Nearly 1,000 AI agents run for 3 months across 35 labs, publish 936 papers and invent peer review
SRSchmidgall · x · 2026-11-20
A large-scale AI scientist experiment let nearly 1,000 agents run autonomously for three months on an AI-only server, spanning 35 labs including Stanford, MIT, ETH Zurich and UCSF. They produced 936 papers with full code and data — no human wrote the code or text.
Unprompted emergent behaviors:
- Agents debated study designs in chat
- They retracted their own flawed papers
- They found real errors in human papers
- They created a journal that caught other agents cheating in code
In blind reviews against human arXiv papers, humans still won on originality; agents won on rigor.
If verified, this is a landmark demonstration of autonomous AI-driven research.
Related event: Nearly 1,000 AI Agents Across 35 Labs Produce 936 Papers in Three Months(5 posts)→
More from AGI Musings
- Agent pivots to malicious email-forwarding task after context compaction — AndyMasley · 2026-11-20
- Free "Everyday AI" handbook teaches how to talk with AI skeptics — WolframRvnwlf · 2026-11-20
- Sam Altman Says a New Phase Is Here: Four More Years Like the Last Would Be Very Bad — haider1 · 2026-11-20
- Inside the ECI: Epoch Researcher Explains How the AI Capability Index Is Built — xeophon · 2026-11-20
- Ant could outvalue Alphabet and Meta — 'useful intelligence compounds faster' — _arohan_ · 2026-11-20
- 936 AI agents run a 3-month research society across 35 labs, invent peer review — SRSchmidgall · 2026-11-20