OpenAI says its artificial-intelligence system has refuted the Exponential Time Hypothesis, a long-standing assumption about the difficulty of solving certain logic problems. The claim has attracted scrutiny from computer scientists, but the disclosed mathematics does not demonstrate a practical way to interfere with Bitcoin mining or steal funds.
What OpenAI is claiming
The result concerns the Exponential Time Hypothesis, or ETH—not Ether, the cryptocurrency. ETH is an assumption that solving some computational problems requires exponential time in the worst case. The specific problem involved is 3-SAT, in which a computer must find true-or-false assignments that satisfy a set of logical conditions.
In a manuscript dated Oct. 17 and revised Oct. 25, OpenAI describes a deterministic algorithm that exactly counts satisfying assignments for fixed-width Boolean formulas in subexponential time and polynomial space. In broad terms, subexponential means the running time grows more slowly than an exponential function as the problem expands.
That claim is significant for theoretical computer science, but its meaning is narrower than suggesting that large, difficult calculations have suddenly become easy on ordinary machines. The result also does not provide a polynomial-time algorithm or resolve P=NP—the open question of whether every problem whose answer can be checked quickly can also be solved quickly.
OpenAI says the result was verified in Lean, a proof assistant used to check formal mathematical statements. However, the scope note for the Lean formalization describes a weaker subexponential bound than the manuscript’s sharpest claim. Researchers arranging an independent review have also emphasized that they are not certifying the result. Their letter states: “Our involvement should not be taken as certification of the claim.”
According to CryptoSlate, Ryan Williams, Russell Impagliazzo and Mohan Paturi were asked to organize an open examination without nondisclosure agreements or compensation. The organizers plan to publish findings as the review develops. CryptoSlate’s own examination of the paper was limited to its introductory pages, 1 through 8, rather than the complete proof.
Why the result does not amount to a Bitcoin attack
Bitcoin’s security involves several separate mechanisms, and the distinction matters. Mining requires miners to search for a block-header hash below a network-set target. Spending coins, by contrast, requires authorization through digital signatures.
Bitcoin uses ECDSA and Schnorr signatures based on the secp256k1 elliptic curve. The relevant security questions include whether an attacker can recover a private key from a public key and whether cryptographic hash functions can be defeated. A claim about worst-case SAT-solving complexity does not, by itself, answer either question.
A practical attack would require more than a theoretical speed-up. An attacker would need a suitable way to encode the targeted cryptographic problem, an implementation of the claimed method and a cost comparison showing that the approach is useful against real keys, signatures or hashes. The number of variables in a SAT formula also cannot simply be treated as equivalent to the bit length of a cryptographic key or hash. The Lean scope measures its bound using the variables and input length of the encoded formula; applying that bound to Bitcoin would require additional work.
Mining and signature security should also not be confused with Bitcoin’s difficulty adjustment. The mining target is recalculated every 2,016 blocks, while that mechanism concerns block production rather than the authorization needed to spend coins. It would not make a private mining advantage irrelevant, either.
Researchers separate this claim from earlier crypto warnings
The debate follows coverage on Oct. 7 of comments by Justin Drake, who urged preparations for “bunker mode” after an earlier OpenAI mathematics release. That warning dealt with a possible future attack, and the earlier announcement had not demonstrated a practical break of ECDSA or RSA.
The Oct. 26 ETH claim is a different result. In a separate post, Williams said he had asked OpenAI to check whether existing cryptosystems had been compromised and relayed that they were apparently safe. That account is second-hand and is not an independent cryptographic audit.
OpenAI has said that practical improvements based on the work would require substantial further research. When asked about benefits developers might see, Williams said he did not know whether meaningful progress would arrive within 12 months or whether no such development would occur.
Moneyiar market picture
Moneyiar data at the stated time showed Bitcoin at $80,430.51, down 0.08% over 24 hours and 1.32% over seven days, as of Oct. 25, 2026, at 22:39 +03:30. Ethereum was $2,342.06, up 0.02% over 24 hours and down 1.77% over seven days at the same time.
Ethena was trading at $0.2093, up 8.19% over 24 hours and 12.35% over seven days, while Celestia stood at $0.5322, unchanged over 24 hours and down 11.85% over seven days. Celestia’s reading was recorded at 22:38 +03:30.
