OpenAI claims to disprove the long-standing Exponential Time Hypothesis in computer science.

Huo Xing Finance reports that on October 27, OpenAI announced that its AI model had “cracked” ETH—a statement that quickly sparked widespread discussion in both the tech and crypto communities. However, it is important to clarify that the ETH in question is not Ethereum, the well-known cryptocurrency, but rather a significant conjecture in computer science that has existed for over two decades: the Exponential Time Hypothesis (ETH). The core question of the Exponential Time Hypothesis is: How hard are problems that appear “particularly difficult to compute”? Take, for example, the SAT problem (Boolean satisfiability), which some may have heard of: given a set of logical constraints, can we find an assignment of values that satisfies all of them simultaneously? As the problem size increases, the best-known methods currently available essentially require exhaustively testing possible solutions, with computational cost growing exponentially with size—commonly referred to as “exponential time.” ETH further asserts that such problems cannot be significantly accelerated; there are no truly efficient “sub-exponential time” algorithms for them. Since its proposal in 1999, this hypothesis has served as a foundational pillar in theoretical computer science. Hundreds, if not thousands, of papers have relied on it to prove that “this problem is hard” or “that algorithm is nearly optimal.” If ETH were disproven, it would imply that many problems we assumed to be “almost impossible to solve faster” might, in theory, have better solutions. OpenAI now claims its model has discovered a deterministic algorithm capable of exactly counting the number of solutions to k-SAT problems in sub-exponential time, with formal verification performed using Lean. If this result withstands expert scrutiny, it would directly challenge the validity of the Exponential Time Hypothesis.