Explainer · updated 21 November 2026

A claimed proof of the Riemann Hypothesis

For 167 years, nobody could solve the most famous riddle in mathematics. On 20 November 2026 the mathematician Levent Alpöge said that he solved it together with Claude, an AI program. This page explains the riddle, the idea, and how sure we can be. You need no mathematics from school to read it.

Status: first checks passed, one check open Other mathematicians ran the proof through a checking program on their own computers. It passed. One last automatic comparison was still open on the morning of 21 November. No person understands the full proof yet.

The riddle

Prime numbers appear without a visible pattern. The Riemann Hypothesis says that there is a strict order behind them.

The claim

A proof that this order is real. A person set the direction. An AI did much of the hard work.

The surprise

The proof is not a text for people. It is a very large computer file that a program checks step by step.
01

The riddle: how are the prime numbers spread?

You can put 12 sweets into equal groups: 3 groups of 4, or 2 groups of 6. With 13 sweets this is not possible. Numbers such as 13 are prime numbers: 2, 3, 5, 7, 11, 13, 17. All other whole numbers are products of primes. This is why people call primes the building blocks of numbers.

Primes become less frequent when numbers get larger. Below 100, one number in four is a prime. Below one million, it is only one number in thirteen. But nobody can tell you by a simple rule where the next prime is.

There is a formula that predicts how many primes exist below a number X. The chart compares this prediction with the true count.

True count and prediction
true count of primesprediction of the formula
Primes up to X25
Prediction30.1
Miss−5.1
Square root of X10
Only the miss: true count minus prediction
misspermitted zone
Move the slider to the right. In the top chart, the two lines become almost one line. The bottom chart shows the small difference. It stays inside the green zone.

The Riemann Hypothesis says: the miss stays inside this zone forever, also for numbers with a thousand digits. The height of the zone is about the square root of X. For X = 1,000,000 the square root is 1,000, because 1,000 × 1,000 = 1,000,000.

A comparison: toss a coin one million times. You expect 500,000 heads. The true result is usually wrong by some hundreds, not by tens of thousands. Pure chance makes small errors of exactly this size. The hypothesis says that primes behave like these coin tosses. They look random, but no hidden force pushes them away from the prediction.

02

The hidden music of the primes

In 1859 the German mathematician Bernhard Riemann found out where the miss comes from. He used a formula with the name zeta function. You can think of it as a hidden sheet of music for the primes. The sheet contains special points, the zeros. Each zero is one tone.

The chart shows what the tones do. One tone alone is a simple wave. Add many tones, and the waves build a staircase. The staircase goes up exactly where the primes are.

The tones build the staircase of the primes
staircase of the primessum of the tones
With no tones, the red line is straight. With more tones, steps appear at 2, 3, 5, 7, 11 and each other prime. Large primes make higher steps. There are also small steps at 4, 8, 9 and other powers of primes.

Now the important question: where are these zeros? You can draw them on a map.

The first ten zeros on the map
0½1position from left to right 01020304050 14.1321.0225.0130.4232.9437.5940.9243.3348.0149.77 centre line

Each red dot is one zero. The number next to it is its height on the map. The height sets the pitch of the tone.

All ten dots are exactly on the red centre line. Riemann guessed: every zero is on this line, and there are infinitely many of them. This guess is the Riemann Hypothesis.

The grey cross shows what must never occur: a zero next to the line.

Computers found the first 10 trillion zeros. All are on the line. But this is not a proof. Zero number 10 trillion and one can still be different.

Why is the line so important? The position from left to right sets how loud a tone is. On the centre line, all tones stay quiet enough, and the miss from section 01 stays in the green zone. One zero to the right of the line is one tone that is too loud. Then the primes are less regular than everybody thinks.

03

Why it took 167 years, and what the new idea is

Mathematicians know a second, more simple world of numbers. It works like a model railway of the real number world: smaller, but with the same rules. This model world has its own primes and its own Riemann Hypothesis.

Solved

The model world

André Weil solved the riddle here in the 1940s. In 1969 Sergei Stepanov found a much shorter solution.

Open since 1859

The usual numbers 1, 2, 3, …

The plan for decades: copy the solution from the model world. It did not work, because one tool does not exist here.

The tool that was missing

The tool is a method to count one point more than one time. The pictures show the idea. A line can go through a point. It can also only touch the point, or stay flat on it. Mathematicians then count the point one, two or three times.

counts 1×goes through
counts 2×touches
counts 3×stays flat
Why this helps: Stepanov’s trick

Stepanov wanted to show that some special points are rare. He had a fixed budget, here 24 slots. If each point counts more than one time, each point uses more slots. Then fewer points can exist.

At most 8 points fit.

The more times each point counts, the smaller the number of possible points. A strict limit of this type solved the riddle in the model world.

The new idea: a long chain

In the model world, the multiple count is easy to get. For the usual numbers, nobody knew how to get it. The note says that a tool from a very different area of mathematics can do the job. Its name is theta lift. Each use of the tool adds one more count. So you use it again and again, as links in a chain.

primes up to 1,000,000: about 1,000 links
The length of the chain is the square root of X. This is the same number as the height of the green zone in section 01.

How long must the chain be? Claude proposed the square root of X. The author writes that this one line will astonish other mathematicians. His own comment in the note: “And yet, what else could it be?”

A limit of this page: the note does not explain the proof. This section gives the idea in the words of the note. The author also writes that he is not an expert in this part of mathematics. Here the AI knew more than he did.

04

Six weeks, one person, one AI

The note lists 20 steps with dates. The work started on 6 October and ended on 16 November 2026.

The 20 steps by date
One dot is one step. Eleven steps are in the first two weeks. Then almost four weeks have no dated step. Eight steps have the date 15 November, a Sunday. One day later the work was complete.
Levent Alpöge is a mathematician at Anthropic, the company that makes Claude. He writes that he prepared for this riddle since his student days. He selected the direction and the ideas to combine.
Claude did much of the technical work, mainly at the end. The author writes that the AI was far ahead of him in that phase and that it made important proposals.
Together they wrote the proof in a form that a computer can check. The author says openly that he does not yet understand all parts of it.
05

Can you trust a proof that no person has read?

The proof is a computer file with about 175,000 parts. A checking program with the name Lean examines each step. Think of a very strict auditor. The auditor does not understand the business. But the auditor recalculates each line and accepts nothing without a reason.

Such a check can still go wrong in two ways. The file can prove a different, easier claim than the real riddle. Or it can contain a hidden shortcut: a step that it assumes but does not prove. So other people must examine these points.

The short answer: the proof is very probably correct. Scientific American uses the words “almost certainly true” in its headline. But checked is not the same as understood. The next task for mathematicians is to find out why the proof works.

06

What this means for you

07

Short glossary

Prime number
A whole number larger than 1 that you cannot write as a product of two smaller numbers.
Zeta function
The formula of Riemann that connects all primes. In this page: the sheet of music.
Zero
A point where the zeta function has the value 0. In this page: one tone.
Lean
A program that checks a mathematical proof step by step. In this page: the auditor.
Generalized Riemann Hypothesis
A larger version of the riddle for a full family of similar formulas. The note says that the proof solves this version too.

Sources