Hardy's lecture
- Maker
- G. H. Hardy
- Date
- 1920
- Form
- Lecture
Riemann guessed that every zero sits on one line. For over a century, people tried to prove even part of that.
The score is the share of zeros proved to be on the line. Scroll down and watch it climb.
There's more in the Run Data Run Deep Dive.
The race up the line
1859 · Bernhard Riemann
A guess, and nothing proved
Riemann wrote that it was "very likely that all zeros are real", which in his notation means all of them on the line, and moved on.
At this point every zero is a grey bar. Each one is somewhere in the strip, and nobody has proved where.
1914 · G. H. Hardy
Infinitely many on the line
Hardy proved that infinitely many zeros sit exactly on the line. With Littlewood he sharpened it in 1921.
Infinitely many can still be a vanishing share of all of them, the way the perfect squares go on forever and get rarer and rarer. The strip lights zeros 1, 4, 9, 16 and so on to show the shape of it.
Six years later Hardy gave this room's lecture: Some famous problems of the theory of numbers.
1942 · Atle Selberg
A share that never shrinks
Selberg showed that a fixed positive share of the zeros lie on the line, however high you go.
The share was small, and the sliver here isn't to scale.
1974 · Norman Levinson
One third
Levinson got to one third, by relating the zeros of zeta to the zeros of its derivative. Refining his method is how most of the next half-century went.
1989 · Brian Conrey
Two fifths
Conrey pushed it to two fifths. Forty percent, and it would sit near there for another thirty years.
2020 · Pratt, Robles, Zaharescu, Zeindler
Five twelfths
Four mathematicians reached five twelfths, 41.6%. From Selberg's sliver to here took 78 years.
Look at the strip. More than half the zeros are still grey.
August 2026 · Claude, checked by Alpöge and Furman
The jump
In one step the bound went from 41.6% to about two thirds. Claude, an Anthropic model, found the argument, and Levent Alpöge and Ralph Furman verified it.
Anthropic said at the time: "We don't expect that the techniques Claude used will lead to proving the Riemann hypothesis."
20 November 2026 · Levent Alpöge, with Claude
All of them, claimed
Three months later Alpöge announced a claimed proof of the whole thing, and of its cousin for Dirichlet L-functions, written in Lean, a language computers can check, so anyone can verify it.
This claim is bigger than a full bar, which could still leave room for a few zeros off the line. Scientific American wrote in August: "Even if Claude had bumped the number all the way to 100 percent, the LLM wouldn't be eligible for the $1 million." The claim is every zero, no exceptions.
Room 9 is about whether it holds up. Skip ahead, or keep walking.