don't you understand
Apparently there are at least TWO proofs made public today of the Riemann Hypothesis, which is a bit of mathematics about prime numbers which you have either never heard about, completely forgotten, or think is really, really, interesting and important. One proof is part of the recent onslaught of computer-generated results of important problems; the other more interesting.
The first to be released is from one Levent Alpöge, who works at Anthropic. Alpöge seems to admit he does not understand his entire proof, which he produced with Anthropic's conversation-simulating product Claude; he writes that, in some parts, "I am sort of paraphrasing a 'slop' summary, which might make zero sense." He is also weirdly emotional in his four-page summary, writing "Claude and I have had a great time together" and "Claude flew, and danced, and sparkled."
Writing that it would have been "inappropriate to attempt to post some massive unreadable travesty of 'slop'," Alpöge has not released a human-readable proof; a draft emitted by the conversation simulator had gone beyond a quarter million pages before Alpöge gave up on producing it. His summary includes a "timeline" of events (which is not in chronological order) suggesting he has been concentrating on this problem since an insight in early October.
Instead of a human-readable proof, Alpöge has released about a gigabyte of computer code. The computer proof is written in a language where some types of logical errors (widely believed to be all types of logical errors) turn into program crashes, and a program that exits successfully corresponds to a logically correct proof.
The second result, by one Alessandro Masullo, seems to have been made public prematurely as a response to Alpöge's announcement. I learned about it from a social media post linking to a blog by someone who is not Masullo, but who knows Masullo and Masullo's collaborators. That blog announcement claims that Masullo's proof is being written without any assistance from computer text generators. Masullo writes in his thousand-page draft that the proof has been a work in progress for a dozen years, and that he had estimated it needed about a year more of expository writing and detail-checking before it would be ready to publish, but that clearly there was reason today to make the incomplete version available.
Masullo's draft (is there a direct link? I got it from the blog above) has lots of "this is a draft" features, like section headings with no text and sections that are just bullet points. But those seem to be mostly in the introductory chapter. The bulk of the document is … well, it's not my field, but it looks quite bulky. The introduction includes about twenty pages of a letter sent to a collaborator last month, which was the subject of a meeting just this past weekend.
One of these two products is a computer program which is the output of another computer program, produced in six weeks, summarized with "a stream-of-consciousness jaunt through the experience of finding all the pieces of the argument through incredible confusion[.]" The other is the product of patient, ongoing work, substantially complete, which was held up by the quality of the explanations the author had yet to write.
The Riemann hypothesis does not have obvious practical uses; even if so, most people have already assumed it is true and that a proof would come eventually. There is money and prestige associated with this result, but it is fundamentally a curiosity-driven research program. The tangible benefit of proving a new result is that the proof typically requires discovering or inventing other new ideas, which will in turn be useful in other contexts. Which of these approaches better accomplishes those goals? I have an opinion.