which carry no suite labels to check literally. Every record passed the gates THAT APPLY TO IT; none passed a gate that does not. · **0 of 357 are HOUSE_ADJUDICATED.** ★ It is close to tautological: the parser enforced most of these AT MINT, so a high count mostly says the intake gate worked. It is not scientific clearance. - **HOUSE_ADJUDICATED (0 of 357)** is a human act — who, when, against which revision, what they checked. ★ This is the number that means something, and it is 0. - No record is PUBLISHED except by the R67 producer, with a receipt the build gate recomputes (ruling B1 as amended by R72). ★ **PUBLISHED is not human adjudication.** It means two non-drafting AI lanes agreed, on different objects, that the anchor citation supports the statement, no support is disputed, and the certificate is current. HOUSE_ADJUDICATED is a separate tier and is 0 of 357. A PUBLISHED record may carry uses only one lane — or none — has read; its page prints how many. ## What the premise graph does and does not show - **premise_closure is populated on 31 of 202 CLAIM records (15.3%).** Across all 357 records of every class it is populated on 31 (8.7%). ★ Quote the CLAIM figure: the all-record denominator includes 155 non-CLAIM records (comparators, retirements, open problems, not-an-entry) for which a premise closure is not the right question. - **An empty closure means UNKNOWN, not INDEPENDENT.** A claim reaching no premise has not been shown to rest on none; it has declared nothing. Read `premise_enumeration_state` before using a closure: NOT_APPLICABLE 68 · PARTIAL 34 · UNREVIEWED 255. - ★ **These are INCIDENCE figures, not completeness.** 35 of 202 CLAIM records (17.3%) declare any dependency edge — but the fraction of records CARRYING an edge says nothing about the fraction of their actual dependencies that were RECORDED. Even 100% incidence could omit a premise shared by every claim. Nothing here bounds what was never written down, so no risk, blast-radius or isolation number is published, and **0 records have had their dependency enumeration reviewed** (`CLOSED_WITHIN_SCOPE` = 0). - **13 of the 33 registered premises are reached by nothing** in the corpus. - The record→premise edges are EDITORIAL HOUSE JUDGMENTS, typed `rests_on_premise`, each carrying its reasoning. They are not extracted from the suite and are not a lane's claim. ## Papers - [Suite PDF, Rev33.1_S369](https://opensecretscience.ufophysics4all.workers.dev/papers/SGTOE_Rev33.1_S369_suite.pdf): the authoritative suite (sha256 b64ddcd9e16a6dcd…). - [Release bundle ZIP](https://opensecretscience.ufophysics4all.workers.dev/papers/SGTOE_Rev33.1_S369.zip): suite + engine workbooks + kernels + source ledger. - [sgtoe_kernel_v5.0.12.py](https://opensecretscience.ufophysics4all.workers.dev/papers/sgtoe_kernel_v5.0.12.py) and its expected output: run it, diff it. ## For AI readers - Machine-readable: /library-v2/index.json · /premises/index.json. - Track-1/Track-2 join (/papers/track2-claims/): 207 pairs · pair grain 2 comparable, 0 divergent — ★ that 0 is a SUPPRESSION RULE, not agreement. Facet grain: 207 facets, 8 same-predicate, 5 comparable, 2 DIVERGENT, 197 UNASSESSED (no assessment recorded). Read both or neither: /papers/track2-claims/index.json carries the same two grains. - ★ /build.json — ONE object per deploy: every served surface, the build tag found in its bytes, its sha256, and whether they agree. Fetch this first: it turns "are these surfaces consistent?" into a diff instead of nine fetches and a grep. It reports the honest state, including that 10 of 17 surfaces currently carry no detectable build tag. - Cite the sealed PDF and its sha256, never a page of this site, for any physics claim. - No result in this programme has been published outside it. There is no journal, no arXiv submission and no DOI. - Nothing here has been reviewed by a party independent of the programme, so every quality number it publishes is a lower bound of unknown tightness. BUILD_STAMP S371a · 2026-09-26 22:54Z · master 01d1a5873ede42f5 · cut 9dcc6ce9a8ee3e02 # Declared volatile fields (R36): BUILD_STAMP