AI-drafted 166-page Navier-Stokes proof is correct but nearly unreadable for humans
Pascallisch · x · 2026-09-26
A report on an AI-drafted proof of the Navier-Stokes problem is making waves: the 166-page manuscript is believed to be technically correct, but it's "not written for humans." Oxford mathematician James Maynard says humanity has so far learned very little from the solution and extracting understanding from the AI proof has proven very difficult. Brown's Javier Gómez-Serrano, who uses AI in his own research, thinks the proof could advance the field after serious rewriting.
Pascallisch builds on this to sketch a future division of labor: math professors ask interesting questions, AI supplies the answers — and since no one can understand those answers, validating and making sense of them becomes part of the human role.
More from AGI Musings
- Jensen Huang dismisses scaling laws, calls AGI "just marketing" in Hard Fork interview — teortaxesTex · 2026-09-26
- By Ricardo's law of comparative advantage, should humans now do busywork for AI agents? — arjunrajlab · 2026-09-26
- AI turns scarcity of production into abundance, and thinking of it as another software cycle misses the point — ericwdolan · 2026-09-26
- The internet gave everyone a printing press; AI gives everyone the newsroom — ericwdolan · 2026-09-26
- The internet made distribution cheap; AI makes creation itself cheap — ericwdolan · 2026-09-26
- Consumer Agents Pose a Google-Level Threat: Who Wins the Next Distribution Battle? — thechrisperry · 2026-09-26