On November 20, 2026, mathematician Levent Alpöge announced that he and Claude had apparently solved the Riemann Hypothesis, one of mathematics’ greatest unsolved problems. In fact, they claim to have solved the stronger Generalized Riemann Hypothesis as well.
The proof? Seventeen million lines of Lean code, a language for formally verifying mathematical proofs. An attempt to render it in ordinary mathematical prose ballooned past a quarter of a million pages.
Let’s assume the proof is correct and will be accepted.
One of the greatest mysteries in mathematics has been settled.
Except nobody really understands why it’s true.
The reaction has been extraordinary. Mathematicians are stunned, some exhilarated, others despairing. Some have even called it the mathpocalypse.
If AI can solve problems like Riemann, what’s left for mathematicians? Have they just watched their defining intellectual achievement become obsolete?
Perhaps the panic rests on a misunderstanding of what mathematicians are actually for.
And perhaps we’ve gotten the analogy between mathematics and science almost exactly backward.
THE WRONG ANALOGY
We usually think of mathematics as something like science, except with much higher standards of certainty.
Scientists propose hypotheses. Mathematicians propose conjectures.
Scientists gather evidence. Mathematicians seek proofs.
In science, sufficient evidence can establish a theory beyond reasonable doubt. In mathematics, a proof establishes a theorem beyond logical doubt.
The analogy seems obvious. Theorems correspond to scientific theories, and proofs to the evidence supporting them.
But I think we’ve matched up the wrong things.
Consider Kepler’s laws of planetary motion.
Drawing on Tycho Brahe’s enormous collection of astronomical observations, Kepler discovered three beautifully simple mathematical laws describing how planets move.
The observations established the regularities. But why should planets behave that way?
Kepler proposed his own explanation, involving a magnetic-like force emanating from the rotating Sun.
He was wrong.
It took Newton to provide a deeper explanation, deriving Kepler’s laws from principles that also explained falling apples, ocean tides, and much else.
The laws were right. Kepler’s explanation wasn’t. Newton’s explained more.
That’s science.
And this is the better analogy with mathematics.
Mathematical theorems are more like empirical laws than explanatory scientific theories.
An empirical law is a concise characterization of a regularity, backed by observations.
A mathematical theorem is a concise characterization of a regularity, backed by a rigorous proof.
Of course, a proof establishes logical necessity, whereas observations cannot. But the intellectual roles are similar.
The Riemann Hypothesis, then, wasn’t an explanatory hypothesis. It was a conjectured mathematical law.
The law has its proof.
But why is it true?
MATHEMATICS HAS UNEXPLAINED LAWS, TOO
Sometimes a proof is also an explanation.
Consider the theorem that the first n odd numbers add up to n². You can prove it algebraically. But arrange those odd numbers into successive L-shaped layers, each completing a larger square, and you can see why it’s true.
The same proof establishes the law and explains it. Mathematics has enjoyed this luxury for thousands of years, making the two jobs easy to conflate.
The distinction itself isn’t new. Aristotle recognized the difference between knowing that something is true and knowing why. Mathematicians have long admired illuminating proofs over those that merely demonstrate.
But appreciating that some proofs aren’t explanatory isn’t the same as recognizing that mathematical theorems occupy the same intellectual role as empirical laws.
Mathematics has always had unexplained laws. It just called them theorems.
And there’s another wrinkle that brings mathematics and science even closer.
MAYBE GRAVITY IS A THEOREM
Sometimes scientific phenomena turn out to be mathematical consequences of something deeper.
Thermodynamics, for example, once looked like a collection of independent physical laws. Statistical mechanics showed that much of it follows from probability and combinatorics.
Noether showed how conservation of energy, momentum, and angular momentum follows mathematically from corresponding symmetries of the physical laws.
These derivations still depend on physical assumptions. But again and again, apparently independent facts about nature turn out to follow mathematically from deeper principles.
For all we know, this continues all the way down.
Perhaps gravity itself is a theorem we haven’t yet recognized as one. Perhaps much of the universe, maybe even the entire thing at some fundamental level, is one gigantic mathematical tautology.
Not every scientific phenomenon need be mathematically inevitable. Our particular visual illusions, for example, depend on the contingent biological machinery evolution built.
Some empirical laws may themselves be theorems in disguise.
MATHEMATICS GETS ITS PARTICLE ACCELERATOR
We got a preview of the coming predicament in 1976, when a computer-assisted proof established the Four Color Theorem.
The result was settled, but many mathematicians were dissatisfied. The proof checked a large collection of configurations without revealing a simple principle explaining why four colors suffice for any planar map.
The mathematical law was established. A satisfying explanation was still missing.
Now imagine AI producing thousands, eventually millions, of rigorously proved mathematical laws. Some proofs elegant. Others running to millions of steps nobody can meaningfully understand.
The theorems themselves might be beautifully simple. Unexpected symmetries, remarkable regularities, surprising connections.
All proved beyond logical doubt.
And almost entirely unexplained.
It’s like science acquiring a powerful new particle accelerator, generating enormous quantities of reliable findings for theorists to explain.
Not that mathematics is becoming empirical. These are rigorous proofs, not simulations.
But mathematics is acquiring science’s enormous gulf between knowing and understanding.
THE SCIENCE OF MATHEMATICS
What will the explanations look like?
Some will surely be new proofs, deriving the original mathematical laws from simpler, deeper principles.
But they needn’t be.
Mathematicians might propose explanatory conjectures. Conjectures not about what is true, but about why.
In fact, there’s already a proposed explanation for Riemann.
It’s called the Hilbert–Pólya conjecture, and I won’t pretend to understand the mathematics. But the basic idea is that Riemann’s mysterious regularity follows from some deeper mathematical structure, of a kind familiar in quantum physics.
There’s even tantalizing evidence for it. The spacing of Riemann’s zeros resembles patterns found in certain quantum systems.
Nobody has established that this hidden structure exists. But notice what’s happening. Mathematicians have proposed an explanation and found independent reasons to take it seriously.
Alpöge and Claude’s enormous proof doesn’t settle whether Hilbert–Pólya is right.
The law is settled. The explanation isn’t.
We might favor one mathematical explanation over others because it’s simpler, unifies more laws, or successfully predicts new mathematical regularities.
Mathematicians already make judgments like these. Bayesian judgments, even.
The proposed mathematical principles are necessarily true or false, but we don’t know which. Uncertainty about necessary truths is still uncertainty.
What’s changing is the scale. Such uncertain explanatory theorizing may increasingly become the central business of mathematics.
MATHEMATICIANS, MEET SCIENCE
Mathematicians have traditionally worn two hats.
One is analogous to the experimentalist’s: discovering mathematical regularities and establishing them beyond doubt. Humans or AIs producing non-explanatory proofs are doing this job.
The other is the explanatory scientist’s: discovering deeper principles that explain those regularities, preferably many otherwise unrelated laws at once.
Sometimes a beautiful proof accomplishes both, concealing the distinction.
But AI may now make the first job so productive that the second becomes the great bottleneck in mathematics.
Which may feel like the end of mathematics. But it could instead be the beginning of mathematicians taking their explanatory role much more seriously.
The machines may increasingly establish what’s true.
Mathematicians will have to figure out why.
Mathematicians were scientists all along.
AI may finally make them realize it.


