P-M4-1: on the free tier, tomo-oi meets both clauses

id:        P-M4-1
filed:     2026-07-17, from spec 2105 doc 10 section 8.5, before the M4 full gate
suite:     core 14 plus the adopted swebench-live cut, fair-tagged denominators only
tier:      free
engine:    tomo-oi against the rival set at gate n
band:      floor: level on aggregated pass rate against the best rival.
           ceiling: effective cost per pass at or under 0.5x of the best rival.
mechanism: the journal's small-n evidence (leaner on every task, level or better on
           grade, under half the rival's tokens) extended to gate n; the leanness is
           structural (code-as-action fresh-per-round) rather than sampled luck, so it
           should survive aggregation.
settled by: the M4 full gate, all arms interleaved, n per the repeat floor.
           go run ./cmd/lab report && go run ./cmd/lab report --suite swebench-live
result:    (open)
verdict:   (open)

This is the first true test of the lab's headline bar, and the first prediction whose miss would be a verdict about the product rather than about the instrument. Both clauses must hold; a cost win at a grade loss is a miss, and the one-scoreboard rule keeps the pass rates published beside the ratio.