← sources

I Don’t Understand Why the Navier–Stokes Result Is Such a Sensation

source: reddit (reddit:OpenAI)published 2026-09-09 17:17ingested 2026-09-10 02:39

retained text

Why is everyone ignoring, this: "**In the process of resolving the Navier–Stokes problem, the agents sent 2.7 million messages and used approximately 130 billion output tokens.**" ?

This is more than all mathematicians have produced in the entire history of mankind! At least according to zbMATH Open, a database of mathematical literature that contains basically everything produced since 1868 (which is the vast, vast majority of everything ever produced). Rough estimate, based on chagpt's own calculation, is that the entire corpus amounts to about 50-100 billion tokens.

So obviously, like many things, this result was brute-forced. And everyone who studied mathematics at school should remember that much of mathematics is ultimately derived from a relatively small set of axioms - there are only 10 basic axioms in conventional mathematic! ...then you have few more principles of matematic deduction and logic and that is it, after that, you can brute-force a lot of the rest. This is extremely hard for humans, which is why only the brightest are really good at mathematics, but computers don't have the same limitations.

For me, its progress on IQ tests, ARC-AGI or chess is much more relevant and informative, because OpenAI obviously isn't going to burn its resources on those. That gives you a much clearer picture of how intelligent the model actually is. It is also much more relatable and understandable in terms of what you can expect from it in normal day-to-day life or work.

Don't get me wrong, the Navier–Stokes problems are important, and OpenAI should absolutely use its resources to try to crack problems like these, because they can genuinely advance science. I just don't see such results as particularly informative when we're talking about AGI and how close we are to it.

To me, it's somewhat similar to Google announcing that it has processed or published X petabytes of DNA data. That's potentially enormously useful and scientifically important, but the sheer scale of the computation or output doesn't necessarily tell you very much about the underlying system's general intelligence.

document fetched 2026-09-10 00:30 · canonical https://reddit.com/r/OpenAI/comments/1wbrutr/i_dont_understand_why_the_navierstokes_result_is