As of: 2026-10-08 23:45 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/v/ai-revolution-openai-just-broke-math/

# OpenAI Just Broke Math With Its Most Powerful AI Yet

AI Revolution, 7 October 2026, YouTube. 90,677 views as of 8 October 2026.
Kind: Review.
Watch: https://www.youtube.com/watch?v=WYpzvsC4qTk

## Why it is here

AI Revolution explainer on the 722 math papers, 'the unreleased model sparks outrage among mathematicians'. ~90k views. Length 13:38.

## Description (written by Gemini from the video)

**Summary**  
This video, produced by the channel *AI Revolution* with voiceover narration and motion graphics, reports on OpenAI’s October 6, 2026 release of 722 AI-written mathematics papers to GitHub. It breaks down the claimed breakthroughs—ranging from the Quasi-Riemann hypothesis to the Unique Games Conjecture—and examines the intense friction, priority disputes, and ethical pushback emerging from the academic mathematics community.

**What is shown**  
- [00:10] The GitHub repository `openai/math` displaying the manuscript collection index of 722 papers across 372 result families.  
- [00:49] The Lean interactive theorem prover website and code interface verifying mathematical proofs.  
- [00:55] Title and abstract of the preprint *"The Quasi-Riemann Hypothesis: A Zero-Free Half-Plane $\Re s > 7/8$"* dated September 30, 2026.  
- [01:29] Clay Mathematics Institute pages covering Millennium Prize problems, including the Navier–Stokes equation and the Hodge conjecture.  
- [01:39] Interface clips of ChatGPT Work and Codex completing tasks with multi-hour thinking runs.  
- [01:53] Sam Altman’s post on X announcing *"Sharing AI progress in mathematics"*.  
- [03:18] Title and abstract of *"The Unique Games Theorem"* manuscript dated September 23, 2026.  
- [04:23] Paper excerpt claiming proof that Catalan's constant is irrational.  
- [04:49] Reporting from *The Washington Post*, *Wired*, and *The New York Times* documenting tensions between mathematicians and OpenAI researchers.  
- [06:54] Excerpts from open statements, including *"A Severe Misalignment of AI in Mathematics"* and recommendations from the Advisory Group on Mathematics and Artificial Intelligence.  
- [09:52] Christian Szegedy’s essay on X analyzing the future trajectory of human mathematicians versus AI systems.

**Claims & numbers**  
- The narrator states that on October 6, OpenAI uploaded 722 mathematical manuscripts grouped into 372 result families to GitHub (`openai/math`) [00:10, 00:35].  
- The narrator states that 235 of those families include formal, computer-checked proofs in the Lean programming language [00:46].  
- The narrator says OpenAI evaluated approximately 4,000 real unsolved math problems, publishing only the significant results [01:10].  
- The narrator states the model began training on August 28, solved a Navier–Stokes problem on September 5 using 10,000 AI agents running for 88 hours, but for this release nearly every result was produced by a single agent with a single prompt averaging ~3 hours of ChatGPT Pro thinking [01:23, 01:35, 01:43].  
- The narrator states the model proved a zero-free half-plane $\Re(s) > 7/8$ for Dirichlet L-functions and uniformly excluded Landau–Siegel zeros [00:55, 02:24].  
- The narrator mentions that roughly 40 mathematicians were convened in a closed-door summit by OpenAI in August [04:49].  
- The narrator reports that the Leiden Declaration on AI in mathematics garnered signatures from over 4,000 mathematicians [09:01].

**Notable quotes**  
- [05:13] Bryna Kra: *"The room felt like extreme excitement and extreme fear at the same time."*  
- [06:27] Sébastien Bubeck (reported quote): *"If you don't want me to be polite, I don't have to be polite."*  
- [08:37] Andrew Sutherland: *"We should demand to see the receipts."*

**Assessment**  
This video is a journalistic summary and commentary aggregating public preprints, news reporting, and community statements. It does not contain an independent hands-on evaluation of the underlying proprietary frontier model, relying instead on OpenAI's repository disclosures, academic preprints, and press coverage.

_Described by gemini-3.8-flash on 2026-10-08 from the video's audio and frames._

## Related

- 2026-10-06: [OpenAI releases 722 AI-written math manuscripts claiming hundreds of open problems](https://postcutoff.com/e/2026-10-06-openai-math-release-722-manuscripts/)

## People in it

- [Sam Altman](https://postcutoff.com/person/sam-altman/), CEO, OpenAI
- [Sébastien Bubeck](https://postcutoff.com/person/sebastien-bubeck/), Researcher, OpenAI