AIToday
Large Language ModelsAI Safety & AlignmentZenn AI/MLPublished: Nov 21, 2026, 16:00 JST

Two Riemann proofs in one day: 17 million lines vs 12 years

Send on LINE
Two Riemann proofs in one day: 17 million lines vs 12 years

3 Key Points

  1. What happened

    Levent Alpöge published a 4-page remark.pdf saying he believes he found a Riemann Hypothesis proof with Claude, formalized in Lean with no assumptions across roughly 175,000 files and 17 million lines.

  2. Why it matters

    The Lean formalization means the proof's correctness can in principle be machine-checked, but Alpöge himself notes the human-readable version is still being written and may contain odd choices, so whether the build actually passes is the open question.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The two proofs map onto two long-standing approaches to the Riemann Hypothesis. One, in the tradition of Weil and Deligne, tries to read the zeros as eigenvalues of a space and force them onto the critical line through positivity; the other, following Stepanov and Bombieri, bounds the count of points with an auxiliary polynomial and squeezes the error down to sqrt(X). Neither approach had been carried over to the integers in 160 years, and the toy version over finite fields was solved in the 20th century with exactly these two strands. What is striking is that the strengths and weaknesses of the two new attempts are mirror images: Alpöge's is machine-checkable but its human-readable form is still being written and may hide strange choices, while Masullo's is humanly readable but its details are not yet settled, with Brian Conrad having spent months reading it and finding it 'looking good so far' despite some gaps being fixed. A broader shift is that being able to confirm correctness and understanding why something is correct are coming apart, and mathematicians ultimately want the latter.

FAQ
What exactly did Levent Alpöge claim?
In a 4-page remark.pdf, Alpöge says he believes he found a proof of the Riemann Hypothesis together with Claude. The proof is formalized in Lean without assumptions, and he says the same method reaches GRH for Dirichlet L-functions.
Has anyone verified the Lean proof builds?
Many reports say the build passes, but the author and Claude did not build it locally because 17 million lines was too large to run on their machine. The absence of sorry, axiom, and native_decide in the repository means the remaining question is whether it truly compiles.
What is different about Alessandro Masullo's proof?
Masullo spent 12 years on it starting from his Stanford graduate years, entirely by hand without AI. The manuscript is titled The Music of the Primes, runs over 900 pages, and claims GRH for Hecke L-functions over all number fields, a broader scope than Alpöge's Dirichlet case.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articlenference, Cedars-Sinai to pool 5M patients' records for AI research