Trevin Chow d22a9b8e7f fix(ce-optimize): show workers why cases fail and cap whole-run spend ...

Judges now give a one-line reason per item, and the worst cases from
judges or per-case harness output reach hypothesis generation and
experiment workers. Workers get a source digest written once instead
of re-reading the corpus each experiment. An optional
stopping.max_total_cost_usd caps model spend for the whole run; the
judge cap still counts scoring only.

2026-09-29 17:50:58 -07:00

..

brainstorms

feat(lfg): route each request to the skill that owns it before shipping (#1702)

2026-09-13 19:41:45 -07:00

dogfood-reports

feat(packs): compound, ground, and enforce domain knowledge via Compound Packs (#1549)

2026-09-09 18:45:01 -05:00

guides

fix(ce-optimize): show workers why cases fail and cap whole-run spend

2026-09-29 17:50:58 -07:00

ideation

feat(ce-pov): add project-grounded verdict skill (#1020)

2026-06-29 15:49:52 -07:00

install

fix(ce-setup): detect the retired Codex tool map (#1577)

2026-08-29 01:58:55 -07:00

plans

fix(test): fail CI test stalls fast and stop orphans holding test output (#1787)

2026-09-28 01:40:47 -07:00

solutions

fix(test): re-run a CI pass wedged by a lost child exit (#1806)

2026-09-29 12:00:04 -07:00

specs

docs(agent-plugins): attribute the 8000-byte skill cap to its real owner (#1511)

2026-08-22 00:32:13 -07:00

development.md

fix(site): make built HTML work at any base path (#1669)

2026-09-09 18:52:14 -05:00

grok-marketplace-submission.md

fix(grok): support compound-engineering as a native Grok Build plugin (#1086)

2026-07-09 10:19:18 -07:00