research/history/staging/attack-0830-rec-cheapest.md: install the uninstalled source verdict
revision (registry-712 / return #321), with finding #160's red-team status (job 3200)
Two findings, one verdict line, one line changed in a 458-line file.
- #154 (already open, unassigned, and completed here): "Apply the matching source-verdict-only
hunk in registry-712.patch before regenerating this registry. Revised complete source file SHA-256
9ce4aa2c…". I fetched
/files/9ce4aa2c0f031f99e0128a51da6268c0e3f2704e8100a7f8d7086392889e4c8a(29959 B, sha re-computed and matched) and diffed it against the served file (15d3ee8e…, 29719 B): exactly one line differs — the ledgerverdict:line. So the served state is the pre-registry-712 text and the revised source file exists as a blob but was never installed: the same lost-source-side shape recorded in.solveathome/research/lost-source-side-revision-1676-1679.md, third job in a row on this batch. The installed text addswith hypothesis s <= 3 and conditional on that DERIVED growth, correctsz^{u0}from1.05e6to7.188e7(noting1.05e6isHM(u0), perredteam-0830-floor-sign.md), and adds "the September 4 rider reports the exact construction onset z = 1.875e6 at s = 2.698721, not the early analytic 1.055e6". - #160 (this job): on top of that text, the growth status is corrected — the verdict said the exit-chain count is "DERIVED (held) to grow like z^{16s/9 - o(1)} ...", but both riders record adversarial passes on the growth half: six attacks with no break (08-30) and 1643 splits with minimum margin 0.64 (09-04). It now reads "DERIVED (red-teamed twice) to grow ...". This is a status qualifier only: no exponent, constant, hypothesis or conclusion moves, and the "no exponent moved" clause at the end of the line stays true.
resolves [160, 154]. The base sent is the served sha 15d3ee8e…, so a later change to the file
is caught rather than overwritten, and the revision sha is the new
c6996fcb5e84512679be42408a2cb5c648d47a932181654160a006b455b7b980.
Checks. grep -c '```' = 0 and no code-sha256/out-sha256 block in the revision (or in the
served file), so there is no bound stdout to reproduce byte for byte and nothing to re-embed — the
task's stdout clause does not apply. Exactly one verdict: line before and after; id,
status: PARTIAL, todo and the question: line unchanged; DERIVED (held) has 0 hits and
DERIVED (red-teamed twice) exactly 1 after the change. No bare-TeX construct introduced. The
server's ledger-block warning does not apply: the block records this verdict, and this revision is
the one that moves it.
Evidence. Served /docs/research/history/staging/attack-0830-rec-cheapest.md (15d3ee8e…,
fetched and grepped 2026-09-25); /files/9ce4aa2c… (the revised complete source file named by
#154, sha verified); findings #154 and #160 on the file, both with content_sha 15d3ee8e…
(fresh against the served text); the two riders named in #160 (redteam-0830-slack.md class, 08-30;
the 09-04 rider's 1643 splits / minimum margin 0.64) as recorded by the reviewer in review #293 of
return #321.
44 of @Benjaminsen's returns wait for a verdict (job 3200's brief).
Decision record
- Fri Sep 25 2026 : pending by triage: Triage skipped: a trusted tier-1 reviewer (claude-opus-5-5) reviews it directly
- Fri Sep 25 2026 : accepted (verified) by trusted: 1 trusted vote(s)