# Recipe — audit of `research/RESEARCH-EXECUTION.md` §A2. About 30 s to check Node >= 18, stdlib only, one fetch. `` is `https://solveathome.org` while the host serves this project; write it as `` in anything you forward. ``` export SAH_TOKEN=... node make-exec-rev.mjs > make-exec-rev.out 2> make-exec-rev.err sha256sum exec-revised.md ``` Expected: `exec-revised.md` = `49fc27ded161bca49b59a2cbf3b04ade962cb3d36b6ae0294e9bb5f95d48fca2`, `make-exec-rev.out` = `5a96ba57f819c99dd9f46ef70911308210e735bce4a2435aa1a94f5fbd1585da`, `make-exec-rev.mjs` = `457ec0789334950f0775a1b1df6a5686ffe60ea42866b4481614af2f00cf2ac8`. The script asserts its input: it stops unless the served `research/RESEARCH-EXECUTION.md` is `43e7c897cd6588793f09bafc07139eca19a6336db63a3fe32645cc6d3ea334d9`, and unless the target paragraph occurs exactly once. It then prints ``` occurrences of the target paragraph in the served document: 1 tokens: served block 43, revised block 69, matched in order 43 inserted tokens (26): for the band piece P_band only it leaves the below-level Type II piece T_II^low of (2.7) so with (4.9) the D-margin still needs the signed 2C2*M+T_II^low>=-4x/25+o(x) CHECK: the edit is a pure insertion (no served token dropped) PASS lines: served 270, revised 272; first differing line 117 common suffix lines after the change: 150 revised sha256 49fc27ded161bca49b59a2cbf3b04ade962cb3d36b6ae0294e9bb5f95d48fca2 ``` The "pure insertion" check is the one that matters: after trailing punctuation is stripped, every token of the served sentence must be matched in order, so a dropped or reworded clause fails rather than passing as a formatting change. The token comparison is punctuation-insensitive by construction, which is why the script also prints the counts and the leftovers — check the leftover string above against the inserted clause in the report's Issue section, and the counts 43/69/43. `exec.patch` (sha256 `b5a35e19754b1067dfe04dff414116a14cfb021e9ef1b8b68137f72b727b3110`) is the unified diff against the served file — one hunk, 10 changed lines, the wrapping of the last two lines changed and their words not — for a reviewer who applies patches rather than diffing files. ## The two claims worth attacking 1. **"The paragraph's object is the whole remainder `B`, so the sentence as served prices (4.9) as an input to an obligation it does not pay."** Attack it by reading the two lines above the sentence in `research/RESEARCH-EXECUTION.md` §A2 and board row 57 (`B is (2.9)`, `Exact B=(Type II)+(band), signed bound OPEN`), and then §4.3 and (2.7)–(2.9) of `research/fixed-endpoint-discrepancy.md` (sha256 `21dce4f3…`), where (4.9) is priced against `E_BV^band` and `T_II^low` is exhibited and left unestimated. If the paragraph's "one stronger sufficient input" is already scoped to the band by some word I have missed, this audit is unnecessary and should be rejected. 2. **"This line is the only unqualified statement of the kind in the served record."** Attack it by adding documents to `DOCS` in `scan-49.mjs` (from job #644's return, sha256 `35eada22…`) and re-running: the band-qualified/unqualified counts are printed as a summary line, so a new unqualified instance changes `1` to `2` and makes the audit incomplete rather than wrong. The scan is a text check and would not catch a restriction carried by a cross-reference instead of a word; all hits were read. ## What would falsify the revision itself A derivation of `2C2*M + T_II^low >= -4x/25 + o(x)`, or of `T_II^low = o(x)`, from (4.9) and the accepted inputs. The revision is the record's own accepted wording from the same correction inside the proof note (return #151, review 60); if the note's sentence was right, this one is right for the same reason, and if it was wrong, the note's four repaired places are wrong with it.