🔍 Read the full analysis: How Strong Is OpenAI’s AI Claim About The Exponential Time Hypothesis? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI says an unreleased AI model produced a proof of an algorithm that would refute the Exponential Time Hypothesis, a major assumption about the difficulty of solving certain logic problems. A Lean formalization has checked a related, slightly weaker result, but independent experts have not yet verified the claim. The work does not resolve P versus NP or imply immediate practical speed gains.
OpenAI said on October 26 that an unreleased internal AI model had produced a proof of an algorithm that would refute the Exponential Time Hypothesis (ETH), a long-standing assumption about how hard it is to solve certain logic problems. The result is not yet independently verified: OpenAI released a preprint and Lean code, and an outside group is set to examine the work.
The preprint describes a deterministic algorithm that exactly counts solutions to k-SAT formulas in subexponential time when k is fixed. For 3-SAT, such a result would contradict ETH. The paper also says it refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses. Those conclusions depend on the proof being correct and on the claims matching the formalized result.
OpenAI says the model found the disproof on October 16 and that the Lean formalization followed the next day. Lean is software that checks whether a proof follows from its stated definitions and rules. The source report counted 3,977 files and about 710,000 lines of code; the formal statement itself is short. But the report says the Lean check covers a slightly weaker bound than the paper’s sharpest runtime claim, which is not included in the formalization.
OpenAI’s October 26 post announced three research results, in mathematics, computer science and physics. For the ETH work, the company said an external group would explain and verify the result, and that the preprint and code were made public at the group’s request. The source report says the group is led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams; Impagliazzo and Paturi were among ETH’s original authors. OpenAI also said it would fund a conference on the result.
THREE RESULTS, THREE RELEASE STRATEGIES
What OpenAI released, what Lean checked, who reviews it
Dates and the physics status are OpenAI's own account. Lean-checked means a machine verified the proof, not that experts have accepted it.
Local smoothing, so Kakeya too
- AI output
- 18 Oct 2026
- Lean
- 19 Oct 2026 · strict ranges
- Public
- AI write-up + Lean proof
- Review
- No group: researchers organize themselves
The Exponential Time Hypothesis is false
- AI output
- 16 Oct 2026
- Lean
- 17 Oct 2026 · weaker bound
- Public
- 67-page preprint + Lean proof
- Review
- Impagliazzo, Paturi (UC San Diego), Williams (MIT)
QCD exists; pure Yang–Mills as sub-result
- AI output
- 9 Sep (Yang–Mills), 18 Sep (QCD)
- Lean
- Yang–Mills existence, 16 Oct (OpenAI)
- Public
- No: the group holds the AI materials back
- Review
- Gross, Douglas, Jaffe, Klebanov, Nekrasov
Before this release
The repository the results sit in
- 6 Oct: more than 700 AI-generated manuscripts published
- 7 Oct: 3 withdrawals after a sign error; 14 other manuscripts revised
- About 42% of top-line results formalized in Lean (repository’s own figure)
- OpenAI calls the manuscripts “raw research data”, not polished papers
Sources: OpenAI, 26 Oct 2026 · Lean note 378 · Lean note 079 · repository history. Status chips: our assessment.
Powered by Thorsten Meyer AI
Why an ETH Refutation Would Matter
ETH is used in theoretical computer science to support claims that particular problems cannot be solved substantially faster than current exponential-time approaches. If this result holds up, researchers would need to revisit work whose hardness conclusions rely on ETH, including some limits used to compare algorithms. That would be a significant change to the field’s theoretical toolkit, not proof that every such result or application is invalid.
The stated algorithm runs in roughly 2 to the power of n divided by the square root of log n, according to the source material. That is subexponential, but it remains far slower than polynomial time. OpenAI says practical improvements would require substantial further research, and the source report’s own calculation suggests the speed-up grows very slowly with input size. The claim does not mean problems become easy in practice, nor does it establish P = NP.
The process is also part of the story. An AI-generated proof followed by a machine-checked formalization offers a possible way to test complex mathematical work. But a formal checker establishes only that a proof encoded in its system follows from its formal assumptions; it does not by itself settle whether the paper’s informal claims are accurately represented, whether the formalization covers the headline result, or how the result should be interpreted. Independent expert scrutiny remains a separate step.
Products Worth Considering
Apple Mac mini Desktop Computer with M5 Pro Chip (2026)
FLEXISPOT 32-Inch Electric Standing Desk Converter with Keyboard Tray
FLEXISPOT EN1 Electric Standing Desk, 48 x 24 in, Black
As an affiliate, we earn on qualifying purchases.
ETH and the AI Proof Process
In k-SAT, a formula is made of rules that each involve no more than k variables. The problem asks whether truth values can be assigned so that every rule is satisfied. A direct search over n variables checks as many as 2^n assignments. Researchers have developed faster-than-brute-force methods, but ETH says 3-SAT still requires exponential time, in the sense that it cannot be solved in time below an exponential bound of the relevant kind.
ETH was proposed by Impagliazzo, Paturi and Francis Zane in 2001. It is stronger than the unresolved P-versus-NP question: refuting ETH would not prove that every efficiently checkable problem can be solved efficiently. The source material also notes that ETH’s status as a broadly accepted belief is disputed. OpenAI describes it as widely used, while computer scientist Scott Aaronson wrote in a 2017 survey that there was no field-wide consensus that ETH is true.
The release follows an earlier batch of AI-generated research manuscripts published by OpenAI on October 6, according to the source report. The repository history later showed three manuscripts withdrawn over a sign error and 14 revised. The repository described about 42% of its top-line results as formalized in Lean. That history gives readers a reason to distinguish an announced result, a formalized proof and independent acceptance rather than treating publication as final confirmation.
What Expert Review Must Establish
The result has not yet received the independent review described by OpenAI. It is not clear from the available material when the outside group will complete its examination, what review process it will use, or whether it will endorse the paper’s central claims. The source report says the group is led by researchers who helped formulate ETH, but their participation is not itself confirmation of the proof.
A key technical gap is the difference between the paper’s sharpest runtime statement and the slightly weaker bound checked in Lean, according to the report. The formalization can support confidence in the result it encodes, but readers still need to know exactly how that result relates to the preprint’s strongest claim. The proof’s correctness, the scope of the formal check and the interpretation of the consequences remain open to scrutiny.
It is also not yet clear whether the result will lead to faster practical algorithms or how much existing ETH-based work would need to change. Any such effects depend on further research and on which results rely on the particular hypothesis being challenged.
Independent Review and Follow-Up
The next major step is the external group’s examination of the preprint and formalization. OpenAI said the group would explain and verify the computer-science result; the source report identifies its leaders as Impagliazzo, Paturi and Williams. A conference funded by OpenAI is also planned, though the material does not give a date or describe its format.
Researchers will need to compare the paper’s runtime claims with what the Lean code actually proves, check the argument independently and assess which ETH-based conclusions are affected. Until those steps are reported, the careful description is that OpenAI claims an AI-produced refutation and that Lean has checked a related formal result—not that the field has accepted ETH as disproven.
Key Questions
What is the Exponential Time Hypothesis?
ETH is the hypothesis that 3-SAT cannot be solved in subexponential time as the number of variables grows. It is used in theoretical computer science to establish limits for algorithms and related problems.
What exactly does OpenAI say its AI proved?
OpenAI says its system produced a deterministic, exact algorithm for counting solutions to fixed-k SAT formulas in subexponential time. For 3-SAT, that would refute ETH. The preprint says it also refutes stronger and related hypotheses, but those claims await independent review.
Has Lean verified the strongest claim in the paper?
Not according to the source report. Lean checks a slightly weaker runtime bound; the preprint’s sharpest bound is not covered by the formalization described there.
Does this prove P = NP or make hard problems easy?
No. Refuting ETH would not resolve P versus NP. The reported algorithm is still slower than any polynomial-time algorithm, and OpenAI says practical improvements would require further research.
When will the result be confirmed?
No completion date for the independent review is provided in the source material. The result should be treated as an OpenAI-announced claim until the reviewers publish their findings and the technical scope of the proof is assessed.
Source: ThorstenMeyerAI.com
