An OpenAI model has refuted the Exponential Time Hypothesis (ETH) by finding a faster method for a classic logic puzzle. A machine-checked proof backs the claim, overturning an assumption researchers trusted for 27 years.

The result does not prove that P equals NP, the larger question behind modern encryption. As a result, the algorithm poses no direct threat to the cryptography securing Bitcoin (BTC) and Ethereum.

Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe! https://t.co/VkfacopR2R — Greg Brockman (@gdb) October 26, 2026

Why Did the Exponential Time Hypothesis Matter?

The hypothesis concerns 3-SAT, a puzzle built from true-or-false switches and rules linking them. The goal is a switch setting that satisfies every rule at once.

Checking a proposed answer is easy. Finding one is hard, because each added switch doubles the possible combinations.

Infographic explaining the Exponential Time Hypothesis as a combination lock where each added switch doubles the combinations / Source: BeInCrypto

Russell Impagliazzo and Ramamohan Paturi proposed ETH in 1999. Despite the shared acronym, it has nothing to do with Ethereum’s ticker.

The hypothesis holds that no algorithm can escape this exponential growth in the worst case.

Researchers treated it as a load-bearing wall. According to OpenAI, hundreds of papers in graph theory, computational biology, and databases built hardness results on it.

How Did an AI Find a Shortcut Experts Missed?

OpenAI’s system produced a natural-language proof on October 16. A day later, the team formalized it in Lean, software that verifies every logical step.

The published algorithm handles inputs with n variables in roughly 2^(n/√log n) steps. That beats the exponential floor ETH predicted, yet remains far too slow for practical use.

OpenAI cautions that practical speedups need extensive further research.

Impagliazzo and Paturi, both at UC San Diego, will now lead the review alongside MIT’s Ryan Williams. In other words, the hypothesis’s authors are vetting its refutation.

Meanwhile, OpenAI claims no scientific credit for the discovery, a stance that follows an earlier Navier-Stokes dispute.

Scott Aaronson, a computer science professor at the University of Texas at Austin, recounted hearing the news.

“My students just came to my office to tell me that an OpenAI model has now refuted the Exponential Time Hypothesis”

Scott Aaronson via Shtetl-Optimized

Does Breaking ETH Mean P Equals NP?

No. P versus NP asks whether every problem with quickly checkable answers can also be solved quickly.

ETH was a stricter claim than P ≠ NP. Therefore, disproving it leaves the larger question open.

Picture a mountain between two valleys. The AI found a shorter path over the ridge, not a tunnel through it.

Greg Brockman, OpenAI’s president and co-founder, made the same point hours after the announcement.

p still probably doesn't equal np — Greg Brockman (@gdb) October 26, 2026

The distinction matters for crypto holders. Bitcoin signatures rely on elliptic curve cryptography, a hard math problem this algorithm does not address.

Benjamin Cowen, a member of BeInCrypto’s Market Intelligence Expert Council, addressedthe concern.

But to be clear, disproving the Exponential Time Hypothesis would not prove P = NP, nor would it mean Bitcoin or Ethereum are compromised.

Exponential Time Hypothesis: What Happens When Machines Outpace Their Makers?

Beliefs about impossibility have collapsed before. In 1960, Anatoly Karatsuba disproved Andrey Kolmogorov’s multiplication conjecture within about a week.

Similarly, Volker Strassen set out to prove standard matrix multiplication optimal in 1969 and found a faster method instead.

What differs now is who finds the shortcut. Machines produce the proofs, while humans work to understand them.

The same release claimed further proofs in harmonic analysis and particle physics.

That pace carries weight for digital assets. Cryptographer Matthew Green has already warned that AI could weaken the math behind wallets.

Cowen framed the wider lesson in market terms. In his view, fields far removed from finance can eventually reshape markets.

For an industry secured by hardness assumptions, each AI breakthrough now doubles as a stress test.