Researcher's open-source paper-review skill flagged ~30 errors in one manuscript proof
RobbWiller · x · 2026-11-09
Stanford researcher @davidhagmann open-sourced a paper-review skill for Codex and Claude Code that audits empirical manuscripts: it lists each issue with a full proposed replacement and generates an interactive HTML report where you select fixes or add comments, then export back to the AI for processing. Repo: github.com/dhagmann/paper-review (MIT).
On a real manuscript proof it flagged 30 errors — major ones (a reported stat belonged to a different DV) and minor ones ("rife with" written as "ripe with") — and with access to raw data it replicates all analyses.
The post also quotes Ethan Mollick: worries about verifying AI output may be outdated, as recent models beat human experts, including PhD specialists, at finding facts and checking claims. Together: AI is now capable of quality-controlling scientific manuscripts.
More from coding & agent
- Together AI: leaderboards are for shortlisting, your production traffic is the real benchmark — togethercompute · 2026-11-09
- Codex bug: switching accounts wipes your Projects list, just like Claude — DevDminGod · 2026-11-09
- Codex Pacer: agent-reported real-time progress tracking for Claude Code — Disastrous_Reward644 · 2026-11-09
- Don't let the training team see your internal benchmarks — just return a score — xeophon · 2026-11-09
- DSH beats Claude Code on transparency as Blender MCP and vork finally work together — teortaxesTex · 2026-11-09
- AI assistant spots S$40 CapitaLand voucher before it expires — gabrielchua · 2026-11-09