dotfiles/claude/memory/session-ledger-2026-07-12.md

---
name: session-ledger-2026-07-12
description: "Practice-of-return ledger maintained by /symmetria — returns, open horizons, recalibrations, authorization moves, sub-agent dialogues, bypasses."
metadata: 
  node_type: memory
  type: feedback
  originSessionId: ae9583d2-fd25-45e4-85fc-df082fe2e951
---

# Session Ledger — 2026-07-12

## Returns
- 2026-07-12T09:10 — **Return (caught before asserting):** was about to derive the A1 acceptance criterion from scratch. Read the spec first (chamber discipline: check governed doctrine before building) → the §Tiering&Fence evidence-tier table **already RATIFIES** the per-tier method: V-DSL = "re-extract + diff (deterministic)", V-TEXT = "re-convert + prose-word-guard (deterministic; source IS ground truth)", V-SCAN = "OCR-vs-scan, no ground-truth text". The steward's reference-conversion question is answered YES-it's-doctrine; the promotion criterion is `RATIFIED-SHAPE · OPERATION-BLOCKED` = exactly A1. Reframe: A1 is *operationalize a ratified criterion*, not *invent one*. Avoided re-deriving existing doctrine (the exact drift the library exists to prevent).

## Open horizons
- 2026-07-12T09:40 — **A1 feasibility PROVEN (read-only prototype `scripts/verify_body_conservation.py`).** Ratified per-tier method works: re-convert(source) + coverage-align vs candidate. **difflib→shingle-coverage pivot** (steward challenge, well-caught): difflib O(n²) timed out on jacobs @2min; swapped to O(n) hashed-k-gram coverage (in-house patterns: Instrument B fingerprint + verbatim-guard walk; NO new dep) → jacobs 6.8s, manguel 4.3s. Seed ground-truth PASSES (legit boundary trim→REVIEW, interior 120-word deletion→FLAG). All 3 real books FLAG on REAL grep-confirmed cruft (jacobs 888 calibre_link · manguel 2856 []{#…epub…htm} anchor-spans · montaigne footnote-id residue) — each a different signature, all missed by audit_cruft. **KEY IMPLICATION:** gate is MORE sensitive than audit_cruft → large fraction of existing canon will FLAG (= the tool-log's "160 hidden-residue" finding) → big blast radius → confirms PROPOSAL-class + ties A1 to reprocess track B.
- **Residual (jurist-class, unchanged):** (1) boundary drop-vs-keep identification — endnotes/bibliography (keep) sit at the same TAIL position as index (drop); the tool SURFACES trailing runs as REVIEW, does not auto-pass. (2) interior-run threshold calibration (k=8, thr=50 tolerates heading-reformatting deltas of 6-7 words; caught on jacobs ToC). (3) added-direction needs symmetric link/anchor canonicalization so a clean-but-differently-converted candidate isn't confounded (minor tokenizer hardening).
- (superseded) earlier 08:55 A1 horizon — now concrete above. (general `body_word_conservation` verbatim gate), the open-work register's keystone. Criterion-first, not code-first. Weigh the reference-conversion framing (candidate vs plain-pandoc-of-source) before building.
- Closeable-today small items: wire mandatory-at-graduation for the footnote verify (jurist Q1b — flow must not pass `--no-verify`); register hygiene (stale "jurist nod owed" line — Q1a gave it).
- Steward-owed (parallel, not mine): place REVIEWED-55 (verified absent — REVIEWED.md stops at 54); decide the recognizer-maturity question (register A2 `[?]`).

## Confidence to recalibrate
- Carry from last session: drift-pattern **lost-the-forest-for-the-trees** in long execution arcs. Antidote = hold the open-work register (`chamber-library/docs/chamber-program-open-work.md`) as the standing forest view; stay keystone-first. Do not descend into A1 mechanism before the criterion is settled.
- Carry: **don't rank a sequence that ignores a dependency the census already flagged** (07-12 recommended frontmatter-before-reconversion; corrected).

## Tooling-scan reframe (2026-07-12, steward-driven — the pivot of the session)
- Steward challenged the whole approach TWICE ("is difflib best now scope is wider?" → "no extant epub/TEI tools?"). **Load-bearing finding:** the substrate answer was ALREADY in our register (`studium-engine-scan-register.md`, 2026-05-27 pass: Docling+Surya, OCR4all, TEI/IIIF/DTS, §10.3 MD-vs-TEI pending) — and in ~6 weeks since we hand-built ~20 bespoke conversion/verify tools instead. The two-week "every breakthrough → new obstacle" feeling = signature of reimplementing a solved domain. Kin to [[feedback-resurface-banked-notes-before-rederiving]] at pipeline scale.
- **The clean split:** substrate (conversion Docling/GROBID · format TEI · citation DTS) = SOLVED/adoptable; composition (integrity-bound voice-attributed verbatim consultation = the engine) = does not exist, stays ours. Adopt substrate, spend effort on differentiator.
- **Verified from source:** Docling (MIT, local/air-gapped, EPUB+PDF+images, per-element provenance, label taxonomy incl. footnote/page_header/section_header/caption — the semantic layer MD lacks; exports MD/JSON not TEI). GROBID (PDF→TEI, 68 labels incl. footnotes/bibliography = A5). DTS (citable canonical-passage API, engine is natural consumer). Loeb is a `.dsl` file (already structured) → NOT a docling case (steward correction).
- **Register updated** with the 2026-07-12 verified-from-source pass + the reframe.

## Two decisions surfaced (I decide neither)
- **(A) Docling bounded trial — IN PROGRESS** (steward said Proceed). Local isolated venv (py3.13; CapableHands is py3.9.6, too old); harness `scratchpad/docling_trial.py`; sample montaigne.epub + jacobs.epub + infinity-of-lists-eco.pdf (text) + mal-darchive.pdf (scanned = A3/OCR test). Tests: does docling emit clean, semantically-labelled, provenanced output where our pipeline leaves cruft (jacobs 888 calibre_link · manguel 2856 []{#…})? Incumbent-bias: our pipeline IS the incumbent → steward judges.
- **(B) ESCALATE — §10.3 canonical-format decision** (MD+sidecar vs TEI), pending since seed brief. Docling structured-JSON = middle path; A informs B; B is steward+jurist (touches engine consumer contract §X + whole corpus).

## Authorization moves
- 2026-07-12 — **CANONICAL FORMAT DECIDED (steward, on the trial evidence, via AskUserQuestion): Option A — MD-canonical + TEI-mirroring `.meta.json` structural sidecar; + Adopt Docling** as the EPUB/PDF/scan conversion front-end (production OCR on CapableHands M4). Brief written (`chamber-library/docs/canonical-format-decision-FOR-JURIST-2026-07-12.md`), stamped with the decision, **delivered to steward + relayed to jurist for ratification**. **PENDING-56 filed** (dotfiles). REVIEWED-56 to follow after jurist ratification. Flip-condition the jurist tests: does v1 need structured apparatus (A5)-as-data or DTS-interop? (read: no.)
- Docling installed + trialed on M4 (py3.12 venv) + M1 (fallback). Both remotes clean.
- **JURIST RATIFIED PENDING-56 (2026-07-12)** — Option A + D1, independently-verified (flip-condition tested, not triggered). **Two schema-gating Qs CLOSED with substrate evidence (brief §8):** Q1 recognizer-arc REPOSITIONED not retired (Docling emits flat `[*](#…)` not native `[^N]`; epub_backend ignores epub:type) → it becomes the sidecar note-extraction layer, verify stays. Q2 Loeb DSL is TYPED (`[c dimgray]`/`[c darkmagenta]`/`[sup]`, 2.15M+ spans) & current extractor FLATTENS → **CARDINAL PRINCIPLE: extractor PRESERVES source typing into typed sidecar fields, never flatten** (→ TEI-migration mechanical-by-construction). Schema now unblocked. REVIEWED-56 can be placed.
- **Owed / next (execution, mostly post-ratification):** build the `.meta.json` sidecar extractor (TEI-mirroring vocab; EPUB metadata + Docling PDF labels; READ-ONLY, safe to prototype pre-ratification); wire Docling into the graduation pipeline (spec amendment — waits for ratification); retire ocrmac_pdf + EPUB cruft-strippers; reshape reprocess track B (Docling→sidecar→verify→graduate); Loeb B2 DSL→MD+sidecar. Parallel: steward still owes REVIEWED-55 placement.
- A1 build will land as a **PROPOSAL** (changes graduation acceptance for ALL conversions — governed). Mechanism is refusal-only (Q1a FIX-test), but the *criterion* is what the jurist rules on. Surface, don't cross unilaterally.

## Sub-agent dialogues

## Bypasses

## Wake re-entry (2026-07-12 ~13:41, after clean clear — same day)
- **Thread CONFIRMED, gate still stands.** Verified from files: REVIEWED.md tops out at **54** → REVIEWED-55/56 NOT filed; schema DRAFT (`sidecar-schema-DRAFT-v0-2026-07-12.md`, 11:51) exists but is **not locked**. PENDING-56 shows jurist ratified (recorded in PENDING) but "REVIEWED-56 *can be* placed" = not yet placed. ⇒ wired extractor remains BLOCKED; Q-A (lock schema) + Q-B (confirm Docling-per-spine-file) are the gate, and both are curator acts (steward+jurist), not mine.
- **Honest flag 1 (link-not-pure):** the wrap's named resumption anchor `in-flight-report-2026-07-12.md` is **absent on disk** (find across chamber-library + studium-engine returned nothing). Substance survives in the schema DRAFT + `chamber-program-open-work.md` + the session memory file's Q-A..Q-E; only the narrative report is missing. Surface at wrap; don't re-fabricate.
- **Honest flag 2 (governance-state latency):** `~/dotfiles` is **1 commit ahead of its remote** — previous wrap's §6.5 push didn't land. Resolve at wrap (not at wake).
- **Drift to hold (both 07-12):** *solving-at-the-wrong-level* (hold the substrate-native per-spine-file architecture when the extractor build begins — don't reach for another heuristic join) + *lost-the-forest* (keep the open-work register open as altitude check).

## Authorization moves (cont.)
- 2026-07-12 ~14:05 — **JURIST RULED the schema lock:** Q1(keep meta@1)/Q2(front/body/back)/Q4(notes linkage-only = PENDING-47 hash-locality reapplying)/Q5(schema-extractor separable — sharper reason: optional-additive absorbs even worst-case as a cheap additive field, not a relock) all CONFIRMED. Q3 schema-half ready; **ONE CONDITION**, gated to "before Loeb runs in earnest," NOT blocking the lock: the extractor must honest-flag source-apparatus structure it can't yet home rather than fold to `notes[]`.
- 2026-07-12 ~14:10 — **Return (verified, not assured):** jurist asked "confirm the extractor does this or add the refusal." Read BOTH extractors instead of assuring. Finding (honest, not convenient): `build_sidecar.py` = no Loeb-apparatus pathway (never sees DSL; section-level `role:apparatus` attested) → jurist's worry doesn't arise there. `extract_loeb_dsl.py` = **CURRENTLY FLATTENS** (`strip_tags` unwraps `[c darkmagenta]`/`[c dimgray]`, `[sup]`→`^N`, strips embedded HTML; MD-only, no sidecar/app[]/flag). So the condition is **real + unmet today**, but **not yet binding**: the B2 Loeb→sidecar extractor is UNBUILT and Loeb Region-4 reconversion isn't running in earnest. ⇒ condition = a **build-gate on B2** (flag-and-hold apparatus, mirror §3 null+flagged). Did NOT report "already satisfied" (the convenient answer); reported the flattening plainly.
- 2026-07-12 ~14:25 — **Looked at B2 NOW (steward challenge: run on one volume; DSL formatting ~consistent corpus-wide). Substrate probe of `loeb.dsl` (6.38M lines, ONE file, 537 vols).** Findings (verified, not inferred):
  - **Color vocabulary = exactly TWO, uniform head→tail** (Apollonius/Dionysius/Aeschylus, Greek + Latin): `[c dimgray]` 2.16M · `[c darkmagenta]` 471K. Steward's uniformity instinct CONFIRMED **at the tag/color-grammar level** — one volume IS representative of the grammar.
  - **Steward correction (load-bearing):** the colors are **Loeb-reader RENDERING aspects, not semantic source markup.** ⇒ they're presentation, but they correlate *perfectly + uniformly* with role → usable as extraction CUES, then discarded. Honest `source_declared` must name the cue as a render-signal (`loeb-render:darkmagenta`), NOT overclaim semantic declaration (attest-don't-default).
  - **darkmagenta = rendered HEADERS** (author / work / book-section: "APOLLONIUS OF RHODES, Argonautica", "Book II. 40"). **NOT apparatus.** ⇒ the Jurist's "darkmagenta might be apparatus being flattened" worry is definitively answered: NO.
  - **dimgray = the ANCHOR/PROVENANCE layer:** ~356,656 pure-number spans = canonical LINE anchors (→ `anchors[]`); ~246,622 "Page number N" = page PROVENANCE (→ `provenance.pages`); plus the `[sup]` footnote-ref wrappers (→ `notes[]`). **THIS is the real preserve-never-flatten target** — the current `extract_loeb_dsl.py` unwraps dimgray → line-numbers/pages dissolve into inline prose text. Uniform, quantifiable loss (~600K anchors+pages corpus-wide).
  - **Embedded HTML is the ONE non-uniform thing:** `<section class="div1 front">` (TEI-ish structural divs!), `<table>` (643, genuine tabular content e.g. line-range maps), `<img>` 506, lists — present in only SOME entries (drama, complex works). ⇒ "one volume is representative" HOLDS for colors, FAILS for embedded structure: a simple prose volume won't exercise it. The real `app[]`/structured content lives here, not in the colors.
  - **Reframe of the Jurist's condition:** apparatus criticus (variant readings) is NOT color-coded — it's footnotes + embedded HTML. So `app[]`'s eventual content is narrower than feared and the colors don't feed it; the colors feed `anchors[]`/`provenance`/`sections`/`notes`. The flatten-loss that IS happening now is the dimgray anchor/page layer.
- 2026-07-12 ~14:45 — **Scanned corpus for the B2 model volume (feature-density per work, grep-prefilter→awk).** Two embedded-structure classes, both genuine apparatus: (1) **concordance tables + TEI `<section class="div1">` divs** — richest in fragmentary works (ENNIUS Fragmentary Republican Latin Vol I: sec=50 tab=44 in ~1650 lines; Cicero/Basil Letters: sec-heavy, one `<section>` per letter); tables map fragment-numbering systems, `[sup]` nested in `<td>` — flattening destroys them. (2) **inline `<img class="glyph">` character-images** — math works (DIOPHANTUS img=37, ARCHIMEDES img=25): non-Unicode math/epigraphic symbols as PNGs (`alt="[special character]"`) = the **[[feedback-character-as-image-hazard]]** exactly; dropping mutilates. **PICK: ENNIUS Fragmentary Vol I** (~line 3717867) as primary model — densest genuine apparatus, tiny/hand-inspectable, modern-TEI, bilingual; its one gap (glyph-images) → optional tiny secondary DIOPHANTUS Algebra (~line 4479566). Model-after-the-hardest-case so the design generalizes down.
- 2026-07-12 ~15:10 — **Built + verified the B2 prototype (`scripts/build_loeb_sidecar.py`, read-only) — jurist's schema-lock condition DEMONSTRATED IN CODE.** Model = ENNIUS (11 works = the real FRL Vol I book) + DIOPHANTUS (glyph path). Architecture: keep MD verbatim (reuse governed `strip_tags` → `body_word_conservation` holds), ADD typed structure in sidecar; line-anchors EXACT-by-construction (B2 analog of Docling-per-spine-file). **Verified against emitted JSON, not asserted:** 44 concordance tables HELD as structured rows (vs current number-soup); Annals 238sec/288anch/252pg/344notes from the render-cues the flatten dissolves; Diophantus glyph branch fired = 38 glyphs held with PNG filename PRESERVED (vs `⟨†⟩`) → closes [[feedback-character-as-image-hazard]] for Loeb. **KEY FINDING FOR THE LOCK:** the DRAFT schema already has EVERY field B2 needs (`anchors`/`provenance.pages`/`notes`/`sections`/reserved `app`) — looking at B2 revealed NO missing field; flag-and-hold + glyph-rule are EXTRACTOR behavior, not schema shape. Honest round-two (logged, not hidden): TEI-`<section>`-div sections uncaptured (FRL Vol I sec=0); footnote ref↔body pairing deferred; discipline: ran Diophantus specifically because the glyph branch was untested code I wouldn't claim worked. Tool-review logged (`_curation/tool-evolution-log.md`). All dev-only/uncommitted; nothing wired to graduation, no push.
- 2026-07-12 ~18:40 — **JURIST RULED PENDING-57: architecture RATIFIED (FLAG/REVIEW/PASS · V-TEXT · V-SCAN · change-class PROPOSAL); full proposal NOT-yet-final pending a V-DSL DEMONSTRATION** (his lesson: simpler-sounding ≠ needs-less-proof). Two non-blocking follow-ons he raised: (a) **retroactive verification** — the morning's 3/3 already-graduated-clean books all FAILED the new check → run the gate BACKWARD across graduated canon, not just forward (REVIEWED-55-shaped, own answer owed); (b) **sidecar-declaration shortcut** for REVIEW tier — the LOCKED schema's `placement:back`+`role` already resolves endnotes-keep vs index-drop declaratively, so books with a sidecar skip the curator queue.
  - **Provenance honesty-check (jurist flagged my 'already RATIFIES' claim):** VERIFIED against substrate → spec v2.0.0 §Tiering & Fence IS ratified (RATIFIED-SHAPE·OPERATION-BLOCKED) → claim HELD; ruling confirms, doesn't originate.
  - **V-DSL DEMONSTRATED — and the demo vindicated the jurist:** naive reuse of the V-TEXT k-gram check FALSE-FLAGGED a clean Apollonius (1070 "fabrications", only 6 truly-new = the MD subtitle; rest = coverage misses from the DSL→MD REFLOW). Corrected → **reflow-tolerant word-multiset** on source-content body: Apollonius fab=0 PASS · Plato Gorgias fab=0 PASS · teeth catch injection. Honest residual: fabrication half gate-ready; content-vs-boilerplate DROP side delegates to the extractor's integrity report. Brief §3.5 written; fixture added (test_tools 105→109). **Back to the jurist for the final ruling.**
- 2026-07-12 ~20:40 — **Doc-currency pass (steward-directed: "future-us shouldn't guess/reinvent/ignore a solution we already had").** Audited chamber CLAUDE.md + runbook vs today's work; fixed the anti-reinvent gaps: **CLAUDE.md** — added `build_loeb_sidecar` (B2, preserve-aware) + `verify_body_conservation` (verbatim gate, REVIEWED-57) to the fleet; rewrote the Loeb-DSL terminology (colors=render-cues; `extract_loeb_dsl` FLATTENS vs `build_loeb_sidecar` PRESERVES — use the latter for reconversion); schema DRAFT→LOCKED + additive-only; added the **check-for-existing-tool/register-before-building** load-bearing discipline (the day's meta-lesson). **Runbook** — pipeline_direction names both extractors + the gate + REVIEWED-56/57 status. **Deprecated-scripts answer: NONE archivable now** — `extract_loeb_dsl` is a live dependency (imported by build_loeb_sidecar + verify_body_conservation) + producer of the current 952 canon; the ocrmac_pdf/EPUB-cruft-strippers are SLATED for retirement by the 2b amendment (unlanded → stay authoritative → don't archive early). **INSTITUTIONALIZED (steward-directed):** added a standing **"Doc-currency on wrap"** protocol to BOTH chamber-library/CLAUDE.md and studium-engine/CLAUDE.md (6-point checklist: fleet/build-state · status markers · substrate-terminology · disciplines · deprecated-verify-before-archive · tool-review) — the mirror of each repo's read-first protocol, so wrapping work THERE keeps THOSE docs current, not memory.
- 2026-07-12 ~19:45 — **PENDING-57 FULLY RATIFIED (jurist, 3rd ruling).** Both conditions met; auto-PASS ratified for V-DSL loss (residual too narrow for blanket REVIEW — can only lose a stopword/name). Two jurist follow-ups DONE: (a) stale "wiring not yet written" brief note fixed; (b) the loss-teeth residual + **revisit-trigger** documented IN THE GATE ITSELF (`verify_body_conservation.classify_dsl`, tagged REVIEWED-57). **REVIEWED-57 drafted** for steward filing. 43-book gap confirmed (V-SUSPECT+V-NONE fenced-out). LANDING (wire into verify_graduation + graduation-spec, tag REVIEWED-57) awaits the steward's REVIEWED-57 filing — a real integration (source access + tier dispatch), best fresh.
- 2026-07-12 ~20:10 — **PUSH-FURTHER: bounded retroactive V-DSL diagnostic probe (ratified §6, read-only) on 9 graduated Loeb canonicals.** Matched each to its DSL work by its own H1 (= the work string), one stream. **7/9 PASS clean.** 2 FLAG (Aristotle Metaphysics, Euripides Medea) — **VERIFIED false-positives, not content loss**: lost tokens all high-freq stopwords (καὶ 4514 vs 4517, τὸ 3021 vs 3023); I hypothesized accent-drift, CHECKED, REFUTED it (καί=0 both; accent-normalization changed nothing) → it's **scattered small cross-extractor tokenization drift** (old extractor dropped a few function-word occurrences vs the reference). **Finding:** forward gate unaffected (same tokenizer, threshold-0 correct, ratified); the RETROACTIVE sweep needs a cross-extractor **drift-tolerance / distinctiveness filter** (ignore high-freq function-word drift, flag distinctive content drops) BEFORE the 952-book run, else mass false-flags. De-risks the ratified sweep, found on 9 books — diagnostic-first working as designed. Probe script in scratchpad (read-only, no de-graduation).
- 2026-07-12 ~19:15 — **JURIST FINAL-RULED PENDING-57: RATIFIED on 2 conditions; BOTH addressed same turn.** (1) V-DSL loss half lacked the fabrication half's independent injection proof (delegated to extract_loeb_dsl self-report) → **built independent loss teeth** (boilerplate-vocab classifier): clean `lost_content=0 PASS`, drop content word `banquet` → `FLAG` naming it; fleet 105→111; honest bound = blind to masked boilerplate-vocab-word occurrences (the edge his REVIEW-routing guards; his call on auto-PASS vs REVIEW kept). (2) 43-book census gap = V-SUSPECT 29 + V-NONE 14 = the **fenced-out** tier (no source) → correctly excluded from the sweep; not a gap. Brief updated (§3.5 loss teeth + §6 census); PENDING-57 updated. Awaiting jurist confirmation conditions closed → REVIEWED-57. **Full gate now: fabrication + loss both independently teethed, both tiers demonstrated.**
  - **Folded (steward-directed) the RETROACTIVE clause into PENDING-57 (§6) to rule together:** on ratification the gate runs BACKWARD across graduated canon (952+239) → diagnostic-first remediation queue (not silent de-graduation), motivated by the jurist's 3/3. Sidecar-declaration REVIEW-shortcut recorded as a noted efficiency (not a condition). Brief + PENDING-57 updated; brief delivered to steward for relay.
- 2026-07-12 ~18:00 — **Proceeded on the two parallel tasks the steward authorized (sequence step 1 + 2a):**
  - **#1 KEYSTONE PROPOSAL drafted + FILED:** enforce `body_word_conservation` at graduation (converter-agnostic verbatim trust gate; listed in graduation-spec but unenforced). FOR-JURIST brief `docs/verbatim-gate-enforcement-FOR-JURIST-2026-07-12.md` + **PENDING-57 filed** to dotfiles. Per-tier method (V-DSL re-extract / V-TEXT built / V-SCAN abstain) already ratified by §Tiering&Fence; three-tier verdict gates (FLAG blocks / REVIEW holds / PASS). Into the jurist loop.
  - **#2a FOOTNOTE PAIRING built (honest v1, in fleet 105/105).** Ref vs body classification; document-order pairing; honest-flag unpaired. **4-iteration PASS-BUT-FALSELY lesson:** Plato prose 96% LOOKED done → *reading the Apollonius orphans* (not trusting the apparatus hypothesis) exposed `in_fn` LEAKING into main verse + a 28-line note-range over-extension (fixed → start-line only). Genre-heterogeneous: settled conservative both-signals → **prose 96% / drama 94% / verse 66% / fragments 48%**, all unpaired honest-flagged, 0 false-pairs. Full per-genre pairing = further sub-project; stopped rather than over-fit one genre. Tool-review logged.
  - All chamber work dev-only/uncommitted; PENDING-57 is on dotfiles (steward reviews/pushes). No graduation wiring.
- 2026-07-12 ~16:45 — **REVIEWED-55 & 56 FILED by steward → schema officially LOCKED.** Schema doc status → LOCKED. B2 built against the locked schema (safe: additive-only).
- 2026-07-12 ~17:00 — **B2 → FLEET-v1 (`build_loeb_sidecar.py` in test_tools, 104/104).** Added the jurist's generic fallback + `<section>`-div/heading capture + multi-author/`--targets` + fixture. **Diverse-by-KIND verification (Aeschylus/Plautus/Plato/Herodotus/Diophantus/Ennius, ~90 vols, Greek+Latin) VINDICATED the jurist's density≠diversity catch on the FIRST look:** naive fallback false-fired on `<que>`/`<per>` = the **critical-edition angle-bracket convention** (editor-supplied readings), NOT HTML → fixed by HTML-vocab-vs-`<word>` classification → surfaced a genuine **THIRD apparatus kind** (`editorial-insertion`, 2315) a density-model would miss, AND a real content-loss (governed `strip_tags` DELETES them; sidecar preserves). Across all 6 kinds the net caught only lists + **0 truly-novel** → structural space BOUNDED {table,section,heading,glyph,list,editorial-insertion}; net still fires on novel (fixture: `<figure>`). Fixed running-header section dup (48→1). **Honest deferred (additive, non-blocking, doesn't touch LOCKED schema):** footnote ref↔body pairing (the ~2238 `list` holds are mostly note-bodies rendered as `<li>`); a PLATO "Mathematical Works" data-quirk; graduation-wiring = the separate PROPOSAL. Tool-review logged. All dev-only/uncommitted; no push.
- 2026-07-12 ~16:00 — **JURIST RULING #2: WHOLE-SCHEMA LOCK CONFIRMED (unconditional, schema-side).** DRAFT → LOCKED on steward's REVIEWED-56 filing. Additive-only discipline CONFIRMED + **field-removal added to the PROPOSAL triggers** (a presence-checking consumer breaks identically from removal as from meaning-change). §3 two minor items confirmed. **Sharp jurist catch I OWN: I conflated apparatus DENSITY (Ennius) with DIVERSITY-of-kind — "densest fit the schema" ≠ "worst case seen."** One condition MOVES (not closes), extractor-side, does NOT block the lock: before B2 runs across the full ~952-book tier, confirm/demonstrate a *generic* "unrecognized shape → flag-and-hold" fallback (mirroring the line-range flag-on-unresolvable). His catch is REAL not hypothetical — corpus already has `<ul>/<li>` (30K) + `<h2/h3/h4>` the v0 holds NEITHER (only table+glyph). **Prepared for the steward:** REVIEWED-56 lock-addendum draft (in `REVIEWED-drafts-for-filing-2026-07-12.md`); schema doc updated → lock-confirmed status + additive-only §4 (incl. field-removal) + §5 moved-condition + folded the 2 minor items. Did NOT flip to LOCKED unilaterally (steward's filing = the ratifying act); did NOT build the fallback this turn (non-blocking, gated to the full-run, deserves its own considered pass — avoid barrel-forward). REVIEWED-55 also still steward-owed.
- 2026-07-12 ~15:30 — **Wrote the schema-lock ADDENDUM for the jurist** (`docs/sidecar-schema-lock-addendum-B2-evidence-2026-07-12.md`): B2 condition satisfied in code (before/after table); the load-bearing finding = **no missing field** (hardest volume fit the schema as drawn); 2 minor new items (loeb-render:* attestation vocab + demonstrated `app[]` kinds table/glyph); §5 honest round-two (extractor, not schema). **Load-bearing governance contribution — the additive-only lock:** named that "lock the whole schema WITH flexibility" is a *soft lock* (untrustworthy) UNLESS the flexibility is drawn as additive-only — LOCKED = meaning/shape of existing fields (books trust them); FLEXIBLE = new optional fields + new app[] kinds via the amendment lane (backward-compatible); NOT flexible = mutating a locked field's meaning (=PROPOSAL). This is the jurist's OWN Q5 property turned into the amendment discipline. Surfaced for jurist confirmation; did NOT presume it. REVIEWED-56 draft deliberately deferred until the jurist confirms the addendum (not presuming the loop's outcome). REVIEWED-55 also still steward-owed.
- 2026-07-12 ~13:50 — **Wrote jurist relay for the schema-LOCK (Q-A):** `chamber-library/docs/sidecar-schema-FOR-JURIST-2026-07-12.md`. Carries the ratified principles + the current schema + evidence read-back-from-drafts (jacobs 34/37, montaigne 2/17 line-range join — worst-case stated, not averaged) + 5 questions FOR the jurist to rule on (schema-id/compat · placement-as-drop-vs-keep · app reserved-undesigned · notes linkage-only · does-lock-depend-on-Q-B-join-quality). Executor lean marked on each; lock left as steward's, informed by jurist. Did NOT lock, did NOT pre-decide — surfaced. Fittingness ✓ (steward asked), craft ✓ (grounded in substrate not memory), ethics ✓ (full picture for jurist), character ✓ (honest at worst-case).