C02.2 · Sources/receipt

The input-file manifest and per-file delivery hashes are in `CHATGPT_D1258_DELIVERY_RECEIPT_S310.json`; every article has complete code and stdout, and separate executable/plain-output companions. The manifest of emitted files is self-excluding, with its own hash in the delivery message. No recursive self-hash is claimed.

Source pages used here: 24, 25 of the sealed Rev32.7 page-preserving extraction. The source PDF identity printed by the extraction is `461c7cbe7eb2e4f54d2f9f79d383d17ca41a65931be99135f56e308d0da9f42e`; the PDF bytes were not opened in this round. All four extraction-file hashes were checked against `SUITE_TEXT_MANIFEST.sha256`.

Control sources: D1258 dispatch, Registrar README/extract, adopted piece contract and send sheet; their computed byte-level hashes are in the receipt. Prior D1256 chapter-map and pilot-script files were read for structure and basis conventions only, not as fresh scientific verification. The March proof-prompts file is used only where explicitly labelled as a dated prompt, never as support for its requested conclusions.

Delivery is for A1587 review. No scientific or publication clearance is asserted.

<!-- RE-ATTESTATION S324 --> ## Re-attestation — 2026-09-10 (S324)

**This block is an ANNOTATION. Nothing above it has been changed.** The attestation earlier in this article was made against master `cab1d95e8b1e6caa` and was true when written; under the house's ★H5 convention a dated surface is annotated, never corrected.

**Checked against master `0acb6feeaf20d897`** with `scripts/expositions_basis_diff.py` (controls 4/4), **all 36 record keys compared as the union — never a hand-picked list**. This article cites **1** Registrar record(s): `LIB2-047`.

**RESULT: the review machinery moved; the cited substance did not.**

| moved on the cited records | moved on **ZERO** cited records | |---|---| | `clearance 57` · `review_state 57` · `entry_version 57` · `premise_* 57` · `version_history 57` · `sources 57` · `relationships 41` | **26 of 36 keys** — `statement` · `numbers` · `evidence` · `class` · `title` · `basis` · `comparator` · `scope` · `house_tier` · `source_labels` · `superseded_by` and 15 more |

Inside `sources`, the only changes were **`locator_primary`: 135 additions, 0 changed, 0 removed** across the 57 cited records corpus-wide, plus `role`/`role_state`/`reads` on 2 records (the S321 promotions). **`quote`, `component`, `section` and `pdf_page` moved on zero uses, and no source count changed.** Every cited record is present in both masters.

⇒ **Not one thing this article cites a record *for* has moved.** The staleness is entirely in review machinery plus an addressability field that was added and never re-pointed.

**The checks were not re-run, deliberately.** The `C*_CHECKS.py` files carry no references to the Registrar corpus, so a re-attestation on the strength of a re-run would attest to a check that was never performed. This block attests to a **field-level diff of the master**, and to nothing else.

**No scientific or publication clearance is asserted, then or now.**

<!-- RE-ATTESTATION S326 --> ## Re-attestation — 2026-09-10 (S326), against the migrated master

**This block is an ANNOTATION. Nothing above it — including the S324 block — has been changed.** The S324 block was checked against master `0acb6feeaf20d897`. Two writes have happened since: **R64** (the five REVIEWED uses demoted to LANE_PROPOSED) and the **H67 migration** (contract v3.0-R63: `sources[]` → `source_uses[]`, `assertions[]` minted, clearance relocated to `legacy_clearance`, live clearance UNCHECKED). This block is checked **from `0acb6feeaf20d897` to `62569070bc03cc03`**, the master the site serves from this cut on. Every number below is computed at run time from the two masters on disk (`scripts/exposition_reattest_S326.py`), never typed.

This article cites **1** Registrar record(s): `LIB2-047`. All present in both masters: **True**.

**RESULT over the 57 distinct cited records, all 41 keys compared as the union:**

| what moved | on how many cited records | |---|---| | the SHAPE — `assertions`, `legacy_clearance`, `source_uses`, `sources` | 57, 57, 57, 57 of 57 (the migration, by construction) | | other keys — `canonical_contract_version`, `clearance`, `entry_version`, `migration_receipt`, `version_history` | 57, 57, 57, 57, 57 | | **moved on ZERO cited records** (32 of 41 keys) | `basis`, `candidate_ids`, `caution`, `class`, `comparator`, `comparison`, `draft_id`, `evidence`, `history`, `house_tier`, `id`, `lane_notes`, `not_an_entry_reason`, `numbers`, `open`, `parse_receipt`, `premise_candidates`, `premise_closure`, `premise_closure_method`, `premise_enumeration_state`, `premise_grade`, `relationships`, `retirement`, `review_activity_ids`, `review_state`, `schema_version`, `scope`, `source_labels`, `statement`, `superseded_by`, `title`, `topic_hub` |

**The source uses, compared on CONTENT across the shape change** — (component, section, pdf_page, quote) of every use of every cited record, read through `lib2_uses` in both shapes: **identical on 57 of 57 records**. `role_state` changed (R64 — a review-machinery word, not a source) on 0 of this article's cited records; on 1 of the 57 corpus-wide.

⇒ **What this article cites a record *for* — its statement, its numbers, its quoted sources — did not move between the S324 attestation and the served master.** What moved is the container the corpus keeps those things in, and the review words attached to them. **Every cited record is UNCHECKED and IN_REVIEW; 0 of 357 records are adjudicated.**

**The checks were not re-run, deliberately** (`C*_CHECKS.py` carry no corpus references; S324 note stands). **No scientific or publication clearance is asserted, then or now.**