What OpenAI’s AI Findings Mean For The Exponential Time Hypothesis

🔍 Read the full analysis: What OpenAI’s AI Findings Mean For The Exponential Time Hypothesis on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

OpenAI said on October 26 that an unreleased AI model had produced a proof of a subexponential-time algorithm for exactly counting solutions to k-SAT, a result that would refute the Exponential Time Hypothesis (ETH) if verified. The company released a preprint and a Lean formalization that a computer can check, but independent expert review is still pending.

The preprint describes a deterministic algorithm that exactly counts solutions for k-SAT formulas, with k fixed, in subexponential time and polynomial memory. For 3-SAT, the authors say this contradicts ETH, which holds that solving 3-SAT requires exponential time in the number of variables. The paper also claims to refute the Strong Exponential Time Hypothesis and three related hypotheses. Those are claims made in the preprint, not conclusions that have already passed independent review.

OpenAI says its system found the disproof on October 16, and that a Lean formalization followed the next day. The source report counts 3,977 files and about 710,000 lines of Lean code; the result’s formal statement is much shorter. Lean can check whether a formal proof follows from its definitions and rules. But the formalization does not cover the paper’s sharpest runtime bound: according to the source report, Lean checks a slightly weaker bound than the preprint states.

OpenAI has made the preprint and code public at the request of the external group it asked to examine the result. That group is led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams; Impagliazzo and Paturi were among ETH’s original authors. The source material says OpenAI will fund a conference on the result. No independent verdict from the group is reported yet.

What an ETH Refutation Would Change

ETH is a widely used assumption in theoretical computer science: researchers rely on it to show that certain problems are unlikely to have substantially faster algorithms. A verified refutation could force a reassessment of results whose hardness depends on ETH or stronger assumptions. The consequences would first be for theory and algorithm analysis, not an immediate change to everyday computing.

The reported algorithm does not make hard problems easy in the practical sense. The source describes its runtime as roughly 2 to the power of n divided by the square root of log n. That remains superpolynomial and can still grow rapidly as problem size increases. OpenAI itself said practical improvements would require significant further research; the source report says its own arithmetic suggests the speed-up grows very slowly with problem size.

The result also does not solve P versus NP. That question asks whether every problem whose answer can be checked efficiently can also be solved efficiently. A subexponential algorithm is not necessarily polynomial-time, and ETH is a stronger hypothesis than P ≠ NP. The Clay Mathematics Institute still lists P versus NP as unsolved, according to the source material.

For AI research, the immediate significance is methodological. OpenAI’s account describes a model producing a result, followed by a machine-checkable formal proof and review by outside researchers. That process may help people test complex mathematical claims, but a formalization that misses the paper’s strongest bound and a pending human review show why machine checking is not the same as independent validation.

Products Worth Considering

Apple Mac mini Desktop Computer with M5 Pro Chip (2026)

AmazonView Latest Price

Apple iPad Air 13-inch (M4), 128GB, Wi-Fi 7, Space Gray

AmazonView Latest Price

Dell UltraSharp 32 U3226QER 4K Thunderbolt Hub Monitor

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

The Hypothesis Behind the Claim

In k-SAT, a formula consists of rules made up of up to k variables or their negations. The task is to determine whether there is an assignment of true and false values that satisfies every rule. With n variables there are 2^n possible assignments, though existing algorithms can do better than checking them all one by one.

ETH was proposed by Impagliazzo, Paturi and Francis Zane in 2001. In the version described by the source, it says that 3-SAT cannot be solved in subexponential time: its runtime must be at least 2^(c·n) for some fixed positive c. The preprint’s claimed subexponential algorithm therefore challenges the central premise directly, if both the algorithm and its analysis hold up.

The source report gives examples of earlier improvements that reduced the base of the exponential, including Schöning’s 1999 algorithm and later PPSZ refinements. Those advances made brute-force search more efficient without removing exponential growth. OpenAI’s claim is different because it says the exponential dependence can be reduced to a subexponential one.

OpenAI presented the ETH work alongside two other research claims, one in mathematics and one in physics, and described the manuscripts as “a form of raw research data.” The source also notes that an earlier October 6 release of more than 700 AI-generated manuscripts was followed by corrections: three papers were withdrawn over a sign error and 14 were revised, according to the repository history. That record makes the distinction between a public AI-produced result and a result that has survived review particularly relevant.

Formal Checks and Review Gaps

The central uncertainty is whether independent experts will accept the proof and its analysis. Lean’s check supports the parts included in the formalization, but it does not validate the paper’s strongest stated runtime bound, according to the source report. The difference between the formalized result and the preprint’s sharpest claim must be resolved before readers can treat the full claim as established.

The expert group’s examination is not complete in the material provided. It is also unclear when it will issue findings, whether it will request changes, and how the result’s implications for papers relying on ETH would be assessed. OpenAI’s announcement is the basis for the dates and description of the model’s work; those details have not been independently confirmed in the source material.

The source gives a formula for the claimed runtime but does not provide enough independently reviewed detail to assess its practical performance. No benchmark showing that the algorithm improves real-world SAT solving is reported. The result should therefore be described as a claimed theoretical refutation under review, not as a demonstrated practical advance or a resolution of P versus NP.

The Expert Review Is Next

The next milestone is the external group’s examination of the preprint and formalization. Its work will need to address both whether the proof is sound and whether the formalized result supports the runtime claim made in the paper. The group’s conclusions could confirm the result, identify errors, or require revisions; no outcome has been reported.

OpenAI says it has agreed to fund a conference on the result, which could provide a setting for wider technical discussion. Until researchers have reviewed the proof, the bounds, and the relationship between the paper and its Lean code, the responsible conclusion is limited: OpenAI has published a substantial, machine-checked research claim that challenges ETH, while its full mathematical status remains unsettled.

Key Questions

What is OpenAI claiming about ETH?

OpenAI says an unreleased AI model produced a proof of a deterministic, subexponential-time algorithm for exactly counting solutions to k-SAT when k is fixed. If the claim holds for 3-SAT, it contradicts the Exponential Time Hypothesis.

Has the proof been independently verified?

Not yet, according to the source material. A Lean formalization has been published and can be checked by software, while an external group led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams is set to examine the work.

Does this solve P versus NP?

No. The reported algorithm is subexponential, not necessarily polynomial-time. The source says P versus NP remains unsolved, and a refutation of ETH would not by itself resolve that question.

Does the result make SAT problems faster in practice?

No practical improvement has been established in the source material. OpenAI said practical gains would require significant further research, and the reported runtime remains superpolynomial.

What is the main gap between the paper and the formal proof?

The source report says the Lean formalization checks a slightly weaker runtime bound than the preprint’s sharpest claim. Experts have not yet reported whether they accept the proof or how they assess that gap.

Source: ThorstenMeyerAI.com