OpenAI unveils three AI-generated research results spanning ETH, quantum field theory and the Kakeya conjecture

OpenAI unveils three AI-generated research results spanning ETH, quantum field theory and the Kakeya conjecture

N
News Editor
2026-10-28 05:04:37
OpenAI said on Oct. 26 that its internal models produced three research results across computer science, physics and mathematics, touching long-standing foundational questions in each field. The company described the documents as "raw research outputs" and said the academic community is being invited to verify, interpret and build on them. In computer science, OpenAI said its system proved the Exponential Time Hypothesis, or ETH, false, and that the result was later verified in Lean. The company stressed that this ETH refers to the complexity-theory hypothesis, not Ethereum. MIT professor Ryan Williams, according to the report, had asked OpenAI to examine whether the result could affect cryptographic systems, and OpenAI concluded with some confidence that it does not imply existing cryptosystems have been broken. In physics, OpenAI said the model produced rigorous existence proofs for pure Yang-Mills theory, quantum chromodynamics and a grand unified theory containing Standard Model gauge fields, though only the Yang-Mills portion has completed Lean verification. In mathematics, the company said its model proved the local smoothing conjecture in all dimensions, which in turn resolves the Kakeya conjecture and the Stein restriction conjecture in all dimensions. OpenAI also said it does not claim academic credit for discoveries generated by internal models, while noting a previously announced $5 million donation to the Simons Foundation to support workshops, conferences and related programs.

OpenAI said on Oct. 26 that its internal models generated three research results in computer science, physics and mathematics, each tied to long-studied foundational questions. The company described the papers as "raw research outputs" and invited the academic community to verify, interpret and extend them.

Computer science: OpenAI says ETH is false

OpenAI said its system proved the Exponential Time Hypothesis, or ETH, does not hold, and that the result was later verified in Lean.

The ETH in this case is the Exponential Time Hypothesis from theoretical computer science, not Ether or Ethereum. The hypothesis has long served as a basic pessimistic assumption in complexity theory: for a certain class of logical problems, no matter how algorithms are improved, worst-case running time still grows explosively with input size, leaving no true shortcut.

If that hypothesis falls, the boundary of what may be computed efficiently has to be reconsidered. OpenAI said the result reopens the question of how much faster such tasks could become, while adding that turning a theoretical advance into better practical algorithms would still require substantial follow-up work.

For crypto readers, the immediate question is whether existing cryptographic systems are affected. Many cryptographic schemes rely on the hardness of specific computational problems. MIT professor Ryan Williams had asked OpenAI to examine that issue. According to his account, OpenAI checked the implications and believes with some confidence that the result does not mean existing cryptosystems have been broken.

Physics: a mathematical foundation for quantum field theory

In physics, OpenAI said the model produced rigorous existence proofs for pure Yang-Mills theory, quantum chromodynamics, or QCD, and a grand unified theory that includes Standard Model gauge fields.

Physicists use quantum field theory to describe the particle world, and its predictions are highly accurate, but its mathematical footing has long been viewed as incomplete. In OpenAI's framing, these results amount to supplying a rigorous foundation for theories that have been used for decades.

So far, only the Yang-Mills portion has completed Lean verification. The announcement did not say whether the QCD portion has been verified in the same way. The original AI-generated proof has not been released publicly. Instead, it is being studied first by a group led by Nobel laureate David Gross, which is writing its own paper, leaving outside researchers unable to test the result independently for now.

Mathematics: local smoothing, Kakeya and Stein restriction

OpenAI also said its model proved the local smoothing conjecture in all dimensions, and from that result resolved the Kakeya conjecture and the Stein restriction conjecture in all dimensions.

The local smoothing conjecture was proposed by Sogge in 1991 and studies how waves behave when they focus. A simple analogy is water waves: several ripples moving toward the same point can briefly stack into a high crest, but that extreme moment does not by itself describe the average behavior over time.

The conjecture seeks precise estimates linking those short-lived extremes with time-averaged behavior. Because the phenomenon is considered fundamental, a proof has consequences for other mathematical problems, including the Kakeya conjecture and the Stein restriction conjecture.

The Kakeya conjecture is often explained through a needle-setting problem. Imagine placing many needles of equal length but zero thickness in space so that every direction has at least one needle. The needles may be translated and may overlap. The question is how small the set containing them can be.

The key notion of size here is dimension. In three-dimensional space, the Kakeya conjecture says such a set must have full three-dimensional dimension; in higher-dimensional spaces, it must likewise have the corresponding full dimension. Dimension is not the same thing as volume, so the problem cannot be reduced to how many cubic meters those needles occupy.

Before this announcement, local smoothing in two dimensions had been proved in 2019, and the Kakeya conjecture in three dimensions had been proved in 2025 by Hong Wang and Zahl. OpenAI said its new result extends the proof of the local smoothing conjecture to all dimensions and, from there, yields the related conjectures in all dimensions as well.

OpenAI says it will not claim academic credit

OpenAI said it does not claim academic credit for discoveries produced by its internal models. It added that researchers who contributed to understanding the results, writing papers and preparing materials should receive recognition.

The company also referred to its previously announced $5 million donation to the Simons Foundation. The foundation will manage and distribute the funds for workshops, academic conferences and dedicated programs aimed at helping researchers understand AI-generated results and pursue further work based on them.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
5900

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.