# TODO.md pruned to its charter after the 2026-08-27/28 wave: five items retired, nine restated, no calibration changed *(2026-08-28. Staging note. Fence: `TODO.md` was the only file edited, this file is the only file created, no git command ran, `research/qc.js` was not run. Every number carried into `TODO.md` came from the note that owns it and kept its rung; nothing was recomputed and nothing was promoted. The `(retired)` edits in §3 are owed by the primary agent, not by this pass, and until they are made `node research/qc.js ledger` will fail on eleven notes.)* ## 0. What this pass did, and what it did not It applied the file's own charter (Chris, 2026-08-14: an attempted item leaves the file, done, refuted or absorbed) against `research/QUESTIONS.md` §1 as it stood after the 2026-08-28 wave. It did not adjudicate any mathematics, did not change any calibration, did not move a number between rungs and did not touch a `Ledger:` line of any item that stayed. Where an executed first move had a verdict, the verdict replaced the move at the rung its note gives it. Length: 552 lines before, 591 after. The file grew because five retirements freed less space than the executed verdicts took, and the verdicts are the thing that stops a question being re-run. ## 1. Retired: five items left the file | item | why it left | where its outcome lives now | |---|---|---| | **11** (three NOW-provable BV prizes) | all three prizes are written as theorems and all three are EMPTY in every computable range: S1/S2 PROVEN at short-note grade once the freshness factor is one class per prime, cap_K PROVEN for q_K = W^{o(1)} needing s ≥ 10.82, first cleared at x = 263. Nothing is left to try. | `thm-mod30-tail.md`, `thm-capK-bv.md`; the empty-range readings are carried into item 8's restatement | | **10** (pin the excess chain's c → 1.074) | the derivation is REFUTED as the mechanism and the value is not; the item's own remaining move read "a mechanism that is not a tail compression, or nothing", which is not a concrete first move. c = 1.05 ± 0.06 stands MEASURED on x = 17..31. | `excess-chain-c.md` | | **1c** (why does c₂′ drift) | executed and ABSORBED, not closed: the 22-term refit is done, the question is PARTIAL, and the only remaining move needs 47#, which the item itself already parked to paper phase. Moved verbatim in substance to the Parked section. | `c2prime-refit-22.md`, `attack-c2drift-01.md`; TODO Parked, first bullet | | **1e** (the maximal law's two objects) | the item's own body said "ANSWERED and RETIRED … this item leaves the file on its next prune". This is that prune. No note names it, so it costs no `(retired)` edit. | `attack-wrongdirection-audit.md` §3.6; `REFUTED.md` row 27 | | **Z5b** (the fold ledger) | its stated open question ("whether any structural property of these columns is FORCED rather than expected") is ANSWERED with fourteen forced constraints, and the item's own last sentence prices what remains as the postulate. ABSORBED into Z2 as one paragraph, at the same rungs, so the object keeps a pointer in the live file. | `fold-ledger-forced.md`; `research/fold-ledger-01/-02-derive/-03-iterate.js`; TODO Z2, the "THE FOLD LEDGER IS THE SAME WALL STATED PER FOLD" paragraph | ## 2. Restated: eleven items stayed and were rewritten on what remains - **Z2** — three changes. The Evans read is recorded as EXECUTED with its verdict (`lit-evans.md`, HELD, and flagged as carrying no ledger block, so `QUESTIONS.md` cannot see it): no reach to y ≈ x^0.28, no sieve-depth coordinate, h averaged with h = 2 inside the permitted exceptional set, no secondary term, no product-form correction available. The depth-axis block is relabelled from a first move to an answer, since the item already carried "ANSWERED as to machinery". The fold ledger arrives from Z5b. What is left as a move is the blind test alone, and it is labelled as the only one. - **Z4** — the head half is recorded as RAN and PARTIAL, with the mod-30 mechanism the old first move named answered NO (class term −0.0001), the exact decomposition, β·CV² controlling rate not limit, and the HL 5.3% pricing. The tail first move is untouched and is now marked as such. - **Z5** — restated on Kourbatov's b, per the brief. The null-side move is recorded as run (survives at z = −3.3, this item's own premises on "location" and n_eff were wrong), the deficit is named as Kourbatov's published b = 1.2597, and the new first move is to pin b, since the two in-house estimators disagree by 15% (1.125 from A against 1.298 from mean z). - **Z6** — the executed 76–82 probe is compressed to a scored pointer; the SEC C2 replication survives as the item. - **Z7** — (3) is now the TAIL field data specifically, Z4's remaining half; (4) β(37) names item X as its consumer and the @37 census that killed all seven registered offset candidates. - **Z0** — reduced to the standing cadence rule plus "currently owed: nothing". The discharged seven-note list is gone. The prereg custody residual is kept and relabelled a standing exposure, since it is not discharged by any pass. - **A** — the stale cross-reference "the zone-side transplant (Z1–Z3)" now reads Z2 alone, with Z1 executed and Z3 retired (its QR refinement CLOSED, `REFUTED.md`). This was the file's only surviving Z3 dependency. - **0** — the ⟨ρ²⟩ first move is recorded as executed before it was posed (proven 2026-08-21, constant 170.88 z^{2s} ln⁸z, and Chebyshev off any second moment diverges from RML, so an ingredient and not a route), and the 2026-08-28 recon's six no-route findings are named with their files. The exact-Λ data point at z = 31/37 is the one move left. - **1d** — the legal zone is CORRECTED to the trusted `K ∈ [1.3946, 11.3568)`, with the old `[1.3555, 7.6394)` named as mis-stated and the derived "6.28 nats" reading dropped with it. The explicit-K route is recorded CLOSED by three mechanisms. What is left is (H-sub-pow) by a fourth mechanism, and nothing names one. - **4** — row 15's banked THEOREM is recorded VOID, not overpriced (Saffari-Vaughan Thm 10 reads x^{6/11+ε} < y and fails at fixed M; its object is a marginal, Ψ_M is joint), covered branches ≤ 8% of the door's mass, no route, no cap-36 line upgraded. - **8** — retitled to the deep-ladder transfer. The shallow half is recorded PROVEN and EMPTY (s ≥ 22.06 against 4.7 @23 and 17.1 @97), the sharp sieve functions CLOSED at finite level, and the old part (c) ANSWERED as the trivial per-block bound with its slack figures. Remaining: the dimension-2 deep-ladder transfer and the untouched prime-comb equidistribution. Untouched, no edit: **D**, **1**, **WHERE TO AIM**, **0b**, **0c**, **X**, **6**, **2**, **9** (already restated on its two open steps on 2026-08-28), **CUSTODY**, the reading queue, the standing compute note, **12**, **13**, the Operational block, and the rest of Parked. The THE TARGET spine and both charter paragraphs are verbatim. ## 3. The `(retired)` edits the gate now needs (owed, NOT applied here) Eleven notes carry a `todo:` line naming a retired item. Each must be rewritten in the form the corpus already uses (`attack-z3-immune-01.md`, `stretch-01.md` both read `todo: Z3 (retired)`). No other part of any note changes. | note | `todo:` now | `todo:` owed | |---|---|---| | `thm-mod30-tail.md` | `11` | `11 (retired)` | | `thm-capK-bv.md` | `11` | `11 (retired)` | | `redteam-0828-engine.md` | `11, 8` | `11 (retired), 8` | | `applied-0828-engine.md` | `11, 8` | `11 (retired), 8` | | `applied-0828-live.md` | `11, 8, Z4` | `11 (retired), 8, Z4` | | `excess-chain-c.md` | `10` | `10 (retired)` | | `attack-c2drift-01.md` | `1c` | `1c (retired)` | | `c2prime-refit-22.md` | `1c` | `1c (retired)` | | `fold-ledger-forced.md` | `Z5b` | `Z5b (retired)` | | `redteam-0828-closures.md` | `10, 1c, 1d, Z5b, 0` | `10 (retired), 1c (retired), 1d, Z5b (retired), 0` | | `applied-0828-closures.md` | `10, 1c, 1d, Z5b, 0` | `10 (retired), 1c (retired), 1d, Z5b (retired), 0` | Ledger ids retired with their items, for the record: **11** carried `Q-mod30-tail`, `Q-capK-bv`, `Q-redteam-0828-engine`, `Q-applied-0828-engine`, `Q-applied-0828-live`; **10** carried `Q-excess-chain-c`, `Q-redteam-0828-closures`, `Q-applied-0828-closures`; **1c** carried `Q-c2prime-drift`, `Q-redteam-0828-closures`, `Q-applied-0828-closures`; **Z5b** carried `Q-ledger-forced`, `Q-redteam-0828-closures`, `Q-applied-0828-closures`; **1e** carried no `Ledger:` line and no note names it. Note that `Q-redteam-0828-closures` and `Q-applied-0828-closures` survive on items 1d and 0, which is why those two notes keep an unmarked `1d, 0`. ## 4. The moratorium block's git bullet The old bullet read "The git remote is fetch-only by three layers (commit b3c340e). Never work around any of them; changing that is a publishing decision and Chris's alone." Chris lifted the three push layers on 2026-08-28 and the repository is on GitHub with both branches at the same commit, so the bullet now records that pushing is routine and no longer a decision, and that the publication moratorium on arXiv, OEIS and the papers stands until Chris says otherwise. Nothing else in that section moved. ## 5. CHANGELOG entry text (for `research/history/CHANGELOG.md`, owed) ``` ## 2026-08-28 — TODO.md pruned to its charter after the proof-shaped wave *(Curation pass under the forward-only charter, run against `research/QUESTIONS.md` §1 as it stood at the end of the 2026-08-28 wave. No claim changed rung and no live question was dropped; what left is attempted material whose records already carry it. TODO.md went from 552 lines to 591, the growth being the executed verdicts that replaced the executed moves. Record: `research/history/staging/applied-0828-todo-prune.md`.)* - **TODO item 11 (three NOW-provable BV prizes) left the file.** All three are written as theorems and all three are EMPTY in every computable range (`thm-mod30-tail.md`, `thm-capK-bv.md`; S2's freshness factor corrected to one class per prime; cap_K needs s ≥ 10.82, first cleared at x = 263). - **TODO item 10 (pin the excess chain's c → 1.074) left the file.** The compressed-tail correction is REFUTED as the mechanism and the value is not: c = 1.05 ± 0.06 MEASURED on x = 17..31 (`excess-chain-c.md`). The item's own remaining move was "a mechanism that is not a tail compression, or nothing". - **TODO item 1c (why does c₂′ drift) left the live queue for Parked.** Executed and absorbed, not closed: the 22-term refit is done, the question is PARTIAL (`c2prime-refit-22.md`, `attack-c2drift-01.md` §3), and the only move left needs 47#, already parked to paper phase. - **TODO item 1e left the file**, as its own body instructed on 2026-08-27. - **TODO item Z5b (the fold ledger) left the file, absorbed into Z2.** Its stated open question is ANSWERED with fourteen forced constraints (`fold-ledger-forced.md`) and what remains, bounding the short-window fluctuation below the mean at every fold, IS the postulate by the item's own sentence, so it is not a separate front. The three embedded producers keep a pointer inside Z2. - **Nine items restated on what remains**, with each executed first move replaced by its verdict at the rung its note gives it: Z2 (Evans read, depth axis relabelled as answered, blind test the only move left), Z4 (head half ran, PARTIAL, mod-30 mechanism NO), Z5 (restated on Kourbatov's published b = 1.2597; first move is to pin b, the two in-house estimators differing by 15%), Z6 (76–82 scored, SEC C2 replication remains), Z7 (item 3 is the tail field, item 4 names item X), Z0 (cadence rule only, custody residual kept as a standing exposure), 0 (⟨ρ²⟩ executed 2026-08-21, six no-route recon findings named, exact Λ at z = 31/37 the one move left), 1d (legal zone CORRECTED to the trusted [1.3946, 11.3568); explicit-K route CLOSED by three mechanisms), 4 (row 15's THEOREM void), 8 (retitled to the deep-ladder transfer; shallow half proven and empty; part (c) answered as the trivial per-block bound). - **TODO item A's cross-reference corrected.** "The zone-side transplant (Z1–Z3)" now reads Z2 alone: Z1 executed 2026-08-22, Z3 retired with its QR refinement CLOSED (`REFUTED.md`). - **TODO's moratorium block, git bullet rewritten.** The three push layers were lifted by Chris on 2026-08-28 and the repository is on GitHub, both branches at the same commit; pushing is routine, and the publication moratorium on arXiv, OEIS and the papers stands until Chris says otherwise. - **Owed by this pass and not applied by it:** eleven notes whose `todo:` line names a retired item need ` (retired)`, listed in §3 of the record file. Until they are made, `node research/qc.js ledger` fails on those eleven. ``` ## 6. What would falsify this pass, and whether that check has run - **The gate has not run.** `node research/qc.js` was out of this pass's fence, so the claim "the file is charter-compliant" is unchecked mechanically, and the eleven owed `(retired)` edits are known to break the ledger check until made. Run the gate after §3 is applied, not before. - **The retirement judgments are contestable on two items.** 1c is PARTIAL, not closed, and Z5b's remaining part is real; both were retired on the charter's ground (attempted, and no legal move left that is not either parked or TPC-strength), not on the ground that their questions are settled. If Chris reads either as still live, the repair is to restore the item and strike its line from §1, and neither retirement changed a rung. - **The Evans, `u2-engine-depth` and `import-rough-anatomy` readings carry no ledger block**, so `QUESTIONS.md` cannot see them and a future brief can re-run what they answer. Z2 now says so in the file; giving them blocks is the real fix and is not this pass's to make.