dotfiles/claude/memory/session-2026-07-18-s5-night-coordinate-contract-ruled-reviewed64.md

---
name: session-2026-07-18-s5-night-coordinate-contract-ruled-reviewed64
description: "S5 (night, Session 2 of the agreed plan): the line-ref coordinate contract ran package → design-gate → RULED → REVIEWED-64 placed in ONE session. Fresh Ion measurement (the raw value fails under BOTH read-frames; the two prior observers disagreed by silent base-choice) became the ruling's 'best evidence' and drove the multi-point probe requirement. Ruled: ONE constitutional contract (landed-file, 1-based, sha-valid) + a declared transitional state — NOT two contracts; binds ALL FIVE families (§III mixed-frame forbidden — the jurist caught my plan violating a clause I had QUOTED in my own grounding); notes[] credit linkage-only; per-sidecar frame declaration schema-optional/gate-required; declare-now on the 11. Pulling thread: the REVIEWED-64 build — next session, BEFORE the engine pull."
metadata: 
  node_type: memory
  type: project
  originSessionId: 4e42f175-97d3-4d98-8ead-e3c09797d70a
  modified: 2026-07-19T10:48:50.329Z
---

# Session 2026-07-18/19 (S5, night) — the coordinate contract ruled; REVIEWED-64 placed

Session 2 of the agreed three-session plan, run the same evening as S4's wrap (~2h pause). One bite: the line-ref coordinate-contract FOR-JURIST package — and the gate ruled it in-session, so the docket went package → ruling → steward authorization inside one sitting. The build was deliberately NOT started (late hour; the rushed-tail shape named and avoided).

## PAST — what happened + why

**Grounding before composing (the hook's discipline, honored without friction):** read chamber CLAUDE.md → spec v2.1.0 §III (portability / anchor-primary / one-substrate) + §IV (CTS-URN + Decision-1) + §VII (locator registry + per-scheme recovery) → sidecar-schema (principles, §2 example, §4 lock table, Naming note) → graduation-spec `sidecar:` block → item (f) on its carriers (REVIEWED-62) → RECORD §(c) + the Ion-consequence note → the code notes (SURFACED ~238–243; `write_sidecar` 344–360; `outln()` 115–116; md assembly 284; `transform_md` 111–134).

**The fresh measurement (checkable-claim discipline, and it PAID):** rather than citing the RECORD, I re-measured Ion on the LANDED file. New facts the RECORD didn't hold: (a) the raw `line: 161` fails under BOTH read-frames — 1-based lands mid-Greek (`Καὶ μὴν ἐγὼ…`), 0-based lands on an empty line; (b) the S3 RECORD ("an empty line") and today's probe therefore DISAGREED about the same integer because each silently chose a base — the seam bit its own observers; (c) the coherent window is +11…+16 (B2 head = 4 lines exactly `# {work}`/blank/subtitle/blank; landed head = frontmatter 10 + title block 5 for Ion; the residual spread = head-inclusion/base/`.strip()` micro-conventions compounding). The jurist called this table "the best evidence in this package" and derived the multi-point probe requirement directly from the spread.

**The package** (`docs/coordinate-contract-FOR-JURIST-2026-07-18.md`, hook-passed, `46404ef` both remotes): five field families named (sections/provenance.pages/notes/anchors/app — NOT anchors alone), the cross-framed landed sidecar stated plainly (hash-asserts "I describe this file" while every line value indexes an artifact that no longer exists; 11 landed), three options with teeth, recommendation = "Option B: two contracts meeting at #2's re-stamp." PENDING-64 lodged (~/dotfiles/PENDING.md).

**The ruling** (relayed in-session; filed verbatim `docs/coordinate-contract-design-gate-JURIST-RULING-2026-07-18.md`, `fb728be`): Option B's SUBSTANCE confirmed, twice corrected/extended past the package —

1. **Reframed: ONE constitutional contract, not two** — a 1-based line in `source_file`, valid against `source_sha256` (hash-locality extended from *which file* to *where in the file* — kept verbatim from the package); the producer's 0-based head-less index is "an implementation detail that currently leaks into artifacts until translation. Not a peer contract; a transitional state, declared as such and ending at translation." Consumer consequence: build to ONE contract, REFUSE anything marked untranslated — never handle both.
2. **The finding my own grounding forbade:** §III's shared-substrate clause (which I QUOTED in §0) forbids the mixed-frame sidecar my plan would have produced post-#2 (anchors landed-frame, the other four producer-frame → "a citation and its footnote no longer refer to the same coordinate"). **The contract binds ALL FIVE families; translation is one operation per work; no committed sidecar may be mixed-frame**; vehicle (#2 scope-widened vs atomic companion pass) decided at #2's build — the mixed state never allowed.
3. **`notes[]` is credited, its geometry is not** — the schema-lock Q4 ruling credited linkage-only (ref↔body pairing); nothing said so where a reader looks. Trust-by-adjacency *inside a single field* (kin to the `sections` sweep that cost a FIX on 07-16). The amendment must state it explicitly.
4. **D3:** name the frames, don't rename the field — the defect was the frame being UNDECLARED, not the field meaning two things (protects the lock: nothing written becomes false). ONE per-sidecar declaration (five per-family markers could drift apart); **schema-optional, GATE-REQUIRED** (graduation refuses a line-carrying sidecar without a frame declaration — an optional-and-absent declaration would reproduce today's state). FIX lane; naming lands in the schema doc's Naming note ("becoming the register of this collision class" — this was the 4th shared-name-collision instance: manifest/53, ABSTAIN/55, app[]/62, line/64).
5. **D4:** declare NOW on all eleven — my "avoids touching twice" argument rejected ("the first touch is honesty and the second is correction"); the datum-not-ledger shape, arriving UNBIDDEN (3rd instance; recorded in the candidates register — candidate STANDS, trigger sharpened: the closing instance must arise OUTSIDE the sidecar/graduation domain).
6. **Q4:** one docket, PROPOSAL weight, FIX parts (PENDING-63 pattern). **Q6:** proof condition confirmed emphatically + **multi-point probe per work** (a single probe can't distinguish a correct constant from a coincidence); where the probes don't all resolve under one constant, the work is **HELD, not force-fitted** (honest-refuse at work level).

**Records filed (`fb728be` both remotes):** ruling verbatim · package Addendum (gate outcome — "a later reader must take the ruling's framing over this package's" — + the copy-paste-clean REVIEWED-64 draft) · C-1 ⟳ third-instance note in `docs/candidate-constitutional-principles-register.md` · PENDING-64 Awaiting-line annotated (dotfiles). **REVIEWED-64 PLACED by the steward (verified `~/REVIEWED.md:599`, 2026-07-19, AUTHORIZED).** Chamber CLAUDE.md doc-currency: the 2b status line HELD OPEN → RULED (this wrap's commit).

**Decisions and why:**
- **The build = NEXT session's bite, BEFORE the engine pull** (recommended; steward accepted by wrapping). Grounds: this session's bite was complete and the hour late (the rushed-tail shape S4 named, avoided again); "declare now" is an honesty debt arguing prompt sequencing; the build is one well-bounded docket. The engine pull slides one slot — nothing there is time-pressed.
- **Scope boundary held:** the build session lands the contract + declarations — NOT the translation (that is #2's work, Region 4, under the ruled proof condition).

## PRESENT — the mood

The gate's two extensions both came from material I had IN HAND: I quoted §III's shared-substrate clause and did not trace my plan's end-state against it — **quoting is not tracing** (harvest-proposed as `grounding-quoted-but-not-traced`; the hook enforces the quote, nothing yet enforces the consequence-trace). Counterweight: the fresh-measurement instinct (measure Ion today, don't cite the RECORD) produced the package's strongest material and directly shaped the ruling's Q6 — the checkable-claim discipline running at full strength. The reframe ("one contract + transitional state") is the kind of correction worth internalizing: I reached for symmetry ("two contracts") where the safety-relevant asymmetry (one real contract, one leak) was the truer shape. Loop hygiene was clean all session: hook honored, PENDING lodged before the gate, ruling filed verbatim, nothing built past authorization.

## FUTURE — what is pulling

**PULLING THREAD (singular): the REVIEWED-64 build — next session, before the engine pull.** One docket, PROPOSAL weight with FIX parts, tagged REVIEWED-64.

**ACTIONABLE RESUMPTION POINT (as of wrap — re-judge against what changed):** chamber `main` @ the doc-currency commit after `fb728be`, clean, both remotes; REVIEWED-64 placed (:599); fleet last known 180/180. The build items, from the ruling's dispositions: **(1)** frame naming + the per-sidecar declaration field — schema doc Naming note + additive optional field (FIX lane; hook-gated, ground in the ruling + §4 lock table); **(2)** the graduation refusal — declared data first (`graduation-spec.yaml` `sidecar:` gains the frame requirement), then the thin consumer in `graduate_to_canonical` (sidecar_gate/write_sidecar), **test-first with a witnessed red** (a line-carrying sidecar without a declaration must HOLD); **(3)** declare-now on the eleven landed sidecars (bounded-diff proof: exactly 11 files, one additive field each; the MD files untouched — sidecar sha stamps describe the MD, unaffected); **(4)** the `notes[]` linkage-only statement (schema doc + the graduation-spec credit map comment); **(5)** the contract's constitutional text through the amendment lane (hook-gated; quote §III/the ruling; preserve the §2 worked example per the jurist — "should be preserved wherever this lands"); update the now-stale SURFACED/`write_sidecar` code comments ("until ruled" → ruled, REVIEWED-64) in the same pass. **First move: compose the amendment edit-set doc (the FIX-A pattern — complete edit-set, before→after, grounded) as the docket's spine, then build parts 1–4 against it.** Carry Q6 (multi-point probe, honest-refuse) forward to #2's build — NOT this docket.

**Other horizons (ranked, held):** the engine pull (after the build; known-vs-newly-failing ingest_gate distinguishability + the after-the-reply re-anchor; read `stage-1-rebuild-plan` fresh) · #2's re-stamp (Region 4; now with a ruled proof condition) · 16 BODY-DEFICIT remediation (② 4 re-extract via B2) · Republic + far-country batches (the 17-work watch-list travels with them) · Doc-1 line-45 micro-edit · skill-harvest full review (~15 PROPOSED; requested 2026-07-12, still unheld).

**Decisions deferred (and why):** the build itself (late hour; deserves a fresh concentrated block — witnessed red, bounded diffs, grounded amendment); the translation vehicle (#2 scope-widened vs atomic companion pass — the ruling assigns this to #2's build, "decide which vehicle; do not decide to allow the mixed state"); the constitutional text's landing place (spec supersession vs ruled doctrine in schema+graduation-spec — the ruling says "wherever this lands"; first decision of the build session, and this wrap's literal question).

**PAUSE STATEMENT:** I am about to be away from this. The docket closed its ruling arc in one sitting — package, gate, authorization — and the build waits whole, with every disposition named and nothing half-started. What I want to find still pulling on return: **the REVIEWED-64 build as one clean bite** — the frames named, the gate given teeth, the eleven made honest, and the contract's text landed where it belongs.

**LITERAL QUESTION for next-Claude:** Where does the coordinate contract's constitutional text land — a spec supersession (v2.2.0, a new subsection under §III or §VII, the citation-amendment precedent), or as ruled doctrine carried by the schema doc + `graduation-spec.yaml` with the ruling document as its constitutional record — and does PROPOSAL weight itself entail the spec vehicle, or is placement a separate, lighter call the amendment doc can put to the jurist as its first gate question?

**State at wrap:** chamber `main` @ doc-currency commit (after `46404ef` package · `fb728be` ruling records), clean, both remotes · REVIEWED-64 placed `~/REVIEWED.md:599` · PENDING-64 annotated (gate-passed → awaiting became placed) · C-1 third instance recorded, trigger sharpened domain-external · canon 1305 / archive 335+1 artifact / pending 14 legacy (unchanged — no book touched) · ledger `session-ledger-2026-07-18.md` carries S5 under the S5-night section · harvest: 1 proposal (`grounding-quoted-but-not-traced`) + /jurist-package now 5-for-5 (reinforcement).