## Why
`tools/build_google_math.py --manifest-only` aborted with `KeyError: 'rows_decoded'`. Its README generator read the gate's legacy keys while the gate it reads now emits per-split keys, so the documented build path was broken for anyone rebuilding after this change.
## What changed
- One boundary adapter, `read_gate_evidence(root, records)`, replaces every direct gate-key access in `write_readme`.
- It understands both schemas: legacy (`rows_decoded`, `segments_compared`, `parquet_roundtrip.rows_checked`, `split_separation`) and per-split (`packed_decode.train/dev/test`, `parquet_roundtrip.rows/exact`, `group_separation`).
- Per-split counters are summed and legacy totals are read as totals; nothing is reinterpreted, so each field keeps the meaning its writer gave it. Fields the legacy gate never measured stay ``None`` and the README omits those clauses instead of inventing values.
- The gate is rejected when its `measured_totals.records` does not match `statistics.json`, so sta…
audit(google-math): re-run the independent publish gate against the delivered files
## Why
The independent publish gate had not been re-run since the rebuild, so `audit/publish-check.json` was written by the build rather than by a separate verification pass. This refreshes that evidence against the files actually delivered in this PR.
## What changed
- Re-runs the gate with new code: 480,605 records, 3,424 groups, 87,033,058 characters, 42 packed rows (14 per split), 613 attributed segments, 9/9 Parquet round-trips, 0 groups spanning two splits.
- Every recomputed total matches `statistics.json` and `validation.json`; 13 declared-number comparisons, 0 disagreements.
- Widens the sampled packed rows from 12 to 14 per split and states the fragment rule precisely: 53 fragments verbatim inside their record, 2 carrying only BOS/EOS or bare whitespace, 4 needing multi-byte edge trimming.
- Updates `audit/publish-check.json` (22,838 to 9,622 bytes), its `manifest.json` entry, and the README paragraph that quotes those counts.
## Scope
Three files under `components/google-m…
feat(dataset): rebuild google-math with the AlphaEvolve and FunSearch constructions
## Why
Three AlphaEvolve and FunSearch repositories were measured by an earlier scan and never decided, so the component had a coverage gap. They are now judged on measured evidence: included where they ship math text, excluded with a named size where they ship code only.
## What changed
- Adds `alphaevolve` (16 verification notebooks over the 15 `maths/` problem directories), `alphaevolve_results` (`mathematical_results.ipynb`) and `funsearch` (four problem notebooks): 21 records, 21 groups, 8,668,170 characters.
- Adds `math_content` to every source and record, from formal_statements, proofs, natural_language_math, generated_qa, constructions, problem_statements, none. The new sources are `constructions`.
- Records `excluded_subpaths` separately from `excluded_candidates`, so a repository is in `included` or `excluded_candidates`, never both. `omega/` (45,511,025 bytes, 44,267,872 of them one `.xz` archive) and the search harnesses are excluded with their measured sizes.
- Rebuilds …