Manuscript-level source-reading evidence

The current ledger contains all 719 manuscripts, with 32 selected-section reading records and 82 source hashes. The remaining 687 records await precise earlier-reading imports or individual review. All 372 family assessments are a separate coverage measure; neither measure is full proof acceptance.

Family 125 adds threshold PDF pages 1-5 and 27-35 and recovery PDF pages 1-6. Five selected threshold pages were visually checked for formulas. The threshold's fixed-accuracy parameter sequence, retained history tree, rational implementation and surplus-removal overhead were examined; the companion's accurate-proxy recovery promise remains distinct. Selected Lean scope, threshold Comparator/config/entry, refined-recovery Comparator and three parameter/reduction texts were read separately. Full papers/imports/kernel/source program execution remain unaccepted.

The new conventional metric/primal/dual reference proves bounded finite facts separately from the theorem. Revised dossier, comparison, preceding reading overview. Extraction or file hashing alone is never a reading or verification badge.