---
title: "Riemann Hypothesis Attempted Twice in One Day: With and Without AI"
description: "On November 20, two independent attempts to prove the Riemann hypothesis appeared almost simultaneously: one using AI and one entirely human. Errors were found in Masullo's work just 8 hours later."
date: 2026-11-21T18:16:25.881Z
lang: en
url: https://xab.info/en/posts/riemann-hypothesis-attempted-twice-one-day
tags: [riemann-hypothesis, artificial-intelligence, mathematics, anthropic-claude]
author: "Alexander Buran"
publisher: "XAB.info"
---

# Riemann Hypothesis Attempted Twice in One Day: With and Without AI

![Blueprints and formulas of the Riemann hypothesis on a board and paper in low light](https://xab.info/media/2026/11/21/gipotaza-rimana-dve-popytki-dokazat-v-odin-den/gipotaza-rimana-dve-popytki-dokazat-v-odin-den-1.webp)

## Key Points

- On November 20, 2026, two independent attempts to prove the Riemann hypothesis appeared almost simultaneously: with AI (Claude, Levent Alpöge from Anthropic) and without it (Alessandro Masullo's GRH proof published by Ravi Vakil).
- Masullo's work was reviewed for several months by Brian Conrad, who initially evaluated the proof positively; it was decided not to delay publication.
- Eight hours after publication, several mathematicians reported errors discovered using various AI models.
- Vakil emphasizes that the proof has not yet been verified and the strategy's correctness is not confirmed: publication does not mean the hypothesis is solved.

On November 20, 2026, the mathematical community witnessed a rare coincidence: nearly on the same day, two independent attempts emerged to solve one of mathematics' major unsolved problems—the Riemann hypothesis. The first involved the use of artificial intelligence, while the second was performed entirely by a human. Both events quickly attracted the attention of specialists and sparked a fresh discussion on the role of AI in mathematical research.

### Two Attempts in One Day

First, Levent Alpöge, a researcher at Anthropic, discussed the work using AI, sharing the process of solving a mathematical problem with the help of the Claude model. On the same day, Ravi Vakil, President of the American Mathematical Society, published a preliminary version of a proof for the Generalized Riemann Hypothesis (GRH) proposed by mathematician Alessandro Masullo. Unlike the first effort, this work was created entirely by a human, without the involvement of neural networks.

### The Human Proof and Preliminary Review

Masullo's work had been reviewed for several months by mathematician Brian Conrad, who provided the author with detailed comments. According to Vakil, Conrad carefully studied the proof and gave it a positive initial assessment. The original plan was to spend about a year jointly refining the text at Stanford and then submit the manuscript for review by a broader circle of experts. However, Vakil decided not to delay publication and to present the current version, expecting that mathematicians would be able to promptly verify the arguments, formalize them, and eliminate minor gaps if necessary.

### Eight Hours to the First Remarks

The situation changed just eight hours after publication: several mathematicians informed Vakil of errors in the proposed proof. Notably, various artificial intelligence models helped discover them. Thus, a paper that the author purposely prepared without AI encountered remarks found with its assistance during the independent review stage. Vakil emphasized that all identified issues must be resolved to the mathematical community's satisfaction before the proof can be verified.

### Publication Does Not Mean Solving the Hypothesis

Vakil specifically noted that he is not yet prepared to claim even the correctness of Masullo's proof strategy. The publication of the manuscript does not mean that the Generalized Riemann Hypothesis has been solved: the work still requires full mathematical verification. The limitations outlined by the author himself right after publication emphasize that this is a preliminary version, not a completed result.

### Context: Why the Riemann Hypothesis Matters

The Generalized Riemann Hypothesis is an extension of the classical Riemann hypothesis, one of the most famous unsolved problems in mathematics, related to the distribution of prime numbers and the behavior of special mathematical functions. The simultaneous appearance of two independent attempts to prove it—using AI and exclusively human methods—became an unusual event for the mathematical community and clearly demonstrated how quickly the practice of verifying complex proofs is changing.

## Fact-Check & Sources

- [Гипотезу Римана попытались доказать дважды с публикациями в один день: с ИИ и без него](https://www.ixbt.com/news/2026/11/21/gipotezu-rimana-popytalis-dokazat-dvazdy-s-publikaciiami-v-odin-den-s-ii-i-bez-nego.html) - Primary source: both publications on Nov 20, 2026, names (Alpöge, Vakil, Masullo, Conrad), 8 hours until errors were found, Vakil's caveats about lack of verification.
- [Гипотеза Римана оказалась законом физики: как математика 165 лет назад предугадала динамические ...](https://www.ixbt.com/live/science/gipoteza-rimana-okazalas-zakonom-fiziki-kak-matematika-165-let-nazad-predugadala-dinamicheskie-fazovye-perehody-v-kvantovyh-sistemah.html) - Contextual source: hypothesis formulated by Bernhard Riemann in 1859, one of the main unsolved problems; used for historical context.

## FAQ

### What happened in mathematics on November 20, 2026?

Almost on the same day, two independent attempts to prove the Riemann hypothesis appeared: work using AI (Claude, Levent Alpöge from Anthropic) and a fully human proof of the Generalized Riemann Hypothesis by Alessandro Masullo, published by Ravi Vakil.

### Is the Riemann hypothesis proven?

No. According to Ravi Vakil, the published manuscript is a preliminary version that has not yet been verified; 8 hours after publication, errors were found in it, and the author does not claim the correctness of the proof strategy.

### What role did AI play in checking Masullo's work?

Several mathematicians who found errors in Masullo's proof used various artificial intelligence models to do so, even though the paper itself was prepared by the author without AI.