Frontier AI Labs Claim They Broke a Foundation of Computer Science

Artificial intelligence models are now rewriting the rules of computational limits. OpenAI released solutions to three blockbuster math problems, including a surprising refutation of the Exponential Time Hypothesis, a durable assumption that governs how scientists understand what is difficult to compute. Anthropic quickly followed with its own refutation of the hypothesis after its mathematician suggested OpenAI based its work on pre-existing research.

The claims suggest AI models have found algorithms that solve tricky logic puzzles in sub-exponential time, moving faster than speeds previously thought possible. If true, the breakthrough could uproot decades of computer science assumptions and force researchers to revisit countless scientific papers built on the hypothesis. The findings arrived as a shock to the scientific community, eliciting a mixture of thrill and dismay.

MIT Professor Ryan Williams previously criticized OpenAI for dropping proof files without warning, calling the release a denial-of-service attack on mathematicians. Yet Williams agreed to collaborate with frontier labs to help interpret the alien-looking parts of the solutions. He organized an open letter with the original authors of the hypothesis to gather feedback and examine AI contributions to the field.

While the labs checked with OpenAI science lead Dan Roberts and found reasonable confidence that current cryptographic systems remain secure, a faster algorithm still calls traditional computational intuitions into question.

Our read is that the rapid release of machine-generated proofs shifts the burden of peer review from traditional academic timelines to a frantic scramble of human interpretation.

Labs are now structuring their disclosures differently across disciplines, turning mathematicians, computer scientists, and physicists into reactive working groups.

Williams notes that AI is forcing researchers to raise the bar for their questions if routine work is now low-hanging fruit.

The unfolding collaboration between human experts and frontier labs will determine whether these alien algorithms represent a genuine scientific revolution or just a chaotic preview of automated discovery.