# 27, The Anchor and the Margin: is self-blindness a threshold or an anchor? (anchor-margin program)

**Program decision document, v1.2 (v1.0 2026-09-05; v1.1 2026-09-23, G26 chain, §7.7; v1.2 2026-09-24, AM-4 lifted to ANCHOR-LEMMA over the order-1 F2cQ SD SPLITL family by the fresh ruling, §1.5 v1.2 block).** Program of record: `26_dream_question_3.md` (sha256 `e7154d995366d042…`; §4 the recommendation, §4.2 the not-decided-at-freeze table, §4.3 the judge's own finding, §5 the mandates M-L1..M-L8 with P-G1..P-G6 + P-G-ctrl, the prime bands and KG-0..KG-7, §6, §7, read in full and applied). Pilot (M-L1): `analysis/anchor_pilot.py` (sha256 `f1028c0da9d58867…`, frozen verdict rules in the docstring; pins exactly 8,200,007 / 8,200,013) → `anchor_pilot_results.md` (`eee3fc0193c243df…`), `anchor_pilot_rows.jsonl` (98 rows; `5d6df83cab4362cc…`); **67/67 checks, exit 0**; three verifier legs consolidated in `analysis/anchor_pilot_verifier.md` (`84f25167d25e86e0…`; leg 1 10/10; leg 2 rigor 37/37 at 109 rows incl. 11 unseen; leg 3 completion `anchor_pilot_verifier_completion.py` `7bd430ec6cbbf985…` 16/16, exit 0, 26.6 s, **KG-0 = PASS under doc 26's KG-5 definition, the owner's ruling of 2026-09-02 confirmed by three paths; verdict on the pilot's record CONFIRMED-WITH-DEFECTS**), the author's reconciliation leg (`anchor_pilot_reconcile.md`, 32/32) and the corrections of record D2–D5, D7, D8 (§5). Spec (M-L2): `analysis/anchor_target_spec.md` **v1.0.1, FROZEN** (plain sha256 `066269d554abd4bb01be9f8e890d3a26416ea23fdeb2f77054b39e1276515e1d`, stamp `d39954e2…`; v1.0 `8ace63b2…` archived byte-identical). Hostile gate G26-0 (M-L3): `analysis/anchor_gate_audit.md` (`bbc15d0be8ea8fe8…`; 44/44, exit 0, 7.0 s; token 8,300,057; **PASS-WITH-MANDATORY-REPAIRS**, M1–M7 and m1–m13 folded into v1.0.1). Mechanism: `analysis/anchor_mechanism.md` (`5b568dd4da046ce5…`; registered pre-code 03:02:39, by the engine author, a disclosed DEVIATION; provenance leg `anchor_mechanism_provenance.md` `882ca86d…`: REGISTERED, MATCH). Engine: `analysis/anchor_lattice.py` (`f06bb93562b3dd27…`; stages A0–A6 + TABLE; python3.12; exact over Q; tokens exactly 8,200,037 / 8,200,061) → `anchor_results.md` (`c4221b4af6e44318…`), `anchor_rows.jsonl` (**790 rows, 338 classes**; `1529452ffcda4536…`), `anchor_state.json`, `anchor_run.log`; ledger **checks 27 (0 failed); verdicts 59 (7 refuted, all reported first)**; seven launches (§5). Verifier legs (disjoint bands): rigor `analysis/anchor_rigor_verifier.md` (`90cbd321f22b9480…` as written at hand-off; tokens 8,300,069 / 8,300,077; **CONFIRMED-WITH-DEFECTS**, status of its machine stages in §5) and physics/steelman `analysis/anchor_physics_steelman.md` (`ae824c595f66f244…`; tokens 8,400,011 / 8,400,019 / 8,400,031; **CONFIRMED-WITH-DEFECTS**; no KG fires by own route). Co-runner (M-L8): `analysis/anchor_corunner.md` (`babec8147ac68003…`; U2 15/15, FORK1 11/11, P3 21/21, second launch identical 47/47 + CONFIRM 11/11; tokens 8,200,109 / 8,200,141; never headline). Corpus: `papers/24_anchor_margin/` (8 arXiv PDFs + `PDF_MANIFEST.txt`; scouts A `SCOUT_MARGIN_ANCHOR_INSTRUMENTS.md` `15ba4ed79333155e…` and B `SCOUT_QRF_OCCUPANCY_TRIGGERS.md` `87cd01e12bfc9d99…`, both 2026-09-05; `perimeter_2058_contributions.md`). **This file supersedes in full the pilot-side data draft that carried its name (sha256 `9782694a3b5049af…`, owner note 2026-09-02); nothing of that draft is cited here as record, every table below is re-sourced from `anchor_pilot_results.md`, the verifier records and `anchor_results.md`.**

**The question (doc 26 §4, dream form, verbatim).** *Is self-blindness a threshold or an anchor? When a universe's own clock cannot read the constraint it sources, is that because its self-sourced energy outruns its own motion above a computable coupling, a margin, derivable, and blind to who routes the charge, or because the constraint role leaves it no zero-energy configuration to read from, an anchor, decided by own routing, and is that two-regime anatomy the whole of "a gravitating observer cannot read itself"?* Operational form (doc 26 §4 in substance, the certificate form corrected to the form of record D2): over the registered docs 12–14 constraint-role family with the coupling extended to ν ∈ Q_{>0} (a FAMILY EXTENSION, rider on every ν ∉ {1, 2} sentence) and the matched Gauss-routed QED clock as the mandatory control, write H(q; ν) − qI = hop + diag(a_x(q)), a_x(q) := νm̃₂q² + (νM̃₁_x − 1)q + νD_x, R_x := Σ_{y≠x}|hop_xy|, **Δ_x^{(R)}(ν) := (νM̃₁_x − 1)² − 4νm̃₂(νD_x + hop_xx − R_x) = −ν²G_x + νb_x + 1** with G_x = 4m̃₂D_x − M̃₁_x² = 4[B(u,u)B(π_x,π_x) − B(π_x,u)²] and b_x = 4m̃₂(R_x − hop_xx) − 2M̃₁_x; the anchor sublattice Z := {x : D_x = 0 and M̃₁_x = 0}; the reading curve C_cell := {(q, ν) : det(H(q; ν) − qI) = 0}; four reading sets (spec §1.1 M1): R_self^{Q,R}, R_self^{Q,Q} (doc 13's R_self), R_self^{cl,R}, R_self^{cl,Q}. Prove class-wide, or type the failure: (G) THE MARGIN, Δ_x^{(R)}(ν) < 0 ∀x ⇒ strict diagonal dominance for every real q ⇒ Lévy–Desplanques ⇒ all four sets empty; explicit ν̄(cell) = max_x [b_x + √(b_x² + 4G_x)]/(2G_x) wherever G_x > 0 ∀x and m̃₂ > 0 (D3); (A) THE ANCHOR, where the margin is silent, the rational reading sets decided by Z (the anchor reading 1/(νm̃₂) exists iff Z ≠ ∅; the quantum survivors the hop-null vectors on Z, the doc-14 bridge with a support structure), Z's shape derived from the two kernel orders (the constraint's slaving order fixes {D = 0}, the clock's routing order fixes {M̃₁ = 0}, M4); (F) THE FOLD, the weak-coupling strata, ν_c, ν_max ≤ ν̄, the sporadic points; the control's three clauses.

**Status of the record.** Scouts A/B DONE (2026-09-05; GROUND-CLEAR, forced riders folded pre-freeze) → PILOT-G PASS (KG-0 ruled) → spec FROZEN v1.0 → G26-0 PASS-WITH-MANDATORY-REPAIRS → v1.0.1 re-freeze → mechanism note registered pre-code → engine 790 rows, 27/27 checks, 59 verdicts (7 refuted) → steelman leg CONFIRMED-WITH-DEFECTS → rigor leg CONFIRMED-WITH-DEFECTS at hand-off grade (R0–R1 complete; R2 at 294/338 classes with 0 mismatches; R2's remainder and R3–R7 NOT COMPLETED, the resumed run launched by this synthesis stalled and was stopped cleanly; typed NOT-COMPLETED-BELOW-CAPS, OWED at G26-0.5) → this document. **G26-1 (closing audit at the untouched reserve ≥ 8,500,007) and G26-2 (re-sweep incl. the ARMED Perimeter trigger, post-2026-10-30) are OWED and block all external quotation (G26-3); a record-integrity gate G26-0.5 (§7.4) is inserted before G26-1 for the generator-side repairs the verifier legs found (none verdict-touching).** The G26-4 pointer-row decision is EXECUTED by this document (§7.1: NONE ENTERED; candidate rows drafted).


**[v1.1 amendment (G26-0.5 executed; rigor leg complete; G26-1 executed, PASS; G26-2 pre-meeting leg executed and re-verified; G26-3 ruled), 2026-09-23, files-of-record supersession.]** The G26-0.5 generator regeneration (the repairs §5 / §7.4 said were OWED; `analysis/g26_0_5_repair_report.md` `278c0b1ed963e310…`, evidence folder `analysis/g26_0_5_evidence/` with `MANIFEST.sha256` `9e7688a75a389a1a…`) supersedes the header shas: engine `analysis/anchor_lattice.py` sha256 `b7f918919d188066781cb2d50dc3fd9ac7378b5acf8b6c0d21e81207a2b88a8c` (the v1.0 text `f06bb93562b3dd27…` preserved as `analysis/anchor_lattice_v1.0.py`; 23 hunks, +383 / −38 lines, every block dated "G26-0.5 (2026-09-22)" and labelled (a)–(e)); results `anchor_results.md` `0101d5fa0d7cd133820e207724b303760aacbbb21d627a0d032704f754056a7f` (new §0.1 "Launch history and verdict flips" and §0.2 "G26-0.5 regeneration record", generator-written); rows `anchor_rows.jsonl` `c07555e117321d334d008facb628bb7a789b6aa899ca492b707a0e1fc2c3491b` (**790 rows, 338 classes**; exactly 24 leaves changed, all `sd_structure.D_zero_is_balanced` False → True at the 24 F2cQ SD SITE rows, the D-S12 stale non-verdict field; 766/790 lines byte-identical); state `anchor_state.json` `8fee0d0bb9f44d4d6edb14db84c02d169be6c94f3c0bd2115036a1aaaaffc413`; run log `anchor_run.log` `8ef14cdd9181eb15…` (append-only: the 47.3 s REGEN segment; the record file a byte-prefix). Pilot record files regenerated and re-pinned with disclosure (D-R1): `anchor_pilot_run.log` `1b40ebf63cf7f289c98d9158e2f3325fb13e9d70f9eeac1e43792c52cc4c80dd`, `anchor_pilot_state.json` `0347c35e8208bfca66065692e5a8038ac8c0741af31da5f560a1fa383d3ce259` (ONE launch of the committed `anchor_pilot.py` `f1028c0da9d58867…`, `--disc`, its own tokens 8,200,007 / 8,200,013; 98/98 verdict-row identity; installed 2026-09-22 21:09:21; the 2026-09-02 originals `e500e04619ff50fc…` / `6d52ee219e17d2ce…` unrecoverable, the pilot's three verdict-bearing files `f1028c0da9d58867…` / `5d6df83cab4362cc…` / `eee3fc0193c243df…` unchanged). Verdict neutrality machine-asserted: 59 VERDICT names with identical CONF / REFUT states (0 flips; the regenerated ledger 92 entries = 33 CHECK + 59 VERDICT, 7 refuted), typed outcome and rung standings identical, every KG fire set identical (KG-1's letter split union-preserving: 6 exact-empty + 6 height-typed + 6 NOT-COMPUTED rows); a fresh single-launch mirror of the final text (mirror3: one launch, 5 workers, 3156 s, rc 0) reproduces every verdict-bearing leaf of all 790 rows (0 SUBSTANTIVE; 737 rows differ in `runtime_s` only). **Rigor leg of record:** `analysis/anchor_rigor_verifier.md` `02c2162002589224e0192bb1b83f37af6649c138e44fc1f22b6ed0a49c0f01b3` (COMPLETE, R0–R7 on disk, E3 done, the mirror comparison reconciled against the REGENERATED canon; **CONFIRMED-WITH-DEFECTS, unconditional**; supersedes the hand-off text `90cbd321…`, kept), `.py` `77234bdbd8142dbb5e318b12e57ef4f152ac4f54120aedc261dee52cc36cb719` (amended from `615bc64512309a83…`, dated fourth-instance disclosure; from `2c29031f…`, dated third-instance disclosure, the R2 stall measured and removed; regression on the isolated pre-regeneration tree `scratchpad/rigor3_tree` reproduced the 20:43 outcome check for check), `_state.json` `1d5426b660ea785ac8b9d2d24bef3dfe6276e8c9cb3a4847163f22b6a89dee5c` (stages R0–R7 `done`, R7 `complete`; 790 rows / 338 classes; ledger 106 entries = 78 CHECK + 28 FINDING; `R7_history` length 1; two launches recorded, 20:13:56 with script `615bc64512309a83…` and 21:50:25 with `77234bdbd8142dbb…`), `.log` `087c77fbda42f3ce…`, `_rows.jsonl` `6018701ce8e8c398…`, `_extra.py` `1e549bc455874bc5…` / `_extra.log` `b4fedcb5a81aabd1…` / `_extra_state.json` `d5c638c21ddcb978…`; tokens 8,300,069 / 8,300,077 re-certified at both resumes, no further draw. **Closing audit of record (G26-1):** `analysis/anchor_closing_audit.md` `7cd12fba2dbb15dac547675df3c7b787fd52739466d1f8937b161b31c53b9af9` (**PASS**, 84/84 own checks; 2 record defects, 0 verdict-touching; adjudicated; generator-written 2026-09-23 04:14:19), `anchor_closing_audit.py` `7aaafa863e096a55dca66d878e0bf9469c72ad299444dd65b46214ce1d75bd00` (stages S0, HP, C1–C9, CS; the audit-design repairs A-1 … A-4 applied in-pass and disclosed), `anchor_closing_audit.log` `c2a0fd1f1342538d…`, `anchor_closing_audit_state.json` `107f9e6a00401b68…` (stages S0 23:46:37 · C1 23:51:17 · C6 02:25:11 · C2 03:51:15 · CS 04:06:13 · C3 04:10:54 · C4 04:10:55 · C5 04:11:10 · C7 04:13:42 · C9 04:13:42 · C8 04:14:19; 84 checks, 0 failed; 71 superseded entries, 7 of them the first-pass FAILs; 6 findings; `primes`, `probe`, four pin tables 64 / 64 / 24 / 24), `anchor_closing_audit_rows.jsonl` `bc9d8784699b640c…` (790 own third-band rows). Steelman `.py` `df0e202cb980bccf…` and `anchor_corunner_common.py` `380ade840bbe90ac…` (pin lists extended by the pilot's two record files; dated comments). Corpus: `papers/24_anchor_margin/` now 10 arXiv PDFs + `PDF_MANIFEST.txt` `00380fd47bfdd087…` (was `c4bc2575f383b2cc…`; two appended lines + one comment line, disclosed), the G26-2 pre-meeting sweep `G26_2_PRE_MEETING_SWEEP.md` `2534dd29d38c20f1ed9c1a227765b98e5a1b004365b8bc6f9d46e7cd7e4595c9` (§0–§10 the leg of 2026-09-22 22:04–22:42; §11 / §12 / §13 the re-verification passes of 23:00–23:10, 23:28–23:32 and 2026-09-23 03:32–03:47) with `scout_scripts/g26_2/` (INSPIRE and arXiv scripts, logs, states, the 2026-09-22 Indico export and timetable, the PIRSA page, the citer diff, `g26_2_state.json`, `g26_2_reverify_2300.json`, `g26_2_reverify_2330.json`, `g26_2_reverify_0330.json`). Spec, mechanism note, doc 26 and every doc-01–26 canon file: unchanged (pins §A). **The extended pin list of record is the union of the engine's 64-file table, the rigor leg's 81-entry table (64 canon + 17 record files) and the closing audit's 64 + 24 tables, 98 distinct files with the sweep's 34 + 16 tables folded in, reproduced with full digests in §A; every entry equals canon at the time of this ruling.**

**[v1.1, 2026-09-23, status line.]** The v1.0 sentence "**G26-1 (closing audit at the untouched reserve ≥ 8,500,007) and G26-2 (re-sweep incl. the ARMED Perimeter trigger, post-2026-10-30) are OWED and block all external quotation (G26-3); a record-integrity gate G26-0.5 (§7.4) is inserted before G26-1 for the generator-side repairs the verifier legs found (none verdict-touching).**" is superseded as follows, the v1.0 sentence standing as the record of what was then owed: **G26-0.5 LANDED (2026-09-22, generator-side, verdict-neutral, machine-verified; items (a)–(e) by the repair leg + owner driver + finalizer, item (f) by the rigor leg's third and fourth instances), DISCHARGED by this amendment; the rigor leg COMPLETE and CONFIRMED-WITH-DEFECTS unconditionally (the §1.5 conditional entry DISCHARGED, no verdict-reversing finding); G26-1 EXECUTED 2026-09-22 23:46 → 2026-09-23 04:14 at the reserve ≥ 8,500,007, PASS (84/84; 2 record defects, both landed, both verdict-neutral; tokens 8,500,007 / 8,500,021 / 8,500,027 consumed; next audit reserve ≥ 8,550,001 designated); G26-2 PRE-MEETING leg EXECUTED (CLEAR-PRE-MEETING; triggers NONE FIRED, re-dated 2026-09-23 03:41 CEST; Perimeter ARMED, export changed 86 → 79) and re-verified three times, the POST-MEETING leg OWED after 2026-10-30; G26-3 RULED, CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING (§7.7): external quotation of every sentence stays blocked until the post-Oct-30 leg discharges.** The record status chain reads: … → this document (v1.0) → G26-0.5 executed → rigor leg complete → G26-2 pre-meeting executed (+ re-verified ×3) → G26-1 executed PASS → G26-3 ruled CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING (v1.1) → [G26-2 post-meeting owed after 2026-10-30] → [G26-3 re-ruled by a dated block].

---

## 1. Verdict up front

### 1.1 Pivot / scoop ruling FIRST (what the literature already had, credited; Scouts A and B, 2026-09-05)

| protected claim | branch | ruling |
|---|---|---|
| **PC-1** the margin certificate, no self-consistent reading in EITHER carrier above an explicit ν̄, by diagonal dominance of the reading operator (M-L4) | **S0, CLEAR, no pivot** | No paper states "self-sourced constraint energy vs hopping ⇒ no self-consistent reading" in either carrier, with or without a certificate (Scout A L1; Scout B §6: 30 arXiv/INSPIRE nets, zero hits). The INSTRUMENT is classical mathematics and is cited, never claimed (Horn–Johnson *Matrix Analysis* 2e §6.1 Thms 6.1.1/6.1.10, §7.2; Varga 2004; Taussky 1949); its prior physics uses are gap/spectral bounds (2006.13938, 1811.11786, 2412.02754, 2208.01187) and one uniqueness certificate (Vilenkin–Zhang 1403.1599), "no prior use as a fixed-point-reading exclusion" is a NOT-FOUND scoping remark at metadata + INSPIRE-full-text grade, never a headline. **Three continuum neighbours force riders, none occupies:** Espíndola–Ali 2605.30423 (a margin-shaped sufficiency theorem for observer backreaction in Euclidean de Sitter, Gaussian saddle control BELOW a susceptibility threshold; admissibility-side, weak-coupling-derived, no carriers, no anchor, no control; same certificate genus, inverse direction, neither implies the other; their "a bound, not the boundary" is the wording precedent for ν̄ vs ν_max, cited as precedent, never corroboration); Arderucio Costa–Sanchez 2602.24177 (clock-mass upper bounds: a validity bound, not a reading exclusion); Schrödinger–Newton self-trapping (1105.1921, 1403.2982: a self-bound state APPEARS above threshold, the opposite consequence); and Yokokura's Constraint-Variance-Identity fragility bound (Perimeter Oct 2026 talk, unposted, the classicality of the background clock at low N_eff, not the existence of a reading at high ν; ARMED-HIGH for G26-2). |
| **PC-2** the anchor, a zero-energy, dipole-blind sublattice Z deciding the rational reading sets; survivors = hop-null vectors on Z; Z's shape by the two kernel orders (M-L5) | **S0, CLEAR** | No "reading ⟺ a zero-(D, M̃₁) configuration exists" criterion, no "survivors = hop-null vectors supported on the zero set", no kernel-order thinning of such a set anywhere (Scout A L2). Höhn–Chataignier–Lock–Mele 2409.06479 (periodic clocks; "Fourier" 0 times) and the third-particle cluster (2607.15351 / 2607.20815 / 2607.21703 / 2608.23698) are cite-and-distinguish; Hausmann's Perimeter no-go form forces the **back-action rider** (theirs is kinematic frame change; Z is a constraint-kernel object). |
| **PC-3** the two-regime anatomy and the carrier split (margin: both carriers agree; anchor: the carriers split) | **S0, CLEAR** | Nobody has computed or stated it (Scout B §6, 30 nets); the nearest carrier typing is Luppi–Kabel–Giacomini–Smirne 2607.05578 (classical misalignment vs quantum reduced-frame effects, a DIFFERENT effect, QM assumed), the **carrier-typing rider** travels with every wall sentence, with the **non-occupancy sentence** (Frauchiger–Renner / Local Friendliness assume QM; QRF covariance results transform descriptions; continuum backreacting-frame constructions build algebras, none decides a self-sourcing frame's reading). |
| **PC-4** the QED control (Gauss-routed clock: the coupling-independent anchor stratum; the SITE fold at √5/20 and margin above 11/16 + 5√5/16; the swap contrast) | **S0, control positioning only** | The control's kinematics is now an active QRF-school object (lattice-QED-with-QRFs: Rothlin–Ferradini–Chen 2604.06149; Pagan Lacambra et al. 2604.06087; Panella and Sartini posters) that asks a different question (codes, edge modes), **control-positioning rider**; their content is never claimed or adjudicated. |

No S1–S5 bin fires. The four Classicality-Adjudicator triggers: **NONE FIRED** (Scout B §1, dated 2026-09-05: 2511.07348 still v1; no GIE data or timeline; no Andriolo pipeline; no ledger-shaped scoop). The Perimeter "Relativity Reframed" trigger (L4) is **ARMED**: Scout A found the Indico contribution export public (86 entries; 61 titled) and graded every contribution, **CLEAR pre-meeting** (no diagonal-dominance margin, no self-sourced-energy-vs-hopping threshold, no zero-energy reading criterion, no carrier split); the post-Oct-30 re-sweep (G26-2) stays mandatory before any wording entry. Premise genre (2608.25800, 0 citers; Carrasco-Martinez 2602.15134, R2 a POSTULATE, 0 citers; Breuer / Wolpert / Szangolies) cite-and-distinguish on every self-observation sentence. **Credited in full:** Lévy–Desplanques, Geršgorin and Cauchy–Schwarz/Gram positivity are evaluated, not re-derived; the content licensed by this program is exactly the pre-registered deltas, the explicit ν̄(cell) table over the registered family, the certificate in both carriers, the anchor sublattice with its hop-null support lemma, the two-kernel-order derivation of Z, the fold constants, and the matched control.

### 1.2 The typed outcome and the answer at record grade

**Outcome type of record (exactly one of the pre-registered types; no upgrade): MIXED**, read as the conjunction of the pre-registered branches the record actually reached, each at its own grade: **MARGIN-THEOREM** on the definite-form population at the registered knobs (the margin theorem DERIVED and role-blind; 246/246 definite-form gravity cells outside the SD SITE sub-family certified per cell at their own knobs, 81/81 of the ν = 3 cells, 14/14 control cells; C1 ≤ 1 and C2 ≤ 2 by exact brackets; the 14 definite-form CANDIDATE-GRAV cells 28/28) **+ MARGIN-EXT at the SD SITE sub-family** (gravity and QED alike; silent below 5/2 under order-2 slaving / 5/4 under Gauss by the DERIVED lower bound H5) **+ ANCHOR-TYPED with a derived sub-scope** (the anchor lemma CK = ker(hop|_Z) at 790/790 engine rows and every verifier row; its ⇒ direction DERIVED at the no-split placements of every even V (Lemma A′) and at the V ≡ 0 (mod 4) split placements (block determinant nonsingular), machine-checked only at V = 6 for the V ≡ 2 (mod 4) split placements, short of ANCHOR-LEMMA over the whole family by exactly that clause; Z DERIVED at both kernel orders as an inertia statement about the constraint-energy form, with the lab-period-5 clause load-bearing) **+ AM-SPLIT STANDS** (the lopsided wall of doc 26 §4.3, sharpened: on every row computed, in and out of family, the certificate TYPE is a function of the inertia of the constraint-energy form alone); **no INVERSION** (KG-4 silent at V = 4 exactly, at V = 6 by mod-p at the band primes, at V = 8 at s = 0 by mod-p incl. capless zero-root certificates at the SD SITE knobs); **no CONTROL-FAILURE** (route G silent at every survivor row; the SG-1 swap contrast reproduced inside the anchor regime at V = 4, 6 and at the unseen V = 8); **KG-0 PASS**. This document is **refutation-led**: P-G3's ≥ 90 % clause and P-G-ctrl (ii) died as frozen at the pilot (both typed MARGIN-EXT); P-G13's sharpness datum, P-G11(b), P-G11(c) and P-G10's dim ∈ {0, 1} clause died at the engine (the last three FORECAST FALSE by the mechanism note before code); three hand predictions at the reserved cells partly missed (U-S2, U-S4, U-S5′); no registered theorem died.

**The answer (one paragraph; every clause typed below).** Self-blindness on the registered family is **both, a threshold and an anchor, and which one is decided by the inertia of the constraint-energy form on lab space, not by the theory**. Where that form is definite on the configurations (every SITE cell, every BC cell, the whole GBc population, 255 of the 273 gravity cells and 18 of the 22 control cells), there is an explicit quadratic-irrational coupling ν̄(cell) above which, at every candidate reading, the constraint energy of the configuration dressed by the clock's own charge exceeds the reading plus the hopping margin, the reading operator is strictly diagonally dominant for every real reading, and no self-consistent reading exists in either carrier, real or rational (Theorem G; derived; uses no property of gravity); at the committed theorem cells and at every definite-form gravity cell outside the SD SITE sub-family the registered knob lies above ν̄, 6.5–180× above the entire real locus at the CANDIDATE-GRAV cells, so the committed self-reading emptiness of docs 13/19/25 at those cells is, for the first time, DERIVED rather than computed [family-extension rider on every ν ∉ {1, 2} clause]. Where the form is singular (the SD SPLIT discretization: 18 gravity cells, 4 control cells, all of them, no third regime in family) the margin is silent at every coupling and the rational reading sets are decided by the anchor sublattice Z of zero-energy, dipole-blind configurations: under order-2 slaving Z is exactly the antipodal uniform pair {0, d−1} at every placement of V ∈ {4, 6, 8} (the period-5 profile's Fourier-symbol cancellation, derived; whichever kernel routes the clock's charge), it carries no hop-null vector, the classical carrier reads 1/(νm̃₂) on it and the quantum rational set is empty by computation; under order-1 slaving (the QED cell) the zero-energy set is the balanced hyperplane, the dipole-blind set a linear condition, and Z carries a hop-null face exactly at the placements doc 13 recorded as reading-keeping (V = 4: s ≢ 1, 2 mod 5; V = 6: s ≡ 1, 2 mod 5, the inversion, now a corollary), with a NEW anchor structure at V = 8 (dim CK = 4 at the 24 face placements, spanned by products of K₄ edge-functions); routing the clock's charge through the order-2 kernel thins a face to its antipodally periodic members (SG-1's kill with a shape; CK = 0 at all 40 placements of V = 8). **The two regimes are separated by the inertia of one quadratic form**, THRESHOLD iff positive definite; NEVER with an anchor iff singular positive semidefinite (which the order-2 kernel produces exactly at lab period 5); the third regime (indefinite, no anchor) exists only out of family (lab periods 3, 4). The margin is blind to who routes the charge; the anchor's face is decided by it only where the constraint leaves a face to decide. Whether this anatomy is "the whole" of doc 13's licensed sentence is **computed at every row of the record and derived at none**: every constraint-role fact of the record (the SG-1 swap contrast, the placement rider, S-SR-D1) sits in the anchor regime and the margin regime sees none of them, and that clause may not be worded as more (spec §2.7).

**Ceiling rider (fused; binding on every sentence).** In-class: exact finite slaved lattices, constraint mechanism (i) only, the registered cells and placements below the stated caps; the coupling axis ν ∈ Q_{>0} is a FAMILY EXTENSION (doc 13's taxonomy was built at integer knobs; admissibility at rational ν as re-registered in spec §1.2); the margin theorem is ROLE-BLIND and is never dressed in gravity words (G26-0(b) / KG-7); doc 13's §3.3 sentence travels only whole-with-riders and is not amended by this program (G26-4 the only door); never "self-observation has a critical coupling" bare (KF-5, and with Espíndola–Ali's threshold running the other way, the bare slogan would be misread as theirs); never "the classical carrier's blindness implies the quantum one" bare; never "requires QM" (carrier splits are typed), never "gravity is classical" (a null anywhere never feeds it); "undecidable" never (N1); no committed cell as an open problem (F-1); NOT-FOUND-BELOW-CAPS never NO-GO; no no-go of any shape (doc 13 §7); no LAW F / LAW G unification (the doc-14 tombstone); no grand-observer drift; "ν̄ is the threshold" never (ν̄ is a sufficient bound; ν_max is the boundary; the ratio runs from 0.98 down to 0.084 across the GBc population); the continuum-threshold, interacting-clock, instrument, back-action, control-positioning, carrier-typing, normalization and premise-genre riders of spec §0.3 on every sentence that engages them; licensing is never posting; defer-until-big governs all external timing.

### 1.3 Per-mandate results (what each mandate produced; grades per spec §2: DERIVED / COMPUTED / NEAR-COROLLARY / TYPED)

- **M-L1, PILOT-G (the pre-check).** 98 rows (23 PILOT-SR rows rebuilt bit-exactly + 29 CANDIDATE-GRAV + 25 two-letter-QED + 15 stratum-cell + 6 bookkeeping-swap rows), 67/67, exit 0, five resumed launches (D1, closed by four byte-identical mirrors). P-G1 CONFIRMED (certificate exists, near-corollary, never a finding), P-G2 CONFIRMED 28/28, **P-G3 REFUTED as frozen** (56/65 = 86.2 % of G-definite rows; the nine failures the SD SITE corner), P-G4 CONFIRMED (ANCHOR-TYPED), P-G5 CONFIRMED, P-G6 CONFIRMED, **P-G-ctrl (ii) REFUTED as frozen at ν = 1** (ν̄_QED = 11/16 + 5√5/16 = 1.386; two real irrational readings at ν = 1; certified at ν = 2), both refutations typed MARGIN-EXT; **KG-0 = PASS** by doc 26's own KG-5 definition (route G silent at every survivor row; the SG-1 swap contrast reproduced), the pilot docstring's stricter any-clause shorthand not governing (owner ruling 2026-09-02, confirmed by three verifier paths); KG-1 fired only at F2c 4 SD SITE (1,) (MULTI-DISCONNECTED; not C1/C2) and its failure was then DERIVED (H5). Corrections of record: D2 (hop_xx in the certificate), D3 (the AM-2 hypotheses G_x > 0 ∀x AND m̃₂ > 0), D4 (V = 4 anchor cells: readings leave Q, not R), D5 (R_max = 2 + κ), D7 (SPLITR ≡ SPLITL), D8 (two near-corollary relabels).
- **M-L2, spec freeze.** The family quantified (§1 of the spec: 273 gravity cells incl. the SPLITR spot, 22 + 14 control cells, placements, ν ∈ Q_{>0} re-registered); the stratification THRESHOLD / WINDOW / NEVER by the Gram sign structure; the ladder AM-0 … AM-INV with bars; caps; the stop deadline 2026-10-25 / wording 2026-11-15 / hard stop 2026-12-15; seven reserved unseen cells; prime bands.
- **M-L3, hostile gate G26-0.** PASS-WITH-MANDATORY-REPAIRS: no route's antecedent decided at freeze in the killing direction; the margin theorem role-blind by a masked-label run (98/98 identical); the O-4 contrast reproduced by an own path; seven mandatory repairs (M1 real vs rational reading sets; M2 the anchor biconditional retired; M3 a false step in Theorem G's (iii) → (iii′); M4 two "own routings" separated, the constraint's slaving order and the clock's routing order; M5 P-G10's live population = six survivor rows + U-S1; M6 P-G14 decided at freeze for ν ≥ 1 at 243 cells; M7 U-S5 replaced by U-S5′/U-S5″) and thirteen minor items, all applied as the v1.0.1 additive amendment; the near-corollary set grew by eight relabels.
- **M-L4, the margin theorem and its census (DERIVED + COMPUTED).** Theorem G written out with (iii′) and machine-checked at every row (the vertex identity; the D2 diagonal; disc ≤ Δ; the Cauchy–Schwarz clause with π_x ∦ u at every configuration, min charged sites 3); the census: gravity 255 THRESHOLD / 0 WINDOW / 18 NEVER, control 18 / 0 / 4; P-G8 246/246 per cell, 81/81 at ν = 3, 14/14 control; ν̄ knob-free with 36 independent GBc values; the SD SITE corner DERIVED (H5, 51 rows); P-G13 ν_max ≤ ν̄ at 184 bracketed classes, its sharpness datum REFUTED (ν_hi/ν̄ ∈ [0.0845, 0.9754] at GBc; 0.92–0.98 at κ = 0, down to 0.08 at κ ≥ 1); the Gram-floor PD certificate (verifier-side sharpening, never load-bearing).
- **M-L5, the anchor lemma and Z (COMPUTED lemma with a DERIVED sub-scope; Z DERIVED at both orders).** CK = ker(hop|_Z) at 790/790 rows (content at the six survivor rows and the 24 U-S1 face placements; a corollary of SG-2 / SG-1 / Z's shape elsewhere); H10 (order-1 zero-energy set = the balanced configurations, D₁ = −S²/NS), H11 (order-2 zero-energy set = {0, d−1} by the circulant eigenvalues W_r = (25/4)csc²(πr/V) − 5/2, W_0 = 0 exactly at lab period 5), H12 (the dipole-blind weights; parity), Lemma A′ and the block lemma, the shadow kernels F_2 = 1, F_4 = 2, F_odd = 0, H7/H8 symmetries, all machine-checked at every placement and both kernel orders and confronted at U-S1; the data lanes (bookkeeping swap 54 rows; the two-clock joint reading; the 272-cell anchor census, 18 nonempty, all {0, d−1}; shadow inclusion 647/647 computed, 143 typed; the 67 doc-14 frames regressed bit-exactly).
- **M-L6, the fold (COMPUTED; one derivation target closed partially).** λ = 0 strata at all 414 GBc rows (H3, near-corollary) and at none of the 125 F2c rows (group coverage: 6132 singleton, 1951 antichain-unique by lemma, 1360 by exact nullspace, the "two-lab + TT degeneracy" derivation target stands as partial); ν_c ∈ [0.0024042, 0.0134860] over 439 gravity strata; μ > 0 everywhere; the exact algebraic ν_max at C2 (0.02813396980546884 at s_c = 8, degree 30; 0.031135408701826732 at s_c = 6, degree 18) inside the Sturm brackets and ≤ ν̄; the letter-0 curve without positive rational root at 105 curves (96 exact, 7 mod-p); the C2 sporadic components typed by the doc-17 engines (genus-0 conics with dense rational points, max ν 270/20021; genus-1 quadratic components carrying 427/18980 and 9777/663655 at height ≤ 64; the (8,4) and (6,3) components NOT-COMPUTED-BELOW-CAPS, p_a ≤ 55 / 28, never "undecidable"); no rational coupling ν ≥ 1 on any component.
- **M-L7, standing sweeps.** Adjudicator triggers NONE FIRED; Perimeter ARMED with the contribution list graded CLEAR; 2608.25800 0 citers; 2607.15351 2; 2605.20330 0; the papers/12–13 maps re-dated with two additions (the lattice-QED-QRF items; Pickup's classical-frames poster); the doc-19 item-5 shadow lane SWEPT and CLEAR (was UNSWEPT).
- **M-L8, co-runners (never headline; §3.9).** (i) Fork-1: COROLLARY BRANCH, the Mediation Boundary is a corollary with three priced magnitudes (the light-crossing phase |V_AB|d/c = 3.3 × 10⁻¹³ rad at BMV; TT content O(⟨n_g⟩); the transfer amplitude ε) and "no shielding" is in-class a theorem of the constraint structure; A-DQ7 closes as a note. (ii) U2: DECIDED-BY-THEOREM under W1 (the double-commutant identity on a non-abelian D₄ lattice lab with a dim-2 irrep on the branch space) and NOT-INSTANTIATED in the committed toys. (iii′) N1 CLOSES BY COMPUTATION (P2 fires at length 1, exactly but at the 10⁻¹⁶ population level; P3 NOT-POSABLE); E3: the Clifford completion makes the face group infinite (P-E3-1 LIVE), 2→1 BLIND, and **a length-2 Clifford-completed word READS the payload with a free label-sector readout** (P-E3-3's registered failure branch), doc 20's W-B verdict is class-relative at the alphabet level; the next p = 3 front is NOT opened (doc 26 §6 item 3's conjunction not met; the auditor ruling at P-E3-0 is handed the exact object).

### 1.4 Registered predictions, fate (declared parity; refutation-led; the near-corollary rows never findings)

| id | as registered | fate | observable |
|---|---|---|---|
| P-G1 (near-corollary) | Δ^{(R)} < 0 ∀x at the 8 BC/SITE pilot rows; fails at every SPLIT row | **CONFIRMED, certificate exists** | max Δ^{(R)} −2180233/100, −771823/25, −1567, −2215/9 at the gravity rows; the judge's hand bounds exact at three rows, a valid bound (−1639/9) at the fourth; some Δ ≥ 0 at all 15 SPLIT rows; three legs |
| P-G2 | route G at the knob at every definite-form CANDIDATE-GRAV cell | **CONFIRMED** 28/28 | uniformity COMPUTED on 8 independent ν̄ values; ω-collapse (H4) and SPLITR = SPLITL (H1) derived |
| **P-G3** | ν̄(C1) ≤ 1, ν̄(C2) ≤ 2, ν̄ ≤ knob at ≥ 90 % of cells | **REFUTED as frozen → MARGIN-EXT** | brackets hold (0.0066982 / 0.0057392 ≤ 1; 0.0296206 / 0.0337119 ≤ 2); 56/65 = 86.2 % under the frozen ROW population (V2-2: 27/30 = 90.0 % per cell); the nine failures = the SD SITE corner, DERIVED (H5); the population re-fixed per cell in P-G8 |
| P-G4 | CK ⊆ Z + a hop-null pattern at every survivor row; none at the CK = 0 QED rows | **CONFIRMED at ANCHOR-TYPED strength** | 98/98 → 109/109 → 790/790 rows |
| P-G5 | Z = {0, d−1} at every SD SPLIT gravity cell, V even, every placement; Z = ∅ at every definite-form gravity cell | **CONFIRMED** (definite-form clause a NEAR-COROLLARY, D8) | 20/30/40 placements; "no third zero configuration" now DERIVED (H11) |
| P-G6 | GBc strata at every candidate cell, none at F2c; μ > 0; ν_c ∈ [0.002, 0.02]; ν_max(C2) ≤ 0.05; ν_max ≤ ν̄ | **CONFIRMED** (existence clause NEAR-COROLLARY H3; μ > 0 an SG-2 corollary) | F2c absence COMPUTED (125 rows); ν_c ∈ [0.002568, 0.013486] at the pilot, [0.0024042, 0.0134860] over the 439 engine strata |
| **P-G-ctrl** | (i) survivor anchor stratum + route G silent; (ii) QED SITE certifies at ν = 1, folds at √5/20; (iii) swap rows lose every hop-null pattern | (i) **CONFIRMED** incl. d = 64; **(ii) REFUTED as frozen at ν = 1 / CONFIRMED at ν = 2**; (iii) **CONFIRMED** (SG-1 corollary) | ν̄_QED = 11/16 + 5√5/16 (64t² − 88t − 1), Sturm 2 at ν = 1, certified at ν = 2, 3; the (4, 0, 5) stratum folds at √5/20; control-side MARGIN-EXT (H5 at order 1: > 5/4) |
| P-G7 (a) SITE THRESHOLD [D]; BC SPLIT G_x > 0 (85 %); (b) no WINDOW (90 %); (c) no third regime [consistency] | | **CONFIRMED** | 126 SITE cells; 129 BC SPLIT cells; windows []; thirds []; SD SPLIT 18 all NEVER with Z ≠ ∅ |
| P-G8 (75 % / 85 % / 80 %) | per-cell census ≥ 90 % outside SD SITE; ≥ 95 % of ν = 3 cells; control ≥ 90 % | **CONFIRMED** | 246/246 = 100 %; 81/81; 14/14 |
| P-G9 | H5 [D]; attainment pattern (85 %); knob-3 certification; Gershgorin clause [sg12 corollary] | **CONFIRMED** | 51 SD SITE rows exact; argmax x = 0 at s_c ∈ {0, 7}, strict at s_c = 4 (order 2), strict at every order-1 placement, the swap SITE rows attained (data); ν̄_0(V) = 2.50147 / 2.50029 / 2.50009 → 5/2 |
| P-G10 (80 %, re-priced) | CK = ker(hop|_Z) at the six rows + U-S1; dim CK ∈ {0, 1} | **lemma CONFIRMED 790/790; the dim clause REFUTED** (FORECAST FALSE by the note) | dim CK = 4 at the 24 U-S1 face placements, a new anchor structure, data |
| P-G11 (a) non-empty face at V = 8 (65 %), orbit choice {0, 3} ∪ {4}; (b) 16-member checkerboard antichain, dim 1 (85 %); (c) off-face N_Z = 0 and face-free Z antipodally periodic; (d) fraction 0.6 | (a) **CONFIRMED** (orbits as chosen; 24/40; 0.6); **(b) REFUTED**; **(c) second clause REFUTED**, first clause TRUE; (d) CONFIRMED | Z = Z_P × Z_N, a 36-member antichain, dim N_Z = dim CK = 4 spanned by products of K₄ edge-functions (F_4 ⊗ F_4); the face-free Z at s ≡ 1, 2 (mod 5) has 18 members (the sectors a = b = ±1, c = ∓2), not an antipodally periodic set; both refutations FORECAST FALSE in the note before code |
| P-G12 (75 %) | no stratum at any F2c cell | **CONFIRMED (COMPUTED)** | 125 rows; 1360 groups decided by exact nullspace, the rest by L1/L2 |
| P-G13 (99 % / sharpness 80 %) | ν_max ≤ ν̄ at every bracketed THRESHOLD row; ν_hi/ν̄ ∈ [0.75, 1) at GBc | theorem clause **CONFIRMED** (184 classes; C2 exact); **sharpness datum REFUTED** | 148 GBc values, min 0.0845, max 0.9754 (C1/C2 stay at 0.92–0.98; the bound is loose by up to ~12× at κ ≥ 1) |
| P-G14 (85 %, live range) | no positive rational root of det(hop + νD) in (0, 1) anywhere; none at any F2c SD cell | **CONFIRMED** | 105 curves; the ν ≥ 1 clause at 243 cells a corollary of sg12 (M6) |
| P-G15 (a) 90 %; (b), (c) consistency; (d) data | bookkeeping-swap Z = {0, d−1} both kernels; joint reading; shadow inclusion; sporadic typing | **CONFIRMED** (a) 54 rows incl. V = 8 (1/9, 1/18, 1/27); (b) exact (1/160, 1/104, 1/34, 1/126, …); (c) 647/647; (d) typed | **the M2 magnitude clause of (a) fails at ν = 3, s ≡ 4 (mod 5)**, D-S2, §3.8 |
| P-G-ctrl′ (i), (iii) regression; (ii) corrected; (iv) H5 at V = 8 QED SITE [consistency]; the V = 8 swap contrast (90 %) | | **CONFIRMED** | swap N_Z = CK = 0 at all 40 placements of U-S1 (|Z_swap| = 8, or 10 at s ≡ 1 mod 5: 6 periodic + accidental members) |
| hand predictions U-S1..U-S6 (mechanism note §7) | | U-S1, U-S3, U-S5″, U-S6 **CONFIRMED**; **U-S2 partial miss** (GBc 16 (1,2,1) s_c = 6: ν̄ = 0.0338267 outside the registered window (0.05, 0.2); every other clause true; GBc 20 SITE (1,1,2) s_c = 8: ν̄ = 0.0162084, all clauses true); **U-S4 partial miss** (ν̄(s = 1) = 0.014705 > 0.01; the low-prior placement slot confirmed: argmax s = 1, argmin s = 5); **U-S5′ partial miss at s_c = 4** (G_min = −16 knob-free, the note's G_min = 0 was a hand error against the record; Sturm 8, factors 6·6·6·14; classical rational set {0, 1/120, 3/8, 8/15}) | every miss recorded, none suppressed |
| the steelman's own registrations (R1.1–R4.1) | | R1.3 **REFUTED** (Z = {0, 15} at dipole shifts t ∈ {5, 10, 15}, not only t = 10, corrected to the commensurability rule); R1.1, R1.2, R1.4, R1.5, R1.6, R2.1–R2.3, R3.1–R3.3, R4.1 CONFIRMED | §3.7 |
| the rigor leg's own unseen cells U-R1..U-R4 (V = 10; C2 and SD SITE at unregistered placements) | registered in its docstring | **NOT COMPUTED** (its stage R5 not reached; the resumed run stalled in R2), OWED at G26-0.5; every miss to be recorded, never suppressed | §5 |

**Near-corollary rows, labelled as such (never findings):** P-G1; P-G5's definite-form clause; P-G6(a)'s existence clause and (b)'s μ > 0; P-G7(c); P-G9's Gershgorin clause; P-G10 at the gravity / swap / CK = 0 rows; P-G11(a)'s period; P-G14's ν ≥ 1 clause at the 243 BC + GBc cells; P-G15(b), (c); P-G-ctrl′(iv)'s H5 clause; P-G-ctrl (iii)'s hop-null clause.

### 1.5 Wording-ladder rulings (spec §5; the merged verifier recommendations; licensed sentences verbatim)

**Ruling of record: rungs AM-1, AM-2, AM-3, AM-5 (in its re-worded inertia form, both kernel orders) and AM-SPLIT are ENTERED; AM-4 is ENTERED at ANCHOR-TYPED strength with its derived sub-scope stated in-sentence (ANCHOR-LEMMA over the whole family NOT entered); AM-0 not reached; AM-INV nothing.** The steelman leg licensed AM-2 and AM-3 at its bar with two wording repairs (D-S4 the mechanism sentence's attribution; "ν̄ is a sufficient bound" kept), ruled AM-4 ANCHOR-TYPED-plus-derived-sub-scope (D-S5), AM-5 DERIVED at both orders "only with the period clause and the commensurability clause" (D-S3), AM-SPLIT STANDS sharpened, AM-INV nothing; the rigor leg's recommendation at hand-off concurred on AM-2 / AM-3 / AM-5 / AM-SPLIT / AM-INV and would have supported AM-4 at ANCHOR-LEMMA grade "in scope" (its R4/R5 census then pending), the merged ruling takes the stricter of the two on AM-4 (no upgrade). The spec's "both verifier legs" conjunct for AM-2 / AM-3 / AM-SPLIT is met by the steelman leg in full and by the rigor leg at R2 grade only (678 own-path rows at 290 classes with 0 mismatches and 0 construction-identity failures, every d = 64 and d = 256 class included, plus the steelman's own-route AM-3 population 30/30 as the interim second path for the C1/C2 brackets); the rigor leg's R3–R7 (own strata / ν_max / letter-0 / C2 bivariate; the AM-2..AM-5 derivation steps incl. the exact H11 check; its own unseen cells U-R1..U-R4 at V = 10; the ladder-bar audit; the mirror comparison) and 44 d = 16 classes of R2 are OWED at G26-0.5 (§5). **Every rung entered above is therefore ENTERED CONDITIONALLY on the rigor leg completing without a verdict-reversing finding, the doc-21 v1.0 pattern (rungs entered conditional on owed record work), and that completion is a listed conjunct of the G26-3 ruling; if it reverses a verdict this document is amended by a dated block.** **External quotation of every rung is BLOCKED (G26-3) until G26-0.5 + G26-1 + G26-2 discharge; sentences travel whole-with-riders only** (the §7.3 stack, spec §0.3's riders, the doc-13 rider stack whenever doc 13's sentence is touched); the four near-corollary rows and the eight relabelled clauses are never findings; nothing here amends any doc-12/13/14/19/25 sentence.

- **S-AM-1 (rung AM-1, ENTERED, census form):** "In this class, at these cells and placements, below the stated caps, with the coupling extended to ν ∈ Q_{>0} (a family extension): at every one of the 246 definite-form gravity cells of the doc-13 taxonomy outside the SD SITE sub-family, every BC cell and every GBc cell, at every registered placement, and at every one of the 14 definite-form control cells outside its SD SITE sub-family, the reading operator H(q; ν) − qI is strictly diagonally dominant at every real q at the cell's own registered knob, so no self-consistent reading exists in either carrier there; at every such cell an explicit quadratic-irrational ν̄(cell) bounds the coupling above which this holds (36 independent GBc values; exact rational brackets; ν̄(C1) ≤ 1, ν̄(C2) ≤ 2); the SD SITE sub-family (9 gravity cells with ν̄ ∈ [2.50, 2.95]; the QED twins 1.386 / 1.339) is uncertified at knobs 1 and 2 and certified at 3, at 2 and 3 for the QED twins, its silence below 5/2 (order-2 slaving) / 5/4 (order-1 slaving) derived. On every computed row the certificate is a function of the Gram sign structure of the discretization alone."
- **S-AM-2 (rung AM-2, ENTERED; role-blind; the D-S4 attribution fused):** "For any finite reading operator of the registered form, a real symmetric hopping with non-negative diagonal plus, at every configuration x, the quadratic a_x(q) = νm̃₂q² + (νM̃₁_x − 1)q + νD_x in the reading q, with self-energy D_x, dipole coefficient M̃₁_x and positive dipole energy m̃₂, whose Gram determinants G_x = 4m̃₂D_x − M̃₁_x² are all positive, there is an explicit algebraic coupling ν̄ = max_x [b_x + √(b_x² + 4G_x)]/(2G_x), b_x = 4m̃₂(R_x − hop_xx) − 2M̃₁_x, above which, at every candidate reading q and every configuration, the constraint energy νB(π_x + qu, π_x + qu) of the configuration dressed by the clock's own charge exceeds the reading plus the hopping margin R_x − hop_xx; the operator is then strictly diagonally dominant for every real reading and no self-consistent reading exists in either carrier, real or rational. At SITE discretizations the Gram positivity is Cauchy–Schwarz (whenever π_x ∦ u); at BC SPLIT discretizations it is a per-cell computed certificate. ν̄ is a sufficient bound, not the boundary." [family-extension rider; continuum-threshold rider incl. Espíndola–Ali 2605.30423, 2602.24177, the SN threshold and the CVI bound; instrument rider (Horn–Johnson §6.1, §7.2); never a gravity word, the theorem holds beyond the registered family as well (order-3 slaving; lab periods 6 and 7; 200 fuzzed instances, §3.7).]
- **S-AM-3 (rung AM-3, ENTERED; scope V3-2 with the census extension):** "At the committed theorem cells C1 (F2c 4 BC SITE (1,), s_c ∈ {7, 4}) and C2 (GBc 12 BC SPLITL (0,2,0), s_c ∈ {8, 6}), at the 14 definite-form CANDIDATE-GRAV cells (28 rows; 8 independent ν̄ values, every one ≤ its knob), and at every one of the 246 definite-form gravity cells of the doc-13 taxonomy outside the SD SITE sub-family at every registered placement, the registered knob lies above ν̄(cell) (exact brackets ν̄(C1) ≤ 1, ν̄(C2) ≤ 2), so the committed emptiness of the self-reading set, quantum and classical, at those cells is derived by the margin rather than computed: the first derived rows of doc 13's self-reading emptiness. The SD SITE sub-family is excepted below 5/2 by a derived bound; the anchor regime (C3 and every SD SPLIT cell) is excluded." [doc 13's sentence unchanged; G26-4 governs any status change; worded as a derivation of the EMPTINESS, never as an amendment.]
- **S-AM-4 (rung AM-4, ENTERED at ANCHOR-TYPED strength; M5 wording; the derived sub-scope in-sentence):** "At every computed row of the record under the primary normalization, the common kernel CK = ker H ∩ ker M₁ that carries a surviving reading 1/(νm̃₂) is exactly the space of hop-null vectors supported on the anchor sublattice Z of zero-energy, dipole-blind configurations, CK = ker(hop|_Z), with content at the six survivor rows of the record (F2cQ 4 SD SPLITL (1,)/(2,) at s_c ∈ {0, 4}; F2cQ 6 SD SPLITL (1,)/(2,) at s_c = 7) and at the 24 face placements of F2cQ 8 SD SPLITL (1,) (dim CK = 4), and a corollary of SG-2 / SG-1 / Z's shape at every other row; the ⇐ direction is a one-line theorem; the ⇒ direction is derived wherever {M̃₁ = 0} is a hop-antichain (Lemma A′, every no-split placement of every even V) and at the V ≡ 0 (mod 4) split placements (the block determinant δ₀(δ₋δ₀δ₊ − 2δ₋ − 2δ₊) nonsingular at every rational ν), and computed at the V ≡ 2 (mod 4) split placements (V = 6); a quantum self-reading survives iff Z carries a hop-null vector." [secondary-normalization exception typed (doc 13 §3.3(c): the readings {−1/25, 0, 1/20} ride the letter −1/5, H9, outside CK); back-action rider.]
- **S-AM-5 (rung AM-5, ENTERED at both kernel orders in the inertia form; the period-5 and commensurability clauses load-bearing):** "On the registered family the anchor sublattice is decided by the two kernel orders through one quadratic form. The constraint's slaving order fixes the inertia of its energy form Q on lab space: negative semidefinite of rank 1 at order 1 (zero-energy set = the balanced configurations, at every lab period), positive semidefinite of corank 1 at order 2 exactly at lab period 5 (zero-energy set = the antipodal uniform pair, by the circulant identity W_0 = (n² − 1)/12 − (n − 1)/2 = 0; at lab periods above 5 there is no anchor and the margin certifies, at periods 3 and 4 the indefinite third regime); the clock's routing order enters only through the dipole-blind hyperplane Σ_j σ_j w_j(s) = 0, which contains the uniform pair iff the dipole's shift is a multiple of the lab period (the antipodal dipole one instance). Hence under order-2 slaving Z = {0, d−1} at every placement of every F2c SD SPLIT cell with V even, derived, invariant under either clock kernel, hop-null-free, where the classical carrier reads 1/(νm̃₂) and the quantum rational set is empty (the real set typed: four real irrational quantum readings at every placement of the V = 4 cells, at every coupling on [1, 3] tested); under order-1 slaving (the matched QED cell) Z carries a hop-null face exactly at the placements s ≢ 1, 2 (mod 5) when V ≡ 0 (mod 4) and s ≡ 1, 2 (mod 5) when V ≡ 2 (mod 4), doc 13's placement rider and its V = 6 inversion as corollaries, with the face's kernel F_{V/2} ⊗ F_{V/2} (dim 1 at V = 4, 0 at V = 6, 4 at V = 8, spanned by products of K₄ edge-functions; the V ≡ 2 (mod 4) split-placement step machine-checked at V = 6); routing the clock's charge through the order-2 kernel thins a face to its antipodally periodic members (SG-1: {5, 6, 9, 10} → {5, 10}; V = 6: the 8-member antichain → ∅; V = 8: dim CK 4 → 0 at all 40 placements). Where the constraint leaves only the pair, no routing can enlarge it." [family-extension rider on every ν ∉ {1, 2} clause; never "own routing thins the anchor" bare; the registered hop-null clause (iii), a 2^{V/2}-member checkerboard antichain of dimension 1, is REFUTED at V = 8 and replaced by the derived F ⊗ F structure.]
- **S-AM-SPLIT (rung AM-SPLIT, ENTERED, the wall, sharpened):** "On the registered family the self-reading anatomy is a wall between (i) a DERIVED, ROLE-BLIND margin regime, every cell whose constraint-energy form is definite on its configurations (255 of the 273 gravity cells, 18 of the 22 control cells), in which all four reading sets are empty above ν̄(cell), and (ii) a COMPUTED, ROLE-DECIDED anchor regime, the SD SPLIT forms (18 gravity cells, 4 control cells; no third regime in the family), in which the rational reading sets are decided by Z: the classical carrier's coupling-independent anchor reading 1/(νm̃₂) is carried exactly on Z; whether it is the only nontrivial rational classical reading is a computed magnitude datum, true at the gravity SD SPLIT rows at ν ∈ {1, 2}, false at ν = 3 at the s ≡ 4 (mod 5) placements (two further rational readings on a rational conic in ν: 3/8 and 8/15 at V = 4, 6/13 and 98/255 at V = 6) and at the QED SD SPLITL rows (rich censuses), a role datum of the classical carrier, never a role claim; the quantum rational set is empty at every gravity anchor row by computation and face-carried at the QED cell; the real sets in the anchor regime are typed and not empty. On the computed rows every constraint-role fact of the record, the SG-1 swap contrast, the placement rider, S-SR-D1, sits in the anchor regime and the margin sees none of them: the margin is blind to who routes the charge; the anchor's face is decided by it only where the constraint leaves a face to decide, and the constraint leaves the uniform pair as a zero-energy set only at lab period 5." [carrier-typing rider 2607.05578; the non-occupancy sentence; "in the registered family (lab period 5)"; family-extension rider on the ν = 3 clause.]
- **AM-INV: NOTHING.** No non-trivial rational quantum self-reading at ν ≥ 1 at any gravity row: complete rational-root sets at d ≤ 18; d = 64 EXACT-Q-FACTORED in the record and re-derived capless at C3's three placements by the steelman's exact degree-128 route; V = 4 exact bivariate curves at all 20 placements (Farey grid height ≤ 40; every component of the factorization); V = 6 mod-p at the band primes (150 jobs); V = 8 at s = 0 (12 jobs; capless zero-root certificates at the SD SITE knobs 1 and 2); NOT-COMPUTED at the 24 uncertified d = 256 rows (typed; D-S11).
- **The dream's third clause, maximal licensed wording (spec §2.7):** *"on the computed rows, every constraint-role fact of the record sits in the anchor regime, and the margin regime sees none of them; the constraint-role facts are inertia facts of the constraint-energy form (order 1 vs 2) at lab period 5; the anatomy is derived as a trichotomy of one form's inertia (definite / singular / indefinite) and computed as a dichotomy on the registered family."* Never "the anatomy IS doc 13's sentence".

**Not licensed at any rung:** "self-observation has a critical coupling" bare; "the margin theorem explains why gravity cannot read itself" (it does not see gravity); "the anchor lemma is proved" (ANCHOR-TYPED with a derived sub-scope); "own routing thins the anchor" bare (the gravity pair is the constraint's zero-energy set, invariant under either clock kernel); "route G is uniform" bare; "ν̄ is the threshold"; "the reading vanishes at ν_c"; "the classical carrier reads iff Z ≠ ∅" as a biconditional about the whole classical set; "the SD SITE cells are unphysical"; "the QED control failed"; any sentence at ν ∉ {1, 2} without the family-extension rider; the S-DEFLATE and slogan shapes of §7.3.


**[v1.1, 2026-09-23.]** The §1.5 clause "**Every rung entered above is therefore ENTERED CONDITIONALLY on the rigor leg completing without a verdict-reversing finding … and that completion is a listed conjunct of the G26-3 ruling**" is DISCHARGED: the rigor leg completed (stage stamps from its state: R2 2026-09-22 20:14:45, 338/338 classes, 790/790 rows, 0 mismatches, 0 construction-identity failures; R3 20:25:46, own strata / ν_max / letter-0 / C2 at 338/338 classes; R4 20:26:26, 660 census jobs; R5 20:27:56, U-R1 … U-R4; R6 20:27:57, caps / KG / ladder bars / wording; R7 21:50:25 `complete`, mirror3 vs the REGENERATED canon: `n_diff_rows` 0 of 790, the only differing field class `runtime_s` at 737 rows, `verdicts_only_in_record` [], `verdicts_only_in_mirror` = the single A5 KG-4 verdict (F-1), `KG_order_only_diffs` = ["KG-1"], typed outcome identical, `mirror_inputs_sha_verified` 66 with 0 mismatch and exactly 2 RE-BASELINED inputs = the pilot's two record files (`1e345b22b32a5c6f…` → `1b40ebf63cf7f289…`, `c8a5c26dc1bb4fb2…` → `0347c35e8208bfca…`), `mirror_pins_identical` true; E3 (21:50:26) 67/67 frames + the stacked recipe at the six survivor rows) **with no verdict-reversing finding**, its ledger holds 78 CHECK entries of which 5 carry `ok = False` (indices 41, 42, 47 = the own-route items D-R12, D-R13, D-R14, adjudicated verifier-side / reporting-side with the record standing at each; indices 86 and 89 = the 20:43 R7 rows / KG checks against the PRE-regeneration canon, kept as `R7_history[0]` with its 24 stale rows and superseded by the 21:50 re-run, all nine checks PASS) and 28 FINDING entries of which 18 are refuted (all provenance / reporting / own registrations, none a verdict of the record). **The closing audit (G26-1) then re-executed the census through a third band and confirmed every licensed sentence without a wording change** (its C7: all six sentences located verbatim; no gravity word in the role-blind theorem sentence; the family-extension rider on every sentence with a coupling clause outside {1, 2}; no banned slogan; its C9: no record defect verdict-touching). **Rungs AM-1, AM-2, AM-3, AM-5 (inertia form, both kernel orders) and AM-SPLIT stand ENTERED unconditionally; AM-4 stands ENTERED at ANCHOR-TYPED strength with its derived sub-scope in-sentence, the "both verifier legs" conjunct of AM-2 / AM-3 / AM-SPLIT is met in full (steelman in full; rigor at 790/790 own-path rows, the own C1/C2 brackets ν̄(C1) ≤ 7722510872643223/2⁶⁰ ≈ 0.0066982 and 6616865162377309/2⁶⁰ ≈ 0.0057392, ν̄(C2) ≤ 2134386784738187/2⁵⁶ ≈ 0.0296206 and 4858396125207049/2⁵⁷ ≈ 0.0337119, the 246/246 per-cell census, G-definiteness at every BC SPLIT cell of the class, 182 theorem-check brackets, KG-3 clean by an own route), and the closing audit adds the third band (338/338 classes, 0 mismatching leaves; the C1/C2 brackets by its own disjoint-node Lagrange + primitive-PRS Sturm route; ν_hi ≤ ν̄ at 184 classes by exact evaluation of its own ν̄ polynomial).** The rigor hand-off's "ANCHOR-LEMMA grade supportable in SCOPE" for AM-4 is WITHDRAWN by the rigor report; the merged v1.0 ruling (the stricter of the two legs) is the ruling of record and the audit's C2 adds a third-path confirmation of the V = 8 structure, not a derivation. **Two data lines join the record (no rung moves; refutation-led):** (i) the M2 magnitude datum "the anchor reading 1/(νm̃₂) is the ONLY nontrivial rational classical reading at every gravity SD SPLIT row" is REFUTED at ν = 3, s_c = 4 at every V of the taxonomy, V = 4: 3/8, 8/15; V = 6: 6/13, 98/255; **V = 8: 24/55, 7/18 (new to the rigor report; present in the record's rows)**, and true at ν ∈ {1, 2} [family-extension rider on the ν = 3 clause]; (ii) the rigor leg's own unseen cell U-R2 (F2c 10 SD SPLITL (1,), order 2, s = 4; `R5["U-R2"]["order-2 s=4"]`): at V = 10, OUTSIDE the doc-13 taxonomy, a family extension with the rider on every clause, the exception appears already at ν = 1 (classical readings 10/23 and 86/225 beside the anchor 8/5175; census (1021, 0, 3, 1024); G_min = −100 against the registered "G_min = 0", the same hand-error pattern as U-S5′'s −16), so S-AM-SPLIT's clause "true at the gravity SD SPLIT rows at ν ∈ {1, 2}" is a statement about the registered family (V ∈ {4, 6, 8}) and is not contradicted, but its extension to V = 10 is FALSE at s ≡ 4 (mod 5), a role datum of the classical carrier, never a role claim. **A third data line, corroborative (the audit's hand probe; family-extension at NS = 24, outside the caps GBc NS ∈ {12, 16, 20}):** at `GBc 24 BC SPLITL kz=(1, 0) s_c=11 order-2`, ν = 2, a cell outside every pilot and engine shape, registered 2026-09-22 23:46:47 and computed 2026-09-23 02:25:11, all 13 registered clauses matched: certificate CERTIFIED, THRESHOLD, ν̄ = 0.007528875824093492 (minimal polynomial `9980048*t**2 - 46449*t - 216`), Z = ∅, dim CK = 0, classical census all-negative, 0 real quantum readings at ν = 2, Gram floor PD, a λ = 0 stratum present; the own construction reproduces the committed one there. **Quotation status of every rung: CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING (G26-3; §7.7), the outstanding conjunct is G26-2's post-meeting leg (after 2026-10-30); external quotation stays blocked until it discharges. No sentence text changes; S-AM-1 … S-AM-SPLIT travel verbatim as printed in §1.5, whole-with-riders only, the rider set as extended in §7.7.4.**


**[v1.2 amendment (AM-4 lift; §7.5 item (4) executed), 2026-09-24. Rung AM-4 re-graded by the fresh ruling. External quotation of every sentence stays CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING (§7.7).]**

The §1.5 ruling "AM-4 is ENTERED at ANCHOR-TYPED strength with its derived sub-scope stated in-sentence (ANCHOR-LEMMA over the whole family NOT entered)", its v1.1 restatement, and the §7.7.3 ladder row "AM-4 … UNCHANGED; does NOT rise to ANCHOR-LEMMA" are superseded as follows:

> **AM-4 is ENTERED at ANCHOR-LEMMA over the order-1 F2cQ SD SPLITL family** (Gauss clock; every even V; every placement; the registered axis ν ∈ Q_{>0}). This is the family that holds every survivor row of the P-G10 populations: the six rows of record and U-S1's 24 face placements. The spec's §2.3 ANCHOR-LEMMA form is met there: ⇒ is derived at every survivor row, and the lemma is machine-checked at the populations and at U-S1. At every other row of the populations the lemma stays exactly what M5 made it: a COROLLARY of SG-2 at the gravity rows, of SG-1 at the swap rows, and of Z's shape at the CK = 0 QED rows. These are consistency rows, never findings, and never derived there by this block.

The basis is typed and refutation-led:

**(i) D-S5 closes as MIS-SCOPED; the lift is a NEAR-COROLLARY, never a finding.**
- The ⇒ direction at the V ≡ 2 (mod 4) split placements never needed the cross-block step. At a split placement, the S-rows of H(ν) split into independent 4-cycle blocks (the block lemma of the mechanism note §4). Every block with a′ ≠ 0 is nonsingular at every rational ν: this is §3.3's "a′² + 4 is a rational square only at a′ = 0", an algebra that never used V mod 4. Every a′ = 0 block has as its kernel the antisymmetric F-checkerboard, which is supported on Z. Lemma A′'s last line then gives hop v = H(ν)v − ν diag(D)v = 0.
- These are pieces committed on 2026-09-05. The ⇒ direction was therefore a near-corollary of them (`analysis/am4_lift_derivation.md` R2 and Theorem A; confirmed by own re-derivation in `analysis/am4_lift_verifier.md` §2.2).
- It runs along the route the mechanism note registered pre-code ("the sign/antichain structure of hop on {M̃₁ = 0}"). No route was chosen after the fact.
- The gap the record carried (v1.0 §1.5 S-AM-4 "computed at the V ≡ 2 (mod 4) split placements (V = 6)"; §3.3 "the one clause short"; §4 "Computed"; §7.3; §7.7.3; rider 7; rigor §10) was never a gap in ⇒. It was the unproved *shape* of N_Z, which is AM-5 content.

**(ii) The cross-block step is derived in its true form (AM-5 content, not an AM-4 premise).**
- At the V ≡ 2 (mod 4) split placements the non-S rows force the coefficient function t(p, n) of v = Σ t(p, n)(e_{(p,n,−+)} − e_{(p,n,+−)}) into F_m ⊗ F_m. Here m = V/2 − 1, and F_m is the joint kernel of the up- and down-shadow maps on the middle layer of the m-cube.
- Hence CK = N_Z = F_m ⊗ F_m ⊗ (e₋₊ − e₊₋), of dimension Cat((V − 2)/4)². This is Theorem B of the derivation. It rests on the sl₂ commutation identity D_{j+1}U_j − U_{j−1}D_j = (m − 2j)·1 of the Boolean lattice, which is classical: cited, never claimed.
- The unified census follows: dim CK = dim N_Z = Cat(⌊V/4⌋)² at the face placements and 0 elsewhere. The face placements are s mod 5 ∈ {0, 3, 4} when V ≡ 0 (mod 4) and s mod 5 ∈ {1, 2} when V ≡ 2 (mod 4).
- Doc 13's placement rider and its V = 6 inversion are the V ∈ {4, 6} instances. Family-extension rider on every V ∉ {4, 6, 8}.

**(iii) Beyond the axis: a matrix statement, never a reading.**
- CK = N_Z holds at every real ν ≠ 0 on the family (Theorem C of the derivation).
- At the irrational couplings ν*(a′) = 10V√(a′² + 4)/(|a′|(a′² − 4)) one pair of a′ ≠ 0 blocks is singular off Z. There the non-S rows next to that sector, together with the off-middle injectivity of the shadow maps, restore the lemma.
- No reading is registered or claimed at irrational ν.

**(iv) Evidence standing.**
- *Derivation leg:* `analysis/am4_lift_derivation.md`, with `analysis/am4_lift_check.py` (`7af3b46e…`) at 156/156. The V = 14 cell was registered 11:45:24, formula-only, and confirmed 19/19 at 11:54–11:55. The V = 18 spot test went 15/15. Its mirror shows 0 differing leaves.
- *Verifier leg:* `analysis/am4_lift_verifier.md` and `.py`. Every step was re-derived. Own routes:
  - own dense Gauss-law objects equal the committed construction (dense at V = 6 and V = 10; per configuration through the committed chain at V = 14);
  - own column-Markowitz exact elimination over Q, cross-checked against sympy at every matrix with ≤ 300 columns;
  - realification over Q at irrational ν.
- *Verifier cells*, all registered before computation (12:35:05 and 13:05:16) with 0 misses:
  - V = 4 … 12 at all 200 placements;
  - V = 14 at all 70 placements, ν ∈ {1, 2}, plus 5/3 at 11 placements;
  - V = 18 at all 90 placements at ν = 2, plus s = 6 at ν = 1;
  - V = 22 at 4 placements (d = 4,194,304, computed by nobody before). At the split placements s ∈ {1, 2, 41}, \|S\| = 739024 falls into 184756 4-cycle components, and ker H[S, S] is exactly the 63504 a′ = 0 blocks, supported on Z. That is Theorem A's content computed, so CK = N_Z there. At s = 0, CK = 0 = Z. The registered dimension 1764 = Cat(5)² is NOT-COMPUTED-BELOW-CAPS (exact cross-system rank estimated at ~1.5 h per rank);
  - 9 irrational couplings.
- *Record:* the bar's machine-check clause is unchanged (790/790 rows; E3 67/67; rigor R4; G26-1 C1/C2).
- *Primes:* none drawn by either leg. Doc 27's reserve ≥ 8,550,001 and the closing-audit reserve ≥ 8,900,029 are untouched.

**(v) Errata and riders entered with this block (none touches a verdict).**
- *(a)* Mechanism note §4 (registered pre-code; the note stays untouched and the erratum travels as a rider):
  - its sentence "the non-S rows … force the cross-block coefficients to be checkerboards … CK = N_Z = the 2^{V/2}-member antichain checkerboard, dim 1" is true at V = 6 only;
  - so is its "Hop-null criterion (iii) refined" clause "dim CK = dim N_Z = 1 for V ≡ 2 (mod 4)";
  - the values are 4 / 25 / 196 at V = 10 / 14 / 18, on |Z| = 72 / 800 / 9800; at V = 22, |Z| = 127008 and the dimension Cat(5)² = 1764 is derived, not computed.
- *(b)* Doc 27 §3.3's paraphrase "the non-S rows force checkerboards on the balanced P′ and N′ patterns" reads "force t ∈ F_{V/2−1} ⊗ F_{V/2−1} (the pair checkerboard at V = 6)".
- *(c)* Mechanism note §4, "for |a′| ≥ 3 a zero needs ν² ∉ Q", reads "needs ν ∉ Q". ν*² is rational at every singular coupling (for example 125/18 at V = 8, a′ = ±6, and 37/36 at V = 14, a′ = ±12). Doc 27 §3.3 already carries the correct reason.
- *(d)* D-S5 (and SYN-3's "overstated by the V ≡ 2 (mod 4) clause") closes as MIS-SCOPED. The engine's `TABLE.outcome` string "ANCHOR-LEMMA candidate: … ⇒ DERIVED over the order-1 F2cQ SD SPLITL family" stays data, verbatim. It is now true in substance, but it was not derived by the engine.

**S-AM-4 (rung AM-4, ENTERED at ANCHOR-LEMMA over the order-1 F2cQ SD SPLITL family; M5 wording kept; supersedes the v1.0 sentence, which stands as the record of the ANCHOR-TYPED ruling):**

> "At every computed row of the record under the primary normalization, the common kernel CK = ker H ∩ ker M₁ that carries a surviving reading 1/(νm̃₂) is exactly the space of hop-null vectors supported on the anchor sublattice Z of zero-energy, dipole-blind configurations, CK = ker(hop|_Z), with content at the six survivor rows of the record (F2cQ 4 SD SPLITL (1,)/(2,) at s_c ∈ {0, 4}; F2cQ 6 SD SPLITL (1,)/(2,) at s_c = 7) and at the 24 face placements of F2cQ 8 SD SPLITL (1,) (dim CK = 4), and a corollary of SG-2 / SG-1 / Z's shape at every other row; the ⇐ direction is a one-line theorem; the ⇒ direction is derived at every placement of every even V of the order-1 F2cQ SD SPLITL family, by Lemma A′ wherever {M̃₁ = 0} is a hop-antichain, and by the block lemma at the split placements s ≡ 1, 2 (mod 5) (every block with a′ ≠ 0 nonsingular at every rational ν; the a′ = 0 kernels the antisymmetric F-checkerboard on Z; then hop v = H(ν)v − νDv = 0); a quantum self-reading survives iff Z carries a hop-null vector."

Riders on this sentence:
- the secondary-normalization exception is typed (doc 13 §3.3(c): the readings {−1/25, 0, 1/20} ride the letter −1/5, H9, outside CK);
- back-action rider (Hausmann–Renner);
- normalization rider (Ardenghi, cite-only);
- family-extension rider on every clause at V ∉ {4, 6, 8} or ν ∉ {1, 2};
- the A-1 labelling note;
- the lemma uses no gravity word; at the gravity rows it is a corollary of SG-2, never derived there.

**S-AM-5, parenthetical only.** "(the V ≡ 2 (mod 4) split-placement step machine-checked at V = 6)" becomes "(the V ≡ 2 (mod 4) split-placement step derived at every V: N_Z = F_{V/2−1} ⊗ F_{V/2−1} ⊗ (e₋₊ − e₊₋), dim Cat((V − 2)/4)², 1 at V = 6 [family-extension rider off V ∈ {4, 6, 8}])". No other word of S-AM-5 changes.

**[v1.2 pointer 1, 2026-09-24.]** **§1.5 ruling paragraph.** "AM-4 is ENTERED at ANCHOR-TYPED strength … (ANCHOR-LEMMA over the whole family NOT entered)" → see the v1.2 block. AM-4 is at ANCHOR-LEMMA over the order-1 F2cQ SD SPLITL family. "Over the whole family" in the sense of every row stays NOT claimed at derivation grade: the corollary rows keep their computed standing.

**[v1.2 pointer 2, 2026-09-24.]** **§1.5 "Not licensed at any rung".** "the anchor lemma is proved" (ANCHOR-TYPED with a derived sub-scope) → it stays banned **bare**. The licensed form is "the anchor lemma CK = ker(hop|_Z) is derived over the order-1 F2cQ SD SPLITL family", with the S-AM-4 riders.


### 1.6 Parity record (findings against expectations, refutation-first)

(i) **P-G3's ≥ 90 % clause and P-G-ctrl's clause (ii) REFUTED as frozen at the pilot**, both against the judge, both on the SD SITE cells the odds table did not single out; both typed MARGIN-EXT and both then DERIVED (H5: the corner is a theorem about the discretization, ν̄ > 5/2 / > 5/4 at every even V and placement). (ii) **P-G13's sharpness datum REFUTED**, ν_hi/ν̄ as low as 0.0845 at the κ ≥ 1 GBc cells (0.92–0.98 at κ = 0): the Geršgorin bound is nearly sharp exactly where the theorem cells sit and loose by up to ~12× elsewhere; "ν̄ is a sufficient bound, not the boundary" is load-bearing. (iii) **P-G11(b), P-G11(c) and P-G10's dim ∈ {0, 1} clause REFUTED at V = 8**, as the mechanism note FORECAST FALSE before code: the face Z is the 36-member antichain Z_P × Z_N with dim N_Z = dim CK = 4 (F_4 ⊗ F_4), the face-free Z has 18 members; a NEW anchor structure, data. (iv) **Three hand predictions at the reserved cells partly missed** (U-S2's ν̄ window; U-S4's ν̄(s = 1) = 0.0147 > 0.01; U-S5′ at s_c = 4: G_min = −16, the note's G_min = 0 was a hand error against the record, Sturm 8, and the classical rational set {0, 1/120, 3/8, 8/15}). (v) **The M2 magnitude clause dies at ν = 3** (D-S2): the "only nontrivial rational classical reading" datum is a knob-arithmetic accident at ν ∈ {1, 2}; the non-anchor readings lie on a rational conic with dense rational points [family-extension rider]. (vi) **The registered hop-null criterion (iii) and the registered "Fourier parity" attribution of AM-5 did not survive as worded**, replaced by the derived F ⊗ F structure and the inertia statement with the period-5 clause; the steelman's own R1.3 ("Z = ∅ at every dipole shift t ≠ 10") REFUTED and corrected to the commensurability rule (Z = {0, 15} at t ∈ {5, 10, 15}). (vii) **Against the engine author (process):** seven launches; three verdicts flipped REFUT → CONF between launches after predicate re-scopings (each confirmed correct by own computation; D-S1); the committed rows carry one stale non-verdict field at 24 rows (D-S12); the S0 stage imported the pilot library from canon and overwrote two pilot record files (D-R1, HIGH as record integrity, verdict-neutral). (viii) **Against the verifier legs (process):** the steelman killed the rigor leg's re-launch by mistake at 08:42 (D-S0); the rigor leg's first instance crashed in its own code (Fraction vs mpf) and its second was handed off mid-R2, this synthesis re-launched it (08:50:33), which completed the four pending GBc 20 / GBcQ 20 classes and then stalled for 24 minutes on the first four V = 4 (d = 16) classes with four workers at 100 % CPU and no completion, and was stopped cleanly (state untouched; typed NOT-COMPLETED-BELOW-CAPS; the likely stall, not diagnosed, no profiler available, is the leg's uncapped sympy `count_roots` reference behind its own Sturm's 240 s cap). (ix) **In the judge's favour:** the leading branch of the odds table was reached (a landed margin at the theorem cells, the anchor typed, the fold census as data); ν̄(C1) ≈ 0.007 as estimated; the two-regime anatomy of doc 26 §4.3 computes exactly as forecast and then sharpens into the inertia statement; every Scout C constant reproduces bit-exactly; the KG-4 hunt at extended caps, all placements and the unseen V = 8 finds nothing.

---

## 2. Literature position

**Scout A (instruments, theorem shapes, the anchor genre, the re-date; `papers/24_anchor_margin/SCOUT_MARGIN_ANCHOR_INSTRUMENTS.md`, 2026-09-04/05; 6 PDFs + the Perimeter export).** **L3 (instrument): CLEAR**, arXiv metadata (abs:Gershgorin ∧ gr-qc/quant-ph/hep-th/hep-lat = 2, unrelated; "diagonally dominant" ∧ clock/frame/self-consistent/WdW = 12, all numerics/control; "Levy-Desplanques" anywhere = 1, an LLM-quantization paper), INSPIRE full text (Levy–Desplanques 5 hits, none a reading obstruction; Gershgorin ∧ PW/QRF/relational 1, a numerical bound), the house corpus (185 PDFs: one Gershgorin eigenvalue bound, 2504.12263); the only Geršgorin hit inside the Page–Wootters genre is a numerical bound on a history-state density matrix (2208.01187); Vilenkin–Zhang 1403.1599 use Lévy–Desplanques for uniqueness of a linear system, the closest in logical role, with no coupling parameter and no exclusion. **L1 (self-energy detuning as an obstruction): CLEAR at theorem shape**, three threshold-shaped continuum neighbours of a different mechanism (§1.1): 2602.24177 (0 citers), the SN self-trapping genre, and Yokokura's CVI (Perimeter talk; no arXiv id as of 2026-09-05, if it posts before any wording entry it is full-texted first and the continuum-threshold rider re-worded). The interacting-clock PW–QRF genre (Smith–Ahmadi 1712.00081; 2304.01263; 2204.11740; 2504.01579; Lock's and Semião's Perimeter items; Mekonnen 2607.24976; Ardenghi 2608.09601; Shimada 2608.16762; Ali 2601.23146; Chen–Giacomini 2606.09344 (7 citers); Doat–Vanrietvelde 2510.13607 (9); Wei 2506.21489 (14); DEHK 2412.15502 (69 citers, the p. 73 open-problem declaration standing, no citer computing a backreacting frame)) asks whether a clock observable exists that makes the conditioned dynamics unitary, never whether a fixed point q = λ(q) exists; the **interacting-clock rider** is forced ("breakdown of the formalism" is theirs). **L2 (anchor / zero-mode genre): CLEAR** (§1.1 PC-2); the Woods et al. accuracy line no-overlap; the lattice-QED-with-QRFs cluster new to the maps (control positioning). Cauchy–Schwarz-type derivations of critical couplings in constraint-role models: CLEAR (the Gram positivity is Horn–Johnson §7.2). **L5 (re-date):** every verdict of the papers/12 and papers/13 maps UNCHANGED; citer counts re-dated; the doc-19 item-5 shadow lane now SWEPT, CLEAR (nearest new relative: Pickup's poster on QRF transformations inside Gaussian/stabilizer subtheories, transformations, not readings); premise genre re-dated (2602.15134's R2 is a postulate; 2608.25800 a classical toy-model epistemic horizon). **L4 (Perimeter, pre-sweep):** the Indico JSON export is public, 86 contributions, 61 titled, 57 with abstracts, every one graded; CLEAR for the house's four shapes; the 18 invited/review slots unnamed until PIRSA posts; the post-Oct-30 target list registered (§7.4 G26-2).

**Scout B (occupancy + triggers; `SCOUT_QRF_OCCUPANCY_TRIGGERS.md`, 2026-09-05; 3 PDFs).** **GROUND-CLEAR for the theorem shape** ("no self-consistent self-reading above a computable coupling, in any carrier, plus a role-decided zero-energy anchor below it") **and for the two-regime anatomy**; nobody has computed or stated it. The one genuine neighbour: **Espíndola–Ali 2605.30423** (full text read; Proposition 2's sufficiency ‖T_P‖² < δ_P for Gaussian saddle control; the spectator / observer / loss-of-control anatomy; the S⁴ benchmark α² ≥ 6), admissibility-side, weak-coupling-derived, no fixed-point equation, no carriers, no anchor, no control; **must-cite and distinguish on every margin sentence** (the G26-0(b) ban is reinforced by it: their theorem is explicitly a de Sitter statement, the house's explicitly not). The DEHK / Chen–Xu / QRF partition-relativity line 2025–26 (Freidel–Kirklin 2604.02228 dressing time, the organizers' backreacting-frame object, "Schrödinger non-ideal" by coherent-state overlaps, no reading exclusion; the third-particle cluster; 2607.05578 reduced-frame channels; 2607.12032 indefinite Born probabilities; Brady 2609.01894; Lacambra–Chatwin-Davies–Honda 2604.06087) computes no two-regime anatomy of self-access on a constrained model. Adjudicator triggers NONE FIRED (§1.1). What does NOT occupy this ground, stated for the record: observer no-go theorems that assume QM (Frauchiger–Renner, Local Friendliness); QRF covariance results; continuum backreacting-frame constructions, the non-occupancy sentence travels with the two-carrier claim.

**Occupied adjacent ground, cite-and-distinguish on every export:** Espíndola–Ali 2605.30423; Arderucio Costa–Sanchez 2602.24177; Giulini–Großardt 1105.1921 and Colin–Durt–Willox 1403.2982; Yokokura (CVI, talk) and Lock (talk), cite as talks until posted; Hausmann–Renner 2607.15351 (+ the Perimeter no-go form); Höhn et al. 2409.06479; Freidel–Kirklin 2604.02228 / 2510.26589; DEHK 2412.15502; Chen–Xu 2511.00622; Luppi–Kabel–Giacomini–Smirne 2607.05578; Castro-Ruiz–Cohen–Barbado–Brukner 2607.12032; Ardenghi 2608.09601 (rider-level, next to doc 13 §3.3(c)); Carrasco-Martinez 2602.15134 and Gonda–Fankhauser–De les Coves 2608.25800 (premise genre); Rothlin–Ferradini–Chen 2604.06149 and Pagan Lacambra et al. 2604.06087 (control positioning); the instrument precedents 2006.13938, 2208.01187, 1403.1599; Schlegel–Kumar–Paterek–Dakić 2605.20330 (classical-carrier genre, 0 citers); Salih 2609.04104 (a watch item for the Fork-1 lane only).

**Method note (typed, affects confidence grading):** both scouts ran with arXiv throttling (HTTP 429 mid-run; Scout B stopped its sweep at job 44/68 and re-ran the lost nets through INSPIRE, whose quant-ph coverage is thinner); INSPIRE is the citer source of record; full text was read for the decisive papers (2605.30423 pp. 1–8, 2604.02228, 2602.24177, 2602.15134, 2608.25800, DEHK p. 73, 2409.06479 grep) and titles/abstracts for the rest, **metadata-plus-INSPIRE-full-text grade**; Google Scholar and the philosophy indices not used; the Perimeter contributed abstracts graded from the export, the invited talks ungraded until named. G26-2 re-runs every lane after 2026-10-30 (§7.4). Watch list standing: Yokokura's CVI posting (ARMED-HIGH); Lock / Semião; Hausmann's no-go form; the 18 unnamed invited talks; 2605.30423's citers (3); 2608.25800 / 2607.15351 / 2605.20330 citers; the Andriolo pipeline and MOVW ref [7] (Adjudicator triggers).

---

## 3. Results

### 3.1 Run record, battery, pins, primes (green before any verdict row)

| item | outcome |
|---|---|
| pilot (M-L1) | `anchor_pilot.py` 67/67, exit 0; 98 rows; five resumed launches (D1) closed by four byte-identical mirrors (92/98 rows byte-identical to the record, the other 6 differing only in cap-typed / label-only fields); three verifier legs 10/10, 37/37, 16/16; reconciliation 32/32 |
| spec + gate | v1.0 frozen 00:43 → G26-0 44/44 (7.0 s) → v1.0.1 re-frozen 01:36 (Appendix D; 46/46 pins) |
| mechanism note | 03:02:39, birth = mtime (written once); pinned at engine A0 and re-hashed at A6 (equal); precedes every engine artifact (`anchor_lattice.py` created 03:26:32) |
| engine (A0–A6 + TABLE) | **checks 27 (0 failed); verdicts 59 (7 refuted)**; 790 rows / 338 classes (A1 241 classes / 669 rows; A6 101 classes / 121 rows); A0: spec sha = sidecar, four verbatim blocks byte-identical (3733 / 25437 / 16802 / 2666 chars), prime gate green, own-toolkit selftest, S0 pilot regression (98 rows identical at every verdict quantity; the 16 verdict items; the P-G5 sweep); the margin stage run with the family label masked (per-row masked-input hashes); campaign 04:04–05:12 in seven launches (header "runtime 3.3 s" = the last segment) |
| canon pins | 62 files (the 46 spec-pinned + the pilot's 41 + the spec) byte-identical before / after every engine stage; 59 (steelman) and 79 (rigor R0) pins likewise; re-verified by the co-runner's CONFIRM pass against 28 pin tables (0 canon mismatches); **exception (record, not canon): `analysis/anchor_pilot_run.log` and `anchor_pilot_state.json` overwritten at 04:10 by the engine's S0 pilot driver (D-R1, §5)** |
| primes | pilot 8,200,007 / 8,200,013; engine 8,200,037 / 8,200,061; co-runner 8,200,109 / 8,200,141; pilot-verifier legs 8,300,009 / 8,300,041; gate 8,300,057; rigor 8,300,069 / 8,300,077; steelman 8,400,011 / 8,400,019 / 8,400,031 (spare 8,400,043 certified); closing-audit reserve ≥ 8,500,007 UNTOUCHED; successor ≥ 8,600,023 fallow; every earlier reserve and the doc-25 bands untouched; nothing decided by a prime anywhere |
| controls | the QED family co-quantified under the same stratification (§3.6); the masked-label margin run (98/98 identical to the pilot at the gate; 790 rows in the engine); the steelman's out-of-family controls (§3.7); KG-3 clean everywhere (Sturm 0 at every certified (row, ν) where computed; the vertex identity and disc ≤ Δ at every configuration; every LDL sample PD; ν_hi ≤ ν̄ at 184 classes; 200 fuzzed instances) |
| caps (typed, printed) | d ≤ 18 exact throughout; d = 64 F(q) mod-p + CRT below height 5,798,310 (engine) / 5,869,003 (pilot verifier) / 5,939,707 (steelman), Sturm NOT-COMPUTED (the steelman's exact degree-128 Sturm alarmed at 900 s); d = 256 F(q) NOT-COMPUTED, ν_max NOT-COMPUTED; the (8,4) / (6,3) C2 components' genus NOT-COMPUTED-BELOW-CAPS; Ostrowski/Brauer sharpening not implemented (never load-bearing) |

### 3.2 The margin, census, ν̄ table, the SD SITE corner, sharpness, exact boundaries

**Stratification (A2; P-G7):** gravity 273 cells → **255 THRESHOLD, 0 WINDOW, 18 NEVER** (the 18 = every F2c SD SPLITL/SPLITR cell at V ∈ {4, 6, 8} × ν ∈ {1, 2, 3}, all with Z = {0, d−1}); control 22 cells → 18 THRESHOLD, 4 NEVER (the F2cQ SD SPLITL cells). 126 SITE cells THRESHOLD by the Cauchy–Schwarz clause (π_x ∦ u at every configuration, every profile has ≥ 3 charged sites); 129 BC SPLIT cells THRESHOLD by computation (G_x > 0 at every configuration, incl. the d = 64 cell U-S6 with G_min = 102446027/135).

**Certification at the knob (A1; P-G8, per cell):** 246/246 definite-form gravity cells outside the SD SITE sub-family; 81/81 of their ν = 3 cells; 14/14 definite-form control cells; the 15 definite CANDIDATE-GRAV cells (14 + the SPLITR spot) 28/28 rows at the knob with C1 ≤ 1 / C2 ≤ 2 by exact brackets. **ν̄ is knob-free** (G_x, b_x per unit ν): the GBc census carries 36 independent values (3 NS × 3 κ × 2 disc × 2 s_c), ω-independent (H4), SPLITR = SPLITL (H1).

| cell class (own routing = order 2; control = order 1) | ν̄(cell) (50 digits in the state file; minimal polynomial where quoted) | certified at |
|---|---|---|
| C1 = F2c 4 BC SITE (1,) s_c = 7 / 4 | 0.0066982104521301681919 (2348293t² − 800t − 100) / 0.0057392156672745021661 (776412t² − 100t − 25) | 1, 2, 3 |
| F2c 4 BC SITE at the 7 unregistered placements s = 1, 2, 3, 5, 6, 8, 9 (U-S4) | 0.014705 / 0.008346 / 0.006451 / 0.005606 / 0.005918 / 0.008077 / 0.009904 (argmax s = 1, argmin s = 5; m̃₂ from 133/80 at s = 1 to 3333/80 at s = 9) | 1 |
| F2c 4 BC SPLITL/R (1,) s_c = 7 / 4 | 0.0074243256 / 0.0062566657 | 1, 2, 3 |
| F2c 6 BC SITE s_c = 7 / 4 (U-S3 at 4) | 0.0011116580 / 0.0013806998 | 1, 2, 3 |
| F2c 6 BC SPLITL/R s_c = 7 (U-S6) / 4 | 0.0011512997 (G_min 102446027/135) / 0.0014424623 | 1, 2, 3 |
| F2c 8 BC SITE s_c = 4 | 0.0005356511 | 1, 2, 3 |
| C2 = GBc 12 BC SPLITL (κ = 0, ·, ω) s_c = 8 / 6 | 0.029620566898399910174 (11125t² + 582t − 27) / 0.03371189525596569154 (148t² + 84t − 3) | 1, 2, 3 |
| GBc 12 BC SPLITL κ = 1 / κ = 2, s_c = 8 | 0.0534749738 / 0.0942458275 | 1, 2, 3 |
| GBc 12 BC SPLITL κ = 1 / κ = 2, s_c = 6 | 0.1885736642 / **0.7565236958** (27/74 + √210/37; the largest CANDIDATE-GRAV value) | 1, 2, 3 |
| GBc 12 BC SITE (0, ·, ·) s_c = 8 / 6 | 0.0245253885 / 0.0269865601 | 1, 2, 3 |
| GBc 16 BC SPLITL κ = 0 s_c = 8; κ = 1 s_c = 6 (U-S2) | 0.0108749404 / 0.0338267499 | 1, 2, 3 |
| GBc 20 BC SPLITL (2, ·, 0) s_c = 8 / 6; κ = 0 s_c = 8 | 0.0512583160 / 0.0910100039; 0.0062207 | 1, 2, 3 |
| GBc 20 BC SITE κ = 1 s_c = 8 (U-S2) | 0.0162084233 | 1, 2, 3 |
| **F2c 4 SD SITE** s_c = 0 / 7 / 4 | **2.5014697242** (272t² − 680t − 1) / 2.5014697242 / **2.9544471594** (256t² − 756t − 1) | 3 only |
| **F2c 6 SD SITE** s_c = 0 / 7 / 4 | 2.5002936513 / 2.5002936513 / 2.7483890268 | 3 only |
| **F2c 8 SD SITE** s_c = 0 / 7 / 4 (d = 256, margin only) | 2.5000932801 / 2.5000932801 / 2.6686493185 | 3 only |
| F2cQ 4 BC SITE (1,) s_c = 7 / 4 (control) | 0.0383492219 / 0.0360843918 | 1, 2 |
| GBcQ 12 / 16 / 20 BC SITE, s_c = 8 / 6 (control) | 1/14 = 0.0714285714 (exactly rational) / 0.0724744871; 0.0503615799 / 0.0503899360; 0.0429613529 / 0.0426180394 | 1, 2 |
| F2cQ 6 BC SITE s_c = 7 / 4; F2cQ 6 BC SPLITL s_c = 7 / 4 (control) | 0.0163873852 / 0.0172626028; 0.0208747931 / 0.0225112584 | 1, 2 |
| **F2cQ 4 SD SITE** (every placement) / **F2cQ 6 SD SITE** (control) | **11/16 + 5√5/16 = 1.3862712430** (64t² − 88t − 1) / 1.3385214790 | 2, 3 (silent at 1) |
| every SD SPLIT row (gravity, QED, swap, bookkeeping), 33 rows in the pilot, 18 + 4 cells in the engine | INDEFINITE Gram (G_min ∈ {0, −16, −36, −64/5, −128, −156/5}); Δ^{(R)} = b_xν + 1 ≥ 1 at the G_x = 0 configurations, NEVER-TYPE, no WINDOW, silent at every ν | - |

**The SD SITE corner is a theorem about the discretization (H5, DERIVED; 51 rows):** at every F2c(Q) SD SITE cell the uniform configuration has M̃₁_0 = 0 at every placement under either clock kernel (the period-5 profile lives on even Fourier modes; the antipodal dipole on odd ones), R_0 = V, hop_00 = 0, D_0 = Vc₅ with c₅ = Σ_period π² = 2/5 (order-2 profile (−1/5, −2/5, 2/5, 1/5, 0)) / 4/5 (order-1 profile (−1/5, 4/5, −1/5, −1/5, −1/5)); hence ν̄_0 = (5/4)(1 + √(1 + 2/(5Vm̃₂))) (order 2) / (5/8)(1 + √(1 + 4/(5Vm̃₂))) (order 1), exact to 10⁻³⁰ at every row, and ν̄(cell) ≥ ν̄_0 > 1/c₅ = 5/2 / 5/4 at every even V and every placement, attained (argmax x = 0) at s_c ∈ {0, 7} and strict at s_c = 4 under order 2 (2.50147 / 2.50029 / 2.50009 → 5/2 at V = 4, 6, 8), strict at every order-1 placement, attained at the swap SITE rows (order-2 dipole on order-1 energy, data). The same threshold is sg12's committed Gershgorin pattern at q = 0 (νVc₅ − V > 0 ⟺ ν > 5/2; the 24 non-strict cells, consistency, never a finding). KG-1's letter therefore fires at the 12 F2c SD SITE rows with ν ≤ 2 (d = 16 exactly; d = 64 height-typed, D-R6/D-S7) and is typed **DERIVED-SILENT**, "the SD SITE sub-family below 5/2", never "cell-accidental" (V2-3); the census is NOT-UNIFORM at the certificate level with its boundary derived, and the Gram-floor positive-definiteness certificate (Sylvester/Bareiss on K = hop_off + diag(α_x), α_x = νD_x + hop_xx − (νM̃₁_x − 1)²/(4νm̃₂)) certifies F2c 4 SD SITE at ν = 2 where route G is silent (six rows), the verifier-side sharpening, never load-bearing.

**The real-locus boundary and the sharpness datum (route G′; P-G13):** ν_max ≤ ν̄ at all 184 bracketed THRESHOLD classes (ν_hi ≤ ν̄ exact); exact algebraic values where the bivariate curve factors, **ν_max(C2, s_c = 8) = 0.02813396980546884** (the boundary root of the (8,4)-factor discriminant, a degree-30 integer polynomial) and **ν_max(C2, s_c = 6) = 0.031135408701826732** (degree 18, the (6,3) factor), both inside the Sturm brackets [160151/5692448, 259449/9221903] / [251389/8074063, 267952/8606013] and ≤ ν̄; C1's curve is irreducible of bidegree (32, 16) at both placements (394 terms) and its exact discriminant stays NOT-COMPUTED-BELOW-CAP (brackets [18657/2876966, 53205/8204356] and [39427/7039325, 20179/3602766] certified by exact Sturm on both sides); the QED SITE cell's ν_max = 1.0773994614031295 (degree 68; the (12,6) factor). **Sharpness REFUTED as registered:** ν_hi/ν̄ = 0.968 / 0.976 at C1, 0.950 / 0.924 at C2, 0.92–0.98 at every κ = 0 BC/GBc cell, 0.777 at the QED SITE cell, **0.316 at the gravity SD SITE cell, and as low as 0.0845 at the κ ≥ 1 GBc cells** (148 values, max 0.9754), the κ-gate's extra hop unit loosens the dominance bound by up to ~12×. The registered knobs lie 6.5–180× above the entire real locus at every CANDIDATE-GRAV row and at C1 (the factor 6.5 at the κ = 2, s_c = 6 rows where ν_max = 0.309).

**The letter-0 curve (P-G14):** det(hop + νD) has no positive rational root at any of the 105 curves computed (96 exact at d ≤ 64, 7 mod-p + CRT at d = 256); its ν ≥ 1 clause at the 243 BC + GBc cells is sg12's corollary (M6); the live range (0, 1) everywhere and every ν at the 27 F2c SD cells is clear.

### 3.3 The anchor, the lemma, Z at both kernel orders, the placement rider, the swap contrast, the new structure at V = 8

**The lemma (A3; P-G10).** CK = ker(hop|_Z) at **790/790 engine rows** (pilot 98/98; pilot verifier 109/109; gate 50 placements; steelman by an own nullspace at the six survivor rows and 50 Gauss placements), with content at exactly the six survivor rows of the record, F2cQ 4 SD SPLITL (1,)/(2,) at s_c ∈ {0, 4} (Z = {5, 6, 9, 10} at s_c = 0 and {3, 5, 10, 12} at s_c = 4: 2-faces of the 4-cube carrying the checkerboard (1, −1, −1, 1); readings 1/4, 1/8) and F2cQ 6 SD SPLITL (1,)/(2,) at s_c = 7 (Z = {7, 11, 21, 25, 38, 42, 52, 56}, an 8-member complementary-pair antichain; readings 2/13, 1/13), and at the 24 face placements of U-S1; at the CK = 0 QED rows Z = {5, 10} (V = 4, s_c = 7) or ∅ (V = 6, s_c ∈ {0, 4}) with N_Z = 0; at every gravity row CK = 0 (SG-2's corollary; N_Z = 0 by H6). The ⇐ direction is a one-line theorem; the ⇒ direction holds by **Lemma A′** wherever S = {M̃₁ = 0} is a hop-antichain with hop_xx = 0 on S (the rows y ∈ S of (hop + νD)v = 0 read νD_y v_y = 0, forcing supp v ⊆ Z; the remaining rows then read hop v = 0), every no-split placement of every even V, and by the **block lemma** at the split placements s ≡ 1, 2 (mod 5), where the two labs with vanishing dipole weight form a 2-cube on which hop|_S is a 4-cycle and the S-rows read (C₄ + diag(δ₋, δ₀, δ₀, δ₊))v = 0 with det = δ₀(δ₋δ₀δ₊ − 2δ₋ − 2δ₊): nonsingular at every rational ν when V ≡ 0 (mod 4) (a′² + 4 is a rational square only at a′ = 0, so no survivor: V = 4 s_c = 7, V = 8 s ≡ 1, 2), singular exactly when a′ = 0, possible only for V ≡ 2 (mod 4), where the block kernel is the antisymmetric F-checkerboard and the cross-block step (the non-S rows force checkerboards on the balanced P′ and N′ patterns) is asserted in the note and **machine-checked at V = 6 only** (D-S5), the one clause short of ANCHOR-LEMMA over the family.

**Z at both kernel orders (H10–H12; derived and machine-checked at every placement of every census cell, both orders).** Configurations σ ∈ {±1}^V; NS = 5V. *Order 1 (Gauss):* π₁ = Σ_j σ_j e_{5j+1} − (S_σ/NS)·1 ⇒ D₁^{SPLIT}(σ) = −S_σ²/NS (the zero-energy set = the C(V, V/2) balanced configurations: 6 / 20 / 70 at V = 4 / 6 / 8; the uniform configurations have D₁ = −V/5) and D₁^{SITE}(σ) = V − S_σ²/NS ≥ 4V/5 > 0 (no zero). *Order 2:* D₂^{SPLIT}(σ) = (1/NS)Σ_r W_r|σ̂_r|² with W_r = (25/4)csc²(πr/V) − 5/2 ≥ 15/4 for r ≠ 0 and **W_0 = −2 + (1/4)Σ_{m=1..4}csc²(πm/5) = 0, the period-5 cancellation, (n² − 1)/12 = (n − 1)/2 exactly at n = 5** (verified exactly by the circulant eigenvalues, sympy, and to 10⁻³⁰ at every configuration) ⇒ D₂ = 0 iff σ is constant: **Z ⊆ {0, d−1} at every V, derived** (P-G5's "no third zero configuration"). *Dipole-blind sets:* M̃₁(σ) = Σ_j σ_j w_j(s) with w_j ∈ {−1, 0, 1} at order 1 (zero exactly at the two antipodal labs when s ≡ 1, 2 mod 5, the split placements) and W_{j+V/2} = −W_j at either order, so the 2^{V/2} antipodally periodic configurations are dipole-blind at every placement under either clock kernel. **Consequences:** gravity SD SPLIT (order-2 energy): Z = {0, d−1} under BOTH clock kernels at every placement of V = 4, 6, 8 (12/12 pilot rows; 54 bookkeeping rows; the P-G5 sweep 20/30/40), the pair is the constraint's doing, not the clock's (M4); QED SD SPLITL (order-1 energy): at s ≢ 1, 2 (mod 5) Z = Z_P × Z_N (balanced on each half-ring block; |Z| = C(V/2, V/4)² if V ≡ 0 mod 4, else ∅), at s ≡ 1, 2 (mod 5) the sectors {a = b = ±1, c = ∓2} (V ≡ 0 mod 4) or the antichain {a = b = c = 0} (V ≡ 2 mod 4). The Z-census symmetries **H7** (Z(s + 5) = rot Z(s); Z(s + NS/2) = Z(s)) and **H8** (Z(3 − s) = complement ∘ reflection Z(s)) hold at every placement of all 22 (cell, order) censuses, so the face census is a function of s mod 5 on the orbits {0, 3}, {1, 2}, {4}.

**The placement rider, derived (AM-5 (iv)):** Gauss faces at residues {0, 3, 4} mod 5 at V = 4 (12 of 20 placements: s ∈ {0, 3, 4, 5, 8, 9, 10, 13, 14, 15, 18, 19}; fraction 0.6) and at {1, 2} at V = 6 (12 of 30; fraction 0.4), doc 13's "V = 4 keeps the reading at s_c ∈ {0, 4} and loses it at 7; V = 6 keeps it only at s_c = 7" is the residue pattern read at three placements; the inversion between V ≡ 0 and V ≡ 2 (mod 4) is the block lemma's a′ = 0 case.

**U-S1 (F2cQ 8 SD SPLITL (1,), d = 256, all 40 placements, both kernels, the reserved order-1 target; hand prediction registered in the note before computation, confronted in A6 and by the steelman's own constructor at 7 placements):** faces at s mod 5 ∈ {0, 3, 4} (24 placements; the orbit choice {0, 3} ∪ {4} as V = 4), CONFIRMED; there |Z| = 36 (Z_P × Z_N, an antichain), **dim N_Z = dim CK = 4**, CK spanned by products f(x_P)g(x_N) of K₄ edge-functions with vanishing vertex sums (F_4 ⊗ F_4; the K₄ incidence matrix has rank 4), reading 1/9 (m̃₂ = 9 = (NS − 4)/4; D₁(uniform) = −8/5; 70 balanced configurations); at s mod 5 ∈ {1, 2} |Z| = 18 (the alternating pair {85, 170} among them), N_Z = CK = 0. Under the order-2 swap kernel |Z_swap| = 8 at s ≡ 0 (6 antipodally periodic balanced configurations + the two accidental members {0, 3, 5, 6} / {1, 2, 4, 7}, as the note predicted), 10 at s ≡ 1 (mod 5), and **N_Z = CK = 0 at all 40 placements, the O-4 contrast at V = 8 (dim CK 4 → 0)**. The registered checkerboard-antichain criterion (P-G11(b)) and the "face-free Z is antipodally periodic" clause (P-G11(c)) died here as forecast; the dim CK ∈ {0, 1} clause of P-G10 died, a NEW anchor structure at V = 8, typed as data.

**The O-4 check (KG-5 trigger 2 absent; M-L3(c) discharged):** Gauss dim N_Z = dim CK = (1, 0, 1) at s_c = (0, 7, 4) of F2cQ 4 SD SPLITL (1,)/(2,) → (0, 0, 0) under the SG-1 swap with Z = {5, 10} at all three placements (the face members 6, 9 / 3, 12 lose dipole-blindness under the order-2 dipole: M̃₁ = ∓10, ∓6; 5 and 10 keep it by parity); at V = 6 s_c = 7 the 8-member Gauss antichain (N_Z = CK = 1) becomes Z = ∅ after the swap (a new datum: empty, not a pair); at V = 8 as above. Route G silent under both kernels at ν ∈ {1, 2, 3} at every survivor and swap row (KG-5 trigger 1 absent: max Δ^{(R)} = 89 / 209 / 2189/5 at the survivor rows).

**Data lanes (never headline):** the bookkeeping swap (order-1 kernel on the gravity SD SPLIT cells, 54 rows): Z = {0, d−1}, the reading 1/(νB₁(u,u)) = 1/(ν(NS − 4)/4) present (1/4, 1/8, 1/12 at V = 4; 2/13, 1/13, 2/39 at V = 6; 1/9, 1/18, 1/27 at V = 8), CK = 0, no nonzero rational quantum reading where computed; the two-clock classical joint reading μ = 1/(ν(m̃₂ᴬ + m̃₂ᴮ + 2B(u_A, u_B))) exact at both uniform configurations of every pair (1/160, 1/104, 1/34, 1/126 at V = 4 ν = 1; 1/1105, 1/608, 1/992 at V = 6 ν = 2), a corollary of S-SR-D1; the 272-cell anchor census at every knob: Z ≠ ∅ exactly at the 18 F2c SD SPLIT cells, always {0, d−1}, empty at every definite-form cell (85 of 91 ν = 1 cells in the pilot's census); the 67 doc-14 frames reproduce dim CK bit-exactly through the {M̃₁ = 0}-column construction (regression); shadow inclusion R_self^{Q,Q} ⊆ R_self^{cl,Q} at 647/647 computed rows (143 typed); **H9 (letter-anchored readings): the engine's lane crashed (an exception string in the results, D-R4/D-S6, typed NOT-COMPUTED there) and the datum was supplied by the rigor leg's own path (E2) and the gate (S6):** at F2cQ 4 SD SPLITL (1,) s_c = 7 the letter −1/5 is carried on {M̃₁ = 0} (dim 1); the primary readings q − 4q² = −1/5 have discriminant 21/5 (irrational); the secondary-normalization readings q − 100q² = −1/5 are exactly {−1/25, 1/20}, doc 13 §3.3(c)'s readings are letter-anchored readings, the anchor lemma's λ = 0 instance generalized; at gravity cells every letter is positive and ≫ 1/(4m₂), so letter-anchored readings need λ ≤ 1/(4m₂), a derivable exclusion at the committed cells, registered as data, never a rung.


**[v1.2 pointer 3, 2026-09-24.]** **§3.3, last sentence of "The lemma".** "… where the block kernel is the antisymmetric F-checkerboard and the cross-block step (the non-S rows force checkerboards on the balanced P′ and N′ patterns) is asserted in the note and machine-checked at V = 6 only (D-S5), the one clause short of ANCHOR-LEMMA over the family" → "… where the block kernel is the antisymmetric F-checkerboard, supported on Z; ⇒ then follows by Lemma A′'s last line, without the cross-block step (v1.2). The cross-block step (the non-S rows force t ∈ F_{V/2−1} ⊗ F_{V/2−1}, the pair checkerboard at V = 6) is derived at every V ≡ 2 (mod 4) as AM-5 content."


### 3.4 The fold (A4/A5; M-L6) and the sporadic components

Strata (hop-null vectors on (D, M̃₁)-constant groups with rational eigenvalue λ): **λ = 0 strata at all 414 GBc rows** (H3: the hop-isolated TT = 0 configuration, both labs in P2, is a hop-null vector inside its own group; the TT = 1 twin a λ = ω eigenvector at ω ≥ 1; near-corollary), plus placement-dependent degenerate-pair strata (s_c = 6, κ = 0: the group {2, 3, 6, 7}); **none at any of the 125 F2c rows** (COMPUTED: 6132 singleton groups excluded by L1, R_x = V > 0, 1951 antichain-unique groups by L2, 1360 groups by exact nullspace; the "two-lab + TT degeneracy is necessary" derivation target stands partially closed); μ > 0 at every gravity stratum (SG-2's corollary); **ν_c ∈ [0.0024042, 0.0134860] over 439 gravity λ = 0 strata** (C2: 0.0134860 at s_c = 8, 0.0105542 / 0.0107191 at s_c = 6; GBc 20: 0.0025684 / 0.0028359; GBc 16: 0.0045900; the κ = 1 unseen rows 0.0046433 / 0.0024042), two orders of magnitude below the knobs [family-extension rider]. Control strata: the (0, 0, 0) anchor stratum at the six survivor rows (ν_max = ∞: the hyperbola qν = 1/m̃₂); (4, 0, 5) at F2cQ 4 SD SITE folding at √5/20 exactly while the cell's full curve reaches ν_max = 1.0774 through its (12,6) factor; (88/15, 0) at F2cQ 6 SD SITE; the negative-μ strata −1/5 (F2cQ 4 SD SPLITL s_c = 7, the letter of §3.3(c); its conic ν = q/(4q² − 1/5) has a real branch at every ν > 0 and rational points where 1 + 16ν²/5 is a rational square: (q, ν) = (1/4, 5), (1/3, 15/11), …, irrational at ν ∈ {1, 2}) and −2/15 (F2cQ 6 SD SPLITL); GBcQ strata (92/3, 40/3) etc.; the QED BC mirrors carry strata too (F2cQ 4 BC SITE s_c = 4: (576/5, 128/5)), strata are not GBc-only on the QED side. **The sporadic components at C2 (P-G15(d), the doc-17 locus engines):** at s_c = 8 the reading curve F(q, ν) (bidegree (36, 18), 499 terms) factors as (2,1)² (4,2)² (4,2)² (8,4)²: the (2,1) factor is the stratum conic ν = 9q/(56q² + 292q + 629), genus 0, rational points dense, carrying the registered points (1, 9/977), (1/2, 3/526), (2, 6/479), max ν found 270/20021; the two (4,2) factors are genus-1 double covers w² = disc_sf(q) carrying exactly the sporadic points (1, 427/18980) and (1, 9777/663655) below height 64 in q (others NOT-FOUND-BELOW-HEIGHT); the (8,4) factor's genus is NOT-COMPUTED-BELOW-CAPS (p_a ≤ 55; one point (1/10, 90/67391) below height 24); at s_c = 6: (2,1)² (2,1)² (4,2)² (4,2)² (6,3)² with the conics ν = 4q/(32q² + 188q + 285) and 9q/(72q² + 414q + 629), the genus-1 factors' points (1, 64917/3250468) and (1, 98613/7501258), and the (6,3) factor NOT-COMPUTED-BELOW-CAPS (p_a ≤ 28). **No rational coupling ν ≥ 1 on any component; every exact ν_max lies on an identified factor and below ν̄.** "Undecidable" never (N1); no committed cell as an open problem (F-1).

### 3.5 The gravity SD SPLIT cells, the anchor regime under own routing, and what stays computed

At C3 (F2c 6 SD SPLITL/R (2,), d = 64) and the F2c 4 and 8 SD SPLIT cells: G_x = 0 exactly at the two uniform configurations and negative at others (indefinite Gram; the NEVER type carried by the uniform configurations with b = 4m̃₂V ≥ 0, m7); route G silent at every ν; Z = {0, d−1} at every placement under both clock kernels; N_Z = CK = 0; the classical anchor reading 1/(νm̃₂) present at every placement (1/40, 1/80, 1/120 at V = 4; 8/1105, 4/1105, 8/3315 at V = 6; 1/330, 1/660, 1/990 at V = 8 ν = 1, 2, 3 [family-extension rider off the knobs 1, 2]); the quantum rational set EMPTY at every gravity anchor row where computed, d = 16 by complete factorization (F(q, 1) with four real irrational roots at every one of the 20 placements: **the readings leave Q but not R**, D4; the steelman's exact bivariate curves show Sturm 4 at ν = 1 and 4 / 8 at ν ∈ {3/2, 2, 5/2, 3} (6 at 3/2 for s ≡ 0, 3 mod 5) at every placement, the real locus of the anchor cell extends along the whole coupling axis tested), d = 64 by the record's complete factorization at C3 (56·72 / 56·72 / 24·24·32·48) and **re-derived capless at all three placements by the steelman's exact degree-128 route (241-digit leading coefficient; rational-root sets {} / {} / {})**, its real-root count NOT-COMPUTED-BELOW-CAPS (900 s alarms), d = 256 NOT-COMPUTED with the mod-p statement at s = 0 (§3.8). The anchor regime's quantum emptiness stays computed, the accepted stop, pre-registered in doc 26 §7 and unchanged here.

### 3.6 The control, anatomy at engine grade (co-quantified; role-blind census)

(i) The six survivor rows carry the (λ, μ, m) = (0, 0, 0) anchor stratum of dimension 1, the coupling-INDEPENDENT anchor q = 1/(νm̃₂) with the committed {0, 1/4} / {0, 1/8} / {0, 1/13} / {0, 2/13} as its ν ∈ {1, 2} points, and route G is silent there at every ν (KG-5 trigger 1 absent). (ii) F2cQ 4 SD SITE: G-definite; (G, b) = (64, 88) at the argmax; **ν̄_QED = 11/16 + 5√5/16 = 1.38627124…** (64t² − 88t − 1) at every placement; Δ_argmax(1) = −64 + 88 + 1 = 25 ≥ 0, so route G is silent at ν = 1 (F(q, 1) has exactly two real irrational readings, Sturm 2, none rational; the Gram floor's first non-positive minor at order 14), certified at ν = 2 (max Δ = −79) and 3; the (4, 0, 5) stratum folds at √5/20; the ν = 1 silence DERIVED (H5 at order 1: ν̄_0 = 1.26238 / 1.25553 at V = 4 / 6, > 5/4); F2cQ 6 SD SITE ν̄ = 1.3385215, the same pattern; doc 13's QC-2 "SITE collapses" is Q-arithmetic at ν = 1 and margin-carried at ν = 2: **control-side MARGIN-EXT, never a control failure**. (iii) The SG-1 swap rows: Z = {5, 10}, N_Z = CK = 0 at all 20 placements of V = 4 (Z = ∅ at V = 6 s_c = 7; |Z_swap| ∈ {8, 10} at V = 8), R_self^{Q,Q} = {0}, six non-anchor classical readings each (census (0, 8, 8)), doc 19's swap-corner pattern reproduced (KG-5 trigger 2 absent). (iv) The QED margin census under the same stratification: 14/14 definite-form control cells certified (GBcQ 12 / 16 / 20 BC SITE at knobs 1 and 2 with ν̄ ∈ [0.0426, 0.0725], GBcQ 12 s_c = 8 exactly 1/14; F2cQ 4 BC SITE 0.0383 / 0.0361; F2cQ 6 BC 0.0164–0.0225); the F2cQ SD SPLITL rows NEVER-TYPE with Z rich under Gauss routing (the NEVER type there carried by Z's members and by G < 0 configurations with b ≥ 0, not by the uniform configurations, m7); the classical control censuses rich ((0, 10, 6, 16) at V = 4 with readings {−3/4, 1/4, 5/4} etc.; (0, 44, 20, 64) at V = 6 with 20 rational-square configurations where Z = ∅), the classical carrier's reading set is a different carrier's set, never the quantum survivor's shadow, never aggregated into a role claim. **A control cell that certifies where its gravity twin does not, or vice versa, is a role-blindness datum, never a role claim** (spec §3.2 (iv)); on the record the pattern is identical on both sides: the SD SITE corner is MARGIN-EXT for gravity (ν̄ > 5/2) and for QED (ν̄ > 5/4), and the SD SPLIT cells are NEVER-TYPE on both sides.

### 3.7 The theorem beyond the family, the steelman's out-of-family controls (lens 1; family-label-free constructor `Lattice(n, V, k_c, k_u, disc, s, t)`, reproducing the committed objects at n = 5, 10/10 rows)

| control | result | reads on the sentences |
|---|---|---|
| lab period n ∈ {3, …, 7}, order (2, 2), SPLIT | the uniform pair is zero-energy **iff n = 5** (W_0(n) = (n² − 1)/12 − (n − 1)/2 = 0 exactly at n = 5 in 2..12); n < 5: D(uniform) < 0, **NEVER-TYPE with Z = ∅, the spec's registered third regime, realized out of family**; n > 5: THRESHOLD (ν̄ = 14.40 / 7.00 at n = 6 / 7, V = 4) | the anchor of the dream question is a period-5 coincidence of the order-2 symbol with the lab spacing (D-S3): the period clause is load-bearing in every AM-5 / AM-SPLIT sentence |
| order (1, 1) SPLIT at every period | D₁(σ) = −S_σ²/NS at every n; the zero-energy set = the balanced set; NEVER-TYPE everywhere | the order-1 anchor is period-independent |
| non-antipodal dipole e_s − e_{s+t}, t = 1..19, at F2c 4 SD SPLIT | Z = {0, 15} exactly at t ∈ {5, 10, 15}, ∅ elsewhere (the steelman's own registration "∅ at every t ≠ 10" REFUTED and corrected): the uniform profile's Fourier support in VZ makes 1 − ω^{−kt} vanish there iff 5 | t | "dipole-blind by Fourier parity" is the t = NS/2 instance of "dipole-blind by commensurability with the lab period", the commensurability clause |
| clock-only control D ≡ 0 | NEVER-TYPE whether M̃₁ is kept or zeroed; with D kept and M̃₁ ≡ 0 the cell stays THRESHOLD (ν̄ = 2.5015 at the SD SITE cell) | the margin is carried by the configuration's constraint energy in excess of the hopping, the clock entering through the vertex deficit, the D-S4 attribution fused into S-AM-2 |
| order-3 slaving (constraint and clock) at period 5 | no anchor (Z = ∅); THRESHOLD at SPLIT and SITE (ν̄ = 25.00 / 4.17 at V = 4) | the margin certifies at a kernel order the family never registered |
| Theorem G fuzzed on 200 random symmetric instances (negative off-diagonals, nonzero diagonals, G_x > 0 by construction) | 0 real roots of F(q, ν) at ν ∈ {ν_hi, 2ν_hi, 10ν_hi} by own Sturm; real roots present below ν_lo/2 at 94/200 instances | the theorem is a statement about matrices; the bound is not vacuous |
| the certificate TYPE vs the inertia of Q_{k_c} on lab space (50 (n, V, k_c, disc) groups) | **THRESHOLD ⟺ Q positive definite** on the computed grid; NSD (order-1 SPLIT), singular PSD (order-2 SPLIT at n = 5, corank 1, kernel = constants) and indefinite (order-2/3 SPLIT at n < 5) all NEVER; no WINDOW anywhere; the TYPE identical under every clock order (ROLE2.1) | the margin/anchor split is the definite / singular split of ONE quadratic form; "blind to who routes the charge" is true of the TYPE beyond the family |
| H5 at V = 8 SD SITE (d = 256, own constructor) | ν̄ = 2.5000932801 at s ∈ {0, 7} (attained at x = 0; = the closed form) and 2.6686493185 at s = 4 (strict); certified at knob 3 only | the record's d = 256 rows reproduced by a second path |
| the swap SITE rows (order-1 energy, order-2 clock) | certified at 2, 3 and silent at 1 exactly as the Gauss SITE rows; H5's closed form holds with c₅ keyed by the CONSTRAINT's order | the certificate does not see the clock's routing order in family either |

### 3.8 The KG-4 inversion hunt (extended caps, all placements, unseen V) and the classical-carrier conic datum

**INV4, V = 4, SD SPLITL and SD SITE, all 20 placements, the EXACT bivariate curve F(q, ν) (bidegree (32, 16); 33 × 17 exact 16 × 16 determinants + Newton interpolation):** rational points searched in the q-direction on the Farey grid of height ≤ 40 (3924 rational q per placement; 1250–3884 settled capless by a 30-prime sieve, the survivors factored exactly), in the ν-direction on the height-≤ 12 grid on [1, 3], and on every component of the exact factorization (ν-degree 1: dense; ν-degree 2: the discriminant-square search to height 64), **no rational point with ν ≥ 1 anywhere at 40/40 placements**; **reducibility datum:** F(q, ν) is REDUCIBLE at every placement of BOTH the SPLITL and the SITE cell with the SAME component shapes, [(12, 6), (20, 10)] at 16 placements and [(6, 3), (6, 3), (6, 3), (14, 7)] at s ≡ 4 (mod 5), every component of ν-degree ≥ 3 (a property of the 4-cube's symmetry, not of the discretization; mechanism not derived); the SITE cell has Sturm(F(q, ν)) = 0 at every ν ∈ {1, 3/2, 2, 5/2, 3} at all 20 placements, **no real quantum reading anywhere on [1, 3] including the interval [1, ν̄) below its margin bound: the derived MARGIN-EXT corner's silence is a looseness of the bound (3×), not a reading; KG-3 clean at 100 (placement, ν) points above ν̄.** **INV6, V = 6 (d = 64), SD SPLITL / SD SITE at s ∈ {0, 7, 4}:** F(q, ν) mod each band prime by 129-node interpolation; 150 jobs (30 q-direction at ν ∈ {1, 3/2, 2, 5/2, 3}; 120 ν-direction at 20 rational q): no rational quantum reading and no rational coupling ν ≥ 1 (5 + 27 capless zero-root-prime certificates; the rest CRT-typed below height 5,939,707); the only exact kernels found are the six chiral-point artifacts (q, ν) = (2, 0) (ν = 0 excluded by registration; kernel dim C(6, 2) = 15, the 6-cube's eigenvalue-2 eigenspace of hop). **INV8, V = 8 (d = 256; NOT-COMPUTED in the record):** F2c 8 SD SITE s = 0 at ν = 1 and ν = 2, **zero-root prime 8,400,019: no rational quantum reading, exact and capless**, the record's NOT-COMPUTED rows at the V = 8 SD SITE cell below its ν̄ = 2.5000932801 carry a capless emptiness certificate at the two knobs where route G is silent; F2c 8 SD SPLITL s = 0 at ν ∈ {1, 2}: CRT-typed, no rational root below height 5,939,707; the ν-direction at q ∈ {1/2, 1/3} zero-root primes, at q ∈ {1, 1/10, 1/330} CRT-typed, at q = 2 the chiral-point artifact (dim C(8, 3) = 56). The steelman's own INV8 predicate did not exclude ν < 1 (its logged R3.3 line reads REFUT on the ν = 0 artifact alone, D-S13, disclosed, log left unedited); the registered R3.3 (ν ≥ 1) is CONFIRMED at every job. **KG-4: silent**, subject to the caps printed (24 uncertified d = 256 gravity rows NOT-COMPUTED; D-S11).

**D-S2, the classical carrier's non-anchor rational readings in the anchor regime lie on a rational conic in ν (data; a role datum of the classical carrier, never a role claim; family-extension rider on ν ∉ {1, 2}).** At knobs 1 and 2 the rational classical set is {0, 1/(νm̃₂)} at every placement of every gravity SD SPLIT cell (the M2 datum, 12/12); **at knob 3 the placements s ≡ 4 (mod 5) carry two further rational readings**, V = 4: configuration 6 (M̃₁ = −36, D = 8, m̃₂ = 40; G = −16) reads 3/8 and 8/15 (120q² − 109q + 24 = 0, discriminant 19²), V = 6: configuration 28 (M̃₁ = −233/2, D = 49/2, m̃₂ = 1105/8; G = −36) reads 6/13 and 98/255, because at those configurations G_x < 0 and the classical discriminant −G_xν² − 2M̃₁_xν + 1 is a rational conic in ν through (0, 1) with DENSE rational points: ν(t) = 2(36 − t)/(t² − 16) at V = 4, ν(t) = (233 − 2t)/(t² − 36) at V = 6; sample rational couplings ≥ 1 with rational non-anchor readings: ν = 23/21 (q ∈ {23/40, 8/23}), 18/13, 11/6, 86/33 at V = 4 and 267/253 (q ∈ {534/1105, 98/267}), 53/44, 263/189 at V = 6, each verified against the exact classical quadratic. Doc 25 rider (e)'s "one real irrational classical reading at C3 s_c ≡ 4 (mod 5)" and these knob-3 rational readings are the same conic read at three couplings; the M2 clause "the only nontrivial rational classical reading" is a knob-arithmetic accident at ν ∈ {1, 2}, and S-AM-SPLIT carries it so. Nothing here touches the quantum sets; nothing feeds "gravity is classical" (doc 25 §7.3; the classical carrier is a different carrier's set).

### 3.9 Co-runners (M-L8; `anchor_corunner.md`; never headline; primes 8,200,109 / 8,200,141 as tokens only; 47/47 records identical across two independent launches)

(i) **Fork-1 corollary check, COROLLARY BRANCH.** For sources coupled linearly to a free field the mass-side reduced state at every t is (direct phase gate) × (dressing overlap), the polaron identity upgraded from final time to all times, machine-verified against brute-force Fock evolution to < 10⁻¹⁴; for the massless 3+1 field the AB-entangling second difference is exactly Θ_AB(t) = V_AB·max(0, t − d/c) (two elementary integrals, verified to < 10⁻¹⁸), so every mass-side protocol sees readings (I) and (II)-direct differ by at most the phase of one light-crossing, **|V_AB|·d/c = 3.3 × 10⁻¹³ rad at BMV**; the interposed field region's physical algebra on H_phys is exactly (functions of the labs' protected charges) ⊗ B(TT) (dim 8 = 2 × 4 on the committed GB model, exact), so an S2-type interposition instrument is either a TT instrument (its share ⟨n_g⟩ ~ 10⁻⁵⁷) or a matter-charge which-path measurement, the same operation in either reading: "no gravitational shielding" is in-class a theorem of the constraint structure, not technology. Doc 03's bound closes the transfer channel only; what it does not cover is typed (the direct-gate reading itself; untrapped constituents; non-static backgrounds; the O(G) edge; continuum lifts; the dressing overlap = A1′). A-DQ7 CLOSES as a corollary note; a cite-and-distinguish rider candidate for Paper B drafted, not entered. (ii) **U2, DECIDED-BY-THEOREM under W1 and NOT-INSTANTIATED in the committed toys.** The committed toy's region multiplet {E₁, D₁, E₂, D₂} is abelian (every irrep 1-dimensional; the momentum charges "kernel-covered, not toy-modeled"); on a lattice lab with a non-abelian D₄ charge algebra (dim A = 13, A′ = 28, Z(A) = 7; two dim-2 irrep blocks; the branch space occupied by the (E, 1) block with isotypic weight 1/2) every charge expectation is identical before and after every interior branch operation (the double-commutant identity; 13/13 basis elements) while the commutant-side data change; two labs: span(A₁′ ⊗ A₂′) = the commutant of the joint charge algebra (16 = 16 = 4²), doc 04's main theorem extends verbatim to non-abelian charges. The "twirl-D front" is a theorem; live only outside W1. (iii′) **N1 closes by computation** (the Cayley flow letter C_H is non-monomial with a negative Gross kernel, minimal entry α + β√3 ≈ −8.66 × 10⁻⁵ exact in Q(√3); P2 FIRES: C_H alone at length 1 separates every registered tt-pair through the free Z-basis label populations, exactly, at the 10⁻¹⁶ population level, typed as data; P3 NOT-POSABLE); **E3:** the group orders corrected (⟨9 constructor faces⟩ = 236,196 = 2²·3¹⁰, the scout's number; ⟨14 faces⟩ = 4·3³¹), F_∂ non-monomial and outside the doc-20 free linear-part group, P-E3-1 LIVE (F_∂·U_G2 has a non-integral charpoly ⇒ an eigenphase with θ/π ∉ Q ⇒ ⟨faces ∪ F_∂⟩ INFINITE; 319 of 582 distinct words ≤ 3 non-integral), P-E3-2 BLIND (20,031 / 20,036 nodes; the capless half a locality theorem), **P-E3-3's registered failure branch fires: dZ-pt-k1 · F_∂ (length 2) reads every registered tt-pair with a free label-sector readout** (the δ-Z faces write payload-dependent phases on the label register; the label Fourier turns them into populations), "doc 20's W-B verdict is class-relative at the alphabet level"; NEGATIVITY-REQUIRED not contradicted (every reading word contains a negative-kernel face); doc 24's F-4 is what is alphabet-relative. The next p = 3 front is **NOT opened** (doc 26 §6 item 3: N1 fails its conjunction; E3's conjunction needs the auditor ruling at P-E3-0, which now has the exact object to rule on, AND a field-carriage conversion found at L ≤ 6, not searched). Never "gravity converts", never "third carrier", never "gravity is a magic factory".

---

## 4. Mechanism at supported strength

The mechanism note (`analysis/anchor_mechanism.md`, registered pre-code 03:02:39 by the engine author, a disclosed DEVIATION, its pre-code status confirmed by the provenance leg by birth times and by the engine's A0/A6 sha pins) registered the AM-2 write-out with (iii′), H5 restated, the zero-energy and dipole-blind sets at both orders (H10–H12), Lemma A′ and the block lemma as the AM-4 route, the F2c stratum-absence lemmas, graded forecasts for P-G7..P-G15 + P-G-ctrl′ (with P-G11(b), P-G11(c) and P-G10's dim clause FORECAST FALSE), and a hand prediction for every reserved cell. The engine confronted every hand fact first (A6); the verifier legs re-derived each by own paths.

**Derived (role-blind; no property of gravity used; confirmed by the gate's masked-label run, the steelman's out-of-family controls and the rigor leg's own-path rows).** *Theorem G (the form of record; spec §2.1 with (iii′)):* for a finite set X, a real symmetric hop with hop_xx ≥ 0, R_x = Σ_{y≠x}|hop_xy|, rationals D_x, M̃₁_x and m̃₂ > 0 with G_x = 4m̃₂D_x − M̃₁_x² > 0 ∀x: (i) the diagonal entry a_x(q) + hop_xx of A(q; ν) = H(q; ν) − qI is a convex quadratic in q with minimum νD_x + hop_xx − (νM̃₁_x − 1)²/(4νm̃₂) at q*_x = (1 − νM̃₁_x)/(2νm̃₂), and "minimum > R_x" is exactly Δ_x^{(R)}(ν) < 0 (the two differ by −1/(4νm̃₂)); Δ_x^{(R)} is a concave quadratic in ν with one positive root ν̄_x, so it is negative for every ν > ν̄ = max_x ν̄_x; (ii) hence A(q; ν) is strictly diagonally dominant for every real q, hence nonsingular (Lévy–Desplanques): R_self^{Q,R}(ν) = ∅, a fortiori R_self^{Q,Q}(ν) = ∅; (iii′) the classical quadratic is the diagonal entry a_x(q) + hop_xx itself (doc 19's simplex carrier; doc 25's closed form with T3 = 4νm̃₂hop_xx), whose discriminant is Δ_x^{(R)} − 4νm̃₂R_x ≤ Δ_x^{(R)} < 0: R_self^{cl,R}(ν) = ∅, the v1.0 step "R_x − hop_xx ≥ 0" is NOT used (it is false at the 90 hop-isolated TT = 1 configurations of the ω ≥ 1 rows; M3); (iv) at SITE forms G_x > 0 whenever π_x ∦ u (Cauchy–Schwarz / Gram; π_x ∦ u structurally, every profile has ≥ 3 charged sites); at BC SPLIT forms G_x > 0 is a per-cell certificate (129/129 cells). Corollaries, never findings: ν_max ≤ ν̄; Z = ∅ at THRESHOLD cells; the near-corollary bound at the committed rows; the Gram-floor PD certificate (strictly sharper, it certifies F2c 4 SD SITE at ν = 2); the exact upper bound of ν_max by monotonicity of the symmetric pencil (route G′). *H5:* the SD SITE corner as a theorem about the discretization (§3.2). *The inertia statement (the steelman's sharpening of the two-regime anatomy; ROLE2, 50 out-of-family groups; H10/H11 exact):* D(σ) = σᵀQ_{k_c}σ with Q the Gram matrix of the slaved single-lab profiles on lab space; THRESHOLD ⟺ Q positive definite on the computed grid; order-1 SPLIT: Q negative semidefinite of rank 1 at every lab period (zero set = the balanced hyperplane ∩ the cube); order-2 SPLIT: Q positive semidefinite of corank 1 with kernel = constants **exactly at lab period 5** (corank 0 at n ≠ 5; the anchor exists because (n² − 1)/12 = (n − 1)/2 has the unique solution n = 5 among n ≥ 2), THRESHOLD at n > 5, indefinite (the third regime, no anchor) at n < 5; SITE at any order positive definite. *Z = (ker Q ∩ {±1}^V) ∩ {Σ_j σ_j w_j(s) = 0}:* the clock's routing order enters only through the weights w_j(s); the uniform pair is dipole-blind iff the dipole's shift is a multiple of the lab period (commensurability; parity = the t = NS/2 instance). *Lemma A′ and the block lemma:* §3.3. *H7/H8:* the SD chain's translation and reflection symmetries of Z. *The F ⊗ F structure:* at the no-split placements Z = Z_P × Z_N and N_Z = F_{V/2} ⊗ F_{V/2}, F_n the kernel of the up/down-shadow maps on the middle layer of the n-cube, F_odd = 0, F_2 = 1, F_4 = 2 (hence dim CK = 1, 0, 4 at V = 4, 6, 8; machine-checked at all 40 placements of V = 8 and by the steelman's own nullspaces at 7).

**Computed (the constraint-role content; role-decided).** The ⇒ direction of the anchor lemma at the V ≡ 2 (mod 4) split placements (V = 6, s_c = 7, two of the six survivor rows): the cross-block step asserted, checked at V = 6 only (D-S5). The F2c stratum absence at the 1360 groups not covered by L1/L2 (P-G12; the "two-lab + TT degeneracy" target). The quantum rational emptiness of the anchor regime (every gravity SD SPLIT row: computed, and the accepted stop). The V = 4 reducibility pattern of the exact reading curves (a property of the 4-cube's symmetry; mechanism not derived). The classical carrier's non-anchor conic (D-S2; data). The sharpness datum's κ-dependence (the κ-gate loosens the dominance bound; a bookkeeping fact of the bound, typed).

**The mechanism sentences that travel (spec §5.2, with the verifier corrections fused; never the polynomial, never the inequality):** *Margin*, at every candidate reading, the constraint energy of the configuration dressed by the clock's own charge exceeds the reading plus the hopping margin (a_x(q) = νB(π_x + qu, π_x + qu) − q identically; the D-S4 attribution, with D ≡ 0 the cell is NEVER-type whatever the clock terms). *Anchor*, a zero-energy, dipole-blind face of configurations carries the surviving reading; the constraint's own energy leaves such a face only where it is slaved at order 1, and routing the clock's charge through the order-2 kernel thins a face to its antipodally periodic members, where the constraint's zero-energy set is already the antipodal pair, no routing can enlarge it and no reading survives. *Control*, the Gauss-routed clock keeps a coupling-independent anchor at the survivor placements (the hyperbola qν = 1/m̃₂) and loses it under the constraint-role kernel; at its SITE cell it folds at √5/20 and is margin-certified above 11/16 + 5√5/16. *The wall*, the margin is blind to who routes the charge; the anchor's face is decided by it only where the constraint leaves a face to decide, and the constraint leaves the uniform pair only at lab period 5.

**What did not survive (refutation-first):** the family-wide "ν̄ below the knob at ≥ 90 % of cells" as frozen by rows (the SD SITE corner; derived); the control's "certifies at ν = 1" (Q-arithmetic at ν = 1, margin at ν = 2); the sharpness datum ν_hi/ν̄ ∈ [0.75, 1) at GBc (0.0845 at κ ≥ 1); the registered hop-null criterion (a 2^{V/2}-member checkerboard antichain of dimension 1, dim 4 at V = 8); "any face-free Z is antipodally periodic" (18 members at V = 8, s ≡ 1, 2); dim CK ∈ {0, 1} (4 at V = 8); the M2 "only nontrivial rational classical reading" beyond ν ∈ {1, 2}; the note's G_min = 0 at U-S5′ s_c = 4 (−16, knob-free, already in the record, a hand error); the ν̄ windows of U-S2 and U-S4; the steelman's own "Z = ∅ at every t ≠ 10"; the v1.0 Theorem G step (iii) (M3); the registered "own routing thins the anchor" attribution (M4, the gravity pair is the constraint's zero-energy set) and the registered "via Fourier parity" wording of the order-1 half (replaced by the combinatorial route: prefix-sum profiles, half-ring lab weights, shadow kernels). **Prediction-grade only (never findings):** that the V ≡ 2 (mod 4) cross-block step derives at every V (the rigor leg's U-R1 at V = 10 is the registered test); that ν_max/ν̄ → 1 at κ = 0 BC/GBc cells as V grows; the V-odd case (outside the registered family); the d = 64 real branches at the knobs; the reducibility pattern of the V = 4 curves at V = 6, 8; the letter-anchored exclusion at gravity cells (H9, registered as data).


**[v1.2 pointer 4, 2026-09-24.]** **§4 "Computed".** Its first sentence, "The ⇒ direction of the anchor lemma at the V ≡ 2 (mod 4) split placements … checked at V = 6 only (D-S5)", moves to "Derived" (v1.2). **§4 "Prediction-grade only":** "that the V ≡ 2 (mod 4) cross-block step derives at every V (the rigor leg's U-R1 at V = 10 is the registered test)" → DERIVED (Theorem B). U-R1's dim 4 = Cat(2)².


---

## 5. Verification record

Six adversarial or second-path legs on top of the pilot's own battery: the three pilot-verifier legs (10/10; 37/37 at 109 rows with 7 unseen rows U-A..U-F; 16/16, consolidated `anchor_pilot_verifier.md`), the author's reconciliation (32/32), the hostile gate G26-0 (44/44 with its own masked-label margin run, own 272-cell census, own O-4 nullspaces, own letter-riding arithmetic), the mechanism-provenance leg (REGISTERED/MATCH), the physics/steelman leg (`anchor_physics_steelman.md`: 24 checks, 2 failed = the two pin checks that fired on the co-runner's own rewrite of its report before one logged re-baseline; 33 verdicts, 5 refuted = its own R1.3 registration, the engine's flips (D-S1), the own INV8 predicate artifact (D-S13) and the KG-4 caps-wording defect (D-S11), a REFUT that IS the finding; no physics verdict of the record refuted; its clean isolated mirror of the final engine text PARTIAL: A0 identical incl. S0's 98 pilot rows, 121/133 A1 rows leaf-identical, 12 differing in one stale non-verdict field), and the rigor leg (`anchor_rigor_verifier.md`; two instances, the first ran R0 9/9, launched the isolated single-launch engine mirror (5 workers, 3084 s, COMPLETE: 59/59 recorded verdicts reproduce with no flip; every verdict-bearing leaf of all 790 rows identical; only `runtime_s` and one stale field at 24 rows differ) and crashed in its own code at R2; the second kept every registration, fixed the comparison, ran R0 (79 pins) and R1 (10 findings, 7 refuted, all provenance/reporting), reached R2 at 294/338 classes = 678 rows with **0 row mismatches and 0 construction-identity failures**, wrote its report as a hand-off, and was killed by the steelman leg's process cleanup at 08:42 (D-S0)). **This synthesis re-launched it (`python3.12 anchor_rigor_verifier.py --stage ALL --resume --workers 4`, 08:50:33) to close the gap before ruling the ladder: the four pending GBc 20 / GBcQ 20 classes completed (3.0 / 3.3 / 40.4 / 40.5 s; logged in `anchor_rigor_verifier.log`, not checkpointed, the checkpoint fires every 10 classes), then the four workers spent 24 minutes at 100 % CPU on the first four F2c 4 (d = 16) classes without completing one, and the run was stopped by the synthesizer at 09:15 so that no process would be orphaned; the leg's state file is byte-identical to its 08:25 checkpoint (sha256 `5897398f…`; R0, R1 done; 678 rows / 290 classes), its log carries the resumed run's five lines. Typed NOT-COMPLETED-BELOW-CAPS: the 44 remaining d = 16 classes of R2 (all F2c 4 / F2cQ 4 classes, never reached by either instance) and the stages R3–R7 are OWED at G26-0.5. The likely stall, not diagnosed (no profiler available on the host), is the leg's uncapped sympy `count_roots` reference (script line 483) invoked when its own Sturm chain exceeds its 240 s cap, on degree-32 polynomials with large coefficients; the leg owner should cap it, resume (the command above), then run `anchor_rigor_verifier_extra.py E3`, and re-issue the report from its hand-off text.**

**All program verdicts CONFIRMED by every leg that reached them:** the margin theorem and its census (gate S1 98/98 masked; steelman lens 1 PASS incl. beyond-family controls and the AM-3 population 30/30 by an own route; rigor R2 own-path rows at 678 rows / 290 classes checkpointed plus the four classes logged by the resumed run, with 0 mismatches); the anchor lemma at every computed row (pilot 98; leg 2 109; gate 50 placements; steelman own nullspaces at the six survivor rows, 50 Gauss placements and 7 U-S1 placements; rigor R2 own RREF); the Z derivations (gate S5/S6; steelman ROLE2/A2 incl. the exact circulant eigenvalues and the inertia classification; rigor E2 and R2); the fold (steelman `_strata.py` reproduces every Scout C constant bit-exactly); the control (three pilot legs; gate O-4; steelman SEG/ROLE2); KG-3 clean by every route; KG-4 silent by every route incl. the steelman's extended hunt; the co-runner's 47/47 records reproduced across two launches. **The pivot ruling (§1.1) CONFIRMED by the steelman's scoop lens (every cite-not-claim line honoured; the results file makes no literature claim; the spec carries every forced rider verbatim).**

**Defect table (merged; complete; none touches a verdict, theorem or certificate value):**

| id | leg | grade | defect | disposition |
|---|---|---|---|---|
| D1 / D1a | pilot | MODERATE (process) | five resumed launches; the committed script text post-dates part of the row computation | closed: four isolated mirrors byte-identical at 98/98 rows; six cap-typed / label-only residues typed (V2-1) |
| D2 | pilot leg 1 (owner-ruled) | MODERATE (registration) | doc 26's registered Δ_x^{(R)} omits hop_xx | form of record with hop_xx; verdicts agree 225/225 with the registered form; verified by four legs |
| D3 / D4 / D5 / D7 / D8 | pilot leg 1 (owner-ruled) | LOW / LOW / LOW / INFO / LOW | AM-2 hypotheses (G_x > 0 ∀x AND m̃₂ > 0); V = 4 anchor readings leave Q not R; R_max = 2 + κ; SPLITR ≡ SPLITL; two near-corollary relabels | all closed and derived (H1, H2, D8; own Sturm 4 at the six V = 4 rows) |
| D6 / D9 / D10 | pilot leg 1 | - | leg 1's return text not on disk; not adopted | not reconstructable; the record's remaining subject matter carried by V2-6/V2-7/V2-8 |
| V2-1 … V2-8; V3-1, V3-2 | pilot legs 2, 3 | MODERATE (V2-1 provenance; V2-2 P-G3's population) … INFO | P-G3's population convention (rows vs cells); KG-1 typing; the sharpness datum; leg-2 reruns; the clause-(i) d > 20 skip; the KG-5 shorthand; leg-2 limits; U5 unexecuted by leg 1; the AM-3 scope (16 → 14 definite cells) | all closed in the spec (P-G8 per cell; DERIVED-SILENT typing; V3-2 scope) |
| M1–M7, m1–m13 | gate G26-0 | MANDATORY / MINOR | real vs rational reading sets conflated; the anchor biconditional refuted on the control side; a false step in Theorem G (iii); two "own routings" conflated; P-G10's population double-counted; P-G14 decided at freeze at 243 cells; U-S5 not unseen; thirteen minor relabels/typings | **all applied as the v1.0.1 amendment before the mechanism note**; no rung moved |
| **D-R1** | rigor (E1) | **HIGH (record integrity; verdict-neutral)** | the engine's S0 pilot driver imported `anchor_pilot` from canon `analysis/` (the isolated copy's directory nested one level too deep) and its module-level writers **overwrote `analysis/anchor_pilot_run.log` (pinned e500e046… → 1e345b22…) and `anchor_pilot_state.json` (6d52ee21… → c8a5c26d…, 'started' = the engine's 04:04:37) at 04:10**; the engine's pin lists omit both files, so its "pins identical" and "no committed file modified" statements are false for them; the pilot's verdict-bearing files (`.py`, `_rows.jsonl`, `_results.md`) carry their pinned shas; the computation was identical (same sha f1028c0d…), the isolation claim false | **G26-0.5:** fix the driver (pass the pilot directory explicitly); add both files to every pin list; restore from backup (Drive version history) or regenerate by a fresh pilot run and re-pin with disclosure; every engine mirror must exclude the two files from sha gating |
| D-R2 = D-S1 | rigor / steelman | MODERATE (provenance / disclosure) | seven engine launches with code edits between them; **three verdicts flipped REFUT → CONF on cached rows after predicate re-scopings** (H10/H11/H12 at the F2cQ SD SITE censuses, a SPLIT predicate misapplied at SITE; P-G9/H5 at 51 rows, a 20-digit vs 50-digit comparison; the P-G9 attainment pattern, the swap rows moved from "Gauss placements" to data, verdict text edited); undisclosed in `anchor_results.md`; header "runtime 3.3 s" | each re-scoping confirmed correct by own computation (rigor R1; steelman SEG own (a)–(c)) and restores the pre-code note's registered wording; **disclosed here**; the fresh single-launch mirror reproduces every verdict; results-writer disclosure at the next regeneration |
| D-R3 | rigor | LOW (process) | the reserved U-S1 cell's letter-0 curve computed at stage A1-letter0 (04:33) before A6 (05:12), IDENTICALLY-ZERO at s = 0 (a face exposed early); the mechanism note (03:02) predates it | disclosed; the hand predictions were not informed |
| D-R4 = D-S6 | rigor / steelman | LOW (engine bug) | the H9 data lane recorded as `{"error": "bad operand type for unary -: 'str'"}` and printed under "data" | typed NOT-COMPUTED in the engine; the datum supplied by own paths (rigor E2; gate S6); one-line fix at regeneration |
| D-R5 = D-S2 | rigor / steelman | LOW–MODERATE (reporting + typing) | the U-S5′ s_c = 4 knob-3 content buried (G_min = −16 knob-free; Sturm 8; the second rational classical pair 8/15, 3/8); the note's G_min = 0 a hand error against the record; the M2 sentence has an exception at a taxonomy knob | typed §3.8 as a classical-carrier role datum on a rational conic, never a role claim; fused into S-AM-SPLIT |
| D-R6 = D-S7(part) | rigor / steelman | LOW (wording) | KG-1 declared to "fire" at 12 SD SITE rows although its letter needs "R_self exactly empty", the V = 6 rows' emptiness is height-typed and V = 8's NOT-COMPUTED | the corner is DERIVED-SILENT; KG-1's letter applies at the d = 16 rows only; over-firing on the conservative side |
| D-R7 = D-S7 | rigor / steelman | INFO (cosmetic) | DEVIATION / NOTE lines printed 7–8× (once per launch); the A0-selftest ledger line ×6; "24 → 12 control rows" | results-writer fix at regeneration |
| D-R8 | rigor | LOW (reporting) | vacuous "hand prediction at every placement True" at the four F2cQ SD SITE census cells (no prediction registered there) | print only where a prediction exists |
| D-R9 | rigor | INFO | the 67 doc-14 frames regressed through the engine's own construction, not "rebuilt through synthesis_lattice" (M-L5's letter) | E3 (stacked recipe at the six survivor rows; CK column at all 67) owed with the finished rows |
| D-R10 | rigor | INFO (verifier-side) | the first instance's crash (Fraction vs mpf); the stopped second mirror | disclosed; mirror-vs-mirror NOT-COMPUTED |
| **D-S0** | steelman | MODERATE (process; another leg interrupted) | the steelman killed the rigor leg's re-launch (PID 20854 + 4 workers) at 08:42 while cleaning up its own pool workers, at R2 class 294/338 | the rigor state checkpointed (R0, R1 done; 678 rows); resumable; resumed by this synthesis, stalled on the d = 16 classes, stopped cleanly (SYN-1); OWED at G26-0.5 |
| D-S3 | steelman | MODERATE (wording; AM-5 attribution) | "own routing thins the anchor to the antipodal pair" is a lab-period-5 coincidence of the order-2 symbol; at periods 6, 7 the same kernel leaves no anchor, at 3, 4 the third regime | the period clause fused into S-AM-5 and S-AM-SPLIT; the dipole clause reads "shift a multiple of the lab period" |
| D-S4 | steelman | LOW (wording; mechanism sentence) | "the clock's self-sourced constraint energy exceeds what its hopping can compensate" misattributes the source (D ≡ 0 is NEVER-type whatever the clock terms) | fused into S-AM-2 and §4 ("the constraint energy of the configuration dressed by the clock's own charge exceeds the reading plus the hopping margin") |
| D-S5 | steelman | LOW (standing; AM-4) | "⇒ DERIVED over the order-1 F2cQ SD SPLITL family" holds at the no-split placements and V ≡ 0 (mod 4); the V ≡ 2 (mod 4) cross-block step checked at V = 6 only | AM-4 entered at ANCHOR-TYPED with the derived sub-scope in-sentence; the V = 10 test registered (U-R1) |
| **D-S8** | steelman | MODERATE (reproducibility; partially closed) | the engine text of record (mtime 05:12:25) post-dates the rows; the author's `--resume` check re-emits cached rows | the rigor leg's first-instance mirror (single fresh launch) reproduces every verdict-bearing leaf of all 790 rows and the 59-verdict ledger (closed at that grade); the steelman's own mirror PARTIAL (133 rows); the full row-by-row identity of a clean run **owed to G26-1** |
| **D-S12** | steelman | MODERATE (record consistency; machine-established) | `anchor_rows.jsonl` is not the output of the committed text at the 24 F2cQ SD SITE rows in the field `sd_structure.D_zero_is_balanced` (the launch-2 predicate's value False survives; the final text and the A3 z-census give True); no verdict quantity differs | H10 at SITE re-sourced from the A3 z-census here (§3.3); rows regeneration at G26-0.5; the closing audit's mirror must complete the row identity |
| D-S11 | steelman | LOW (caps wording) | the results' KG-4 verdict line omits the d = 256 cap (24 uncertified d = 256 gravity rows NOT-COMPUTED) | cap printed in §3.8 and §1.5; INV8's mod-p statement at s = 0, ν ∈ {1, 2} the partial closure |
| D-S9 / D-S13 | steelman (own) | INFO | R1.3 registered "Z = ∅ at every t ≠ 10", REFUTED and corrected (commensurability); the INV8 predicate does not exclude ν < 1 (the ν = 0 chiral-point artifact) | reported as refuted own predictions; logs unedited |
| D-S10 | steelman | INFO (inherited) | the rigor leg incomplete at the time of the steelman's write; the co-runner rewrote its own report (re-baselined once) | for the record |
| O2 | provenance leg | - | the S0 overwrite of the two pilot record files, flagged not judged | = D-R1 |
| SYN-1 | synthesizer | MODERATE (record completeness) | at the time of this synthesis the rigor leg's machine stages R2 (44 classes) and R3–R7 had not completed on disk; its report's "CONFIRMED-WITH-DEFECTS" is provisional on R3–R7 by its own header | the resumed run (§5) closed 4 of the 48 pending classes and stalled on the d = 16 classes; **OWED at G26-0.5**: R2's 44 d = 16 classes (with a cap on the `count_roots` reference first), R3–R7, the `_extra.py E3` frame check and the re-issued report; every rung of §1.5 is entered conditionally on it; a verdict-reversing finding there amends this document by a dated block |
| SYN-2 | synthesizer | INFO (display) | the engine's results header quotes "runtime 3.3 s" and the pilot's "12.7 s" (the last segments); the pilot's 2,130 s of row work and the engine's ~68-minute campaign are in the logs | display only |
| SYN-3 | synthesizer | INFO (wording) | the engine's typed-outcome line carries "ANCHOR-LEMMA candidate: … ⇒ DERIVED over the order-1 F2cQ SD SPLITL family", overstated by the V ≡ 2 (mod 4) clause (D-S5) | ruled §1.5: ANCHOR-TYPED with the derived sub-scope |

**Ladder recommendations of record:** steelman, AM-2 LICENSED at its bar with D-S4; AM-3 LICENSED in substance (rigor items then owed); AM-4 ANCHOR-TYPED with the ⇒ direction derived except at the V ≡ 2 (mod 4) split placements; AM-5 DERIVED at both orders, LICENSED only with the period and commensurability clauses; AM-SPLIT STANDS sharpened; AM-INV NOTHING; quotation BLOCKED until G26-1 + G26-2. Rigor (at hand-off, provisional on R3–R7), AM-2 / AM-3 enterable once the steelman lands, with (iii′), the SD SITE exception derived and "ν̄ a sufficient bound"; AM-4 ANCHOR-LEMMA "supportable in SCOPE"; AM-5 DERIVED at both orders with the registered "Fourier parity" wording replaced by the combinatorial mechanism and (iii) replaced by F ⊗ F; AM-SPLIT STANDS as a complete dichotomy on the 273-cell taxonomy; AM-INV nothing; the dream's third clause computed only; G26-4 none. **§1.5 is the merged ruling** (the stricter of the two on AM-4; the steelman's period and attribution clauses fused everywhere).


**[v1.1, 2026-09-23, dispositions executed; none verdict-touching.]** All owed items in the §5 table LANDED at G26-0.5 (generator-side; outputs regenerated, never hand-edited; the verdict-row identity proof and the regression record in `analysis/g26_0_5_repair_report.md`; the rigor report §9 carries the same dispositions from the verifier's side; the closing audit §4–§6 re-established D-R1 and D-S12 from independent evidence and confirmed their repairs):

| id | disposition (dated) |
|---|---|
| **D-R1** (HIGH, record integrity) | **LANDED (a), 2026-09-22**, explicit `--pilot-dir` S0 driver; the overwrite reproduced in isolation on the v1.0 text (`before_tree`: `1e345b22b32a5c6f…` → `ab69eaa75c3319ae…`, `c8a5c26dc1bb4fb2…` → `78db7449368b7427…`) and absent on the repaired text (`after_tree`; mirror3's new check `[PASS] S0 (G26-0.5 D-R1 guard)`); pin lists extended (engine 62 → 64, steelman 59 → 61, co-runner 42 → 44, rigor 79 → 81); the two files regenerated by ONE pilot launch (98/98 verdict-row identity; differing leaves cap-typed / runtime only; the regenerated rows / results preserved as `g26_0_5_evidence/pilot_regen2/` `78b612e5c8415be4…` / `270f83b215afaef4…`) and installed 21:09:21 with disclosure in `anchor_results.md` §0.2; the originals unrecoverable (E1 stays REFUT as history); every engine mirror pins all 64 files. **CAUGHT INDEPENDENTLY by G26-1 (C5, 2026-09-23 04:11:10):** four pre-engine leg states pin the ORIGINAL shas (`e500e046…` / `6d52ee21…`) while the rigor leg's on-disk record carries the OVERWRITTEN ones (`1e345b22…` / `c8a5c26d…`); the mechanism reproduced from the v1.0 text in an isolated tree (the subprocess re-derives PILOT_SCR from the inherited ANCHOR_SCRATCH, nests the pilot directory one level too deep, falls through to the script directory and truncates its run log); the repair confirmed (verdict-bearing pilot files never touched; the regenerated state 98/98 verdict-identical under the record's own pilot-compare gate, re-read from the committed engine text, A-4) |
| D-R2 = D-S1 (MODERATE, provenance) | **LANDED (c), 2026-09-22**, `anchor_results.md` §0.1 lists the 7 launches and the 3 flipped verdicts with their per-launch states (parsed from `anchor_run.log`); the header's "runtime 3.3 s" replaced (SYN-2 closed). Residues typed, reporting only: (i) **F-1**, the committed text emits 60 VERDICT lines in one launch, the record's ledger holds 59; the missing `KG-4 check on the C2 components: no rational point with ν* ≥ 1 found on any factor below the height caps` is [CONF] with identical detail in the record's own run log (launch 2, line 402) and in every single-launch mirror (launch 2 crashed in A6; the resumed launches never re-emitted the cached A5 stage); a TABLE-only regeneration cannot restore it; **G26-1's C4 carries it explicitly** (the single-launch mirror's one extra verdict = exactly F-1, state True, flips nothing, A-3); (ii) six verdict DETAIL texts differ between the multi-launch record and a single launch (P-G10 "790 rows" vs "669 rows"; the anchor census "18 of 273" vs "17 of 272"; P-G15(c) "647 computed, 143 typed" vs "606 / 63"; KG-1 row order; U-S5′ tuple vs list; P-G7(c) 18 vs 17), the record's A3 was re-done after A6, a single launch's A3 precedes A6; the same CONF / REFUT state at every one |
| D-R3 (LOW) | disclosed; no repair possible; the hand predictions were not informed |
| D-R4 = D-S6 (LOW) | **LANDED (d), 2026-09-22**, `A3.H9` computed at regeneration: letter −1/5 carried on {M̃₁ = 0} with dim 1; primary readings irrational (disc 21/5); secondary readings {1/20, −1/25}; the crash dict kept as `record_value_before_regen`; equal to the rigor leg's E2 datum and the gate's S6; no verdict reads it |
| D-R5 = D-S2 (LOW–MODERATE) | typed at doc-27 grade (§3.8, S-AM-SPLIT); **extended** by the V = 8 pair 24/55, 7/18 and by U-R2 (§1.3 above); INFO residue: the results' §0 still does not surface the s_c = 4 pair (next regeneration) |
| D-R6 = D-S7(part) (LOW) | **LANDED (e), 2026-09-22**, KG-1's fire set = the 6 exact-empty d = 16 rows, with the 6 height-typed d = 64 rows and the 6 NOT-COMPUTED d = 256 rows carried as typed fields (`TABLE.KG["KG-1"]`: `fires_at` 6, `height_typed_rows` 6, `not_computed_rows` 6; union = the record's 12); verdict state unchanged |
| D-R7 = D-S7 (INFO) | **LANDED (e), 2026-09-22**, deviations 21 → 3, notes 7 → 1, the A0-selftest ledger line once (86 raw → 81 distinct) |
| D-R8 (LOW) | **LANDED (e), 2026-09-22**, "hand prediction: NONE REGISTERED at this cell (not evaluated; D-R8)" at the four F2cQ SD SITE census cells |
| D-R9 (INFO) | **CLOSED by E3, 2026-09-22 21:50:26** (`_extra_state.json` `E3_frames`: n 67, missing [], mismatch []): the 67 doc-14 frames' committed dim CK = the own rows' at 67/67; the stacked [H; diag M₁] recipe at the six survivor rows (dim 1; supports {5,6,9,10} / {3,5,10,12} / {7,11,21,25,38,42,52,56}), M-L5's "rebuilt through synthesis_lattice" letter met |
| D-R10 (INFO, verifier-side) | disclosed; **mirror-vs-mirror determinism stays NOT-COMPUTED** (instance 1's mirror deleted with the scratchpad; mirror2 stopped), G26-1's C4 re-derived the row-by-row identity of the ONE clean launch of record (mirror3) from the mirror tree without reading `mirror3_compare.json` (53 rows byte-identical, 737 `runtime_s`-only, 0 substantive, D-S8 discharged at the grade the §7.4 row specifies) and launched no second mirror; a second fresh launch is a residue of the next generator regeneration, gated by nothing |
| **D-R11** (INFO, verifier-side; corrects this document) | the R2 stall of 2026-09-05 was the SYMBOLIC Berkowitz determinant of `F_poly_own` in the symbol q (> 280 s CPU on one d = 16 row), **not** the uncapped sympy `count_roots` named "likely" in §5 and §1.6 (viii), measured by the rigor leg's third instance (probe logs in `scratchpad/rigor3/`); the route replaced by the companion-pencil Berkowitz over Q (the same exact polynomial, validated at 7 record rows incl. every d = 16 cell shape); §5's attribution and §7.4's "(a cap on its sympy `count_roots` reference first)" are superseded by this line wherever quoted (erratum rider, §7.7.4) |
| **D-R12** (INFO, engine reporting) | the engine's letter-0 `deg` field is the nominal d at the four F2c SD SPLIT classes whose curve has degree d − 4 (V = 4: deg 12, factors [1×6, 2, 2]; V = 6: deg 60); its own `factor_degrees` sum to the true degree; own curves agree at 88/88 on factor degrees and rational roots, no positive rational root anywhere (P-G14 unaffected; no verdict reads `deg`); field corrected at the next regeneration (rigor ledger check index 41) |
| **D-R13** (INFO, verifier-side) | the rigor leg's own bivariate route at C2 s_c = 8 derived a boundary (0.013486) from the quadratic-in-q factor only, the genus-0 conic's fold (= the stratum ν_c datum), while the record's ν_max 0.02813396980546884 sits on the degree-(8, 4) factor; typed NOT-COMPUTED by that route; the engine's ν_max CONFIRMED by the own Sturm bracket [1685/59898, 1082/38455] (check index 42) **and by G26-1's C3** (0 real roots at ν_hi = 1082/38455 by a disjoint-node route; a sign change at ν_lo = 1685/59898; every factor-discriminant boundary root certified by a sign change of its own minimal polynomial, 4/4 at s_c = 8 and 5/5 at s_c = 6) |
| **D-R14** (INFO, wording + a NEW datum) | the engine's verdict "H7 / H8 … at every placement of every census cell" covers its 22 (cell, order) censuses, F2cQ 8 SD SPLITL under the order-2 (swap) kernel is excluded by code, invisibly in the text; there the reflection-complement symmetry maps Z(s) → Z(2 − s), not 3 − s; H7 holds; the swap Z = the 6 antipodally periodic members + 2 (or 4 at s ≡ 1 mod 5) accidental zero-sum members (\|Z\| = 8 / 10); dim N_Z = dim CK = 0 at all 40 placements (unchanged; SG-1's kill stands; G26-1's C2 swap contrast: dim CK = 0 at every tested face placement by its own route); no rung reads H8 under the order-2 kernel; wording "22 censuses" at the next regeneration (check index 47) |
| D-S0 (MODERATE, process) | **CLOSED**, the interrupted leg resumed (2026-09-22 20:13:56, adopted by the owner after its launcher died on a usage limit, `g26_drive.log`) and completed 20:43:59, rc 0 |
| D-S3 / D-S4 / D-S5 (wording) | unchanged, fused in §1.5 (the period-5 and commensurability clauses; the D-S4 attribution; AM-4's derived sub-scope, now with the V = 10 instance added to the machine-checked set and no derivation) |
| **D-S8** (MODERATE, reproducibility) | **DISCHARGED at the §7.4 grade**, the row-by-row identity of ONE clean launch of the final text against the REGENERATED canon is on the record by two independent comparisons: rigor R7 (2026-09-22 21:50:25; 790/790 verdict-bearing leaves identical, 59/59 verdicts no flip) and **G26-1 C4 (2026-09-23 04:10:55; re-derived from the mirror tree: same 790-row population in the same order, 53 byte-identical, 737 `runtime_s`-only, 0 substantive; every record verdict reproduced with the same state, the one extra = F-1)**; the mirror-vs-mirror residue stays D-R10's |
| **D-S12** (MODERATE, record consistency) | **LANDED (b), 2026-09-22**, the 24 stale `D_zero_is_balanced` leaves regenerated from the committed text (the ONLY leaves that changed; `G26_0_5.D_S12_rows_changed` lists the 24 keys; verified by four independent diffs: the finalizer's, the rigor leg's, the engine's own identity gate, and **G26-1's C5 (2026-09-23): exactly 24 leaves differ from the `.pre_g26_0_5` backup, all that field, all False → True, 766/790 lines byte-identical, row order unchanged; the audit's own third-band recomputation returns True at all 24**, at these order-1 SITE cells D_x = V − S_x²/NS > 0 at every configuration, so the zero-energy set is empty and the clause holds vacuously) |
| D-S11 (LOW, caps wording) | **LANDED (e), 2026-09-22**, `TABLE.KG.KG-4.caps` (the 24 uncertified d = 256 rows) printed with the KG-4 line |
| D-S9 / D-S13 / D-S10 | for the record, unchanged |
| O2 | = D-R1, landed |
| **SYN-1** (MODERATE, record completeness) | **DISCHARGED, 2026-09-22**, R2's 44 d = 16 classes (+ the 4 GBc 20 / GBcQ 20 classes: 48 classes / 112 rows by the amended route, agreeing with the record on every leaf) and R3–R7 complete on disk (R2 20:14:45, R3 20:25:46, R4 20:26:26, R5 20:27:56, R6 20:27:57, R7 21:50:25), E3 done, the report re-issued from the hand-off text |
| SYN-2 (INFO) | closed at regeneration (the header now carries the campaign start, the last computation stage and the regeneration segment) |
| SYN-3 (INFO) | ruled §1.5, unchanged (the engine's `TABLE.outcome` string of record still carries "ANCHOR-LEMMA candidate: … ⇒ DERIVED over the order-1 F2cQ SD SPLITL family", data, verbatim, identical in the record, the regeneration and the mirror; the ruling overrides it: ANCHOR-TYPED with the derived sub-scope) |

**The closing audit's own findings (G26-1; `anchor_closing_audit.md` §4 / §6; none a record defect beyond the two above):** **A-1** (AUDIT-DESIGN, repaired), the first C2 pass built 0 K₄ edge-function products because it kept only anchor configurations with ONE up-site per half, which matches 0 of the 36 members of Z (every member carries TWO up-sites per half and IS an edge of K₄); rebuilt with an own edge labelling: the four products are supported on Z, dipole-blind, in ker(hop\|_Z) and ker H(ν), rank 4, span CK exactly, S-AM-4's "dim CK = 4" and S-AM-5's "F_{V/2} ⊗ F_{V/2} … spanned by products of K₄ edge-functions" are REPRODUCIBLE under a registered labelling, no wording change; **A-2** (AUDIT-DESIGN, repaired; ×3 at C1, C2, C2′), the first C3 pass demanded the own Gram-floor LDL certificate be PD at ν_hi, which exceeds the record's claim ("ν̄ is a sufficient bound, not the boundary"; ν_hi sits STRICTLY below ν̄ at every theorem row, sharpness 0.9682601194820663 / 0.976009494735019 / 0.9499069801622525 / 0.923665730873715); repaired to the audit's own real-root count at ν_hi (0 real roots at all four targets) and supplemented by the positive check that the certificate DOES hold at a rational immediately above the own ν̄, at ν = 1, at ν = 2 and by a PD Gram floor at the registered knob; **A-3** (AUDIT-DESIGN, repaired), 60 mirror verdicts vs 59 record verdicts = exactly F-1; re-scored as a name → state comparison, none missing, none flipped; **A-4** (AUDIT-DESIGN, repaired), 92/98 under a runtime-only exclusion → 98/98 under the record's own pilot-compare gate (the cap-typed leaves: the time-capped ν_max bracket endpoints at the four theorem rows, the record's bracket lies INSIDE the regenerated one at all four, and the d = 64 Sturm label / q-grid depth at two rows, both typed absences). **Findings of this adjudication on the audit (INFO; verdict-neutral; for the record):** **G3-1**, a fifth check re-specification is in the audit's `superseded` list but not in its §4 table: at 04:12:41 the C7 self-scan of the audit's own report flagged two hits (the two banned tokens, present in its own quoted ban list), the check was re-specified to exclude quoted mentions ("a slogan named in quotes is a mention, not an assertion", the doc 27 §1.5 convention, stated in its §7) and re-run at 04:13:42 → PASS (`report_ban_scan`: 39,269 chars, hits []); the report file was written three times (04:12:42 "1 failures, verdict FAIL"; 04:13:43 and 04:14:19 "0 failures, verdict PASS"), the last being the report of record; the superseded FAIL entry is kept (7 FAIL entries in `superseded`, the report's §4 counts "six" for the C1–C5 pass), a mention-vs-assertion re-specification of a self-referential check, no record content involved; **G3-2**, the audit script's pre-repair revisions are not preserved as files (only the final text `7aaafa863e096a55…`, mtime 04:14, is on disk); the superseded check entries and the disclosures are, so every repair is auditable from the state, not from a diff, disclosure only, no action.

**Other audit findings of the chain (none verdict-touching; for the record):** the repair report's **F-2** (the engine's pin tables of record froze `27_anchor_margin.md` `9782694a3b5049af…` and `README.md` `f8361df6a3342f0b…` at their 04:04 drafts, both finalized by the owner later on 2026-09-05, `eb97522a2e93e50d…` / `9d7ae323cec05273…`; every table self-consistent; one sentence in §6), **F-3** (the repair harness derived the scratch directory as `<tree>_scratch` while the rigor launcher used `mirror3_scratch`, the driver's `pilot_outputs_in_isolated_dir: []` open item and the `tree_pins.json` stop, both resolved; harness left as is), **F-4** (a first `pilot_regen` launch cut off at ~19:42, never compared or installed; `pilot_regen2` is the regeneration of record), **F-5** (the driver's final `pins` call never ran because of the step-6 stop; the finalizer produced `after_shas.txt`), **F-6** (= item (f), discharged). The G26-2 pre-meeting leg reports NO defect of the record; its §11 re-verification corrects three round figures of its own §9 ("60 / 55 / 78" → 56 / 54 / 77, no lane affected) and adds one method fact; its §13 corrects its own processes bullet in-pass (a sibling G26-1 process was running at its exit, not launched, not touched, not waited for by it; the D-S0 lesson honoured); its two re-wording candidates are applied as rider wording in §7.7.4 (Lock / Semião / Cepollaro; the CVI clause). **The chain of G26-3 drafts:** 22:59 (`80adc7e0668353d2…`) → 23:25 (`1cdea3ce39c7ef17…`, STILL-BLOCKED, G26-1 not executed) → this file (CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING, G26-1 PASS); each supersession disclosed, the superseded texts preserved.


**[v1.2 pointer 5, 2026-09-24.]** **§5.** D-S5 row: "CLOSED, MIS-SCOPED (v1.2)". SYN-3: the same. Add V-1 (the ν² slip) and V-2 (the V ≡ 2 (mod 4) dim-1 sentence) as INFO erratum rows on the mechanism note, verdict-neutral.


---

## 6. Ledger

- **Bands and primes (every base certified by sympy `isprime` AND trial division at every leg start; route-disjoint; consumed ledger of docs 08–26 respected):** PILOT-G **8,200,007 / 8,200,013** (mod-p root-count pins; the B5 nonzero-determinant certificate); pilot-verifier legs 1–3 **8,300,009 / 8,300,041** (rigor band; the second never drawn upon at 272/272); reconciliation, spec and provenance legs none; gate G26-0 **8,300,057**; engine **8,200,037 / 8,200,061** (spares 8,200,079 / 8,200,091 certified, not drawn); co-runner **8,200,109 / 8,200,141**; rigor leg **8,300,069 / 8,300,077**; steelman **8,400,011 / 8,400,019 / 8,400,031** (spare 8,400,043 certified; the small-prime sieve 1009–1213 for INV4's rigorous zero-root implication); **closing-audit reserve ≥ 8,500,007 UNTOUCHED** (8,500,021 / 8,500,027 / 8,500,057 certified; spare 8,500,061); **successor band ≥ 8,600,023 fallow** (8,600,029 / 8,600,041 certified); every earlier reserve (≥ 6,600,001; ≥ 7,300,001; ≥ 7,700,071; ≥ 8,100,073) UNTOUCHED; the doc-25 bands (7,800,041+ / 8,000,009+) fallow; composites of record never drawn (8,200,003; 8,300,003; 8,400,001; 8,500,003; 7,700,003; 8,100,007). Committed-engine internals (SG 1000403 / 1000409; synthesis 2800001 / 2800003) exercised only inside the committed constructions. Nothing was decided by a prime anywhere in this program.
- **Caps (spec §6.1, honoured and printed):** F2c V ∈ {4, 6, 8}, GBc NS ∈ {12, 16, 20}; the exactness ladder by d (d ≤ 18 exact; d = 64 F(q) mod-p + CRT below height 5,798,310 with the Gram-floor PD at certified rows and Sturm NOT-COMPUTED; d = 256 F(q) and ν_max NOT-COMPUTED); ν̄ to 50 digits with exact brackets; the locus engines at their inherited caps (genus of the (8,4)/(6,3) components NOT-COMPUTED-BELOW-CAPS; rational points to height 64 / 24); machine budget ≤ 12 engine-hours (the campaign ~68 minutes); every unbounded loop capped with 60 s progress logging; the C1 exact discriminant NOT-COMPUTED-BELOW-CAP (brackets certified); the steelman's degree-128 exact Sturm alarmed at 900 s (typed); Ostrowski/Brauer not implemented.
- **Canon pins:** 41 (pilot) / 46 (spec) / 62 (engine) / 59 (steelman) / 79 (rigor R0) / 42 (co-runner) files byte-identical before and after every leg; the co-runner's CONFIRM pass agrees with every pin table of every leg on every file both pinned (28 tables, 0 canon mismatches); **no committed canon of docs 01–26 written by any leg.** Record files of THIS program changed after their pinning: `anchor_pilot_run.log` and `anchor_pilot_state.json` (D-R1, by the engine's S0 at 04:10, the only record-integrity breach); `anchor_corunner.md` (by the co-runner's own second launch, re-baselined once by the steelman); `anchor_pilot_verifier.md` (by its own leg 3 append, disclosed); `anchor_rigor_verifier.log` (five lines appended by the resumed rigor run, §5; its `_state.json` unchanged, `5897398f…`). **No supersession of any prior canon this program.** New canon created: `analysis/anchor_*` (pilot 5 + zcensus 2 + verifier 12 + reconcile 4 + spec 4 + gate 5 + mechanism 2 + engine 5 + rigor 7 + steelman 16 + co-runner 8), `papers/24_anchor_margin/` (2 scout reports, 8 PDFs + manifest, the Perimeter export, scout scripts + state), and this document.
- **Dated deviations / amendments:** the spec's v1.0.1 additive amendment (Appendix D; 2026-09-05); the gate's report filename `anchor_gate_audit.*` vs the spec's `anchor_hostile_gate.*`; the mechanism note authored by the engine author (the mechanism-author leg had not run) and registered pre-code; the engine's seven launches with three verdict flips (D-S1) and one stale non-verdict field (D-S12); the engine's S0 overwrite (D-R1); the rigor leg's two instances and the steelman's interruption of the second (D-S0); the GBc SPLITR spot (0,2,0) computed in addition (273 gravity cells); the reservation honoured at the class level for the ω families of U-S2; the H9 lane crash (typed NOT-COMPUTED); the results-writer duplications.
- **Files of record (sha256 prefixes):** doc 26 `e7154d99…`; pilot `f1028c0d…` / results `eee3fc01…` / rows `5d6df83c…` (98) / run.log now `1e345b22…` (pinned e500e046…, D-R1) / state now `c8a5c26d…` (pinned 6d52ee21…, D-R1); pilot verifier `84f25167…` (+ leg-1 `.py` 7d4291f4…, leg-2 `.py` d106cb48…, leg-3 `.py` 7bd430ec… / `.log` a63de9c8… / `_state.json` add4f60d…); reconcile `f16da830…`; spec v1.0.1 `066269d5…` (stamp d39954e2…; v1.0 8ace63b2… / ba62706f…); gate `bbc15d0b…`; mechanism `5b568dd4…`; provenance `882ca86d…`; engine `f06bb935…` / results `c4221b4a…` / rows `1529452f…` (790) / state / run.log; rigor `.md` `90cbd321…` (hand-off text) / `.py` 2c29031f… / `_state.json` `5897398f…` (unchanged by the resumed run) / `.log` `e776e4e6…` (five lines appended by the resumed run) / `_extra.*`; steelman `.md` `ae824c59…` / `.py` / `_state.json` / `_rows.jsonl` / `_mirror.*` / the five side scripts + JSONs; co-runner `.md` `babec814…` / `_state.json` / four scripts; scouts `15ba4ed7…` / `87cd01e1…`; papers/24 README `175fd0e4…` (the pilot's lanes note); this document (v1.0). Full digests in each leg's state file.
- **Process:** every background run of every leg waited for or handed off with a resumable state (the one interrupted leg disclosed, D-S0, and resumed by this synthesis, §5); watchdogs on every unbounded loop; 600000 ms timeouts honoured; python3.12 only; no README / memory / SUBMISSION_STATUS / paper edit by any leg (the owner folds).


- **[v1.1] Files of record superseded (G26-0.5 regeneration + the rigor leg's completion + G26-1 + the G26-2 pre-meeting leg, 2026-09-22 / 2026-09-23):** as §1.1 above, engine `b7f918919d188066…` (v1.0 `f06bb93562b3dd27…` preserved as `anchor_lattice_v1.0.py`); results `0101d5fa0d7cd133…`; rows `c07555e117321d33…` (790); state `8fee0d0bb9f44d4d…`; run.log `8ef14cdd9181eb15…` (append-only); the pre-regeneration quartet preserved as `*.pre_g26_0_5` (= the v1.0 pins `c4221b4af6e44318…` / `1529452ffcda4536…` / `ef84491f0f3ec1c4…` / `31cba877e82b61de…`, re-hashed by this leg, §B); pilot run.log `1b40ebf63cf7f289…` / state `0347c35e8208bfca…` (regenerated, D-R1; the "run.log now 1e345b22b32a5c6f… / state now c8a5c26dc1bb4fb2…" clause of the v1.0 line is discharged); rigor `.md` `02c2162002589224…` / `.py` `77234bdbd8142dbb…` / `_state.json` `1d5426b660ea785a…` / `.log` `087c77fbda42f3ce…` / `_rows.jsonl` `6018701ce8e8c398…` / `_extra.py` `1e549bc455874bc5…` / `_extra.log` `b4fedcb5a81aabd1…` / `_extra_state.json` `d5c638c21ddcb978…` (the v1.0 line "rigor `_state.json` unchanged `5897398f…`" superseded); **closing audit `.md` `7cd12fba2dbb15da…` / `.py` `7aaafa863e096a55…` / `.log` `c2a0fd1f1342538d…` / `_state.json` `107f9e6a00401b68…` / `_rows.jsonl` `bc9d8784699b640c…`**; steelman `.py` `df0e202cb980bccf…`; `anchor_corunner_common.py` `380ade840bbe90ac…`. New records this gate cycle: `g26_0_5_repair_report.md` `278c0b1ed963e310…`; `g26_0_5_repair.py` `5e7637251ab78477…` (harness; read-only comparison tool); `g26_drive.sh` `aaa053511714f473…` / `g26_drive.log` `8f4c1135d1c4f3b5…`; `g26_0_5_evidence/` (56 files, 9.8 MB, `MANIFEST.sha256` `9e7688a75a389a1a…` without self-hash); `papers/24_anchor_margin/G26_2_PRE_MEETING_SWEEP.md` `2534dd29d38c20f1…` (§11 appended 23:09, §12 23:32, §13 2026-09-23 03:48; the pre-addendum text `c78f850fca78ffef…`, the §11 text `82950060ccb4e34b…`); `2512.14834.pdf` `3b026d345a5552ad…` (614,428 B) and `2609.09931.pdf` `9b7d970d686f2487…` (371,998 B), title-verified; `PDF_MANIFEST.txt` `c4bc2575f383b2cc…` → `00380fd47bfdd087…`; `scout_scripts/g26_2/` (the two scripts with logs and states `d6bce17b6eeafef4…` / `a462057d90aa8f24…`, the confirm-nets JSON / log, the citer diff `184c83bda37c5f1a…`, the Indico export `e86b52588d52560e…` and timetable `52344cb8b9f13b42…`, the PIRSA page `8df9e4efba61ba7f…`, `g26_2_state.json` `02df38076785b797…`, `g26_2_reverify_2300.json` `ee4c487282a60986…`, `g26_2_reverify_2330.json` `5d26791457fcdea5…`, `g26_2_reverify_0330.json` `0310a57fcb829eef…`); `g26_3_amendment_draft.md` (this ruling, applied as this v1.1; the 22:59 draft `80adc7e0668353d2…` and the 23:25 draft `1cdea3ce39c7ef17…` superseded, the latter preserved as `analysis/g26_3_amendment_draft_v2_2026-09-22_2325.md`). `papers/manifest.txt` NOT appended (pin `a0a621107229ba86…` unchanged; the owner may mirror the two lines).
- **[v1.1] Prime ledger, the closing-audit reserve CONSUMED by G26-1; a new reserve DESIGNATED.** Drawn from the reserve ≥ 8,500,007 by the closing audit (2026-09-22 23:46:37, S0): **8,500,007 / 8,500,021 / 8,500,027**, mod-p rank and root tokens only (the own strata scan's mod-p nullity at two reserve primes; the d = 256 mod-p rank certificates); **nothing was decided by a prime**; each certified prime by sympy `isprime` AND trial division at draw, and re-certified by this leg (trial division AND sympy; not a draw). Certified, not drawn: **8,500,057 / 8,500,061** (spares). **NEXT untouched audit reserve of THIS program, DESIGNATED by the audit: ≥ 8,550,001, 8,550,001 / 8,550,007 / 8,550,011 / 8,550,023** (above every draw of this chain, strictly below the fallow successor band; each certified both ways at designation and re-certified by this leg), the designated consumer of any further audit of this program (a re-audit after the next generator regeneration; never a verifier band). **The successor band ≥ 8,600,023 (8,600,029 / 8,600,041 certified) stays reserved fallow for the NEXT program and is never this chain's.** No other leg of this chain drew a prime: the regenerations re-drew the committed tokens inside isolated trees only (pilot 8,200,007 / 8,200,013; engine 8,200,037 / 8,200,061; spares 8,200,079 / 8,200,091 certified, not drawn); the rigor leg's tokens 8,300,069 / 8,300,077 re-certified at the 20:13:56 and 21:50:25 resumes, nothing further drawn (R7 draws none); the sweep, its three re-verification passes and the three invocations of this adjudication drew nothing. Every earlier reserve (≥ 6,600,001; ≥ 7,300,001; ≥ 7,700,071; ≥ 8,100,073) and the doc-25 bands untouched; composites of record never drawn (8,500,003 re-checked composite by this leg).
- **[v1.1] Canon pins:** 76 (repair baseline 19:13) / 64 (engine `G26-0.5-before` = `-after`) / 81 (rigor `after` = the 21:50 resume) / 64 + 24 (audit `before` = `after`, `record_before` = `record_after`) / 34 + 16 (sweep and each re-verification pass) / **98 distinct across twelve tables (this ruling, before and after)** files byte-identical before and after every leg of the chain; the writes are exactly the disclosed regenerations (§1.1), the rigor leg's own files, the closing audit's own five files, the two PDFs + the manifest append, the sweep's §11–§13 addenda + its three re-verification JSONs, and the report / draft files; **no committed canon of docs 01–26 written by any leg; no decision document, README, SUBMISSION_STATUS, memory or paper edited (the owner applies this v1.1).** F-2 recorded: the engine's record pin tables froze the 04:04 drafts of doc 27 and the README; each table self-consistent inside its run. The rigor leg's `drift_since_R0` (its state) lists exactly the sanctioned drift since its 2026-09-05 R0: doc 27 draft → v1.0, README, the engine text, the four regenerated record files, the two installed pilot files. The closing audit's C8: all 64 pinned files unchanged by the audit and equal to their recorded sha256; no record file of this program written by it.
- **[v1.1] Dated deviations / amendments (joining the list):** the G26-0.5 engine amendment (2026-09-22, items (a)–(e), 53 marker lines; two dry runs (revisions `f9265e84db13ba8d…`, `368593f9819a79fd…`) + before / after A0 batteries + mirror3); the rigor verifier's third-instance amendment (`2c29031f…` → `615bc64512309a83…`: the F(q) route, capped references, checkpoints per class, the mirror wait) and fourth-instance amendment (`615bc64512309a83…` → `77234bdbd8142dbb…`: the 81-file pin list, the record-driven RE-BASELINED typing, list-order-insensitive KG comparison, `R7_history`), both verifier-side, dated in the docstring, regression-tested; the steelman's (`034363b8043579dd…` → `df0e202cb980bccf…`) and the co-runner's (`fd91cafd9fa2ea27…` → `380ade840bbe90ac…`) pin-list amendments (dated comments); the owner-synthesized `mirror3_tree/tree_pins.json` (build-time shas; `__note`); the abandoned first `pilot_regen` launch (F-4); the PDF manifest append; the Indico export change (86 → 79) as a literature datum; the sweep's three re-verification passes (verified and did not re-run); **the closing audit's four in-pass audit-design repairs A-1 … A-4 and the C7 self-scan re-specification (G3-1), each re-run in full with the superseded entries kept verbatim in its state; its C2 re-run (`--rerun C2`, 03:42–03:51) after the A-1 labelling repair; its report written three times, the 04:14:19 text the report of record**; this v1.1 (its draft issued three times, each superseding the last, the superseded texts preserved).

---

## 7. Consequences

### 7.1 Pointer-row ruling (G26-4): **NONE ENTERED, the default holds; candidate rows drafted**

The material-change rule, quoted from the house precedents that govern it: *"adding any pointer row would be a fresh registration decision, not an entitlement of this run"* (doc 15 §6.1 + v1.1); *"No committed law in docs 09/15/16 changed status; the pointer-row trigger did not fire"* (doc 17 §7.1); docs 19 / 21 / 23 / 24 / 25 §7.1 (NONE ENTERED on the same rule); and the spec's own G26-4 clause: *"default NONE ENTERED … The one condition that could change doc 13's status: AM-3 ENTERED at theorem grade after G26-1 + G26-2, … entered as a pointer row into §4, never as an amendment of the §3.3 sentence."* AM-3 is ENTERED by this document but its quotation is BLOCKED behind G26-0.5 + G26-1 + G26-2 (§1.5), so the spec's condition is not yet met in full; the ruling is explicit per target:

- **Doc 13 §3.3 (the licensed causal constraint-role sentence with riders (i)–(v) + (a)–(e)), NO ROW, and no landed rung of this program changes its status.** A landed AM-3 does NOT touch it: AM-3 derives the EMPTINESS at the definite-form cells at their knobs (a fourth certificate route delivering what the SG rows computed) and says nothing about the constraint-role clause, which stays licensed at computed strength with its riders, G26-0(b) forbids the margin theorem from carrying any of its content. A landed AM-5 does NOT amend it either: it gives the placement rider (b) a mechanism (the residue pattern of the dipole-blind hyperplane and the face kernels F ⊗ F) and gives rider (c)'s secondary-normalization readings a name (letter-anchored readings, H9), rider CONTENT candidates, drafted below, never sentence changes.
- **Doc 13 §4 (the mechanism paragraph: "class-wide at m1 grade, a real no-go would need the p-adic/Newton-polygon route at slaving-denominator primes"), NO ROW NOW; the ONE candidate that would change a doc-13 status.** With AM-3 entered, a fourth route exists and delivers at the committed theorem cells and at every definite-form gravity cell outside the SD SITE sub-family: the emptiness there is DERIVED at the registered knob, a material change to the status of that paragraph's "a real no-go would need …" clause (it needs no p-adic route at those cells; it still says nothing at the anchor cells, where the emptiness stays computed, and it is not a no-go of any shape). Candidate row drafted, NOT entered, **a fresh registration decision at or after G26-1 + G26-2, when AM-3's quotation unblocks:** *"[anchor-margin run, doc 27 §1.5 S-AM-3] at C1, C2, the 14 definite-form CANDIDATE-GRAV cells and every definite-form gravity cell outside the SD SITE sub-family (246/246 per cell at the registered knobs) the self-reading emptiness in both carriers is DERIVED by the role-blind margin certificate (Lévy–Desplanques on the reading operator; ν̄(cell) an explicit quadratic irrational, exact brackets ν̄(C1) ≤ 1, ν̄(C2) ≤ 2); the SD SITE sub-family is excepted below 5/2 by a derived bound; the anchor cells stay computed."*
- **Doc 13 §3.3 riders (b) and (c), NO ROW; rider-content candidates drafted, NOT entered:** *(b) "[doc 27 §3.3] the size-dependent placement rider is the residue pattern s mod 5 of the order-1 dipole-blind hyperplane against the balanced zero-energy set: faces at s ≢ 1, 2 (mod 5) for V ≡ 0 (mod 4), at s ≡ 1, 2 (mod 5) for V ≡ 2 (mod 4); at V = 8 the face carries dim CK = 4 (F_4 ⊗ F_4)";* *(c) "[doc 27 §3.3, H9] the secondary-normalization readings {−1/25, 0, 1/20} are letter-anchored readings q − σ²m₂q² = −1/5 carried on {M̃₁ = 0} outside CK; the anchor lemma is the λ = 0 instance."*
- **Doc 14 §2.3 (the bridge survivor ⟺ dim CK ≥ 1), NO ROW.** The support structure CK = ker(hop|_Z) holds at 790 + 109 + 98 rows and at every placement of V = 4, 6, 8, with the ⇒ direction derived on a sub-scope, corroboration with a mechanism, not amendment; the spec's own condition for a doc-14 row ("only if AM-4 reaches ANCHOR-LEMMA") is not met. Candidate row drafted, NOT entered: *"[doc 27 §3.3] the bridge's kernel is the space of hop-null vectors on the anchor sublattice, CK = ker(hop|_Z) at every computed row of docs 13/14/25 and this program; dim CK = 4 at the V = 8 face placements (a new anchor structure); the two-clock joint reading at the uniform configurations is exact."* Doc 14's tombstone (§6.3) untouched.
- **Doc 19 §3.5 item 5 / §7.6 (the classical shadow; the census-strengthening rider), NO ROW.** Doc 19's sentences are GBc-scoped and remain true; the 18/18 negative-discriminant census at C2's placements is now a corollary of route G at ν = 2 there (the classical quadratic's discriminant is Δ^{(R)} − 4νm̃₂R_x ≤ Δ^{(R)} < 0), and doc 25's S-SR-D1 scope candidate is re-carried with the conic datum (§3.8), rider-content candidate drafted, NOT entered: *"[doc 27 §3.2, §3.8] the item-5 census at the backreacting GBc placements is derived by the margin at ν ≥ ν̄(cell); on the SD SPLIT sub-family the classical carrier keeps the anchor reading 1/(νm̃₂) at every placement and, at ν = 3 at s ≡ 4 (mod 5), two further rational readings on a rational conic."*
- **Doc 25 (S-SR-0, S-SR-D1, S-SR-D2), NO ROW; S-SR-D1's mechanism is now the order-2 half of AM-5 (derived at every V by H11), which is corroboration of a data finding, not amendment; its rider (e) is the ν = 2 point of the D-S2 conic.** Doc 26, the decision document; this file is its report (§7.6 answers its §7); the owner folds.
- **Paper G (doc 13's GO paper, unassembled), NO TEXT; NO AMENDMENT OF ITS CLAIM SET.** Its claim set is "exhaustive, nothing wider is claimable" and nothing in it changed: claim (3) (the R2 collapse at 16/16 SR coverage with the causal sentence) is corroborated at a stronger grade at its definite-form cells and untouched at its anchor cells; claim (4) (the 14-cell QED control with the common-kernel invariant and the placement rider) gains a mechanism (the anchor lemma's support structure; the residue pattern), discussion material at Paper G's own gate process, a fresh decision there, never an edit by this program. **Papers A/B/C/F/M/S, 21a, NO TEXT CHANGE** (Paper B: the co-runner's light-crossing-phase rider candidate drafted, not entered).

Any future entry of the drafted candidate rows is a fresh registration decision at or after G26-1 + G26-2.


**[v1.1, 2026-09-23.]** The one condition that could change a doc-13 status, "**AM-3 ENTERED at theorem grade after G26-1 + G26-2**" (spec §7.4; §7.1), is **STILL NOT MET IN FULL: AM-3 is ENTERED unconditionally at theorem grade (the rigor conjunct discharged) and G26-1 is now PASS, but "G26-2" is discharged only in its PRE-MEETING leg; the POST-MEETING leg is owed by the calendar (after 2026-10-30).** Ruling: **NONE ENTERED, the default holds, unchanged; the condition is PENDING G26-2's post-meeting leg, not pending G26-1 any more.** The doc 13 §4 candidate row (the fourth route that delivers the emptiness at C1, C2, the 14 definite-form CANDIDATE-GRAV cells and the 246-cell definite-form population, excepting the SD SITE sub-family below 5/2 by a derived bound, the anchor cells staying computed) stays drafted-not-entered; it is the one row the post-meeting block (§C) must rule on as a fresh registration decision, ENTERED into doc 13 §4 or NOT ENTERED with reason, never as an amendment of doc 13 §3.3. The doc 13 §3.3 rider-content candidates (b), (c), the doc 14 §2.3 candidate (its own condition "AM-4 reaches ANCHOR-LEMMA" still not met, AM-4 stays ANCHOR-TYPED), the doc 19 §3.5 item 5 / §7.6 candidate and the doc 25 lines are unchanged. Three data facts are added to the candidates' evidence (not to any sentence): the V = 10 order-1 face census and swap kill (U-R1), the V = 10 classical-carrier datum at ν = 1 (U-R2), and the audit's NS = 24 hand probe (13/13), all family-extension data. Any future entry is a fresh registration decision at or after the post-meeting G26-2. Doc 13 §4's mechanism paragraph itself (`13_selfgrav_observer.md` `56808d0997540c33…`, byte-identical to its pin in every table of the chain) is untouched.

### 7.2 Paper assessment (Paper G amendment candidate vs a new note)

**Neither is triggered now; the natural home is NEW, and it is not Paper G.** (i) **Not a Paper G amendment:** the margin theorem may not carry a gravity word (G26-0(b)), so it cannot be Paper G content by construction; what Paper G receives is corroboration (the derived emptiness at its claim-(3) definite-form cells at their knobs) and mechanism for its claim (4), discussion material at its own gate process, not an amendment of its claim set (the reach, the taxonomy and the causal sentence are unchanged). (ii) **A new note is the honest shape, at methods/data grade, when policy calls:** *"a diagonal-dominance margin for self-consistent readings of constrained lattice clocks, with its anchor: a role-blind certificate in both carriers above an explicit coupling, sharp to 2–8 % at the theorem cells; the singular case of the constraint-energy form as the zero-energy anchor sublattice with a derived support lemma; a matched Gauss-routed control; the fold constants"*, a certificate-grade note in the doc-21a/05a pattern, self-contained, citing docs 13/14/19/25 for the objects and Espíndola–Ali 2605.30423 as the inverse-direction continuum precedent, the SN / 2602.24177 / CVI distinctions in its discussion, Horn–Johnson for the instrument; the co-runner's three closures (Fork-1 corollary; U2 theorem; the E3 exhibit) are riders for Papers B / A / S at their own gate processes, never headline. (iii) **Timing:** defer-until-big governs (standing policy; a direct scoop or user ask revisits, Yokokura's CVI posting and the Perimeter output are the named watch items); no paper action is gated now; G26-0.5 + G26-1 + G26-2 precede any export in either home. Doc 26 §7's DECISIVE compositions (Scout C's Composition A with doc 16's clockability locus; the wall-shaped cross-carrier ingredient; the continuum cite-and-distinguish reading for Paper G's discussion) remain unclaimed and unchanged in status.

### 7.3 Never-claimables (this program's, on top of the inherited stacks, doc 12's scoping bans, doc 13 §7, doc 14 §6.3, doc 19 §7.3, doc 25 §7.3 whole, doc 22 §6.7's ban list, doc 26 §4.1's riders and §7, all verbatim in force)

- **The margin theorem is never a gravity theorem** (G26-0(b) / KG-7): never "the margin theorem explains why gravity cannot read itself"; never "the certificate sees the constraint role" (on every computed row, in and out of family, it does not); never "self-energy detuning" without the continuum-threshold rider; never "self-observation has a critical coupling" bare (KF-5); never "ν̄ is the threshold" (a sufficient bound; the boundary ν_max sits at 0.92–0.98 of it at κ = 0 and as low as 0.084 of it at κ ≥ 1); never "the reading vanishes at ν_c" (the fold is the stratum's; the off-conic and real loci reach beyond it).
- **The anchor:** never "the anchor lemma is proved" (ANCHOR-TYPED with a derived sub-scope; the V ≡ 2 (mod 4) cross-block step is checked, not derived); never "own routing thins the anchor" bare (M4: at every gravity SD SPLIT row the pair is the constraint's zero-energy set under either clock kernel; the licensed thinning instance is the QED cell); never AM-5 without the period-5 clause and the commensurability clause (at lab periods 6, 7 the same kernel leaves no anchor; at 3, 4 the third regime); never "the classical carrier reads iff Z ≠ ∅" as a biconditional about the whole classical set (only the anchor reading is iff; the QED rows and the ν = 3 conic read off Z); never "the SD SITE cells / the SD SPLIT discretization are unphysical" (registered discretizations; their looseness / indefiniteness are reported properties); never "the anchor is an epistemic horizon".
- **Carriers and nulls:** never "the classical carrier's blindness implies the quantum one" bare (it needs the hop margin and fails at every anchor); a certificate in the margin regime never feeds "requires QM" (it is where the carriers AGREE); a null in the anchor regime, the KG-4 silence, or the classical conic never feeds "gravity is classical" / "the classical carrier can read itself" bare (doc 25 rider (c) is the licensed physics reading); never "the QED control failed" (clause (ii) is a knob-scoped MARGIN-EXT datum); never "the two-regime anatomy is the whole of doc 13's sentence" (§1.5's maximal wording only); never "route G is uniform" bare (THRESHOLD-TYPE cells at the certified knobs, census printed).
- **Shapes:** no no-go of any shape (doc 13 §7); "undecidable" never (N1); no committed cell as an open problem (F-1; every unfinished typing is NOT-COMPUTED-BELOW-CAPS with its caps printed); NOT-FOUND-BELOW-CAPS never upgrades to NO-GO; no LAW F / LAW G unification (the doc-14 tombstone); no grand-observer sentence (KR-6); "a universe cannot measure itself" never; "Gödel from Gauss law" never; "the QRF school's non-ideal frame is our margin" and "Espíndola–Ali found our threshold" never; "the QRF school already showed" never; the co-runner's exhibit never "gravity converts" / "third carrier" / "gravity is a magic factory".
- **Sentences at ν ∉ {1, 2} carry the family-extension rider without exception** (the fold, the sporadic points, the knob-3 conic, the unseen knob-3 rows, the beyond-family controls); the near-corollary rows and the relabelled clauses are never findings; the refutations of this program are reported first wherever its results are quoted.
- **Per-export obligations:** G26-0.5 + G26-1 + G26-2 before any quotation (G26-3); every sentence whole-with-riders (spec §0.3 + this section + the doc-13 rider stack whenever doc 13's sentence is touched; the doc-14 compatibility sentence next to any continuum backreacting-frame citation); the must-cite blocks of both scouts; INSPIRE primary for citer enumeration; defer-until-big; licensing is never posting.


**[v1.2 pointer 6, 2026-09-24.]** **§7.3, the anchor bullet.** "never 'the anchor lemma is proved' (ANCHOR-TYPED with a derived sub-scope; the V ≡ 2 (mod 4) cross-block step is checked, not derived)" → "never 'the anchor lemma is proved' bare (the licensed form: derived over the order-1 F2cQ SD SPLITL family, with its riders); never derived at a gravity row (SG-2 corollary, computed); never 'the V ≡ 2 (mod 4) face is a dimension-1 checkerboard' beyond V = 6; never the real-ν statement (Theorem C) as a reading". The rest of the bullet is unchanged.


### 7.4 Gates

| gate | content | status |
|---|---|---|
| **G26-0** (hostile gate; doc 26 M-L3) | per-route decided-at-freeze check; circularity (the masked-label run; the vocabulary audit); the O-4 check; the §7.5 checklist | **EXECUTED, PASS-WITH-MANDATORY-REPAIRS (2026-09-05; `anchor_gate_audit.md` 44/44); M1–M7 + m1–m13 applied as spec v1.0.1** |
| **G26-0.5** (record-integrity repairs; inserted by the verifier legs, the doc-21 G21-0 pattern) | generator-side, verdict-neutral: (a) D-R1, fix the S0 pilot driver, extend every pin list by `anchor_pilot_run.log` / `anchor_pilot_state.json`, restore or regenerate the two files and re-pin with disclosure; (b) D-S12, regenerate `anchor_rows.jsonl` from the committed text (the 24 stale `D_zero_is_balanced` fields) and complete a clean single-launch mirror of the final text (D-S8); (c) D-S1, disclose the three verdict flips in `anchor_results.md`; (d) D-R4/D-S6, the H9 lane; (e) D-R7/D-S7/D-R8, the results-writer duplications, the KG-1 letter, the KG-4 caps line (D-S11); (f) (f) the rigor leg's completion, R2's 44 remaining d = 16 classes (a cap on its sympy `count_roots` reference first), R3–R7, the `_extra.py E3` frame check, and its report re-issued from the hand-off text (SYN-1) | **OWED, blocks G26-1** |
| **G26-1** (closing audit at the untouched reserve ≥ 8,500,007) | third-band re-execution of the census through own-route second paths (own profiles by dense elimination; own hop from the configuration basis; own Δ in the D2 and the vertex form; own CK; own strata; own Sturm at d ≤ 18; own Gram-floor Bareiss at d = 64); the reserved unseen cells by a third path (U-S1's face census; U-S3 / U-S6); the C1/C2 brackets by an independent route; ν_hi ≤ ν̄ at every bracketed row; the row-by-row identity of a clean run of the final engine text (D-S8); must catch D-R1 and D-S12; a hand probe registered pre-engine | **OWED, gates external quotation** |
| **G26-2** (scoop re-sweep incl. the **ARMED Perimeter trigger**) | after 2026-10-30: PIRSA C26010, Indico 2058's published timetable, INSPIRE `cnum:C26-10-26.5`, the 18 speakers' + 5 organizers' feeds Nov 2026 – Jan 2027; Scout A §5.3's target list (Yokokura's CVI, full-text first if posted, the continuum-threshold rider re-worded; Lock; Semião; Hausmann's no-go form; Panella; Cepollaro; Suleymanov; Pickup; Mekonnen; Chen; de la Hamette; the 18 unnamed invited talks); citer sweeps of 2605.30423, 2608.25800, 2607.15351, 2605.20330, 2602.24177, 2602.15134, DEHK, 2510.13607, 2606.09344, 2506.21489, 2409.06479; the grep targets; the papers/12–13 maps re-dated; the four Adjudicator triggers re-checked (doc 26 §6 item 1 governs if one has fired); INSPIRE primary | **OWED (armed, not fired; CLEAR pre-meeting on the public contribution export)** |
| **G26-3** (external quotation of S-AM-1 … S-AM-SPLIT) | whole-with-riders only (spec §0.3 + §7.3 + the doc-13 rider stack) | **BLOCKED** until G26-0.5 + G26-1 + G26-2; then a dated amendment of this document rules it conjunct by conjunct (the doc 19 §7.6 / doc 21 §7.6 pattern); licensing is never posting |
| **G26-4** (pointer rows) | the material-change rule | **RULED THIS DOCUMENT: NONE ENTERED** (§7.1; five candidate rows drafted-not-entered; the doc-13 §4 row is the one that would change a status, at or after G26-1 + G26-2) |
| **G26-5** (standing) | per-export re-sweep inherited permanently (G19-6 / G20-4 / G21-5 / G22-2 / G24-2 / G25-2 carried); the M-L7 named sweeps; method notes binding (INSPIRE primary; arXiv throttling) | **STANDING** |


**[v1.1 amendment (G26-0.5 executed; rigor complete; G26-1 PASS; G26-2 pre-meeting executed + re-verified; G26-3 ruled), gate rows of record, 2026-09-23; v1.0 rows above preserved.]**

| gate | outcome |
|---|---|
| **G26-0** hostile gate | Unchanged, **EXECUTED 2026-09-05, PASS-WITH-MANDATORY-REPAIRS**, M1–M7 + m1–m13 applied as spec v1.0.1 |
| **G26-0.5** record-integrity repairs | **EXECUTED, DISCHARGED 2026-09-22.** (a) D-R1 landed (driver `--pilot-dir`; pin lists 64 / 61 / 44 / 81; the two pilot files regenerated by ONE launch, 98/98 verdict-row identity, installed 21:09:21, disclosed in results §0.2 and the repair report); (b) D-S12 landed (24 sanctioned leaves; 766/790 lines byte-identical) and D-S8's clean single-launch mirror of the final text complete (mirror3: 790/790 verdict-row identity, 59/59 verdicts, 0 SUBSTANTIVE; its rows / state / results / run.log preserved as `g26_0_5_evidence/mirror3/` `9d9e3831570d01bb…` / `de2d97baf19b0385…` / `18faa99ac343db9c…` / `8fb1ff3db683d609…`); (c) D-S1 disclosed (results §0.1, generator-parsed); (d) H9 lane computed; (e) results-writer deduplication, the KG-1 letter split (union-preserving), the KG-4 caps line, the D-R8 clause; (f) the rigor leg's completion, R2's 48 remaining classes by the amended F(q) route (the stall measured: D-R11), R3–R7, E3, the report re-issued (`anchor_rigor_verifier.md` `02c2162002589224…`). Verdict neutrality machine-asserted at every step (rows 0 verdict-field differences; ledger 0 flips; outcome, rungs, KG sets identical); before / after A0 batteries on the v1.0 and the amended texts; two `--regen` dry runs; canon pins identical before / after every launch; evidence in `g26_0_5_evidence/` (`MANIFEST.sha256`); report `analysis/g26_0_5_repair_report.md`. **Both record-integrity items re-established independently by G26-1 (C5) and their repairs confirmed** |
| **G26-1** closing audit | **EXECUTED 2026-09-22 23:46:36 → 2026-09-23 04:14:19 at the untouched reserve ≥ 8,500,007, PASS** (`analysis/anchor_closing_audit.md` `7cd12fba2dbb15da…`; script `7aaafa863e096a55…`; state `107f9e6a00401b68…`; 790 own rows `bc9d8784699b640c…`). **84/84 own checks; 2 record defects (D-R1, D-S12), 0 verdict-touching; 4 audit-design items (A-1 … A-4) repaired, re-run and disclosed (+ the C7 self-scan re-specification, G3-1); no licensed sentence touched, no rung downgraded, no dimension / lemma / bracket / verdict of the record moved, no wording change recommended.** Item by item against the v1.0 row: third-band re-execution of the census through own-route second paths, 338 classes / 790 rows leaf-by-leaf against the REGENERATED canon, **0 mismatching leaves**, construction identity (own profiles by dense elimination with the mean-zero gauge as an extra row and rook pivoting; own hop from the configuration index dict; own forms as explicit NS × NS Gram matrices; own Δ in the D2 AND the vertex form cross-identified; own CK / N_Z by integer fraction-free elimination + mod-p rank certificates at d = 256; own Gram floor by exact LDLᵀ; own F(q) at d ≤ 18 by Lagrange interpolation at a DISJOINT node set; own primitive-PRS Sturm; own strata by mod-p rank at the reserve primes confirmed exactly on the group columns) at every class, 5 identity routes × 338 classes with 0 failures; own strata at 338/338 classes (355 strata rows, 182 fold ν_c certified from their own minimal polynomials); the reserved unseen cells by a third path, U-S1's face census (24 faces, dim CK = 4 on each and 0 off, residues {0, 3, 4} mod 5, \|Z\| = 36 as edges of K₄ on each half, own F₄ of dimension 2, the four products in ker(hop\|_Z) spanning CK, swap contrast dim CK = 0), U-S3 / U-S6 at d = 64 (ν̄ exact-certified to 10⁻²⁰ by a sign change of the own minimal polynomial, `14267620*t**2 - 144*t - 27` / `103271798*t**2 - 1638*t - 135`; G > 0; certified at the knob; Gram floor PD; Z empty; classical census all-negative; U-S6 G_min 102446027/135 identical); the C1/C2 brackets by an independent route (upper end: own real-root count 0 at ν_hi = 16/2467, 551/98366, 1082/38455, 1449/46534; lower end: an exact sign change of F(q, ν_lo) at ν_lo = 623/96078, 274/48925, 1685/59898, 2273/73011 with the own Sturm count 2 = the record's); ν_hi ≤ ν̄ at every bracketed row (184 classes, by exact evaluation of the own ν̄ polynomial; STRICTLY below at the four theorem rows, and the certificate confirmed AT ν̄ and above, at ν = 1, at ν = 2 and by a PD Gram floor at the registered knob); the row-by-row identity of the clean run of the final text (D-S8: re-derived from the mirror tree, `mirror3_compare.json` never read, 53 byte-identical / 737 `runtime_s`-only / 0 substantive; 59 verdicts same state; the one extra = F-1); D-R1 caught (four pre-engine states pin the originals; the mechanism reproduced from the v1.0 text in an isolated tree); D-S12 caught (24 leaves from the `.pre_g26_0_5` backup; own recomputation True at all 24); the hand probe registered 2026-09-22 23:46:47 and computed 2026-09-23 02:25:11 at the fresh cell `GBc 24 BC SPLITL kz=(1, 0) s_c=11 order-2`, 13/13 clauses matched [family-extension: NS = 24 outside the caps]; the wording bans (C7) PASS on the six sentences and on the audit's own report; canon pins 64 identical before / after, record pins 24 unchanged. **Primes: 8,500,007 / 8,500,021 / 8,500,027 drawn (tokens only); 8,500,057 / 8,500,061 certified not drawn; next reserve ≥ 8,550,001 designated.** Residues typed, gating nothing: mirror-vs-mirror determinism NOT-COMPUTED (D-R10; no second launch); D-R12's `deg` field, D-R14's "22 censuses" wording, D-R5's results-§0 residue and F-1's 60th verdict at the next generator regeneration. **The G26-1 conjunct of G26-3 is MET** |
| **G26-2** scoop re-sweep | **PRE-MEETING leg EXECUTED 2026-09-22 (22:04–22:42) and RE-VERIFIED three times (23:00–23:10; 23:28–23:32; 2026-09-23 03:32–03:47), CLEAR-PRE-MEETING** (`papers/24_anchor_margin/G26_2_PRE_MEETING_SWEEP.md` §0–§13): the four Adjudicator triggers re-checked and dated, NONE FIRED (re-dated 2026-09-23 03:41 CEST by a full isolated INSPIRE re-run + an arXiv window validated by controls); 45 citer enumerations (INSPIRE `refersto:recid:`, 250-cap; membership identical at 03:41), 56 INSPIRE + 54 arXiv author feeds, 22 INSPIRE full-text / title nets, 23 arXiv nets, 14 abstract-level confirmation nets + 1 control (state-file counts): 0 kills, 0 forces-rider, 0 scoops of S-AM-1 … S-AM-SPLIT; two new full-text items, both cite-and-distinguish (2512.14834 v2, the first INSPIRE citer of 2605.20330, classical-carrier genre; Nandi 2609.09931, Paper B discussion only); the Perimeter trigger ARMED and CLEAR pre-meeting with a new datum: the public Indico export CHANGED 86 → 79 (Lock's talk, Semião's and Cepollaro's posters, Rothlin's, Giovanakis's, Rastgoo's, Diaz's, D'Esposito's and Aguilar Gutierrez's contributions removed; Astwood and Brzic added; Cummings and Mele re-typed to talks; id-set, titles and types identical at every re-export through 03:44); the 18 invited / review slots still unnamed (timetable 56 entries, 0 presenters); PIRSA C26010 0 talks; INSPIRE `cnum:C26-10-26.5` 0; Yokokura's CVI still unposted (ARMED-HIGH); the papers/12–13 maps re-dated, verdicts unchanged; the 13 watch-id citer totals identical at 22:04, 23:06, 23:30 and 03:41 (2605.30423 = 4, 2608.25800 = 0, 2607.15351 = 2, 2605.20330 = 1, 2602.24177 = 0, 2602.15134 = 0, 2412.15502 = 70, 2405.00114 = 92, 2510.13607 = 9, 2606.09344 = 7, 2506.21489 = 15, 2409.06479 = 25, 2511.07348 = 20). **POST-MEETING leg OWED after 2026-10-30, itemized (sweep §8):** (1) PIRSA C26010, every recording's title + abstract (+ video where thin) of the 5 review, 13 conference, 19 contributed and the lightning talks, grep targets incl. "diagonal", "Gershgorin", "fixed point", "self-reading", "critical coupling", "variance identity", "fragility", graded per Scout A §5.2; (2) Indico 2058's published timetable and contribution export re-diffed against the 2026-09-22 files (86 → 79 → ?), the 18 slots named, the nine removed contributions checked for re-appearance; (3) INSPIRE `cnum:C26-10-26.5` (record 3150071) as proceedings land; (4) the 18 speakers' + 5 organizers' feeds Nov 2026 – Jan 2027 on arXiv (7–11 s gaps; 429 recovery) and INSPIRE `de >= 2026-10-26`, the framing-scoop trigger of doc 26 §6 item 2 re-checked; (5) Scout A §5.3's target list as amended, Yokokura's CVI (full-text first if posted; rider re-worded), Lock, Semião, Cepollaro (withdrawn, feeds only), Hausmann's no-go form, Panella, Suleymanov, Pickup, Mekonnen, Chen L.-Q. de la Hamette, Brzic (new), the 18 talks once named; (6) the citer sweeps of the eleven watch ids + DEHK's pair re-run (INSPIRE primary; `refersto:recid:`, never `refersto:arxiv:`) and diffed against `g26_2_inspire_state.json` (the 250-cap enumeration; the cheapest full re-date is an isolated copy of `g26_2_inspire.py` with `SINCE` edited, ~254 s), incl. 2512.14834 v2's own citers; (7) the grep targets and the theorem-shape nets re-run over 2026-09-22 → the sweep date (both scripts, `SINCE` / `WIN` edited; the arXiv window control `all:"black hole"` ∧ window > 0 before trusting any zero); (8) the papers/12–13 maps re-dated; (9) the four Adjudicator triggers re-checked and dated (doc 26 §6 item 1 governs if one has fired), `2511.07348`'s version, the MOVW feeds, GIE data / timeline, Andriolo, the Theorem W / M nets; (10) then the G26-3 re-ruling by a dated block (§7.7; template §C), UNBLOCKED on CLEAR, riders on forces-rider hits, downgrade on a kill, the order of record on a fired trigger, together with the G26-4 fresh registration decision on the doc-13 §4 row. **The G26-2 conjunct of G26-3 is MET for the pre-meeting leg and OWED for the post-meeting leg** |
| **G26-3** external quotation | **RULED 2026-09-23, CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING** (§7.7): every conjunct MET (Adjudicator triggers NONE FIRED, ruled first; G26-0.5 neutral; rigor CONFIRMED unconditionally; G26-1 PASS at the reserve; no pre-meeting kill; whole-with-riders discipline available and extended) except G26-2's post-meeting leg (calendar); **external quotation of S-AM-1 … S-AM-SPLIT stays BLOCKED until the post-2026-10-30 leg discharges**; then a dated block (§C) rules UNBLOCKED whole-with-riders (or re-rules on a kill / a fired trigger); licensing is never posting |
| **G26-4** pointer rows | Unchanged, **RULED IN v1.0: NONE ENTERED** (§7.1; re-checked §1.6 above: the doc-13 §4 condition pending G26-2's post-meeting leg only, AM-3 at theorem grade, G26-1 PASS); the drafted candidate rows remain a fresh registration decision at or after the post-meeting G26-2, to be ruled in the same dated block |
| **G26-5** standing | **STANDING**, watch list UPDATED (§7.7.4): Yokokura's CVI posting (ARMED-HIGH; full-text first); the nine removed Perimeter contributions and Brzic; 2512.14834 v2's citers; 2605.30423's citers (4); 2608.25800 (0) / 2607.15351 (2) / 2605.20330 (1) citers; the Adjudicator triggers; method notes binding (INSPIRE primary, `refersto:recid:<recid>` for citers, `refersto:arxiv:` returns the whole index; `abstracts.value:` for abstract nets, `abstracts:` is silently invalid; `de` is the earliest date, so union with an arXiv `submittedDate` window; arXiv 429 after ~40 queries at 3.5 s, recovery 300 s cool-down at 11 s; the arXiv API index carried the 2026-09-22 announcement by 03:40 CEST on 09-23; Indico / PIRSA need a browser User-Agent; python `urllib` on this Mac has no CA bundle, the scripts use an unverified context, curl with a browser UA works; Semantic Scholar a lagging secondary, never a count source) |

### 7.5 Next calculation

When policy calls, in order: **(1)** G26-0.5 (the generator-side repairs above; verdict-neutral; the regenerated rows must be byte-stable on a fresh isolated single-launch run of the final engine text); **(2)** G26-1 at the untouched reserve; **(3)** G26-2 after 2026-10-30, then the G26-3 ruling by dated amendment; **(4)** the residue of this front, none of it headline: the V ≡ 2 (mod 4) cross-block step of the block lemma as a derivation (the rigor leg's V = 10 cell U-R1 is its registered test; a landed derivation lifts AM-4 to ANCHOR-LEMMA over the family, a fresh ruling); the F2c stratum-absence derivation at the 1360 computed groups (M-L6's target); the exact discriminant of C1's irreducible (32, 16) curve (the evaluation/interpolation route); the letter-anchored exclusion at gravity cells (H9; data); the reducibility pattern of the V = 4 reading curves at V = 6, 8 (data); the third regime (lab periods 3, 4) and the period > 5 THRESHOLD cells as a beyond-family data lane (never a taxonomy extension without re-registration); **(5)** the successor front per doc 26 §6: the charged pair on I-CR as this front's second wave on the same GBc cells (the reader supplies the anchor the charged clock destroyed, the anchor lemma's pair form; its afternoon pre-check as registered), unless the Perimeter output fires the framing-scoop trigger (doc 26 §6 item 2) or an Adjudicator trigger fires (§6 item 1); the co-runner's E3 exhibit goes to the P-E3-0 auditor ruling (doc 26 §6 item 3) and opens nothing by itself. The observer line (docs 12–14, 19, 25, this) is now residue-free apart from the gates and the anchor half's remaining derivation: both mandated theorem shapes landed at their honest grades, the forecast refutations were recorded first, and the dream question has its answer at record grade, threshold AND anchor, split by the inertia of one form.


**[v1.1, 2026-09-23.]** Item (1) G26-0.5, **DONE 2026-09-22** (the regenerated rows byte-stable at the verdict-row level on a fresh isolated single launch of the final text: mirror3; the only non-byte-stable leaf class is `runtime_s`). Item (2) G26-1 at the untouched reserve, **DONE 2026-09-22 / 2026-09-23, PASS** (tokens 8,500,007 / 8,500,021 / 8,500,027 consumed; next audit reserve ≥ 8,550,001). Item (3) G26-2 post-meeting after 2026-10-30, then the G26-3 re-ruling by dated block, **OWED (calendar); the only open gate work of this program.** Item (4) unchanged, with one datum added: the V ≡ 2 (mod 4) cross-block step now has TWO machine-checked instances (V = 6, V = 10) and no derivation, the lift of AM-4 to ANCHOR-LEMMA needs the derivation and a fresh ruling, never more machine checks; the residues owed at the NEXT generator regeneration (no new gate): D-R12's `deg` field, D-R14's "22 censuses" wording, D-R5's results-§0 pair, F-1's 60th verdict in the ledger, a second fresh launch for mirror-vs-mirror determinism (D-R10), such a regeneration is re-audited at the new reserve ≥ 8,550,001. Item (5) unchanged. **Order of the owed work: G26-2 post-meeting (after 2026-10-30) → G26-3 re-ruled + G26-4's fresh registration decision in one dated block.**


**[v1.2 pointer 7, 2026-09-24.]** **§7.5 item (4).** Its first clause is DONE (2026-09-24: the derivation, `am4_lift_derivation.md`, and the verifier, `am4_lift_verifier.md`; the fresh ruling by this block). The rest of item (4) and item (5) are unchanged.


### 7.6 Doc 26 §7's honesty section, answered point by point

- **"Is this formulation-level?", No, confirmed by the run.** Every definition was a computed inequality or set on the engines' objects; the first computation split three ways exactly as pre-registered (near-corollaries, P-G1, the μ > 0 and swap clauses, plus the eight relabels the gate added; genuine computations, the census, the ν̄ table, the anchor lemma, Z's shapes, the fold; derivation targets, Theorem G, Cauchy–Schwarz on definite forms, Z at both orders, the F2c absence), and the derivation targets landed at the grades of §1.5. The word "anchor" survives as more than S-SR-D1's classical sense: the lemma holds everywhere and the anchor is the singular case of the constraint-energy form.
- **"The derivation is on the role-blind side."** Confirmed as priced, and then less lopsided than feared: the margin is derived and role-blind; the anchor's SHAPE is derived at both kernel orders (the inertia statement), its support lemma derived on a sub-scope, and its quantum emptiness computed. The export is the lopsided wall doc 26 said the house must be prepared to publish, with the anchor side carrying more derivation than the odds table gave it (Z-thinness derived at SD SPLIT: priced ~50 %, landed; the anchor lemma at the survivors: priced ~80 %, landed at computed strength with a derived sub-scope).
- **"ν̄ may sit above the knobs somewhere."** It does, at the SD SITE sub-family only, gravity and QED alike, and the corner is a theorem about the discretization (> 5/2 / > 5/4 at every even V); everywhere else the knobs sit 6.5–180× above the entire real locus. The wording caps at AM-2 there and reaches AM-3 on the whole definite-form population outside it, doc 26's MARGIN-EXT / MARGIN-THEOREM split, located and derived.
- **"The busiest adjacent field could brand the reading first."** It has not (Scout B: nobody has computed or stated the anatomy; the Perimeter contribution list CLEAR); it has produced the nearest neighbour (Espíndola–Ali's inverse-direction sufficiency certificate) and a threshold WORD-shape (Yokokura's CVI), both cite-and-distinguish, both riders, neither the object. The delta doc 26 named (explicit ν̄, Z, the anchor lemma, the QED control, the fold constants) is on the record at certificate grade with the Perimeter re-sweep armed.
- **"A thin theorem, again."** The mechanism sentences travel and the inequality does not; the sharpness datum says the Geršgorin bound is within 2–8 % of the boundary exactly at the theorem cells and loose (up to ~12×) where the κ-gate adds a hop unit, reported as such, with the Gram-floor sharpening typed non-load-bearing; the anchor mechanism travels with its shape (the inertia trichotomy; faces thinned to antipodally periodic members; the F ⊗ F kernels). The D-S4 correction of the mechanism sentence's attribution (the configuration's constraint energy, not "the clock's self-sourced energy") is fused so that the sentence says what the D ≡ 0 control shows.
- **"The fold is a family extension."** Every ν ∉ {1, 2} sentence carries the rider; the physics reading changed from absolute to strong-coupling as doc 26 said (the real locus of every CANDIDATE-GRAV cell confined below 0.31 with the knobs at 1–2), and the classical conic datum (knob 3) is family-extension data of the classical carrier, nothing here moves doc 13's sentence.
- **"The doc-13 never-claimables bind."** No no-go; doc 13's sentence untouched (G26-4 NONE ENTERED); the anchor regime's quantum emptiness stays computed (the accepted stop, pre-registered); the parity declaration honoured, every refutation reported first, including the mechanism note's own forecast-false clauses and hand errors, the engine's flips, the steelman's own refuted registration and predicate artifact, and the record-integrity breach.
- **"After three stops, was FIELD the honest pick?"** Yes, on the record: the front's first rung could not come back empty and did not; every higher rung was open and each landed or died honestly; the fourth stop (KG-0 / KG-1 at C1 or C2, priced ~8 %) did not occur; the DUTY alternative (the Adjudicator) keeps its promotion clause dormant (no trigger fired), and the charged pair is the registered second wave.

---


### 7.7 The G26-3 ruling (external quotation), v1.1 amendment, 2026-09-23

*(Numbered 7.7 because v1.0's §7.6, "Doc 26 §7's honesty section, answered point by point", is preserved; the task brief called this the new §7.6; the doc 19 §7.6 / doc 21 §7.6 pattern is followed.)*

#### 7.7.1 The bar, ruled conjunct by conjunct (from v1.0 §1.5 / §7.3 / §7.4 and the house rule of this chain)

| conjunct | ruling | basis |
|---|---|---|
| the four Classicality-Adjudicator triggers, ruled FIRST (doc 26 §6 item 1) | **NONE FIRED (dated 2026-09-22; re-dated 23:06, 23:30, 2026-09-23 03:41 CEST)**, the order of record unchanged | sweep §1 + §11 + §12 + §13; `g26_2_reverify_0330.json`: `2511.07348` still v1 (published = updated 2025-11-10); INSPIRE cc 20 (Nandi 2609.09931 [FT], Salih 2609.04104, neither MOVW); the MOVW, GIE-collaboration, Andriolo-group and §5.3 feeds 0 since 09-21 on both indices; no GIE data or timeline (MAST-QG "at least a decade"); no ledger-shaped Theorem W / M scoop |
| G26-0.5 landed verdict-neutral (items (a)–(f)) | **MET** | repair report §0 / §3 / §5 / §6: 790/790 rows with exactly the 24 sanctioned leaves (`G26_0_5.D_S12_rows_changed`, 24 keys); 59/59 verdicts 0 flips (engine ledger 92 = 33 CHECK + 59 VERDICT, 7 refuted); outcome / rungs / KG sets identical; pilot 98/98 verdict-row identity by ONE launch; mirror3 790/790 verdict-row identity, 0 SUBSTANTIVE; the D-R1 guard PASS in isolation; every engine pin table (64 files) = canon now (re-hashed by this leg, 0 mismatches); item (f) by the rigor report; **D-R1 and D-S12 re-established by G26-1's C5 from independent evidence, repairs confirmed** |
| the rigor leg completed CONFIRMED, its conditionality on the entered rungs lifted | **MET, DISCHARGED** | rigor report §0 / §1 / §3–§8 / §10 and its state: stages R0–R7 `done`, R7 `complete`; R2 790/790 rows, 338/338 classes, 0 mismatches, 0 construction-identity failures; R7 (21:50:25) `n_diff_rows` 0, `verdicts_only_in_record` [], `verdicts_only_in_mirror` 1 (the A5 KG-4 verdict, F-1), `KG_order_only_diffs` ["KG-1"], outcome identical, 66 inputs sha-verified with 0 mismatch and 2 RE-BASELINED = the pilot's two files, `mirror_pins_identical` true; mirror ledger checks 23 (0 failed), verdicts 60 (7 refuted); E3 PASS/PASS; ledger 78 CHECK (5 `ok = False`: indices 41, 42, 47 verifier-side / reporting-side; 86, 89 the historical pre-regeneration R7 entries), 28 FINDING (18 refuted, all provenance / reporting / own registrations); U-R1 … U-R4 recorded with two misses (data); **no verdict, theorem or certificate value touched; no rung reversed**; the 81-entry pin table = canon now (re-hashed, 0 mismatches) |
| G26-1 closing audit PASS at the untouched reserve ≥ 8,500,007 | **MET, PASS** | `anchor_closing_audit.md` (`7cd12fba2dbb15da…`) + its state (`107f9e6a00401b68…`): 84 checks, 0 failed; 6 findings = A-1 … A-4 (audit-design, repaired, re-run, superseded entries kept) + D-R1 + D-S12 (record, verdict-neutral, repairs confirmed); every item of the v1.0 G26-1 row PASS (§1.7); C9: no record defect verdict-touching, no licensed sentence downgraded, no dimension / lemma / bracket / verdict moved; C7: all six sentences verbatim, wording bans hold; C8: 64 canon pins + 24 record pins unchanged; primes 8,500,007 / 8,500,021 / 8,500,027 drawn as tokens, 8,500,057 / 8,500,061 spares, next reserve ≥ 8,550,001 designated (all re-certified prime by this leg); the hand probe registered 23:46:47 before computation, 13/13 at 02:25:11; the audit's four pin tables = canon now (re-hashed by this leg, 0 mismatches); G3-1 / G3-2 INFO only |
| G26-2 found no kill against any sentence, PRE-MEETING leg | **MET (pre-meeting; re-verified ×3)** | sweep §0 / §2 / §4 / §11 / §12 / §13: 0 kills, 0 forces-rider, 0 scoops on 45 citer enumerations, 22 + 23 + 14 nets, 56 + 54 author feeds; two cite-and-distinguish additions; the papers/12–13 maps unchanged in verdict; the pre-meeting silence index-validated (no QRF / PW abstract posted 09-05 → 09-22 on either index; the 09-21 → 09-24 window 0 on 15 arXiv nets with the window control 42 / 106 > 0); the 03:41 full isolated INSPIRE re-run membership-identical on 45/45 ids |
| G26-2, POST-MEETING leg | **OWED by the calendar (after 2026-10-30)** | the meeting has not happened; ten items (§1.7 row) |
| whole-with-riders discipline available | **MET, extended** | spec §0.3 + §7.3 + the doc-13 rider stack (i)–(v) + (a)–(e) whenever doc 13's sentence is touched + both scouts' must-cite blocks, extended by §7.7.4 |

**ALL CONJUNCTS MET EXCEPT G26-2's POST-MEETING LEG → CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING: external quotation of every sentence STAYS BLOCKED until the post-2026-10-30 leg discharges.** Per-sentence rulings (sentences verbatim from §1.5; no text change; each row states the evidentiary standing so that the post-meeting block can rule by reference):

#### 7.7.2 Per-sentence rulings

| sentence | ruling 2026-09-23 | evidentiary standing (G26-0.5 · rigor · G26-1 · G26-2 pre-meeting) | outstanding | riders specific to it (on top of the base set) |
|---|---|---|---|---|
| **S-AM-1** (rung AM-1, census form) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING**, external quotation blocked until then | G26-0.5: no census field regenerated (the 24 leaves are a control-side consistency flag at F2cQ SD SITE rows; 246/246, 81/81, 14/14 untouched) · rigor R2 / R6 own: P-G8 246/246 per cell, 14/14 control; the 9 SD SITE gravity cells certified at knob 3, uncertified at 1, 2 (H5 own at 51 rows, closed form to 1e-30; > 5/2 own / > 5/4 Gauss); KG-1's uncertified G-definite rows all SD SITE (18; 6 exactly empty at d = 16); the own ν̄ at C1 and the pilot's 8 independent GBc values reproduce to 7 digits · **G26-1 C1 / CS: 338/338 classes, 790/790 rows, 0 mismatching leaves (ν̄ by EXACT quadratic-surd comparison over Q, the printed 20- and 50-digit values certified by a sign change of the minimal polynomial; the type; the classical census; the anchor reading; PD); 355 strata rows and 182 fold ν_c re-derived** · G26-2: no occupant of the census shape; premise-genre citers 0 | G26-2 post-meeting | family-extension (ν ∈ Q_{>0}); instrument; Espíndola–Ali (every AM-1 / 2 / 3 sentence); "on every computed row … the Gram sign structure of the discretization alone" stays a statement about the registered discretizations |
| **S-AM-2** (rung AM-2; role-blind; D-S4 fused) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING** | G26-0.5: no theorem row changed · rigor R2 own: the vertex identity and disc ≤ Δ at every configuration; the Cauchy–Schwarz clause (π_x ∦ u; G_x > 0 at every SITE row); own exact LDLᵀ PD at every certified row d ≤ 64 and two d = 256 rows; the mod-token recomputation at every d ≤ 64 row; R3 P-G13 182/182 (Sturm ≥ 1 / 0; ν_hi ≤ own ν̄); KG-3 clean by own route; the steelman's masked-label / out-of-family controls stand · **G26-1 C3: ν_hi ≤ ν̄ at all 184 bracketed classes by exact evaluation; ν_hi STRICTLY below ν̄ at the four theorem rows and the certificate confirmed AT ν̄ (a rational immediately above the own ν̄), at ν = 1, at ν = 2, Gram floor PD at the knob, A-2 records that the certificate is NOT claimed below ν̄ ("a sufficient bound, not the boundary" load-bearing); C7: no gravity word in the quoted sentence** · G26-2: diagonal-dominance / Lévy–Desplanques / Geršgorin-as-reading-exclusion nets 0 on both indices at every re-check; 2605.30423 citers 3 → 4 (2609.17669 no-overlap); 2602.24177 0; the CVI unposted | G26-2 post-meeting | family-extension; continuum-threshold incl. Espíndola–Ali 2605.30423, 2602.24177, the SN threshold and the CVI bound (its clause stays "unposted", re-worded against the posted version the day it posts, full-text first); instrument (Horn–Johnson §6.1, §7.2); never a gravity word, the theorem holds beyond the registered family (order-3 slaving; lab periods 6, 7; 200 fuzzed instances; the audit's NS = 24 probe [family-extension]) |
| **S-AM-3** (rung AM-3; scope V3-2 + census extension) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING** | G26-0.5: neutral · rigor: the own C1/C2 brackets (§1.3 above; ≤ 1 / ≤ 2), the 14 definite-form CANDIDATE-GRAV cells (+ the SPLITR spot) certified at every registered placement, 246/246 per cell, G-definiteness at every BC SPLIT cell of the class, the sharpness data own; U-R3 (C2 at the unregistered placements s_c ∈ {2, 4, 5}: certified at knob 2, ν̄ = 0.0398157 / 0.0296206 / 0.0316232 < 0.1, Z = ∅, one λ = 0 stratum each) · **G26-1 C3: the C1/C2 brackets by a THIRD independent route, own Lagrange interpolation at a disjoint node set + own primitive-PRS Sturm: 0 real roots at ν_hi at all four targets, a sign change of F(q, ν_lo) on an exact rational q-interval at all four, the own Sturm count at ν_lo = the record's (2); the C2 factor-discriminant boundary roots 4/4 and 5/5 certified by sign changes of their own minimal polynomials; C7: "gravity" in the quoted sentence only as the doc-13 TAXONOMY SCOPE, which G26-0(b)'s letter permits** · G26-2: no kill; doc 13's sentence untouched by anything found | G26-2 post-meeting | doc 13's sentence unchanged, whole-with-riders (i)–(v) + (a)–(e); G26-4 governs any status change (the doc-13 §4 row is the post-meeting block's fresh decision); worded as a derivation of the EMPTINESS, never as an amendment; family-extension; Espíndola–Ali |
| **S-AM-4** (rung AM-4, ANCHOR-TYPED; M5 wording; derived sub-scope in-sentence) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING; grade UNCHANGED (ANCHOR-TYPED)** | G26-0.5: the H9 lane (the secondary-normalization exception's datum) now computed in the record (= E2 / gate S6); no verdict · rigor: Z / dim N_Z / dim CK / the lemma / the S-antichain flags identical at every row incl. the six survivor rows and the 80 U-S1 rows; R4 660 (cell, placement, order) jobs, CK = ker(hop\|_Z) at every placement; the block lemma symbolic, scope V ≡ 0 (mod 4); U-R1 (V = 10, all 50 placements, d = 1024), the lemma at all 50; **the V ≡ 2 (mod 4) split-placement step machine-checked at V = 6 and V = 10, derived at none, ANCHOR-LEMMA "in scope" WITHDRAWN**; E3 67/67 · **G26-1 C1: dim CK / N_Z / the lemma at 790/790 rows by own integer fraction-free elimination (+ mod-p rank certificates at d = 256); C2: U-S1's 24 faces with dim CK = 4 and 0 off, the F₄ ⊗ F₄ basis reproducible under a registered labelling (A-1), the swap contrast dim CK = 0, a third-path confirmation, not a derivation** · G26-2: anchor / zero-energy / zero-mode nets 0; 2409.06479 citers 25 unchanged; 2607.15351 citers 2 unchanged | G26-2 post-meeting | secondary-normalization exception typed (doc 13 §3.3(c): {−1/25, 0, 1/20} ride the letter −1/5, H9, outside CK); back-action rider (Hausmann–Renner); normalization rider (Ardenghi cite-only); family-extension rider on any V = 10 mention (U-R1 is out of family); the labelling note of A-1 (the V = 8 anchor configurations carry two up-sites per half, edges of K₄) |
| **S-AM-5** (rung AM-5, both kernel orders, inertia form) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING** | G26-0.5: the 24 regenerated leaves are the H10 consistency flag at the F2cQ SD SITE rows (order-1 zero-energy set = balanced, stale False → True); the H10 / H11 / H12 verdict CONF unchanged · rigor: H11 EXACT by sympy (λ_r = W_r/5, W_0 = 0 exactly, W_r = (25/4)csc²(πr/V) − 5/2; normalization corrected 2026-09-22 from W_r/NS); H10 / H12 PASS; H13 / H14 (dim F_2 = 1, F_4 = 2, F_6 = 5, F_odd = 0); the order-1 face residues {0, 3, 4} at V = 4 and {1, 2} at V = 6 (doc 13's rider); U-R1 at V = 10: residues {1, 2}, fraction 0.4, \|Z\| = 72, dim N_Z = dim CK = 4 at the 20 split placements, Z = ∅ at the 30 others; the swap kernel: N_Z = CK = 0 at all 50; D-R14's datum (H8 under the order-2 kernel at V = 8 maps Z(s) → Z(2 − s); no rung reads it) · **G26-1 C5: the order-1 zero-energy clause re-derived at the 24 SITE rows (D_x = V − S_x²/NS > 0, the set empty, the clause vacuous, the regenerated True); C2: the residues {0, 3, 4} mod 5 at V = 8, \|Z\| = 36, F₄ dim 2, products spanning CK; C1: Z at 790/790 rows** · G26-2: parity / Fourier / inertia-decided-reading nets 0 | G26-2 post-meeting | family-extension rider on every ν ∉ {1, 2} clause and on every V ∉ {4, 6, 8} clause; the M4 bare form of the thinning clause never travels alone; the period-5 and commensurability clauses load-bearing; the registered hop-null clause (iii) REFUTED at V = 8 and replaced by F ⊗ F (as printed); the A-1 labelling note |
| **S-AM-SPLIT** (rung AM-SPLIT, the wall sharpened) | **CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING** | G26-0.5: KG-1's split union-preserving; KG-4's caps line now printed (the 24 uncertified d = 256 rows NOT-COMPUTED, the sentence's typing) · rigor: P-G7 255 THRESHOLD / 18 NEVER (all F2c SD SPLIT, Z ≠ ∅) / 0 WINDOW by own route, no third regime in family; KG-5 clean (route G certifies no survivor row; the swap contrast reproduced inside the anchor regime); the classical-carrier datum re-derived: true at ν ∈ {1, 2} in family, false at ν = 3 at s ≡ 4 (mod 5) at every V ∈ {4, 6, 8} (V = 8: 24/55, 7/18 new), false at ν = 1 at V = 10 s ≡ 4 (mod 5) (U-R2, out of family) · **G26-1 C1: the certified-set type at 338/338 classes by own ν-set analysis (255 / 18 / 0 reproduced); the classical census and the anchor reading at every row; C2: the swap contrast at every tested V = 8 face placement** · G26-2: the carrier-typing anchor 2607.05578 citers 2 unchanged; 2605.20330 citers 0 → 1 = 2512.14834 v2 (cite-and-distinguish on the classical-carrier clause; the anchor regime's classical reading never glossed with their "hybrid") | G26-2 post-meeting | carrier-typing rider 2607.05578; the non-occupancy sentence; "in the registered family (lab period 5)"; family-extension rider on the ν = 3 clause; the U-R2 data line travels with the classical-carrier clause as a V = 10 family-extension datum; 2512.14834 v2 next to 2605.20330 |
| AM-INV | **NOTHING** (no sentence) | KG-4 silent by own route at every computed row (d ≤ 64 exact / capped; d = 256 typed); the steelman's R3.1–R3.3; the audit's C1 at every row and its NS = 24 probe (0 real quantum readings at ν = 2) | - | every EMPTY row prints its caps |
| the dream's third clause (spec §2.7 maximal wording) | **with the sentences: CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING; computed at the rows, derived at none** | unchanged by every leg | G26-2 post-meeting | never "the anatomy IS doc 13's sentence" |


**[v1.2 pointer 8, 2026-09-24.]** **§7.7.2, the S-AM-4 row.** Grade ANCHOR-LEMMA over the family (v1.2). The ruling stays CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING. The evidentiary standing adds both legs of 2026-09-24.


#### 7.7.3 The ladder, final (unconditional)

| rung | v1.0 §1.5 | after the rigor leg and G26-1 (this ruling) | why |
|---|---|---|---|
| AM-0 | not reached | not reached | no KG fires by any route (rigor own; audit third band) |
| AM-1 | ENTERED conditionally | **ENTERED** | census re-derived 246/246 · 81/81 · 14/14 by the rigor leg's own route; 338/338 classes by the audit's third band |
| AM-2 | ENTERED conditionally | **ENTERED** | bar (1)–(5) met in full: the derivation machine-checked by the engine, by the rigor leg's own paths at 790/790 rows and by the audit's own Δ in both forms at 790/790 rows; G26-0(b) PASS (gate, rigor R6, audit C7); KG-3 clean; U-S3 / U-S4 / U-S6 realized; both legs + the audit |
| AM-3 | ENTERED conditionally | **ENTERED** | AM-2's bar + the per-cell census + the own C1/C2 brackets at the rigor band + the audit's brackets by a third route at the reserve + G-definiteness at every BC SPLIT cell of the class + the near-corollary rows labelled + worded as a derivation of the EMPTINESS |
| AM-4 | ENTERED at ANCHOR-TYPED (derived sub-scope) | **ENTERED at ANCHOR-TYPED (derived sub-scope), UNCHANGED; does NOT rise to ANCHOR-LEMMA** | the V ≡ 2 (mod 4) split placements: machine-checked at V = 6 and V = 10, derived at none (rigor §10: "a computed confirmation at one more V is not a derivation"; the hand-off's "in SCOPE" withdrawn); the audit's C2 confirms the V = 8 F₄ ⊗ F₄ structure by a third path and derives nothing; the lift = §7.5 (4), a derivation + a fresh ruling |
| AM-5 | ENTERED (inertia form, both orders) | **ENTERED** (inertia form, both orders; period-5 + commensurability clauses) | H11 exact; the block lemma; F ⊗ F; U-S1 confronted by three paths (engine, steelman, audit), U-R1 by the rigor leg [family-extension rider at V = 10] |
| AM-SPLIT | ENTERED (sharpened) | **ENTERED** (sharpened; a complete dichotomy on the 273-cell taxonomy by two own routes) | P-G7 own (rigor); the type at 338/338 classes (audit); the classical-carrier datum typed with its ν = 3 and V = 10 exceptions |
| AM-INV | NOTHING | **NOTHING** | KG-4 silent by every route |

Wording of every rung as printed in v1.0 §1.5; no sentence text changes; the closing audit recommends none.


**[v1.2, 2026-09-24, the AM-4 ladder row (Part B of the fresh ruling).]**

| rung | v1.1 (2026-09-23) | v1.2 (this block) | why |
|---|---|---|---|
| AM-4 | ANCHOR-TYPED (derived sub-scope); does not rise | **ENTERED at ANCHOR-LEMMA over the order-1 F2cQ SD SPLITL family; the corollary rows unchanged (SG-2 / SG-1 / Z's shape)** | The bar's derivation clause is met by a near-corollary of pieces committed 2026-09-05 (Theorem A; D-S5 mis-scoped), along the route registered pre-code. The step the record named is derived in its true form (Theorem B). The machine-check clause is met by the record (790/790; E3 67/67) and re-met by two legs at V = 4 … 18; the ⇒ content is also computed at V = 22. Quotation is gated as every rung (G26-2 post-meeting). |

**[v1.2 pointer 9, 2026-09-24.]** **§7.7.3, the AM-4 row.** As in Part B.


#### 7.7.4 Rider set of record (travels with EVERY quotation once unblocked; binding on any internal citation now)

**Base (unchanged):** spec §0.3 in full (in-class; family extension; role-blindness G26-0(b); doc 13's sentence whole-with-riders; the never-bare list; the continuum-threshold, interacting-clock, instrument, back-action, control-positioning, carrier-typing, normalization, premise-genre riders; the non-occupancy sentence; the Perimeter rider; parity) + §7.3 in full + the doc-13 rider stack (i)–(v) + (a)–(e) whenever doc 13's sentence is touched + the doc-14 compatibility sentence next to any continuum backreacting-frame citation + both scouts' must-cite blocks + the standing per-export re-sweep (G26-5). **Additions of this cycle (2026-09-22 / 2026-09-23):**

1. **Must-cite block adds** 2512.14834 v2 (Schlegel–Dakić–Del Santo, "Entanglement without Quantum Mechanics: Operational Constraints on the Quantum Signature", v2 2026-09-16; PDF `3b026d345a5552ad…`, title-verified) next to 2605.20330, the classical-carrier genre's standard for which entanglement signatures a classical carrier can host; cite-and-distinguish on the doc-19 item-5 shadow lane and on S-AM-SPLIT's classical-carrier clause; the anchor regime's classical reading is never glossed with their "hybrid" (theirs is a phase-space statistics statement; the house's a computed fixed-point set on a constraint), and 2609.09931 (Nandi; PDF `9b7d970d686f2487…`) next to 2608.30075, Paper B discussion only (LOCC premise by citation, no ledger, no mediation definition; not a Theorem-W scoop).
2. **Interacting-clock rider re-worded:** doc 27 §2's "Lock (talk), cite as talks until posted" → "Lock, cite the abstract as submitted to the Perimeter meeting (contribution withdrawn from the public list by 2026-09-22) until posted"; likewise Semião and Cepollaro (posters withdrawn); **Brzic** (Sorbonne; "Quantum controlled Dyson operator from temporal reference frames", poster) joins the interacting-clock list, cite as a poster until posted; Scout A §6 rider (2) reads accordingly.
3. **Continuum-threshold rider:** the CVI clause stays worded "unposted" (Yokokura, Perimeter Oct 2026); re-worded against the posted version the day it posts, full-text first (Scout A rider (8)); until then ARMED-HIGH.
4. **Perimeter rider (M-L7):** the public Indico export changed 86 → 79 (nine removed, two added, two re-typed; id-set, titles and types stable at every re-export from 22:04 to 2026-09-23 03:44); Scout A §5.3's target list is amended as in the G26-2 row; the 18 invited / review slots unnamed; no wording entry above the record before the post-2026-10-30 re-sweep; the framing-scoop trigger (doc 26 §6 item 2) not visible pre-meeting.
5. **Family-extension rider extended to every V ∉ {4, 6, 8} and NS ∉ {12, 16, 20} clause** (the rigor leg's U-R1 / U-R2 at V = 10 and the audit's hand probe at GBc NS = 24 are out of the doc-13 taxonomy / the registered caps): the V = 10 face census, swap kill and classical-carrier datum, and the NS = 24 certificate datum, are family-extension data, never taxonomy statements.
6. **Classical-carrier data rider (S-AM-SPLIT):** the "only nontrivial rational classical reading" datum is true at ν ∈ {1, 2} on the registered family; false at ν = 3 at s ≡ 4 (mod 5) at every V ∈ {4, 6, 8} (V = 4: 3/8, 8/15; V = 6: 6/13, 98/255; V = 8: 24/55, 7/18) [family-extension]; false already at ν = 1 at V = 10, s ≡ 4 (mod 5) (U-R2: 10/23, 86/225) [family-extension], a role datum of the classical carrier, never a role claim.
7. **Anchor-lemma grade rider (S-AM-4):** the V ≡ 2 (mod 4) split-placement step is machine-checked at two instances (V = 6, V = 10) and derived at none; the "proved" wording of §7.3's anchor bullet stays banned; D-R14's datum (H8 under the order-2 kernel at V = 8: Z(s) → Z(2 − s)) is no sentence's content; **the A-1 labelling note**, at the V = 8 face every anchor configuration carries two up-sites per half (an edge of K₄), and the F₄ ⊗ F₄ basis clause is reproducible only under that labelling.
8. **Sufficiency rider (S-AM-2 / S-AM-3; the audit's A-2):** the margin certificate is claimed at and above ν̄ and never at ν_hi (the upper end of the ν_max bracket, which lies strictly below ν̄ at every theorem row, sharpness 0.968 / 0.976 / 0.950 / 0.924); "ν̄ is a sufficient bound, not the boundary" is load-bearing in every quotation; the endpoint's grade is BRACKET-CERTIFIED (a real-root count), never a PD certificate.
9. **Erratum riders (§5 v1.1):** (i) the R2 stall attribution "the leg's uncapped sympy `count_roots` reference" (v1.0 §5, §1.6 (viii), §7.4 (f)) is superseded by D-R11 (the symbolic Berkowitz determinant in q) wherever quoted; (ii) the pilot's two record files carry the regenerated shas `1b40ebf63cf7f289…` / `0347c35e8208bfca…` (the 2026-09-02 originals `e500e04619ff50fc…` / `6d52ee219e17d2ce…` unrecoverable), disclosed wherever the pilot record is pinned; (iii) the v1.0 §6 lines "run.log now 1e345b22b32a5c6f… / state now c8a5c26dc1bb4fb2…" and "rigor `_state.json` unchanged `5897398f…`" are superseded by §A; (iv) the sweep's own round figures "60 / 55 / 78" read 56 / 54 / 77 (its §11); (v) the v1.0 §6 clause "closing-audit reserve ≥ 8,500,007 UNTOUCHED" is superseded by the consumed / designated ledger of §1.5.
10. **Watch list (G26-5) updated:** Yokokura's CVI posting (TOP); the nine removed Perimeter contributions' re-appearance and Brzic; 2512.14834 v2's citers; the Adjudicator triggers (`2511.07348`'s version; the MOVW feeds; GIE timeline; Andriolo; Theorem W / M nets); 2605.30423 (4), 2608.25800 (0), 2607.15351 (2), 2605.20330 (1), 2602.24177 (0), 2602.15134 (0) citers; the lattice-QED-QRF cluster (2604.06149 6; 2604.06087 9); Salih 2609.04104 (Fork-1 lane only).


**[v1.2 pointer 10, 2026-09-24.]** **§7.7.4, rider 7.** "the V ≡ 2 (mod 4) split-placement step is machine-checked at two instances (V = 6, V = 10) and derived at none; the 'proved' wording of §7.3's anchor bullet stays banned" → "the ⇒ direction at the V ≡ 2 (mod 4) split placements is derived: a near-corollary of the block lemma and Lemma A′. The cross-block step is derived (Theorem B). Both are machine-checked at V = 6 and 10 (record), 14 (all placements, two legs) and 18 (all placements at ν = 2; spot at ν = 1). At V = 22 the ⇒ content is checked at four placements and the dimension is NOT-COMPUTED [family-extension off V ∈ {4, 6, 8}]. 'The anchor lemma is proved' stays banned bare (pointer 6)." D-R14's datum and the A-1 labelling note are unchanged.

#### 7.7.5 What this ruling does NOT do

No sentence text changes; no rung moves (AM-4 stays ANCHOR-TYPED by the record's own finding; the audit's third path adds confirmation, not derivation); no pointer row enters (G26-4 stands as ruled; the doc-13 §4 condition is pending the post-meeting G26-2 and is that block's fresh decision); the mechanism note stays untouched (registered pre-code); no prime is drawn by this ruling (the reserve was consumed by G26-1, the next reserve ≥ 8,550,001 is designated and untouched); D-R12 / D-R14's wording fixes, D-R5's results-§0 residue, F-1's 60th verdict and the mirror-vs-mirror residue (D-R10) remain owed at the NEXT generator regeneration with no new gate; **nothing is externally quotable today**, every sentence is CONDITIONALLY-UNBLOCKED-PENDING-G26-2-POST-MEETING and external quotation stays blocked until the post-2026-10-30 leg discharges; licensing is never posting, defer-until-big governs all timing, and every future quotation event owes the G26-5 re-sweep first. The post-meeting block (§C) is a template, not a ruling.

---

*Compiled 2026-09-05 by the synthesizer from: doc 26 §4–§7 (whole); the pilot record and its three verifier legs (+ reconciliation); the frozen spec v1.0.1 (whole, incl. Appendix D); the hostile gate; the mechanism note and its provenance leg; the engine's results (§0–§11, whole), rows and ledger; the rigor leg's hand-off report, log and state (and the resumed run launched by this synthesis, §5); the physics/steelman leg (whole); the co-runner (whole); both scout reports (whole); docs 13 §3.3/§4/§7, 14 §2.3, 19 §3.5, 25 §1.5/§7 for the instrument and the pointer-row targets; doc 21 for the house format. Synthesizer-side computation: sha256 of every file of record before and after; the reconstruction of the rigor leg's pending classes from its state; no prime drawn; no engine executed beyond the resumed rigor run; no committed file of any earlier doc written; the pilot-side draft of this file superseded in full; README / memory / SUBMISSION_STATUS left for the owner. House discipline: every quoted number exact (Fraction / integer polynomial / sympy over Q) or typed as display; every claim tagged by the legs that confirmed it; NOT-FOUND-BELOW-CAPS never upgrades to NO-GO; the near-corollary rows never findings; licensing is never posting; defer-until-big governs all external timing.*