Claimed subexponential 3-SAT algorithm attributed to OpenAI and Anthropic models
Article timeline
The articles in this story. Times are ET.
Andrew CurranXAndrew Curran speculates Anthropic's Mythos 5.5 wrote a paper refuting the ETHAIAndrew Curran says a paper refuting the Exponential Time Hypothesis lists Claude as its only author, and that an internal Claude research model found the core algorithm on October 16 with 14M output tokens and no human input. He speculates the model is Mythos 5.5, ahead of a planned Fable 5.5 release, and that Anthropic is holding further discoveries for its IPO. These claims are unverified conjecture.
DeedyXOpenAI and Anthropic models independently break the Exponential Time Hypothesis for 3-SATAIDeedy says OpenAI and Anthropic models independently solved 3-SAT with the same 2^(n/√log n) bound on October 16, undermining the 27-year-old Exponential Time Hypothesis. The author states the result does not prove P = NP but is faster than expected. The attached image shows papers titled The Exponential Time Hypothesis Is False from OpenAI and A 2^(n/√log n) Time Algorithm for 3-SAT from Anthropic.
- Andrew CurranXAnthropic's internal Claude model reportedly refutes the Exponential Time Hypothesis
AIAndrew Curran reposts a claim that an internal Claude research model found a deterministic algorithm for 3-SAT running in 2^o(n) time, refuting the Exponential Time Hypothesis. The quoted post says the core algorithm was found on October 16 with 14M output tokens and no human input, and that Anthropic researchers helped refine the paper's presentation. The paper says the result was formalized in Lean 4.
Sebastien BubeckXOpenAI model reportedly finds 3SAT subexponential-time algorithmAISebastien Bubeck calls a result placing 3SAT in subexponential time, 2^{n / log^eps(n)}, one of the biggest surprises of his 20 years following theoretical computer science. The post is a reaction to a Dan Roberts announcement of three new discoveries in mathematics, computer science, and physics, linked to an OpenAI page on building scientific understanding with academic communities.