# Manuscript-level source-reading evidence

The [current ledger](checkpoints/manuscript-ledger-2026-10-09-v12.json) covers all 719 manuscripts, with **33 selected-section reading records and 83 source hashes**. The remaining 686 await precise earlier-reading imports or individual review. The 372 family assessments remain a separate coverage measure; neither measure means full proof acceptance.

Family 121 adds PDF pages 1-8, 34 and 43, plus the conclusion paragraph on page 44. Pages 7, 8 and 34 were visually checked. Selected scope/config/Comparator/Definitions, the complete AccuracyCutoff and RawBandMain entry establish the reviewed theorem shape and explicit large-input guards. The necessary ell threshold alone excludes practical document sizes. Full paper/imported formal closure/kernel/source-program acceptance remains open.

The new existing-library and guard controls are independent finite evidence, not execution of that approximation algorithm. [Dossier](opportunities/008-long-document-edit-comparison/2026-10-09-v2.md), [baseline report](prototypes/edit-distance-comparison-2026-10-09-v1/report.md), [preceding overview](MANUSCRIPT-REVIEW-2026-10-09-v11.md). Extracted or hashed files alone never receive a reading badge.