# OpenAI's ETH refutation

AI & tech · Heat 6.5 / 10 · 4 shows · 29m airtime

On October 27, 2026, OpenAI announced three results in computer science, math and physics, including a claimed refutation of the exponential time hypothesis via a deterministic sub-exponential algorithm for counting satisfying assignments of fixed-width CNF formulas. Russell Impagliazzo, Ramamohan Paturi and MIT's Ryan Williams are leading the public review of the disproof, while the evaluation group examining the ETH claim has signed an NDA and is unpaid. The math result was released in full with prompts and outputs; the physics result stays private as incomplete, with David Gross to lead the Yang-Mills write-up. OpenAI's math advisory group publicly recommended open access, and named physicists called the handpicked review group a disaster. The refutation remains a claim rather than an established…

Page: https://boilpoints.com/point/openais_eth_refutation_3896966387

## Segments

- [Tech Brew Ride Home](https://boilpoints.com/u/techbrewridehome), 2026-10-27, 1:25 to 5:24: **Covers OpenAI's claimed refutation of ETH**. Covers OpenAI's October 27 release of three results, including a claimed refutation of the exponential time hypothesis via a sub-exponential 3SAT algorithm. Notes the external shepherds and that the AGMAI panel recommended the day before against designating first recipients of output. New angle: OpenAI claims a deterministic sub-exponential 3SAT algorithm refutes the exponential time hypothesis.
- [Daily Tech News Show](https://boilpoints.com/u/dailytechnewsshow), 2026-10-27, 14:16 to 23:29: **Adds the NDA and unpaid evaluators**. Walks through the ETH claim and adds the terms OpenAI attached to it: the evaluation group signed an NDA and is unpaid, OpenAI will release raw data and fund a conference, and David Gross will lead the Yang-Mills write-up. Then a long discussion of credit and the existential reaction among mathematicians. New angle: The ETH evaluation group signed an NDA and is unpaid; David Gross will lead the Yang-Mills write-up.
- [TBPN](https://boilpoints.com/u/tbpn), 2026-10-27, 22:46 to 26:53: **Names the ETH disproof reviewers**. Reads a long explainer on the ETH claim and names the reviewers: the hypothesis's three proposers plus MIT's Ryan Williams are leading the public review. The rest is a recap of the announcement and what a disproof would mean for the field. New angle: Russell Impagliazzo, Ramamohan Paturi and MIT's Ryan Williams are leading the public review of the ETH disproof.
- [The AI Daily Brief](https://boilpoints.com/u/theaidailybrief), 2026-10-27, 1:34 to 5:13: **Walks all three October 27 results**. Walks through all three of OpenAI's October 27 results - local smoothing in every dimension, the ETH disproof, the partial Yang-Mills existence proof - and rounds up expert reaction. Adds how OpenAI handled release this time: full prompts for the math result, a working group for the CS result, and the physics kept private. New angle: The math result was released in full with prompts and outputs; the physics result stays private as incomplete.
- [TBPN](https://boilpoints.com/u/tbpn), 2026-10-27, 16:34 to 21:14: **Explains 3SAT and the Lean-checked proof**. Walks through the ETH claim with a whiteboard explainer: what 3SAT is, why the exponential blowup was treated as a law of nature, and that the proof was checked in Lean. It stays inside the announcement and the review already on the record, then pivots to speculation about unshared commercial work.
- [The AI Daily Brief](https://boilpoints.com/u/theaidailybrief), 2026-10-28, 8:24 to 11:36: **Adds the advisory group's open access push**. Follows up on the disclosure fight: OpenAI's own math advisory group publicly recommended open access, the appointed physicists defended holding the Yang-Mills result back, and named professors called the handpicked group a disaster. New angle: OpenAI's math advisory group publicly recommended open access, and named physicists called the handpicked review group a disaster.