OpenAI's AI discoveries in math, computer science and physics: what's verified

OpenAI says its models made three discoveries in math, computer science and physics, including disproving the Exponential Time Hypothesis. None is independently confirmed in the reports we read; an MIT expert says his team is still verifying the ETH proof.

Published October 28, 2026 5:51am EDT · 3 min read

Add ScoviAI on Google

OpenAI CEO Sam Altman speaking at TED (photo: Steve Jurvetson, CC BY 2.0, Wikimedia Commons)· 3 min read

Key points

Announced Monday, October 26, 2026 across harmonic analysis, complexity theory and quantum field theory
OpenAI says its models proved the local smoothing conjecture in every spatial dimension and disproved the Exponential Time Hypothesis
MIT's Ryan Williams says his team is still verifying the ETH proof
OpenAI calls the outputs 'raw research data' and uses three expert groups to review them
Math and CS data is public now; the physics group is delaying release

OpenAI says its AI models have produced three scientific results, in harmonic analysis, complexity theory and quantum field theory. An MIT complexity expert says his team is still verifying the proof of one of them, and in the sources we read, none of the three has been independently confirmed.

OpenAI announced the results on Monday, October 26, 2026, according to Al Jazeera's report. This article is based on that report; OpenAI's own post (its announcement) was not readable for us at the time of writing, so we attribute every claim to Al Jazeera's account of it.

What exactly is OpenAI claiming?

Per Al Jazeera's summary of the announcement:

Harmonic analysis: OpenAI's models proved the local smoothing conjecture, which Al Jazeera describes as a central proposition of this branch of mathematics, for every spatial dimension. Harmonic analysis is the study of breaking complicated signals into simple waves. Per Al Jazeera, earlier proofs covered only a single case, three dimensions; we could not check that detail ourselves.
Complexity theory: the models disproved the Exponential Time Hypothesis (ETH), which relates to the time it takes to solve complex computational problems.
Physics: the models provided a "mathematically rigorous" proof concerning quantum chromodynamics, the theory of how quarks and gluons interact. The exact scope of this claim is not spelled out in the report.

OpenAI describes the outputs as "raw research data" that outside scientists need to turn into understanding, so these are claims under review, not settled results.

Why would the ETH result be a big deal?

Ryan Williams, a computational complexity expert at MIT, told Al Jazeera that "the ETH has been widely believed true since the late 1990s and a great deal of theoretical computer science was built on it being true." He called a confirmed refutation a major development, but also said: "We are working to verify and examine their proof and determine if it holds up."

Williams said the result would not have real-world applications in the short term, but could lead to many "future real-world benefits for society." He described the proof as proposing "a mysterious new method" for solving a variety of fundamental problems more efficiently.

What do mathematicians say so far?

Alex Iosevich, a University of Rochester professor who works in harmonic analysis, called it a "historic development" and said OpenAI appeared to have "cracked" the conjecture. That is one expert's early reaction, not a formal verification.

How are the results being checked?

OpenAI says it asked three expert groups to act as "shepherds." The groups working on the mathematics and computer science results chose to make the AI-produced data public immediately. The physics group is holding back its release so it can write its own interpretation first.

OpenAI itself acknowledged trade-offs: choosing initial shepherds "shapes who gets involved first and creates asymmetry of information," and it invited feedback on broadening participation.

Is this part of a trend?

Al Jazeera notes it follows other AI-lab science claims. In September 2026, per Al Jazeera, Anthropic announced a new enzyme said to resemble CRISPR, which drew mixed reactions from scientists. For other lab results on our site, see Claude's scorpion-venom protein binders and Claude's formal verification of 123 libraries. OpenAI is also pushing into life sciences, as in its Orchestra Bio acquisition.

What does this mean for you?

For now, little changes in your daily life. Williams says the ETH result would have no real-world applications in the short term, though he sees possible long-term benefits.
Treat headlines cautiously. A lab announcing a proof is not the same as the math community accepting it. Review can take time and may find gaps.
Watch the verification. If the proofs hold up, expect follow-up work from complexity and analysis researchers.

Sources

OpenAI's announcement (OpenAI; not directly read by us)
AI makes discoveries in mathematics, computer science, physics, OpenAI says (Al Jazeera)

Source: OpenAI

ScoviAI Team is the ScoviAI newsroom. We write every article ourselves from the original announcements and link our sources. Spotted a mistake? Tell us and we'll fix it.