# Evidence for job 3734: `research/attack-prior-art-last-ground.js` (#143, #144, #2578, #2579, #2580) | what | sha256 | bytes | |---|---|---| | served text (base) | `87d9d6b693f6cce5153885ff743be1a3b91a59fafd297e17c057c560df4a08e0` | 60807 | | pre-embed text (all code edits, old block) | `fe4776d11be933dce8fef8c19794e546efba2f00f826bdb80b08513939b4eeac` | 60973 | | **revised text (this return, re-embedded once)** | `a1f52aaf015db57937c121268a075fab451ca3d31b1bccb09e7212f80e457d20` | 61520 | | **the quoted 2026-08-18 block** (filed with #1674, re-attached here) | `f89e053c492e9b8237fc79459acf61d0fbcc349be1ac3ed0fecc8a4b62aaf42c` | 27910 | ## The one embedding run (performed, not described) node research/qc/embed.js --timeout 480 --volatile "live-API crawl: zbMATH, OpenAlex, Crossref, MathSciNet, Open Library, Google Books" research/attack-prior-art-last-ground.js 487 lines of output, 130.5 s code-sha256 cffcf5ac85113a99... out-sha256 8ed7ba4bb4b9d4ea... (CHANGED from the previous embed) superseded 6 of 41 figures in the replaced block not reproduced (first: 269, 1965669, 187, 1015) Run in a scratch mirror (`research/qc/{embed.js,tailfmt.js}` fetched and byte-matched) **after** every code edit, so the `superseded:` line describes the *served* 2026-08-18 block — the value #2578 predicts verbatim. No second embed: an intermediate one would have overwritten that record. ## Verification (`--check` on the revised file) volatile live-API crawl: zbMATH, OpenAlex, Crossref, MathSciNet, Open Library, Google Books (observed 2026-09-25; a re-run will not reproduce out-sha256) code-sha256 matches body matches out-sha256 -- the pasted block is bit-honest out-sha256 not checked by re-running: this tail is volatile ## Per finding | finding | change | |---|---| | **#144** | the progress line is now `console.error(...)`, so it leaves stdout. `embed.js` records stdout only (default `--streams`), so the artifact no longer carries `network calls: N elapsed: T s`. | | **#143** | point 3's "the honest repair is this note rather than a replacement snapshot" and the `--force` sentence are gone; the block was re-embedded with `--volatile`, never `--force`. | | **#143** | the stale-`code-sha256` point 2 — and the `qc.js embeds` reports `tail-does-not-belong-to-this-code` remark — are gone, because the re-embed refreshed it. | | **#2580** | point 2 is one past-tense statement: the 2026-08-18 block is quoted at the file above, the block was re-embedded on 2026-09-25 with `--volatile`, the command is given with `--timeout 480`, and `--force` is excluded. The duplicated "and quote the 2026-08-18 figures..." clause and the untrue "the guard refuses a re-embed whose block changes" claim are dropped. Points renumbered 1–3. | | **#2578** | all code edits first, then **one** embed from the served block (above); the banner names the quoted block and its file. `--timeout 480` is in the recorded command because this crawl is ~130 s, over embed.js's 120 s default. | | **#2579** | READINGS reading 7 now reads "returned 187 on 2026-08-18 -- a live count, 19 in the 2026-09-25 block below -- and included Ziller-Morack". Made **after** the embed: the writer preserves everything from the READINGS banner onward, so this needs no second embed. | ## The quoted 2026-08-18 figures The 2026-08-18 block is quoted verbatim in `reports-attack-prior-art-last-ground-block-2026-08-18.txt` (sha256 `f89e053c492e9b8237fc79459acf61d0fbcc349be1ac3ed0fecc8a4b62aaf42c`, filed with return #1674) and that file is attached to this return as well, so the citation in the banner resolves from either. Its header matches the served block exactly (`code-sha256 86b326e5...`, `out-sha256 f85b022f...`, `embedded 2026-08-18`, `elapsed 123.5 s`). ## Unresolved obligations 1. The re-embed is **live data**: `out-sha256` is unreachable by construction and the block is a snapshot, now of 2026-09-25. `body-lines` 487. 2. Findings **#144** and the rest of #2578/#2579/#2580 close only on acceptance of this revision; #190-style remainder findings on this path are not touched here. 3. The pre-embed and revised texts are both in this run's `work/` (`file.pre-embed.js`, the mirror). 4. 16 of @Benjaminsen's returns wait for a verdict; nothing for the person to do.