Throughout 2025 and 2026, 3Blue1Brown interviewed 12 mathematicians and scientists about this hypothetical scenario, before Levent Alpöge announced his proof with Claude.
Here is what they said:
Kevin Buzzard:
“You press a button and you get this billion-line-long proof of the Riemann Hypothesis, and it's just like, let A be this function, let B be this function, let C be this number, let D be this function. It's just pages and pages, but no explanation of why any of these things are being used. That, to me, would be funny.”
Ken Ono:
“Hundreds, if not thousands of papers are written assuming the Riemann Hypothesis is true. And that's a brand new toolkit. If you could prove the Riemann Hypothesis, there'll be a quantum leap in knowledge in mathematics.
As a romantic in terms of mathematics, what you describe would be horrible. I would want a human being involved in the proof, and I'd like to believe that the proof is beautiful. And if it isn't beautiful, I don't think the search for a beautiful conceptual proof would be any less important.”
Elliot Glazer:
“I personally would get a very sick enjoyment out of that. I'm a logician whose research tends to be in metamathematics, so I find the idea of inscrutable proofs a fascinating concept.”
Jordan Ellenberg:
“If there's a Lean proof of the Riemann Hypothesis that's 20 billion tokens long and nobody can read it, that's not a proof of the Riemann Hypothesis, and I just don't care that that exists.
I wouldn't consider that dangerous. I would just be like, okay, too bad, hopefully someday we'll figure out the proof.”
Leonardo de Moura:
“This is amazing, even if nobody can understand initially. Like I said, a proof is not static.
People would start breaking it into pieces, isolate fragments. People would invest a lot of time trying to understand these objects. If they have high confidence that the proof is correct, they will spend time trying to understand it.
I think it would be a great day for math if it just happens.”
Terence Tao:
“I do feel that fully autonomous one-shot approaches are not the right approach for these problems.
[...]
I can see one of these problems being solved by smart humans assisted by some extremely powerful AI tools, but the exact dynamic may be very different from what we envision right now. It could be a collaboration of a type that just doesn't exist yet.”
Daniel Litt:
“If you're in this weaker situation where the AI produces a Lean proof that's poorly exposited or not exposited, but not intrinsically impossible to understand, that would be a huge deal.
Many mathematicians would drop everything and do everything we could to translate it into something human-readable.
I see no reason why an AI that's superhuman at math research wouldn't be able to help with that expository task.”
Adam Brown:
“We have a notion that it's useful to prove something. We have a notion that it's useful to understand something. And typically in the history of mathematics, those two things have gone together.
Now we're at the stage where they may come separately, where we may receive scrolls of ancient wisdom, notes from the gods telling us that things are true without really being very clear at all on why they're true.
[...]
These things, as well as being superhuman provers, will also be superhuman explainers.”
Alex Kontorovich:
“If an AI system spits out a million-token proof of the Riemann Hypothesis and says, 'I got it,' I don't go, 'Yay, okay, let's move on.'
I go, 'What's the answer? How do you do it? Tell me.' I have to understand. That's why I'm doing this. I'm not doing this for any reason other than understanding.”
Johannes Schmitt:
“I'm really hoping that at some point DeepMind just comes out and says, 'Hey, our model proved the Riemann Hypothesis. Here's the Lean proof.'
I'm confronted also with this kind of skepticism, like, will these models really make progress? And I'm actually always on the side of, I think you should wait and see, or you should give them a few more months.”
Ankur Moitra:
“It does seem to me that certain types of problem-solving are going to become cheaper. We're going to value them a little bit less over time.
Now, there can still be very challenging problems that, when you solve them — Riemann, you know, boy, what an amazing set of tools that would be to solve such a question. So maybe these amazing landmark problems will all still survive. Who knows?”
Jared Duker Lichtman:
“Such a Move 37 moment might come when a machine produces the proof of the Riemann Hypothesis or a Millennium Problem, or even comes up with a deep theory that really synthesizes all these different fields in one fell swoop.
I would consider that to be maybe a Move 37 for mathematics.”
Full video: https://t.co/Fsa68t1dEA