Artificial intelligence is increasingly being used to tackle problems that have challenged researchers for decades. OpenAI has now reported a particularly ambitious result: its AI systems produced a mathematical argument that refutes the Exponential Time Hypothesis (ETH), a longstanding assumption about how efficiently certain computational problems can be solved. The company says the result was formalized using Lean, a proof-assistant system that helps verify mathematical arguments.
OpenAI published the announcement on October 26, 2026, as part of a broader effort to work with academic researchers on understanding AI-generated scientific results. Rather than presenting the announcement as a completed scientific consensus, the company says it is involving external experts to examine the findings, explain their significance and help the research community evaluate the work.
What Is the Exponential Time Hypothesis?
The Exponential Time Hypothesis is an important assumption in theoretical computer science. In simplified terms, it proposes that certain Boolean satisfiability problems, particularly 3-SAT, cannot be solved by an algorithm whose running time grows subexponentially with the number of variables. The hypothesis has been used to establish conditional lower bounds for many algorithms and to understand the limits of efficient computation.
If the hypothesis is false, some of those conditional results may need to be reconsidered. That does not automatically mean every difficult computing problem can be solved quickly, nor does it establish that P equals NP. Instead, the significance depends on the exact theorem, its assumptions and how it changes the existing understanding of algorithmic complexity.
What OpenAI Says Its AI Discovered
According to OpenAI, its system generated a natural-language argument refuting ETH on October 16, 2026. The company says the result was subsequently formalized in Lean on October 17. Lean is a proof assistant that allows mathematical statements and their proofs to be expressed in a formal language and checked by a computer against a set of logical rules.
Formalization is an important step because it can help identify gaps in a mathematical argument that might be difficult to spot in ordinary prose. However, interpreting a formal result still requires care: researchers need to understand precisely what theorem was encoded, what assumptions it uses, and whether the formal statement corresponds to the intended problem. Independent review is therefore central to determining how significant the result really is.
Why Independent Researchers Are Being Involved
OpenAI says it has asked external researchers to examine the result and help explain its implications. The group is led by computer scientists Russell Impagliazzo of the University of California, San Diego; Ramamohan Paturi of UC San Diego; and Ryan Williams of the Massachusetts Institute of Technology. OpenAI says it has released the AI-generated write-up and a Lean proof in a public GitHub repository to support scrutiny and further analysis.
The company is also experimenting with different ways to publish AI-generated scientific results. In its announcement, OpenAI described separate review and collaboration arrangements for mathematics, computer science and physics. The approach is intended to let domain experts examine the work, explain its context and help the wider research community assess what the findings do—and do not—establish.
Why This Could Matter for AI Research
If the ETH result withstands independent examination, it could affect research into algorithm design and the theoretical limits of computation. Many computer-science results rely on ETH or related assumptions to show that certain algorithms are unlikely to achieve particular performance guarantees. A valid refutation could prompt researchers to revisit some of those conclusions and explore whether new algorithmic approaches are possible.
The broader AI implication is equally important. Mathematical reasoning is often difficult to evaluate because a convincing explanation can still contain a subtle error. Combining AI-generated arguments with formal verification and independent expert review offers one possible workflow for testing whether an AI system can contribute useful results to advanced research. It also provides a way to distinguish a model's confident claim from a result that has survived rigorous scrutiny.
This Is Not a Solution to Every Computing Problem
The announcement should not be confused with a solution to the P versus NP problem, one of theoretical computer science's most famous open questions. ETH is a separate hypothesis about the time complexity of particular computational problems. Refuting ETH would have important implications, but it would not by itself prove that all problems whose solutions can be checked efficiently can also be solved efficiently.
It is also important to distinguish formal verification from broad scientific acceptance. Lean can check a proof relative to its formal definitions, axioms and underlying system, but researchers still need to establish that the formalized statement captures the intended claim and understand its consequences. Peer discussion, reproduction and expert analysis remain essential parts of the scientific process.
What Happens Next?
OpenAI says the external collaboration will examine the result and make its findings available to the research community. The public release of the proof materials gives specialists an opportunity to inspect the argument, test its formalization and determine whether the conclusion changes existing knowledge about computational complexity.
For now, the development is best understood as a major AI research claim that is undergoing scrutiny. If independent experts confirm its validity and significance, it could become an important example of AI-assisted mathematical discovery. Until that process progresses, the most responsible conclusion is that OpenAI has reported a potentially consequential result—not that the foundations of computer science have already been rewritten.
