This change only alters presentation. It touches nothing under content/ and does not edit AGENTS.md, the citation tool found no added identifiers, and no schema, CI check or gate logic is loosened. The governance.ts edit only adds a tie-break to the sort order, and panel.test.ts now points at the moved verdict page rather than dropping its assertion. The labels stay honest. StandingPanel still reads "an AI-generated assessment · not reviewed by a human", standingInWords names only what models did, and a test pins that the standing glosses never mention a human. A seat that judged an earlier version is now labelled "judged an earlier version" instead of being shown as concurring. Nothing is removed from the record: the rationale, every run and the full changelog move to /cases/<slug>/record/, each gate verdict gets its own page with every seat's reasoning open, and the new seat-tally table adds to inspectability under §3.15. The change also removes false or stale statements: the retired "until a named human endorses them" promise, the dead /panel links and the retired literature watch. The DECISIONS entry records the founder's direction with its date and channel, as the 2026-09-09 amendment requires. Two small imperfections do not break a rule: the claim-by-claim split summary is hidden whenever any seat is stale, and the method page lists unresolved and presently_untestable in a line although the constitution puts them on one rung. Both are worth tidying later.

The change does not edit AGENTS.md, alter content records or assessments, add citations, or weaken the publication gate. It relocates rather than suppresses provenance: complete run records, refusals, changelogs, and seat-by-seat gate reasoning remain publicly accessible on dedicated pages, while stale panel checks are explicitly identified as judging an earlier version. Verdict-move summaries are derived from successive adopted assessments and preserve the complete edition history elsewhere, including reversals, so append-only assessment history is not replaced. AI authorship, ratification standing, lack of human review, and disagreement remain visibly labeled, and the decisions entry stamps the founder direction with its date and session channel.

The change is an editorial and presentation refactoring that splits the verbose operational/run history onto dedicated subpages (`/cases/[slug]/record/` and `/operations/gate/[pr]/`) while keeping direct links and recent history visible on the primary case page. No content files under `content/` or citations are altered, and `AGENTS.md` is untouched. The new UI components and domain derivations accurately preserve epistemic labels, explicitly clarifying that ratifications are independent multi-model AI checks rather than human reviews in accordance with §4, §7, and §3.15.

The diff is presentation-only: UI routes, components, CSS contrast, meta/sitemap/feed, and pure derivations (moves, lede, glosses, gate tallies). It does not edit AGENTS.md, content ledgers, assessments, editions, or protocols, and adds no citations. AI assessments stay labeled as AI; standing copy and tests refuse human-review language; the home/method pages drop retired human-endorsement and /panel claims rather than inventing provenance. Record/run detail moves to /cases/<slug>/record/ and gate pages without hiding disagreement or weakening §3.15 checks. Omitted-files and citation-verification sections are empty, so nothing material is hidden from review.

This is a presentation-layer change: no content/ records, assessments, editions or citations are touched (citation verification found no added identifiers, and no run records are carried), and AGENTS.md is unedited. The UI additions are epistemically honest — the Lede component keeps the whole text in the page behind a disclosure (tested: lede plus rest equals the whole text), verdict moves are derived from the append-only editions/assessments rather than a new editable record, and the full changelog and run history remain public on the new /record/ and /operations/gate/ pages, so nothing is hidden or weakened; §3.14's visible revision histories are preserved, with per-day netting on the home feed balanced by the complete per-edition history one link away. The new glosses and standing wording are checked by tests to never claim human review, consistent with §4's labeling rules, and the method page's corrections remove retired promises (human endorsement, /panel) rather than adding misattributions. The seat tally and its commentary in the operations page derives mechanically from harvested arbiter records with tests, and the DECISIONS entry records the founder's 2026-09-29 direction with date and channel as §3.15 permits. Nothing here fabricates, exposes confidence material, edits the constitution, or loosens a check.