The line you're quoting comes from a RuntimeWire write-up dated October 25, 2026. It reports that Scott Armstrong, a PDE mathematician (NYU Courant Institute and Sorbonne Université), called OpenAI's Navier-Stokes construction 'unbelievable' and said it may be beyond what a human could have devised, though not beyond human understanding. RuntimeWire That wording closely echoes Greg Brockman's praise from the same week, so the article may be framing Armstrong's reaction against Brockman's. I could not open Armstrong's full post, so I can't confirm his exact reasoning.
His own posts in search results are more critical on the technical side. He cites a new paper by Constantin, Ignatova and Vicol arguing that OpenAI's blow-up construction cannot produce an analytic forcing term, so it does not work with zero forcing, the formulation (A) of the Clay problem. Armstrong on X
For context, OpenAI's claim (September 2026) is a finite-time singularity for the forced Navier-Stokes equations, with a Lean formalization. OpenAI Critics note that the forced version is the one the Clay statement allows, but not the harder unforced case. Science Whether the proof and its Lean formalization fully match is still being disputed, so treat the claim as unsettled.
🕒 As of October 2026
🏛 The council's answer (Claude, Llama · 35% sure)
Nobody here could establish why Armstrong said it. The earlier answer reports that a RuntimeWire piece (Oct 25, 2026) quotes him calling OpenAI's Navier-Stokes construction 'unbelievable' and saying it may be beyond what a human could have devised, though not beyond human understanding. The most that can be said is that he seems to mean the construction is so intricate or unexpected that people would be unlikely to have found it unaided. That is a plausible reading, not a confirmed one. The earlier answer admits it never opened his full post, so his actual reasoning is unconfirmed.
On checking the earlier answer: it was right to separate the praise from his technical criticism. A mathematician can admire a construction and still say it doesn't settle the Clay problem. It was also right to flag that the proof and Lean formalization are disputed, and that the forced-equation case is not the unforced Clay problem. The Clay point is general knowledge that Claude confirmed.
It was weaker in a few places. Its idea that the article is echoing Greg Brockman's praise is speculation. It gave no direct evidence that the quote and the Constantin-Ignatova-Vicol criticism come from the same source. Its specific claims (the date, the quote, the paper, the Lean dispute, the links) could not be independently verified by Claude, which has no knowledge of these 2026 events. Llama accepted them at face value. Treat them as reported but unverified until you read the RuntimeWire article and Armstrong's posts.
Where they agree
- Armstrong's actual reasoning for the 'beyond human invention' remark is not confirmed.
- The praise and the technical criticism are not necessarily in conflict.
- The status of the proof and its Lean formalization is disputed or unsettled.
Where they differ
- Whether the earlier answer's specific claims can be trusted
- Claude: It cannot verify any of the specifics (quote, date, paper, links) and calls the Brockman-echo theory unsupported speculation.
- Llama: The earlier answer correctly reported the statement and context, with high confidence, citing the same links.
- Better supported: Claude's caution is better supported. Llama only repeated the earlier answer's sources and did not independently check them.
What none of them could settle
- Whether the RuntimeWire article and the quote exist as described
- Armstrong's actual reasoning
- Whether the cited URLs are real
- The status of the Constantin-Ignatova-Vicol paper and the Lean formalization dispute
Sources they gave
- runtimewire.com/article/scott-armstrong-says-openai-s-navier-stokes…
- x.com/scottnarmstrong/status/2100937920576094569
- openai.com/index/navier-stokes-solution/
- science.org/content/article/how-ai-math-breakthrough-ignited-contro…
Each AI's own answer
- Claude · 30% sure: I can't verify this. I have no reliable knowledge of an OpenAI Navier-Stokes result from 2026, a RuntimeWire article, or Scott Armstrong's statements about it. My training data does not cover these events, so I can't confirm or refute them. What the earlier answer seems to get right: Scott Armstrong is a real PDE mathematician, and he is affiliated with the Courant Institute (NYU) and Sorbonne Université, as I understand it. The Clay Millennium problem concerns the unforced Navier-Stokes equations, so a blow-up for a forced version would not solve it. That is a general point I'm fairly confide…
- Llama · 80% sure: Scott Armstrong, a PDE mathematician, said OpenAI's Navier-Stokes result may exceed human invention because it is 'unbelievable' and potentially beyond what a human could have devised. However, Armstrong's own posts are more critical, citing a paper by Constantin, Ignatova, and Vicol that argues OpenAI's blow-up construction cannot produce an analytic forcing term. The context of OpenAI's claim is a finite-time singularity for the forced Navier-Stokes equations with a Lean formalization, which is still being disputed. The previous answer correctly reported Armstrong's statement and provided co…
_AIs from different companies answered separately; Claude combined their answers. Check anything important: they can all be wrong together._
