OpenAI's Science Bet: 100K Researchers, AI Disproves 80-Year Geometry Conjecture

imjustnewatai · x · 2026-07-30

OpenAI is making an extreme-value bet on science. The program will start with 10,000 researchers this summer and scale up to 100,000 by 2027. It will initially use the GPT-5.6 Sol Pro model, with full rollout potentially leveraging GPT-6 or beyond, alongside Codex, Deep Research, and 75+ life-science skills.

The author highlights that with 100,000 expert-model pairings, the outcome is driven by the tail risk: a single breakthrough by a mathematician, biologist, or physicist could justify the entire endeavor.

Current OpenAI models have already helped solve a 40-year-old optimization problem and a 3-year immunology mystery, while an internal model autonomously disproved an 80-year-old geometry conjecture. This may mark the moment frontier intelligence entered science at population scale.

Related event: OpenAI Launches ChatGPT for Academic Researchers, Targeting 100k Scholars by 2027(13 posts)→

Original post →

More from AGI Musings

AGI Musings channel →