AI Reports · Brookings · 2026-10-29 · Blog / commentary
Spending it faster: What a 10,000-agent proof means for AI’s borrowed expertise
An unverified proof of the Navier-Stokes problem, which OpenAI said an internal model produced using roughly 10,000 parallel agents over 88 hours, shows AI excelling at "normal science" within frameworks humans built long ago. The author argues that AI can generate results at industrial scale, but understanding them still depends on trained people, so the shortage of qualified evaluators now limits what science delivers. The piece urges Congress to reverse a proposed 40% cut to NSF graduate fellowships and fund paid verification teams, since no institution exists to check machine-generated science outside pure mathematics.