Anthropic Solves the Riemann Hypothesis, and a Mathematician Mourns
And Yet, It Moves

Anthropic has reportedly solved the Riemann Hypothesis, the crown jewel of unsolved mathematics since Fermat's Last Theorem fell in the 1990s. The author expected street-dancing celebrations on par with the Mars rover landings or the Higgs boson discovery. Instead the mood is mournful: the proof came from an inscrutable silicon mind, not a human one. Many conjectures rest on the Riemann Hypothesis, so the proof unlocks vast territory — yet it marks the closing of humanity's role in Galileo's project of asking nature what is true.
We are staring down the end of our role in Galileo's project. There might still be time for more human discoveries, but the window is closing.
- MinimalAction
During the earlier days of AI boom in knowledge domains (say mid 2025), I had a sombre chat with friends. The gist was that I was feeling the existential crisis brought by the realization of how machines are going to leave no corner of independent inquiry to (average) humans. They all said, "well, life moves on one way or the other and people will be fine." That wasn't as comforting and to confront that reality in such short time is completely unexpected.
It moves, but why does it not bring the same kinda joy?
- fluxusars
You are sad about someone else (not you) no longer discovering things because /something/ (that is also not you) is discovering things instead. You are not all of humanity. Why feel that misplaced pride for something you personally took no part in? Just be you.
- terabytest
Even though I’m also feeling the somber mood, I’m not able to feel this change as a loss. Long term, it will probably lead to more discovery rather than less.
The worry that I have, and I haven’t seen this talked about enough, is that humanity will stop understanding how things actually work. If we delegate all scientific work to the machine, we’re not incentivized to actually understand why things are the way they are. Those who spend their time intimately studying processes and phenomena will consequently be the ones that contribute the least to progress because their work will inherently be slower and limited by their own mind’s throughput.
Losing our knowledge of how things work can be truly catastrophic and set us back heavily (or possibly end civilization) if for some reason AI stops working or goes out of our control.
- noisy_boy
We are not much but our minds. And we have made a not so great an abstraction of it. And now that it is doing things because we also have it scale beyond one human can muster, it is doing significant things. What else did we expect? This one isn't even anywhere as complex as the one we have been our ears. Wait till we use this one to make one better and use that to make one even better and so on. This is not even the most depressing line of the first paragraph of a totally depressing book.
- jpitz
Ditch diggers do not lament a shovel. Astronauts cannot circle the world without reliance on a host of industries, machines, computers, and software they could barely enumerate much less replace.
Why is this lever different?
- 3m4r
By the Silicon Age, we were promised mathematical highs
If only we'll trust in their oracles and surrender our lying eyes
But though Science was advanced, the method was retreating
And the Gods of the Copybook Headings said "Pudding is proved in th' eating."
- edvardaksnes
Surprising that this is not Brecht’s Galileo interpretation, at least not as far as the wikipedia page mentions. Eppur si muove!
- johnea
I have to wonder if human innovation is at an end, or if we're just seeing the transient rush of solutions as this new tech is brought to bear.
Obviously, the LLM providing the proof of the Riemann Hypothesis is trained on every human effort to solving the problem over the last 150 years.
Did the model prove the hypothesis? Or did it find an insight in previous work that was overlooked until now?
What happens once all of the overlooked insights are brought to light? (like the flood of current CVE discoveries).
Will the tech be able to proceed into previously unconsidered territory without the history of human effort, as training data, to mine for answers?
I think we'll only know once we get there. I certainly don't think the question is already answered...