Research notes

Dated notes from building opendrs: what was tested, on which repositories, and what changed as a result. Each note lists its sources. The experiment log has the benchmark runs behind them.

• Where the dev loop’s time goes, 22 Oct 2026: A measured audit of check/test/eval wall time across one working session on opendrs: the git-spawn test fixture, full-suite skill evals, and the npm run check inner loop, with the fix and the numbers for each.

• What Anthropic’s code-review write-up means for opendrs, 20 Oct 2026: Research note on ‘How the Claude Code team does code review’ (claude.dev, 2026-10-15): a checked summary, a practice-by-practice map against opendrs’s docs and code, opportunities with cost/risk/measurement, three strategic bundles, what the article implies for the chosen verification pass, open questions, and a verification log.

• Field test: opendrs 0.4.0 on six new repositories, 19 Oct 2026: opendrs 0.4.0 run against about 8 merged PRs each on psf/black, cli/cli, actix/actix-web, vitejs/vite, fastlane/fastlane and vapor/vapor: captain-fact quality, score saturation, language coverage gaps, module-boundary fit and performance, with a ranked fix list.

• How AI code reviewers judge pull request risk, 18 Oct 2026: How Claude Code, Superpowers, PR-Agent and other AI code reviewers judge pull request risk, the heuristics they share, and ten ranked changes for opendrs.

Updated 24 Oct 2026 · Edit this page on GitHub