Garry Tan bbcc882c8e feat(eval): BrainBench core — adapters, metrics, harness, scoreboard, governance ...

Cross-harness memory conformance suite (Cathedral 2): one shared Reflex
pipeline driven by three seam adapters (openclaw=production, claude-code +
codex=contract with exported wire types), four metric suites (know-to-ask +
false-fire, push P/R, write-back via the production pipeline, continuity),
source-isolation gating at zero, intrusion diagnostics, sealed-gold loader
with strict validation + corpus hash, fail-fast hermetic seeding, ONE PGLite
per run with resetTables between fixtures, and the main-baseline compare
gate (same-hash count-aware / corpus-bless modes, justification flow).
Eleven new metrics registered in the glossary (doc regenerated).

2026-06-12 12:06:03 -07:00

..

METRIC_GLOSSARY.md

feat(eval): BrainBench core — adapters, metrics, harness, scoreboard, governance

2026-06-12 12:06:03 -07:00

SEARCH_MODE_METHODOLOGY.md

v0.42.25.0 fix(pricing): unify chat-model pricing into one canonical source; add Opus 4.8 (#1819) (#1827)

2026-06-03 23:00:11 -07:00