GoKawiilResearchers report being outpaced by AI agents on unpublished math and science problems
Mathematicians including Tristan Buckmaster of New York University say that after working for about a year on a fluid-dynamics puzzle related to the Navier-Stokes problem, OpenAI announced its agents had solved it the day after Buckmaster posted his own progress. Scientists say this is one of at least two recent cases in five weeks where AI companies reached or announced results on questions researchers were already pursuing, prompting some to limit sharing unpublished work with AI tools.
GoKawiil's interpretation of the reporting above, not reported fact.
The pattern suggests AI agents are becoming capable enough to independently reach research conclusions, which could reshape incentives around how and when scientists share early-stage work. Geoffrey Irving of Resolution argues scooping will increasingly come not from AI reading unpublished manuscripts but from AI simply solving problems faster, a claim that if true could pressure norms of academic credit and openness.
- Mathematicians report AI agents reaching results on problems they had already been working on for months.
- OpenAI announced a solution to a Navier-Stokes-related puzzle a day after a researcher posted his own progress.
- Some scientists are now restricting use of AI tools to avoid unpublished work being scraped or surpassed.
Source: nature.com — Glickman, 2026-10-08
Published there as: “Will AI scoop your science? Some researchers see a gloomy future”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.