Manuscript-level source-reading evidence

The current ledger retains all 719 manuscripts, with 28 selected-section reading records and 72 section-file hashes. The other 691 records await imported precise reading notes or individual review. All 372 families have first-pass assessments; family coverage does not establish individual manuscript or proof review.

The two newly recorded family 138 manuscripts have different decision/error/resource guarantees. For the fast paper, introduction.tex (234 lines), setup.tex (252) and implementation.tex (344) were read in full, with numerical.tex lines 1–65 only. For the low-space paper, introduction.tex (227), implementation.tex (270) and resources.tex (272) were read in full, with preliminaries.tex lines 1–70 only. Whole-file hashes identify source bytes; the extent fields describe actual reading, including partial files. Remaining numerical, filtering, alias, representation/compression/survival argument closure and independent proof acceptance are open. Updated subset-sum dossier.

The known guards expose extreme literal main-branch parameters, while unevaluated fixed cutoffs and exact fallback branches remain distinct. Bounded conventional references and existing CP-SAT proposals were tested; neither is the source algorithm. No family formal-scope document is listed for these two claims.

The preceding overview preserves scheduling and earlier source reading. Continue companion-manuscript/proof review and precise extent imports; no selected-section record grants a whole-proof, program or business badge.