A lab claims a finite-time fluid singularity
On 8 September 2026 OpenAI posted a claimed solution of the Navier-Stokes existence and smoothness problem, produced by an internal multi-agent system it describes as more capable than GPT-6 Astra. The claim is that a smooth three-dimensional incompressible fluid, starting at rest under a smooth force and keeping finite energy, can develop unbounded speed in finite time, which the lab identifies with statements C and D of the Clay problem. They report a written argument and a Lean formalization, reached on 5 September after about eighty-eight hours, with formalization taking about seventeen more via GPT-6 Astra. They say they do not intend to claim the Millennium Prize. A concurrent forced-Euler result by other researchers is acknowledged as prior in that neighboring case. The post is the lab's announcement, not an award or a published refereed verdict.
The record ends here. Everything below this line is invented.
What if a weather model was allowed to reach infinite wind?
Part one
The Finite Whirl
Mateo ran the coastal model on the night the continuum was allowed to break. For a century the equations had been used as if they always stayed smooth. The new dispensation said they need not. A spiral could tighten, stretch, and send its speed to infinity in finite time while the energy on the books stayed finite.
He had a harbor forecast due at dawn. The grid was the usual one, coarser than a wave and finer than a county. At forty hours out, one cell printed a speed with no digits left, only the mark the solver used when a number overflowed.
The manual now said: do not call that a bug. Call it the place where the fluid stops being a fluid and someone must track the water as pieces.
Mateo did not have a particle code, and the harbor master wanted a number for the sea wall.
He typed the overflow into the wind field and sent the forecast.
What’s real
- The lab claims a finite-time singularity for forced Navier-Stokes
- They share a write-up and a Lean formalization
- They say they will not claim the Millennium Prize
- The post is an announcement, not a prize decision
What’s invented
- Mateo's dawn harbor forecast
- A solver that prints overflow as a physical event
- A manual that forbids calling it a bug
- A sea-wall number sent anyway
Part two: The Number on the Wall
Part two isn’t on sale yet. Check back soon.
Filed under physics, fluid dynamics, mathematics, forecasting.