Claude-discovered algorithm breaks 3SUM and APSP speed barriers, Columbia/MIT paper shows
rohanpaul_ai · x · 2026-11-22
A Columbia/MIT paper reports Claude found, in one 16M-token session with no human input, a new thin-matrix-product algorithm that yields O(n^1.9992) 3SUM and O(n^2.9995) APSP — refuting the 3SUM and APSP hypotheses and related conjectures. Main results were formally verified in Lean 4. Lesson: treat "known hard" limits as open targets for AI search.
Related event: Claude-Helped Algorithm Breaks Decades-Old Speed Limits for 3SUM and APSP(2 posts)→
More from Research
- Anthropic releases FlashProve with Lean formalization, proving 400B MoE inference on one H100 — jiqizhixin · 2026-11-22
- MiniMax says internal model broke NIST candidates and sped up cryptanalysis solvers up to 10000x — teortaxesTex · 2026-11-22
- a16z: LLMs update predictions in mathematically precise ways — but AGI needs continual learning — vishalmisra · 2026-11-22
- Riemann Hypothesis "human proof" called wrong, with §12.3 the fatal gap — aran_nayebi · 2026-11-22
- Musk: superintelligence's Riemann proof removes conditions from dependent theorems — elonmusk · 2026-11-22
- Lean's Riemann Hypothesis proof reads like Hamlet written by AI — soumitrashukla9 · 2026-11-22