How mathematicians spent 167 years closing in on the Riemann Hypothesis, and how Claude entered the story in 2026.
Last updated November 20, 2026
The Riemann Hypothesis has a longer paper trail than almost any problem in mathematics. Below is the story in five eras, from Euler's first link between primes and an infinite sum, through a century of partial progress, to the 2026 results produced with Claude. Human milestones carry dark dots, results produced with Claude carry terracotta dots, and the unverified November 2026 claim is marked in gold.
Human mathematics
Results produced with Claude
The November 2026 claim (unverified)
Context
Before the hypothesis (1737 to 1858)
The idea that a smooth function could encode the primes starts with Euler, a century before Riemann.
Euler links primes to an infinite sum
Leonhard Euler shows that 1 + 1/2^s + 1/3^s + … equals a product taken over all prime numbers, the first bridge between a smooth function and the primes.
Gauss guesses how dense the primes are
A teenage Carl Friedrich Gauss notices that primes near x occur with frequency about 1/ln x, the seed of the prime number theorem.
Riemann's conjecture (1859 to 1901)
One short paper sets the agenda for the next century and a half.
Riemann's only paper on number theory
Presented to the Berlin Academy, On the Number of Primes Less Than a Given Magnitude extends ζ(s) to the complex plane, ties the primes to its zeros, and remarks that all non-trivial zeros very likely have real part ½.
The prime number theorem
Jacques Hadamard and Charles de la Vallée Poussin independently prove that the number of primes up to x is about x / ln x, by showing ζ(s) has no zeros on the line Re(s) = 1.
Hilbert's eighth problem
David Hilbert puts the Riemann Hypothesis on his list of 23 problems for the 20th century.
Von Koch's equivalence
Helge von Koch proves that RH is equivalent to the best possible error term in the prime number theorem.
A century of partial progress (1914 to 1999)
Nobody proves the whole hypothesis, but the share of zeros known to lie on the line keeps rising, and the analogue over finite fields is settled.
Infinitely many zeros on the line
G. H. Hardy proves that the critical line contains infinitely many zeros.
Riemann's hidden calculations
Carl Ludwig Siegel finds in Riemann's unpublished notes a powerful method for computing zeros, now called the Riemann–Siegel formula.
RH over finite fields, for curves
André Weil proves the analogue of the Riemann Hypothesis for curves over finite fields, with the full proof published in 1948.
A positive proportion
Atle Selberg proves that a positive fraction of all non-trivial zeros lie on the critical line.
Stepanov's elementary method
Sergei Stepanov finds an elementary route to Weil-type results using auxiliary polynomials that vanish to high multiplicity. Alpöge says his 2026 argument descends from this idea.
Zeros behave like random matrices
Hugh Montgomery's pair-correlation work, and a chance conversation with physicist Freeman Dyson, reveal that the spacings between zeros match eigenvalues of random matrices.
One-third, and Deligne
Norman Levinson proves that at least one-third of the zeros lie on the line. Pierre Deligne completes the proof of the Weil conjectures, the Riemann Hypothesis analogue for varieties over finite fields.
Two-fifths
Brian Conrey raises the proven share of zeros on the line to at least 40%.
The modern era (2000 to 2025)
A million-dollar prize, a famous false alarm, and steady technical gains.
A $1 million prize
The Clay Mathematics Institute names the Riemann Hypothesis one of seven Millennium Prize Problems.
A famous claim that did not hold
Sir Michael Atiyah presents a claimed proof at the Heidelberg Laureate Forum. The mathematical community does not accept it.
Five-twelfths
Kyle Pratt, Nicolas Robles, Alexandru Zaharescu and Dirk Zeindler show that more than 5/12, about 41.7%, of the zeros lie on the line.
Trillions of zeros checked
David Platt and Tim Trudgian rigorously verify every zero up to height 3 × 10¹², more than 12 trillion zeros, all on the line.
Guth and Maynard
Larry Guth and James Maynard improve Ingham's 1940 bound on how many zeros can lie far from the line, the first major advance on it in about 80 years.
The year of Claude (2026)
AI systems start producing research-level mathematics, and the Riemann Hypothesis becomes a target.
The Jacobian conjecture
Alpöge shares a Claude-generated result on the Jacobian conjecture. Anthropic later describes it as a disproof.
Take a real stab
Anthropic staff member Jarred Sumner asks an unreleased Claude to attempt the Riemann Hypothesis itself. Its first session tries about 650 ideas; none work.
Over two-thirds of zeros on the line
Anthropic announces that Claude, coordinating about 60 subagents, raised the proven share of zeros on the line from 41.6% to 67.2%, with a Lean-verified proof.
The paper
Alpöge and Ralph Furman post arXiv:2608.13637, crediting the discovery of the proof to Claude. It also shows at least five-sixths of the zeros are distinct.
AI proofs under scrutiny
OpenAI's claimed proof for the Navier–Stokes problem sparks debate over whether machine-made proofs give mathematicians real understanding.
A human simplifies Claude
Live Science reports that Youness Lamzouri of the University of Lorraine found a shorter, more conceptual proof of the 67.25% result, and showed at most 11% of zeros can be neither simple nor on the line.
First step of the claimed proof
According to Alpöge's note, the work that led to the full claim begins with a Stepanov-style determinant.
Iterated theta lifts
Per the note, chains of theta lifts of length about √X supply the missing multiplicity.
GRH
The final entry on the note's timeline: the Generalized Riemann Hypothesis.
The announcement
Alpöge posts his note and a 14-part Lean archive. Independent verification begins.
What 167 years of attempts tell us
Three patterns stand out in the Riemann Hypothesis timeline.
Progress came in fractions. Mathematicians could not prove that every zero lies on the critical line, so they proved that more and more of them do: a positive share in 1942, a third in 1974, two-fifths in 1989, five-twelfths in 2020, and more than two-thirds with Claude in 2026. Each step needed a new idea, not just more computing power.
The strongest evidence came from neighbouring worlds. Weil's and Deligne's proofs of the Riemann Hypothesis over finite fields showed that the statement is true in a closely related setting. Alpöge says his new argument descends from Stepanov's finite-field method, which makes the November 2026 claim a direct heir of that tradition.
Famous claims have failed before. Atiyah's 2018 announcement is the best-known recent example, and many lesser-known proofs have been posted and forgotten. That history is why mathematicians insist on verification before celebration, and why the Lean formalization matters so much this time.