New project aims to use AI agents to audit and fix errors in every scientific paper ever published
ValerioCapraro · x · 2026-11-18
Valerio Capraro (University of Milano-Bicocca) is launching an ambitious research project: "Fixing humanity's scientific knowledge." It's inspired by physicist Casey Handmer's AI audit of 50,000+ theoretical astrophysics papers, which flagged 872 possible errors.
The vision:
- Go beyond finding errors — AI agents review each article and produce a revised version that corrects the mistakes
- Scope spans every scientific paper ever published, across all disciplines
- Explicitly not about shaming authors or seeking retractions, but giving current and future scientists firmer foundations
Capraro is recruiting researchers from every discipline; participants can fill out a survey or email him to claim fields to cover.
More from AGI Musings
- MLCB takeaway: agentic science is booming, with plenty of open questions — jmschreiber91 · 2026-11-18
- Why I get why he left Anthropic: choosing participation over retreat as AI transforms everything — danfaggella · 2026-11-18
- OpenAI's Will Depue: "We're in the f***ing singularity" — willdepue · 2026-11-18
- Coinbase CEO Brian Armstrong: 99.999% of transactions will be agent-to-agent in 10 years — davidsenra · 2026-11-18
- Inside Bezos's $41B secret physical AI startup Prometheus: ~250 staff, co-led by Bezos — _fastcompany · 2026-11-18
- Tim Sweeney on AI job fears: same teleprompter script as the tractor panic 120 years ago — liu_mingyu · 2026-11-18