While grabbing coffee today, I ran into Daniel Litt who passed along the news that RH seems have been proven. I do not follow news sometimes for a few hours, to get some work done. The news had escaped me earlier in the morning. But here are some links from the Scientific American. The announcement was by Levent Alpoge, who like Daniel Litt had been undergraduate math student here. At the moment (morning of Friday, the 20th of November) there seems not yet been much beef around except for the announcement. [Update at noon of November 20: I asked GPT 6.1 whether any details are available. And it https://github.com/CocoSgt/riemann-hypothesis-lean#readme].
[Update morning November 21: a blog entry of Aaronson has a good new name for AI math: stochastic parrots. ]

The RH had been the most important open problem in mathematics, no doubt. I myself always suspected that the result is actually false. Here is the link to the Clay math page. Interesting are some of what one can find right now. (Friday (Nov 20) morning ). I myself liked most the probabilistic approach (youtube short). It had been mentioned also in the talk of Peter Sarnak (youtube short) last spring.
Alpoge already proved the Jacobean conjecture the last summer. I had this as a homework problem in Math 22 in 2018). Levent Alpoge now proved using Claude the Riemann hypothesis. Gosh, this again had been a homework problem (PDF) for Math 154 2025 … Seriously, we should not assign such problems any more even not as a joke! Harvard students (or former students) seem to take such assignments serious, and solve them ….
Below are the four pages of the historic event. I myself actually feel beside the excitement also more anxiety. Unlike the messy and human-made and ugly Navier Stokes problem, the Riemann hypothesis is a god made problem, It would be studied anywhere in the universe. On no other planet, one would look at Navier Stokes in the same form than here as it is a numerical scheme to approximate a complicated n-body problem. But RH is different. It is a problem which (if life should exist on an other planet) would be studied there too. We even study the object in basic calculus courses (youtube short) like math 1b which covers it as p-series.
See the video from a Math 1b class which mentions RH.. And a video from last semester in Math 154 (probability theory). And there was the Zeta flight from 2020 on google earth. It had been an iteration from a zeta flight from 2013, when Google earth was still much more primitive.
A human readable paper would exceed 250’000 pages if written out. The announcement of Alpoge is only 4 pages long:




Update: evening of November 20: already some lean double checks (Sutherland) on a youtube video of the AI channel “Dirac”. The formal verification seems to be ok and there seem not to be any obvious technical flaws in the program. Sutherland at MIT has started also to rerun it with a minimal lean variant. There is now also a hurried announcement of a proof of Alessandro Masullo for a proof of GRH. See this movie (unfortunately also one of these AI generated slop news videos). Just as a side remark: Human generated news of course should have preference. but this is a fast time.
AI slop channels like this Dirac youtube channel obviously have an advantage as they just need to run a script based on some news outlet and get a movie out. Also the AI summary now shows both news. Obviously, the AI channel Dirac takes things directly from what the machines know. A chilling time for humans indeed, not only mathematically, but also journalistically.