v0e claims/ Extension — Decision-Session Kickoff (OPS-019)

Ledgered 2026-07-28 as OPS-019 (NEXT.md). Provenance: schema-changelog v0d.33/v0d.34 (PR #119) — the v0d.34 measurement established the empirical case. To start the session, paste:

Read wiki/.audit/v0e-claims-extension-kickoff.md and run the decision session it describes.

The prompt

Task. Decide, with the user, whether and how to build the deferred claims/ + type: claim extension — dedicated pages for claim entries that have outgrown the register's row format. This is a decision session, not a build session: the deliverable is a go/no-go plus a written design, and at most a one-entry pilot. Nothing migrates in batch here.

Load first.

  1. CLAUDE.md §Claims Register Format — v0d.34 rewrote the opening paragraph: the outgrow test is functional (row-function), the numeric marker is retired, and dedicated pages remain forbidden under v0d.
  2. wiki/schema-changelog.md v0d.33 + v0d.34 — why the marker fell.
  3. Re-run the measurement fresh (the register moves); per-entry words across Evidence + Counterpressure/Limits + Payoff, distribution + top tail. At 2026-07-28 it read: 303 entries; 200 over the old 300-word marker; median 449; 20 entries above 2,000 words, 12 of them above 3,000; top eight 3,098–4,002 (pbp-as-pivot-from-language-monograph-to-late- ontology 4,002; language-as-missing-case-of-lateral-universal 3,956; deformation-coherente-mp-coinage 3,394; ontologie-de-la-chair-misnomer 3,354; phenomenal-invariants-replace-resemblance-and-denotation 3,241; bukharin-as-mp-exemplar-phenomenological-political-ethics 3,170; malraux-metamorphosis-precedes-mp-advent 3,153; sedimentation-pluralizes-by-practice-type 3,098). The fattest are the most-juried — growth is process-mandated (refuter objections written into Counterpressure; contested expansions; Status History accumulation).

Resolve with the user, in order:

  1. Go/no-go and threshold. Does the tail fail the row-function test badly enough to justify a new page type? Name the alternatives honestly: leave as-is (selective consumption still works — Summary table + grep); or trim the tail by routing evidence depth downward (v0d.34's overflow route: extraction notes, traceably-anchored .audit/ artifacts) with no new type.
  2. The link contract — the load-bearing risk. Inbound [[claims#slug]] anchors resolve to the H2 in claims.md; the slug-is-canonical promise and the slug-uniqueness invariant depend on it. Options: (a) an extracted entry leaves a permanent H2 stub in the register (anchor survives; stub carries lifecycle header + pointer to the page); (b) repo-wide link migration (heavy churn, dead-link lint, breaks the promise). Recommend (a) unless the user argues otherwise.
  3. What stays in the register. Working proposal to test: the register remains the single source of lifecycle truth — Status, Confidence, Status History (the v0d.29 lifecycle grammar and the Atlas observatory replay contract read only claims.md) stay in the stub; the page carries Claim/Evidence/Counterpressure/Payoff depth. If Status History were to move to pages instead, the Atlas contract and claims_lifecycle_backfill.py both need extension — check both before deciding.
  4. Machinery inventory (do this before writing the design). Grep scripts/, .claude/, and CI for claims.md and claims# and produce a per-consumer disposition table: normalize_claims_table.py (Summary + index generation), consolidation_pregate.py, claim_retirement_queue.py, lint-mechanical.py (register integrity, Claim-type vocabulary, lifecycle syntax), refresh_baselines.py pins, the Paper-A fixtures, and the write paths of audit Harvest, weave, consolidate, and speculate-signoff ratification. Every consumer gets an explicit answer: reads stub, reads page, or extended.
  5. Page format. type: claim frontmatter (which fields mirror the entry header; how entry Confidence: maps to page confidence:); subfolder wiki/claims/; whether claim pages count in Total pages or separately like claim entries (decide explicitly — index counts line and baselines move either way); tags and wikilink conventions.
  6. Migration discipline. The v0d rule says bodies are append-only and never relocated; extraction-to-page is a relocation and must be defined as a new sanctioned operation — with its own Status History line (decide whether the lifecycle grammar needs a non-transition event form for it), one commit per migrated entry, Rule 19 before each.
  7. Versioning. Lifting "do not create dedicated claim pages under v0d" is the v0d/v0e boundary as written. Decide v0e.0 vs. a continued v0d.NN; either way the prohibition text, the register-format paragraph, and the changelog move in the same commit (OS-parity discipline).
  8. Pilot. If go: migrate exactly one entry end-to-end — recommend pbp-as-pivot-from-language-monograph-to-late-ontology (the 4,002-word max, which also carries a queued revision set; check OPS-015's notes) — run every gate, and stop for user review before any batch.

Constraints that bind regardless: Rule 21 (quotation integrity — pages quote verbatim-or-point exactly as entries do), Rule 19 (baselines lockstep), Rule 20 (single writer; branch + PR), slug canonicity, the append-only log, and the v0d.24 promotion routing — a migration never changes a status.

Deliverable. A schema-changelog entry draft (the design), the machinery-disposition table, and — only on the user's go — the one-entry pilot PR.