← Back to feed

AI #189: New Math

L5 · ResearcherResearchLessWrong AI· 10/8/2026

Represents a fundamental advance in AI reasoning that could reshape mathematical research and AI development approaches.

AI Summary

OpenAI solved 90 of the top 500 open math problems using only 3 hours of Pro-level compute per question on average, marking a major breakthrough in mathematical AI capabilities.

Excerpt

The big drop of this week was not a new AI model. It was instead the biggest day (so far!) in the history of mathematics, as OpenAI dropped solutions to 90 of the top 500 open math problems, along with many others, reached on an average budget of three hours of Pro-level compute per question. This was kind of a big deal and I plan to cover it tomorrow. We did see Claude Haiku 5.5, which looks promising given it is only $0.10/$0.50. My week was largely spent at The Curve. The entire conference wa

Read Original
0 upvotes · 0 downvotes · 1 min read

Related Articles

L5 · ResearcherResearchTowards Data Science
Your Model's MSE Is Lying to You III: Time Series Diffusion

Extends time series forecasting beyond Gaussian assumptions by introducing diffusion models to capture complex uncertainty shapes through multi-step sampled rollouts.

L5 · ResearcherResearchHacker News
OpenAI Withdraws 3 Math Papers

OpenAI withdrew 3 mathematics research papers due to a sign error in a key argument, while revising 14 others with proof repairs and corrections.

L4 · DeveloperResearchTowards Data Science
AI Agents Beat PyTorch: Writing Faster CUDA Kernels

AI agents generated CUDA kernels that achieved 1.57× speedup over PyTorch in matrix multiplication, but the bigger finding was about benchmarking rigor and detecting flawed tests.

L5 · ResearcherResearcharXiv
RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing

Researchers introduce RECAST, a learned framework that formulates evidence construction as sequential decision process over retrieval and computation operations.

L5 · ResearcherResearcharXiv
Decentralized SGD under Heavy-Tailed Noise: Optimal Convergence Rates and the Role of Gradient Clipping

Researchers prove that clipped decentralized SGD achieves optimal convergence rates under heavy-tailed noise conditions while maintaining linear speed-up across agents.