# Coherence read, 2026-08-30: the landed notes against the live layer **Status: HELD.** A coherence read, not a red team. No mathematics re-derived. Each finding quotes both sides with line numbers and proposes a replacement graded APPLY or ASK. ## 0. Counts 27 distinct findings. By kind: 18 DISAGREEMENTS (D1-D18), 3 STALE FIRST MOVES (S1-S3; S3 = D7), 6 RUNG INFLATIONS (R1-R6; R1 = D1, R2 = D14), 8 ORPHAN rows (O1-O8; O2-O4 restate D7, D8, D3-D4; O7-O8 are correctly held and need nothing). By grade: 15 APPLY (D1, D2, D3, D4, D5, D6, D7, D8 for G2-STATE, D9, D11, D14, D15, D17, S1, O1), 12 ASK (D10, D12, D13, D16, D18, S2, R3, R4, R5, R6, O5, O6). By live file: `G2-STATE.md` §0 four (D1, D2, D8, D9), `README.md` §Status five plus a missing paragraph (D2, D3, D5, D6, D17; O5), `TODO.md` eight (D7, D11, D12, D13, D14, D15, D16, S1; S2 after the in-flight note lands), `anchored-calm.md` one with two sentences (D4), `REFUTED.md` one (D17), `CHANGELOG.md` two (D16, O1), `GLOSSARY.md` one shared (R4). Nothing in any note's mathematics is proposed for edit; D18 is the one place a landed note contradicts the record rather than the reverse. Ledger ids: none missing. Gate: 251/251 audit checks, selftest fires on all planted defects, one gate section red (§7). ## 1. What was read, and when All times CEST, 2026-08-30. Live layer, read as it stood: `TODO.md` 09:04 (837 lines), `README.md` §Status 09:04 (lines 68-210), `research/G2-STATE.md` §0 09:05 (lines 17-244), `research/REFUTED.md` 09:05 (71 data rows; rows dated 2026-08-29 at lines 94-95, 2026-08-30 at lines 96-98), `research/anchored-calm.md` 09:05, `research/history/CHANGELOG.md` two newest entries 09:05 (lines 15-147), `research/QUESTIONS.md` §1 and the §2 rows for the new ids 09:05, `research/GLOSSARY.md` entries for the skeleton (372-384) and Var/E (404-425) 09:07 and 09:10. Landed notes, each in full, 09:06-09:08: `attack-0829n-{rml-proof,hsubpow-K,X-upper,doubling-bridge,parity-dstar}.md`, `attack-0830-{rec-cheapest,doubling-killrun,varE-identification,anchored-ladder-17,buchstab-deep,head-remainder,tail-derivation,record-mechanism}.md`, `measure-0830-delta-reader.md`, `blind-0830-quadpoint-31607.md`, `verify-0830-usup-convention.md`, `recon-0830-rec-killrun.md`, `decide-0830-skeleton-door.md`. Supporting records opened only to adjudicate a specific disagreement: `paper/proposals/prop-exact-fold-L.md` (grade line), `research/history/staging/lemmaV-sup-extension.md` lines 440-445 and 500, `session-0829-summary.md` 118-135, `applied-0829-refuted.md` 279-281, `zonegap-03-score.md` ledger, the five `recon-0828-*.md` ledger verdicts, `paper/variance-note.md` (grep for 0.611), `research/natal-cap-36-skeleton-door.md` 30-40, `research/SEARCH-CONVENTIONS.md` (grep). The six in-flight notes were not opened. No note's mathematics was re-derived; where a note and a live sentence disagree on a number the record behind both was read and is cited. Line numbers below are the file's own at the read time; the orchestrator's concurrent edits to `TODO.md`, `CHANGELOG.md` and `REFUTED.md` may have shifted them. ## 2. DISAGREEMENTS (live sentence contradicts a landed note) | # | live sentence (file:line, quoted) | landed note (file:line, quoted) | replacement | grade | |---|---|---|---|---| | D1 | `G2-STATE.md`:57-61, under **PROVEN**: "Per-fold `L` is a statistic of the old gap word alone, exactly ... `paper/proposals/prop-exact-fold-L.md`." | `prop-exact-fold-L.md`:3 "**Grade: PROPOSAL**", :44 "PROPOSAL. The mathematics is in better shape than the grade suggests and the grade is held down by four things"; `attack-0830-doubling-killrun.md`:299-303 "`prop-exact-fold-L.md` §2 grades itself PROPOSAL, with the equality lemma proven in a staging record, verified at 36 cells and nine diagonal folds, and never through the live layer or a second reader" | Move the bullet out of the PROVEN list into a one-line rider after it: "PROPOSAL grade, not proven here: per-fold `L` as the longest alternation-legal window with the 3/2 constant a Perron root (`paper/proposals/prop-exact-fold-L.md` §2, equality lemma proven in a staging record, verified at 36 cells, no second reader)." | APPLY | | D2 | `README.md`:82-83 "67 closed routes are indexed one line each in `research/REFUTED.md`"; `G2-STATE.md`:158 "**DEAD.** 67 routes (recounted 2026-08-28)" | `REFUTED.md` as read 09:05: 71 data rows (`grep -c '^\| '` = 73 minus header and rule); rows 94-95 dated 2026-08-29, rows 96-98 dated 2026-08-30 | README: "71 closed routes are indexed"; G2-STATE: "71 routes (recounted 2026-08-30: 67 on 08-28, two added 08-29, three 08-30)". README:195 "every one of the 67 closed routes was re-derived" describes the 08-29 audit and stays. | APPLY | | D3 | `README.md`:140-141 "its all-x form is open behind a door that reaches about a tenth of the mass" | `decide-0830-skeleton-door.md`:8 "the door as named removes at most 0.0102 from a G30_agg whose open part is 0.094 to 0.126", :40-41 "the −10.8% was a subsample artifact with the wrong sign, and the 'tenth' is the @13 value alone"; `REFUTED.md`:96 (row added 2026-08-30) | "its all-x form is open, its named door closed as a route on 2026-08-30 (it governs under one percent of the mass from @17 on, `research/REFUTED.md`)" | APPLY | | D4 | `anchored-calm.md`:46 "its all-x form is open behind a door that reaches a tenth of the mass"; :61 "its door is named, analytic, and measured open" | same file :40 "the named door is CLOSED as a route (2026-08-30, `research/REFUTED.md`)", :67-68 "the door is a closed route in `REFUTED.md`"; decide note as in D3 (its edit list :144-146 named :40 and :59-64 and missed :46 and :61) | :46 "its all-x form is open, its named door closed as a route (2026-08-30), and"; :61 "its door is named, analytic, and closed as a route, since the branches" | APPLY | | D5 | `README.md`:194-195 "the sealed 1e12 zone-gap forecast scored 8 hits and 0 misses" | `TODO.md`:386-387 (Z7) "`zonegap-03-score.md`: ten sealed rows, ten HIT, zero MISS"; `zonegap-03-score.md` ledger verdict "All ten sealed rows score HIT and none miss, after a second pass ..."; commit ca3a154 | "the sealed 1e12 zone-gap forecast scored ten hits of ten after a second pass made the last two rows scorable" | APPLY | | D6 | `README.md`:196 "27 clauses corrected" | `session-0829-summary.md`:125 "25 clauses rewritten (`applied-0829-refuted.md`)"; `applied-0829-refuted.md`:280 "Twenty-five row clauses were corrected in place"; commit 406ad44 "25 clauses corrected". The 27 is the summary's SOUND-NARROWER count at :119. | "25 clauses corrected" | APPLY | | D7 | `TODO.md`:325-326 (Z5 title) "b has no mechanism in either corpus"; :370-371 "the residual, 0.78 to 0.97 of b, sits at 3.2 to 4.2 ensemble sd and has no mechanism in either corpus"; :372-373 "First move now: the mechanism question is unchanged and is the item." | `attack-0830-record-mechanism.md`:8 "The residual is carried by the twin-prime gap law AT HEIGHT and by nothing else tried ... carries 95.7 to 99.8 percent of the residual"; :389-392 "The item's coordinate is therefore not 'b has no mechanism in any model this corpus can build' but 'b is the finite-height gap law of twin primes, which this corpus can measure to 1e11 and cannot derive; the top height band is the open falsifier'"; `CHANGELOG.md`:69-74 records it; the item body carries only the ledger id (:391) | Title: "it is Kourbatov's b, and b is the twin-prime gap law at height, which is measured and not derived." After clause (d): "THE MECHANISM IS LOCATED, NOT DERIVED (2026-08-30, `attack-0830-record-mechanism.md`, HELD, PARTIAL): a new sieve to 1e11 measures the twin-gap law at height under-dispersed (CV² 0.7276 → 0.9295 over seven decades, far-tail log-slope 1.0638) and, fed into the record null with no parameter, it carries 95.7 to 99.8% of the residual (d b_z = 1.0370 and 1.0806 against 1.0833, 2,000 paired replicates, MEASURED on a model); the tile's period-wide law and Kourbatov-Wolf's k = 1 bin count overshoot 4 to 7× and are closed as posed; 'computable from Var/E' is false as stated (0.62 against 1.06 by family). Not height-stationary; the top band's b = 0.937 against 1.69 to 1.98 predicted on 17 records is the open falsifier; per-band sigmas owed." First move: "derive the gap law at height, or run the per-band sigmas (20 min re-embed) and the 1e12 sieve (about 2 h) that test the stationarity." | APPLY | | D8 | `G2-STATE.md`:137 "**HEURISTIC, held behind two open steps.**"; :142 "the CRT-mixed lags a divisor-distribution question about y-smooth divisors of C(C²−4)"; `TODO.md`:722 "the CRT-mixed lags reduce, by an exact identity, to the distribution of y-smooth divisors of C(C²−4) above 2L, a Ford/Hooley-convention question not yet searched" | `attack-0830-varE-identification.md`:27-29 "The brief's two open steps are one", :101-105 "The divisor-distribution statement of `varE-theta2-proof.md` section 6 is for `n > 2L` only. The band `n <= 2L` carries `82 %` of the measured mixed remainder"; :8 "the one inequality left is the uniform o(1) equidistribution of y-friable squarefree integers in progressions to moduli up to y^(4/5)". README:186 already reads "the identification half is the one open step". | G2-STATE:137 "held behind one open step"; :142 "the CRT-mixed lags a class discrepancy whose dominant part (82% of the measured remainder at x = 19) sits below 2L in the unbalanced range and is a friable-integers-in-progressions question (`attack-0830-varE-identification.md`, HELD)". TODO:722: trim the "above 2L, a Ford/Hooley-convention question" clause to "(the record's n > 2L display covers ~18% of it, see below)". | APPLY (G2-STATE); ASK (TODO 9, since :730-752 already carries the correction and the item may prefer one paragraph) | | D9 | `G2-STATE.md`:205-208 "proving `Ĝ(2s) ≤ C₂·Ĝ(s)` on the base-2 power chain for any `C₂ < 2^{β₂} = 19.2455` is trap-free (custody data force ...) and every provable point beats β₂." | `TODO.md`:417-418 "TRAP-FREE FOR THE ALL-s FORM ONLY (scoped 2026-08-27)"; `attack-0829n-doubling-bridge.md`:59-64 "The eventual form `∀k ≥ k₀` is excluded on purpose ... every `C₂ ∈ (3.1034, 4)` would be data-consistent and TPC-implying"; the K*-product and fold-composition bridges CLOSED (`REFUTED.md`:94, :98) | "is trap-free in its all-s form only (the eventual form is TPC-implying below C₂ = 4, TODO D) and every provable point beats β₂; the two bridges tried, the K*-product and the per-fold composition, are closed (`REFUTED.md`, 2026-08-29/30)." | APPLY | | D10 | `G2-STATE.md`:176-182 "Lemma V's missing piece is the quantifier ... Deshouillers–Iwaniec 1982 Theorem 12, Maynard's Lemma 6.12 and Pascadi's 2026 Corollary 18 all require ... If both were smooth, either theorem would clear Lemma V outright." | `attack-0829n-rml-proof.md`:325-333 "(ii) The consumer cannot take a smooth g either: Brüdern–Fouvry's pointwise inequality needs λ⁻ ≤ 1_{rough} ≤ λ⁺, which Möbius-signed Rosser supports supply and a smooth profile does not. (iii) Granting the smoothness anyway, the published bound prices at exponent 5.0907 > β₂"; :145-155 names the open arrow REC(s, u₀); `REFUTED.md` row "manufacturing the smooth profile" | The paragraph is the 08-20 state. A rewrite around REC(s, u₀) is a judgement call: at minimum strike "either theorem would clear Lemma V outright" and append "The open arrow is now written with its quantifiers as REC(s, u₀) (`attack-0829n-rml-proof.md` §3, HELD); the smooth-profile form is disconnected from it at both ends (§4.5), and an adverse derived floor is on file (`attack-0830-rec-cheapest.md` §4, HELD, not adversarially checked)." | ASK | | D11 | `TODO.md`:154 (Z2) "root u* = 3.565845, 1/u* = 0.280438" | `blind-0830-quadpoint-31607.md`:44-48 "`u2-engine-depth.md` §2 and `import-rough-anatomy.md` §3 both record 3.565847 on their own ω grids; the producer's own grid (step 10⁻⁵, tail 1 SEC B) gives 3.5658466. The brief's value is off in the seventh digit" | "root u* = 3.565847" | APPLY | | D12 | `TODO.md`:231-232 "64 HIT and 12 MISS, every miss on the naked asymptotes as pre-declared" | `blind-0830-quadpoint-31607.md`:252-256 "Every MISS is on OLD or NEW: their R2 rows at all four bands, as pre-declared in §2 ..., and their R3 rows at B9 and B10, which the dry run had shown at B6–B8 and §2's expectation line did not list separately." | "every miss on the two naked asymptotes (their R2 rows at all four bands, as pre-declared; their R3 rows at B9 and B10, not listed in the pre-declaration)" | ASK | | D13 | `TODO.md`:238-240 "exact-Mertens comparator above the data by 3.6 to 8.8 se at B11/B12 while the ω-form is not" | `blind-0830-quadpoint-31607.md`:265-269 "C2's are 1.6, 1.1, 0.4 and 2.8 se from zero (B9–B12) ... C2 (ω at finite u) does not, except marginally at B12." | "while the ω-form sits at 0.4 and 2.8 se" | ASK | | D14 | `TODO.md`:739-740 "Henriot Cor. 2 (2012, with the 2014 erratum, read at the page)" | `attack-0830-varE-identification.md`:355 "PROVEN given the theorem as read at the text layer of arXiv v1 plus the erratum ... PARTLY: v1 and the erratum read; the MPCPS published text NOT read"; campaign rule "text extraction is not reading" | "Henriot Cor. 2 (2012, read at the text layer of arXiv v1 with the 2014 erratum; the published MPCPS text not read)" | APPLY | | D15 | `TODO.md`:727-729 "variance-note's abstract, §7, GLOSSARY and README still carry 0.611 and are queued for correction once the paper edit is Chris's." | `README.md`: no occurrence of 0.611 (grep 09:09); `GLOSSARY.md`:416-419 carries it only as "refuted as an inference"; `paper/variance-note.md`:51 "$0.611$ is refuted as an inference", §11 (:822-888) is the refutation, §7 has no 0.611 | Delete the sentence, or: "variance-note §7 still carries the older 'u = 2 exactly' reading pending Chris's paper edit (its abstract and §11 refute 0.611; GLOSSARY and README no longer carry it)." | APPLY | | D16 | `TODO.md`:618-619 "(Maier–Pomerance, o(1) unconstrained; the corpus's 'δ ≈ 2 known' was wrong)"; `CHANGELOG.md`:43-44 "the brief's 'δ ≈ 2 conjectured for the one-class control' was wrong" | `measure-0830-delta-reader.md`:87 "**BRIEF IMPRECISE.** What is conjectured (Maier–Pomerance 1990 ...) is `Y(x) ≪ x(log x)^{2+o(1)}` ... Neither fixes `δ = 2`"; the corpus's own wording was `attack-0829n-hsubpow-K.md`:231 "The control's conjectured `δ` is `2 + o(1)`", never "known" | TODO: "(Maier–Pomerance, o(1) unconstrained, so the 08-29 note's 'conjectured δ = 2 + o(1)' fixed neither δ nor its finite-reach value)"; CHANGELOG: "was imprecise" for "was wrong" | ASK | | D17 | `REFUTED.md`:43 "measured rising at all eight steps over nine levels on a flat ~2.05 per-prime factor"; `README.md`:81 "it rises at all nine levels on a flat ~2.05 per-prime factor" | `verify-0830-usup-convention.md`:120-122 "The factor is 2.0516, the geometric mean of nine ratios over ten levels (line 442)"; `lemmaV-sup-extension.md`:440-442 lists nine ratios 13→17 ... 43→47, "geometric mean 2.0516 over all nine" | REFUTED: "rising at all nine steps over ten levels (z = 13..47)"; README: "at all ten levels to z = 47" | APPLY | | D18 | `README.md`:168 "a five-field recon of about seventy-five candidate applications" | `recon-0830-rec-killrun.md`:86-88, :368-373 "'~75' matches no line" (it sums "21 + 10 + 15 + 13 across the four `recon-0828-*` notes" = 59) | The live sentence is right and the note's brief-error clause is wrong: the five `recon-0828-*` ledger verdicts read 21 (sieve) + 10 (rough) + 15 (covering) + 13 (jacobsthal) + 16 (farfields) = 75; the note omitted farfields. No live edit; annotate the HELD note's §0 and §4 item 1 at the red team. | ASK | ## 3. STALE FIRST MOVES (TODO first move already ANSWERED/CLOSED in QUESTIONS.md) | # | TODO first move (line, quoted) | QUESTIONS.md row (line, status) | replacement | grade | |---|---|---|---|---| | S1 | `TODO.md`:241-245 (Z2) "FIRST MOVE NOW, and the only one left on this item: any larger run is a BLIND TEST of the zero-parameter law (forecasts y* ≈ 336 at 31607, ≈ 642 at 1e5, now to be restated on u·ω(u) = 2; fresh prereg + width audit first, per v2's rule)." | `QUESTIONS.md`:31 `Q-quadpoint-blind-31607-0830` ANSWERED: "on 8363 new anchors 10009 <= Q <= 100003 the 76 sealed rows score 64 HIT and 12 MISS"; the note's :246-249 scores y*(31607) = 313 and y*(100003) = 631 against exactly those forecasts. The item's own :230-240 records the run; the first-move sentence was not rewritten. | "FIRST MOVE NOW: none by decade run. The fourth decade is scored (above); the fifth is sealed and in flight (`blind-0830-quadpoint-c1c2.md`, `Q-quadpoint-c1c2-0830` OPEN) and decides only the C1-versus-C2 preference; no run of this kind at any reachable Q separates 0.280438 from 0.280730 (`blind-0830-quadpoint-31607.md` §6). Coordinate with item 8 (its (a) is CLOSED; (b) untouched)." | APPLY | | S2 | `TODO.md`:616-620 (item 1) "One line blocks it: W = 1.308e16 > 2^53 in the CRT anchoring product. First move: BigInt or double-double on that single path; then decide if a sixth residual collapse is worth ~13× the @41 runtime." | `QUESTIONS.md`:104 `Q-at43-bigint-0830` PARTIAL: "The engine side is done and VERIFIED (five paths promoted to BigInt, @7..@37 and a 0.248% slice of @41 reproduced digit for digit); the @43 point is NOT run, because the measured extrapolation is 576 h of eight cores" | The note (`engine-0830-at43-bigint.md`) is IN FLIGHT and was not opened; the ledger block already answers the first move as posed and prices the run at 576 core-hours × 8, far above the 4-hour rule. When it lands: first move becomes "Chris's go on 576 h of eight cores, or a segmented redesign; nothing else". | ASK (after landing) | | S3 | `TODO.md`:372-373 (Z5) "First move now: the mechanism question is unchanged and is the item." | `QUESTIONS.md`:54 `Q-record-mechanism-0830` PARTIAL: "The residual is carried by the twin-prime gap law AT HEIGHT and by nothing else tried" | See D7. | APPLY | Not stale, checked: item 0's (a) and (b) (the z = 41 sup is in flight), 1d's (a) (unrun by design) and (b) (marked RAN), A's (a) @29 (priced, not run) and (b), D (parked, none), Z4 (verification in flight), 8 (b) (untouched), 9 (the smooth-numbers-in-APs search is `recon-0830-smooth-aps.md`, in flight). ## 4. RUNG INFLATION (HELD quoted without HELD; measured quoted as derived) | # | live sentence (file:line, quoted) | what the record says | replacement | grade | |---|---|---|---|---| | R1 | `G2-STATE.md`:57-61, PROPOSAL listed under PROVEN | D1 | D1 | APPLY | | R2 | `TODO.md`:739-740 "read at the page" for a theorem read at a text layer | D14 | D14 | APPLY | | R3 | `README.md`:79-82 "`u_sup`, its by-product, was raised and closed the same day: it rises at all nine levels on a flat ~2.05 per-prime factor, `research/REFUTED.md`" | `verify-0830-usup-convention.md`:198-207 "Of its closing disjunction 'plateau above beta2, or divergence', the divergence disjunct is REFUTED at fixed s (u_sat ≤ 2s + o(1)); the 'plateau above β₂' disjunct remains MEASURED only"; :169-174 the per-prime-factor extrapolation's antecedent is "PROVEN false at fixed s". The README sentence gives the geometric rise as the closure and omits that the rise cannot continue. | "(`u_sup`, its by-product, was raised and closed the same day: measured rising over ten levels on a ~2.05 per-prime factor, bounded at fixed s by z^{2s+o(1)} and so plateauing above β₂ rather than diverging, `research/REFUTED.md`, `verify-0830-usup-convention.md`)" | ASK | | R4 | `TODO.md`:721 (item 9) "residuals ≤ 0.002 at nine exact points from x = 13"; `G2-STATE.md`:140 "residuals ≤ 0.002 at nine exact points from x = 13"; `GLOSSARY.md`:422 "residuals ≤ 0.002 from x = 13" | `attack-0830-varE-identification.md`:405-408 "(d) TODO item 9's 'residuals <= 0.002 at nine exact points from x = 13' is the Monte-Carlo-model residual; against the exact model the `x = 13` residual is `+0.00291` (`varE-theta2-step.md` section 6, `redteam-0828-varE.md` section 6)" | "residuals ≤ 0.002 at nine exact points from x = 13 against the Monte-Carlo model (+0.0029 at x = 13 against the exact model)" in all three places | ASK | | R5 | `TODO.md`:470-472 (A) "floors 36@11/115@13 at K = 0 reaching truth 45/307 at K = 8/28" | `attack-0830-anchored-ladder-17.md`:37-40 "The brief states truth 'at K = 8 and K = 28' at @11/@13; that is the ascending-pool depth. The exact minimal pool is 4 and 21 (PROVEN lower bound by sole killers, VERIFIED upper bound by the greedy pool)" | "reaching truth 45/307 at ascending K = 8/28 (minimal pools 4/21)" | ASK | | R6 | `TODO.md`:706-707 (item 8) "0.9044 against a model 0.9586 at q = 4001 full depth, counts ~10³" | `attack-0830-buchstab-deep.md`:268-269 "on counts of 216 to 557" (its own §0 item 7 says "counts of order 10³"; the table :259-266 is the record) | "counts of 216 to 557" | ASK | HELD wording checked and present: every 0829n/0830 note the live layer cites in `TODO.md` carries HELD or "no adversarial pass" at the cite (Z2 :217, :227-231, :241; Z4 :296, :306; D :436, :455; A :484; 0 :527, :544, :548; 1d :603; 8 :702; 9 :730), `README.md`:206 ("All five records are HELD"), `REFUTED.md`:94-98 ("HELD, no adversarial pass yet" on four of the five rows; row 96, the door, carries none and its record is likewise unreviewed: ASK whether to add the tag). `G2-STATE.md` §0 cites none of the eighteen notes. ## 5. ORPHANS (ledger id missing from a TODO Ledger line; consequence not applied) Ledger ids: all eighteen landed ids and the four in-flight ids that already carry blocks sit on their items' `Ledger:` lines (`TODO.md`:249 Z2, :323 Z4, :334, :391 Z5, :467 D, :496 A, :562 0, :614 1d, :621 1, :711 8, :753 9; `Q-recon-0830-rec-killrun` on both 0 and D; `Q-skeleton-decide-0830` under "4 (retired)", six blocks flipped as `CHANGELOG.md`:28 says, grep 09:10). None missing. | # | consequence | where it was owed | state at read | replacement | grade | |---|---|---|---|---|---| | O1 | `CHANGELOG.md`:37-39 "Also caught: `prop-exact-fold-L.md` self-grades PROPOSAL while `G2-STATE.md` §0 lists it under PROVEN (see below for the resolution)." | the same entry, below | no resolution appears in the entry (:40-77); `G2-STATE.md`:57-61 unchanged | Apply D1 and replace the parenthetical with "(resolved: the bullet moved out of §0's PROVEN list, `coherence-0830.md` D1)" | APPLY | | O2 | `attack-0830-record-mechanism.md` verdict | `TODO.md` Z5 body | only the ledger id added (:391); body contradicts the note (D7) | D7 | APPLY | | O3 | `attack-0830-varE-identification.md` "one open step", the n ≤ 2L band | `G2-STATE.md`:137, :142 | not applied (D8) | D8 | APPLY | | O4 | `decide-0830-skeleton-door.md`:144-146 edit list (ii) | `anchored-calm.md`:46, :61; `README.md`:141 | :40 and :63-66 applied; :46, :61 and README:141 missed (D3, D4) | D3, D4 | APPLY | | O5 | the 0830 morning wave (thirteen notes) in `README.md` §Status | `README.md`:68 is dated 2026-08-29 and :198-206 carries the evening wave; nothing for the morning wave | absent | One paragraph after :206, in the evening paragraph's register: item 4 demoted, D's fold composition and 8(a) closed as routes, REC's cheapest point does not close and carries an adverse HELD floor, Z2's fourth decade blind 64/12, Z4's ensemble halves derived and anchored halves HL-strength, Z5's residual located in the measured gap law at height, δ's sign unreadable at reach 79, the `u_sup` convention adjudicated; all HELD; no exponent moved. Also update the heading date. | ASK | | O6 | `recon-0830-rec-killrun.md`:343-351 "Rows this note would add to `SEARCH-CONVENTIONS.md` §1, not applied under the fence" (the `Y(x)` / third-sieving-stage row; the oscillation-theorems row) | `research/SEARCH-CONVENTIONS.md` §1 | not present (grep 09:11 for "sieve out the whole", "third sieving", "oscillation theorems": none) | add the two rows as the note words them, both with the direction warning | ASK | | O7 | `attack-0830-rec-cheapest.md`:419-422 candidate `REFUTED.md` row "HELD until the pass" | `REFUTED.md` | correctly NOT added; `TODO.md`:548-556 carries it as "NEW AND ADVERSE, HELD, NOT ADVERSARIALLY CHECKED" | none; recorded so the row is not added by mistake before the red team | none | | O8 | `attack-0830-varE-identification.md`:376-398 five-file correction list; `attack-0830-tail-derivation.md`:338-344 the "R + 1/2" and "7.5"/"0.5" corrections; `attack-0830-head-remainder.md`:303-326 the pooling corrections | the named records | correctly HELD pending `verify-0830-record-defects.md` (in flight), as `TODO.md`:317-318 and `CHANGELOG.md`:61-63, :64-68 say | none | none | ## 6. Checked and found consistent Read sentence against record and found to agree (numbers, rungs, quantifiers): `TODO.md` Z2 :205-229 against `attack-0829n-X-upper.md` (7.82 → 10.13, 1.78 → 2.53 × 4.39 → 4.01, s_max 1.81 → 1.96, flat 378100.0 from K = 17 to 46, 5.18 → 5.55×, 0.8415 → 0.9591) and `attack-0829n-parity-dstar.md` (1.097 ± 0.012, null within 0.025); Z2 :230-240 against `blind-0830-quadpoint-31607.md` except D12, D13; Z4 :296-322 against `attack-0830-head-remainder.md` and `attack-0830-tail-derivation.md` (R + 5/2, R + 3, 0.6007, 5.600 → 6.034 against 6.05, 0.6306 → 0.7532, t/R 1.0228, the six half-decade windows within 1.6 se, the four owed records), and the tail note's proposed §8 text is applied nearly verbatim; D :435-467 against `attack-0829n-doubling-bridge.md` (K*(16) = 17, 18 against 8, ≥ 14 at 64, ≥ 24 at 128, sup 6.6364, slack 1.00 to 1.33, Cesàro) and `attack-0830-doubling-killrun.md` (product, 2^N, rungs 16/32/64, sup w/a 0.8295, the PROPOSAL rider); A :482-492 against `attack-0830-anchored-ladder-17.md` (K* = 27, 1732 of 1739, 1543 = 0.887, the five-level table 0/0/2/10/27 and 0.40/0.62/0.73/0.81/0.89) except R5; 0 :527-556 against `attack-0829n-rml-proof.md` (six arrows, one open, 2s + o(1), 5.2974, margins 2.8e2 to 6.3e2, slope 2.766 ± 0.212 at 1.1 and 5.8 se) and `attack-0830-rec-cheapest.md` (1.030992 + 3η, 2.1332 − η/2, z/(2.13 ln z), 16s/9 = 4.7088, 0 sign violations to z = 73, 0.012 → 0.35 over 601 → 5e5) and its rider 7 ("Lemma V re-enters as ASK not import; fixed-smooth-profile binds", TODO :505-506); 1d :569-610 against `attack-0829n-hsubpow-K.md` (zone [1.3946, 11.3568), base 82 for K ≥ 3, 5.607, the sign lemma, 10.35 and 0.39 nats, 10 of 81 at K = 11) and `measure-0830-delta-reader.md` (four chain readers killed, RD −0.38 ± 0.14) except D16; 8 :701-707 against `attack-0830-buchstab-deep.md` (F₂/f₂ = 3.668, negative floor at every K, 0.007 to q = 1009, 0.9044 against 0.9586) except R6; 9 :730-752 against `attack-0830-varE-identification.md` (one statement, Kloosterman fractions, −0.49 ln y, 82% below 2L, ~18%, n ≤ L·ln^{2+o(1)} y, y^{4/5}, the X2 defect 2.3×, five-file list HELD) except D14 and D8's :722. `REFUTED.md`:94-98, the five new rows, against their notes' own §0 verdicts and the notes' proposed wording (the door row is the decide note's :129 verbatim; the parity row's "5 HIT 1 MISS" matches :260; the killrun row's "7 of 14", "20.83×", "0.8295" match :39-42, :154, :183-185). `REFUTED.md`:42 "Opera de Cribro 6.18": consistent with `recon-0830-rec-killrun.md`:378-381, which says the number is a result and not a section and that "the corpus's row writes it without the sign"; the brief to this pass misread the note (no edit; optional "result 6.18" for clarity). `CHANGELOG.md` 0830 entry :22-77 against the thirteen notes: consistent except O1 and D16. `CHANGELOG.md` 0829-evening entry :91-144 against the five notes and `verify-0830-usup-convention.md`: consistent; the rml note's :241-256 and :440-443 and `lemmaV-sup-extension.md`:500 carry the adjudicated wording. `GLOSSARY.md`:372-384 (skeleton) carries the full-prime shares and the closed door; `natal-cap-36-skeleton-door.md`:37 carries the corrected 90.8% to 100.9%; `fekete-1d.md`:170 carries "with δ ≥ 0". `QUESTIONS.md` §1/§2 rows for the eighteen ids match the notes' blocks (generated; nothing to fix). `anchored-calm.md`:8, :40, :63-69 consistent with the decide note. Two things this pass could not settle and did not try to: whether `session-0829-summary.md`:119 "27 SOUND-NARROWER" or `applied-0829-refuted.md`:279 "30 SOUND-NARROWER, 70 verdicts over 67 rows" is the count (both staging, neither live; D6 rests on the clause count both give as 25); and the in-flight notes' consequences (S2, O8), which wait for their landing. ## 7. qc.js --full Run twice, 09:14-09:18 and 09:18-09:23 (the second captured to the scratchpad). Result both times: `FULL GATE FAILED: qc checks`, with every section clean except LEDGER, which reports three `ledger-todo-unlisted` findings, all on notes that landed during the run and are not among the eighteen read here: `TODO.md`:727 item 9 does not list `Q-recon-0830-smooth-aps` (OPEN) and `Q-verify-record-defects-0830` (ANSWERED); `TODO.md`:251 item Z4 does not list `Q-verify-record-defects-0830`. Those are the orchestrator's pending `Ledger:` additions for two of the in-flight notes, not defects in this pass; no finding names `coherence-0830.md` or `Q-coherence-0830`. `audit-numbers.js` 251/251 checks passed; `selftest.js` fires on every planted defect; refs, quotes, crosslinks, scripts, transfers, calibration, absence, sourcing, embeds, widths, provenance and search-convention all clean. Once the two ids are added to items 9 and Z4 the gate should be green; not re-run here.