OpenAI cracks Millennium Prize Problem with 10,000 agents, at an estimated $15M cost
nordicinst · x · 2026-09-12
The Guardian reports OpenAI claims its latest model has cracked a Millennium Prize Problem—a $1M puzzle that stumped mathematicians for decades. The feat was unlike typical mathematical breakthroughs: the near-trillion-dollar company unleashed an estimated 10,000 autonomous AI agents on the problem, at a reported cost of around $15 million.
The speed of the achievement shocked the field. Prof Colva Roney-Dougal of St Andrews said she publicly claimed months ago AI wouldn't do anything remarkable soon—"three months later, I'm totally wrong." Mathematicians are now asking what remains for human researchers when works in progress can be hoovered up by AI, and how to train the next generation when tough assignments fall at the press of a button.
More from AGI Musings
- Anthropic alignment lead puts >10% odds on AI killing humanity within a decade — beglen · 2026-09-12
- Fruit Fly Brain Simulation Solves Rubik's Cube, Igniting Consciousness Debate — sebkrier · 2026-09-12
- p(doom) debate: 50% risk routes through institutional failure, not alignment progress — jd_pressman · 2026-09-12
- Thought experiment: what if ASI finds humanity boring and just leaves? — pafagaukurinn · 2026-09-12
- Roko argues human expected utility hinges on P(Banksian Culture), not alignment — RokoMijic · 2026-09-12
- Riley Ralmuto Posts the Connor Leahy Interview Video Behind His Viral Critique — RileyRalmuto · 2026-09-12