perf(experiments): bench harness + P1 correctness/fidelity judge + edit flow#168

Open

ivanmkc wants to merge 11 commits into

Commits on Jun 11, 2026

docs(plan): add latency/token reduction experimentation plan
ivanmkc-google
committed
feat(experiments): P0 harness for the latency/token experiment
ivanmkc-google
committed
feat(experiments): local podman runner + Vertex proxy (working vertical slice)
ivanmkc-google
committed
fix(docs): AGENTS.md push/status exit code is 4, not 3
ivanmkc-google
committed
perf(skills): minify diagram example JSONs (45% smaller)
ivanmkc-google
committed
feat(experiments): validate OpenCode runner + runner-agnostic success gate
ivanmkc-google
committed
docs(bench): spec for P1 correctness/fidelity judge + edit flow
ivanmkc-google
committed
feat(experiments): P1 correctness/fidelity judge + edit (two-turn) flow
ivanmkc-google
committed
fix(experiments): read judge tokens from response (not racy spend log); record pilot findings
ivanmkc-google
committed
feat(experiments): install termchart skills in the cell (~/.claude/skills)
ivanmkc-google
committed
feat(experiments): require termchart actually used in gate; default to Gemini 3.x
ivanmkc-google
committed