ResearchAnthropicConfirmed
Story 7 of 12 in the Tuesday 27 October editionAnthropic publishes Claude paper claiming a sub-exponential 3-SAT algorithm refuting ETH
A preprint credited to Claude presents a deterministic algorithm counting 3-SAT solutions in sub-exponential time, backed by a Lean 4 formalisation.
Anthropic has published a research preprint titled “A 2^o(n) Time Algorithm for 3-SAT”, authored by Claude with human research co-authors, accompanied by a complete Lean 4 formalisation in the anthropics/formal-math GitHub repository. The paper presents a deterministic algorithm capable of counting 3-SAT solutions in 2^O(n log log n / sqrt(log n)) asymptotic runtime, directly contradicting the widely held Exponential Time Hypothesis (ETH).
The Exponential Time Hypothesis conjectures that 3-SAT requires 2^Omega(n) time in the worst case, serving as a cornerstone for conditional lower bounds across theoretical computer science. Anthropic stated that the core structural reduction was discovered during autonomous mathematical exploration by an internal Claude research model.
The authors emphasised that the result is strictly asymptotic and does not provide practical speedups for finite SAT instances. The Lean 4 formalisation allows mathematicians and complexity theorists to inspect every inductive step of the reduction independently, establishing complete machine-checked validity across the formalised mathematical assertions.
Sources
- PrimaryAnthropic formal-math repository: 3-SAT algorithm and Lean 4 formalisationgithub.com/anthropics · Official GitHub repository
- OfficialHexagon Math: A 2^o(n) Time Algorithm for 3-SAThexagonmath.org · Research paper preprint
The rest of the day
Read the whole day01ModelGoogle DeepMind
Google DeepMind unveils Genie 4 real-time interactive world model
Built on the Gemini family, Genie 4 generates interactive 3D simulations on the fly with natural-language steering, native audio and multi-agent dialogue.
02Open sourceLiquid AI
Liquid AI releases LFM3 open-weight model family with 256K context
Six 2B-class non-transformer models span text, vision, audio and decision tasks, pre-trained on 22T tokens and optimised for single-GPU fine-tuning.
03Open sourceinclusionAI
Ant Group’s inclusionAI releases Ling-3.1-flash-1019 under MIT licence
The 560B mixture-of-experts model activates 25B parameters per token, provides a 1M context window and scores 44 on the Artificial Analysis Index.
04Open sourceBaidu
Baidu open-sources Unlimited OCR 2 and MarathonDocBench under MIT licence
A 3B document vision model uses shifted recursive sliding window attention to parse multi-page manuscripts without truncation.
Spotted an error? Email aiforsuccess.in+compass@gmail.com or tag @ai_for_success on X






