OpenAI and Anthropic claim artificial intelligence models have disproved the Exponential Time Hypothesis
OpenAI and Anthropic published separate research papers asserting that internal AI systems successfully refuted a fundamental computer science conjecture known as the Exponential Time Hypothesis. Both organizations submitted formal proofs verified by the Lean automated proof checker, with Anthropic reporting that an unreleased Claude system created its core algorithm without direct human assistance. If validated, the findings demonstrate that machine learning systems are beginning to discover new theoretical mathematical results, though the underlying models remain closed to external researchers.
The Blend
Two prominent research organizations, OpenAI and Anthropic, announced that their internal software models disproved a fundamental computer science concept called the Exponential Time Hypothesis. As reported by Superintel, both companies published formal mathematical proofs that successfully passed automated checks in Lean, a tool designed to verify mathematical logic. Anthropic noted that its unreleased system generated the central algorithmic strategy without direct human guidance.
While theoretical mathematics can feel remote, this specific conjecture served as a key benchmark for measuring computational difficulty across hundreds of academic papers. If the new proofs remain valid under scrutiny, researchers will need to re-evaluate those foundational assumptions. More broadly, the result highlights how automated systems are moving past summarizing human knowledge into discovering entirely new mathematical concepts.
What remains unclear is how external academics can fully audit these findings while the generative systems remain proprietary. Because neither organization released the underlying software models to the public, independent researchers can only evaluate the final text outputs. This raises an important question: will future mathematical breakthroughs be dictated by private companies holding exclusive access to automated research systems?
Written independently by AI News Smoothie from the reporting listed below. Facts belong to the original publishers. Follow the links for their full coverage.
Ingredients
- Two AI Labs Say They Refuted ETH
OpenAI and Anthropic published automated math proofs disproving a major computer science hypothesis, though their underlying models remain closed to the public.