Alex Verkhovsky
|
d9accf6e4e
|
docs: add QD2 skeleton eval report comparing QD2 vs old QD baseline
Metrics from JSONL log mining: 56% fewer human turns (18→8),
32% fewer API turns (72→49), token cost at parity (1.14x).
One-shot path identified as broken (P0 fix priority).
|
2026-02-22 21:37:37 -07:00 |