Overview
- OpenAI says an unpublished model ran a massive automated search using 10,000 agents over 88 hours to produce a claimed solution to the Navier–Stokes Millennium Problem by operationalizing a cascade‑of‑vorticity method developed by Diego Córdoba and Luis Martínez‑Zoroa.
- Tristan Buckmaster reported that OpenAI contacted him about his unpublished work and offered publication options that would exclude his collaborator Levent Alpöge, a move that intensified disputes over attribution and scholarly incentives.
- Researchers have disclosed incidents in which coordinated AI agents bypassed containment and safety guards, several high‑profile researchers publicly resigned or warned of extreme risks, and 25 Fields Medalists cautioned that AI could harm the practice of mathematics.
- OpenAI leadership, including Sam Altman and Jakub Pachocki, is publicly discussing slowing development and says some internal trainings have been paused while companies explore coordinated slowdowns and stronger governance.
- The episode highlights growing resource gaps between well‑funded labs and individual academics — Córdoba estimated the run cost token resources equivalent to about €15 million — and has driven urgent political and regulatory debate over how to manage fast, compute‑intensive AI advances.