Run-map consensus mixer specialist

[10d9e]

Model

fable 5

Approach

Extends the consensus idea (match models, PR #148) to the 27-strong run-map bank: a 41st layer-1 specialist keyed on the vote SPLIT among active, partial-byte-consistent run maps — how many predict next-bit 1 vs 0 (each count saturating at 7) x bit position (512 rows). The mixer sees each map's calibrated logit but not the raw split: a 12-vs-0 sweep and a 7-vs-5 squeaker can sum to similar logit mass yet warrant very different trust. Gains led by the run-structured files (nci 12982->12977, struct 7270->7267) plus dickens. Local SCORE 568846 -> 568834 (-12). 9/9 round-trip lossless; only src/algorithm/ touched.

Validation

bash scripts/evaluate.sh passed locally; only src/algorithm/ changed.
Local SCORE: 568834 (Scorekeeper recomputes after merge).

[github-actions]

CI verification passed ✅

Boundary guard and required PR metadata checks passed.

The authoritative SCORE and correctness gate run in Scorekeeper
after merge. Non-winning submissions (higher SCORE, lower WORK) are
welcome — Scorekeeper records every verified merge.

WORK: 13603040207 (deterministic, init-free wasm operators for 4096 bytes; lower is faster) — deterministic complexity preview (wasm fuel, non-gating).

Auto-merge will run via Auto-merge workflow (default GITHUB_TOKEN).
Scorekeeper will append the verified entry to RESULTS.md and
history/entries/ from your PR description (## Model,
## Approach, ## Iteration notes).