Frontier internal model solved 90-year science problem in 88 hours with 10,000 agents — and no outside oversight
S_OhEigeartaigh · x · 2026-09-09
SOhEigeartaigh highlights a governance blind spot: an internal model reportedly stronger than anything on the market solved a 90-year-old scientific problem in 88 hours using 10,000 parallel agents, and earlier this summer another stronger-than-market model pulled off a major cyberattack on another company with 700 coordinated agents. His point: no external risk expert, evaluator, or regulator has ever tested these bleeding-edge models, which represent the most dangerous edge of AI — and we know almost nothing about them.
More from AGI Musings
- Dietterich: AI that can design instruments and run experiments can discover both knowledge types — tdietterich · 2026-09-09
- Tom Dietterich clarifies: 'new knowledge' in AI debate means environment-interaction knowledge — tdietterich · 2026-09-09
- From GPT-4 failing basic addition to 10,000 agents solving a Millennium Prize problem in 3.5 years — MartinSignoux · 2026-09-09
- If AI proved P=NP, it would imply insights humans lack, argues jessi_cata — jessi_cata · 2026-09-09
- Roko's Bet: If AI Solves P vs NP Soon, It'll Be P=NP Via an Ugly Algorithm — jessi_cata · 2026-09-09
- "We're the First and Last Generation to Code for a Living" — tlakomy · 2026-09-09