Mathematics is beautiful. Definitions expose just the right level of detail and structure, and then, through a chain of immaculate deduction, we are led from propositions and lemmas to the pièce de résistance: the theorem. In just three years of undergraduate study, I’ve been fortunate enough to encounter many examples of this: Gauss’s Theorema Egregium, the classification of covering spaces, and the central limit theorem, to name a few. But beyond the formal chain of deduction lies something more profound. Mathematics is an art full of elegance, surprise, and inevitability.
I am in good company. Hardy, in A Mathematician’s Apology, writes that “beauty is the first test: there is no permanent place in the world for ugly mathematics”.1 Poincaré goes further: the scientist “takes pleasure in it because it is beautiful”.2 And Russell saw in mathematics the combination of “godlike freedom, with the sense of inevitable destiny”.3
Over the course of my undergraduate degree, AI progress in mathematics has been remarkable. When ChatGPT first appeared in November 2022, there was a striking asymmetry between its dexterity with words and its capability at mathematics.4 A model that could produce fluent prose could fail at basic arithmetic.5 At the time, it felt almost blindingly obvious that a machine would never understand the intricate structure and logic of mathematics.
Progress did not care what we thought. In 2024, Google DeepMind announced AlphaGeometry, which solved 25 of 30 Olympiad geometry problems.6 In 2025, Gemini Deep Think reached gold-medal standard at the International Mathematical Olympiad (IMO).7
Surely, we thought, models would plateau: IMO problems were still not real problems. Then came the slew of Erdős problems8 —still, we said, not real math. The Jacobian conjecture, open for 87 years, fell to Anthropic’s Claude Fable 5.9 It was announced almost casually, in a tweet. That captured the absurdity of the last two years: progress had been so fast that results that would once have felt like science fiction barely moved us at all. Somewhere along the way, an inversion happened: first I was much smarter than the models; now they were much smarter than I.
On September 8, 2026, came the final, shattering blow: OpenAI announced that an internal model had produced a paper claiming to resolve the Navier–Stokes smoothness and existence problem: in an incompressible fluid under a smooth external force, a blow-up existed in finite time.10 Despite having seen rumours circulating the previous day11, this came as a huge shock. I remember, unable to think very much at all, taking the train to Kamakura and running along the coastline in the rain—I just wanted to get away from it all.
But mathematics remained beautiful. Gauss’s Theorema Egregium, the classification of covering spaces, and the central limit theorem had not changed because OpenAI had resolved Navier–Stokes. So why was I grieving? Was it not absurd that the proof of a single theorem made me want to run away from it all?
I wrote to an older professor, an expert in harmonic analysis whose work concerns the Navier–Stokes equations, about my worries. I told him that I grieved to see our art being taken over by machines and behemoth AI companies looking for their next marketing stunt. Greater still was the loss of identity: the skill I had spent years cultivating was suddenly no longer rare, and perhaps no longer valuable either.
His reply contained a thought that struck me deeply. “Beautiful things, poetry or Mathematics, remain beautiful even though machines can copy us in uncanny ways. Contemplation of beauty is always marvellous,” he wrote, “but most of us would not be content with pure contemplation.”
Why, though, are we not content with pure contemplation?
And here, I feel, lies the ugly truth about mathematics, and one that I refused to recognise for so long: mathematics is as much about ego, pride, and ambition as it is about beauty.
Even in my own case, my reasons for starting mathematics were far from a monastic pursuit of beauty. I was good at it, and I loved being good at it. At school, there was nothing quite like the exhilaration of doing well in math competitions and math Olympiads: math, hard math, was something that I was not just good at—but much better at than others. It flattered my ego.
I am hardly alone in this. Hardy made a similar reflection, writing that his motives for doing mathematics were “far from noble”: “I wanted to beat other boys, and this seemed to be the way in which I could do so most decisively”.12 John Conway, upon discovering the surreal numbers, described spending six weeks thinking not only how beautiful the mathematics was but “how clever you are, John, to have discovered it”.13 And Timothy Gowers, in a recent blog post, wrote that his “main motivation for becoming a mathematician was the dream that I would solve unsolved problems—the more famous the better”; perhaps, he wondered, solving problems had simply been a way of “getting respect from a small group of peers.”14
AI begins to separate these two aspects of mathematics. For most of its history, beauty and achievement were intertwined: great mathematical work was both an encounter with something beautiful and the culmination of the mathematician’s own ambition, effort, and creativity. AI threatens the latter far more than the former.
I found this thought deeply reassuring. One day soon, machines may surpass us in all manner of intellectual tasks. But one thing they can never take away from us is our capacity to contemplate beauty—not just in math, but in music, literature, and more—nor our capacity to fall deeply into thought. Mathematics does not become less beautiful because machines are better at it than us.
And yet there is something unsatisfying about this reassurance. It is too simplistic to label all that is threatened here “ego”. We want to do more than just gaze at mathematical beauty. We want to take part in this grand game ourselves: to struggle through hard problems, spend months or years in deep thought, and leave our mark on this centuries-long human endeavour. To be human is to want to achieve great things, and this is much more than ego, hubris, or vanity.
In my reply, I told the professor that if I had the humility to let go of my ego, what I loved most about mathematics remained intact.
A few days later, he wrote back. He agreed that there was truth in what I said, but added one final qualification:
“We might be called to let go of our ego, but there is a thin line between ambition (in its positive meaning) and pride (in its negative meaning).”
Received 2 November 2026.
- Hardy, A Mathematician’s Apology, §10. ↩︎
- Poincaré, Science and Method, p. 22. ↩︎
- Russell, letter to Gilbert Murray. ↩︎
- OpenAI, ChatGPT announcement. ↩︎
- Yuan et al., ChatGPT and arithmetic. ↩︎
- Google DeepMind, AlphaGeometry. ↩︎
- Google DeepMind, Gemini at the IMO. ↩︎
- Feng et al., Gemini and Erdős problems. ↩︎
- Alpöge, Jacobian counterexample announcement. ↩︎
- OpenAI, Navier–Stokes announcement. ↩︎
- Buckmaster, statement on the Navier–Stokes developments. ↩︎
- Hardy, A Mathematician’s Apology, §29. ↩︎
- Conway, interview on surreal numbers. ↩︎
- Gowers, on the Fields medallists’ letter. ↩︎

Add to the discussion