OpenAI starts a science blog for AI research and methods
Original title: Our new blog for science
The first posts are meant to cover results, how scientists use models, and open questions about checking and attribution.
OpenAI said on November 15, 2026 that it is starting a science blog for scientists to share discoveries made with AI and the methods behind them. Planned topics include research results, practical guidance on scientific computing and experiment planning, and researcher perspectives on training and attribution. The post cites recent OpenAI-linked work: an internal system’s September proof addressing the Navier–Stokes problem, October mathematics results including a quasi-Riemann hypothesis estimate, and a PerturbAI and Scripps Research analysis of mouse brain-cell data with thousands of GPT-Rosalind agents in 32 minutes. It also notes César de la Fuente’s team uses ChatGPT and Codex to design antibiotic-candidate algorithms that are then made and tested.
Unconfirmed: Claims about the Navier–Stokes proof and mathematics results are OpenAI’s own and are not independently confirmed here.
- GPT-Rosalind agents analyzed data from millions of mouse brain cells in 32 minutes.
- An internal system produced a Navier–Stokes existence and smoothness proof in September.
- October math results included a sharper estimate of primes below a given number.
- de la Fuente’s group uses ChatGPT and Codex, then synthesizes and tests molecules.
- Vendors
- OpenAI
- Organizations
- PerturbAIScripps Research
- AI models
- GPT-Rosalind
From the original article
4 min read at openai.comThe OpenAI science blog shares new findings, practical methods, and firsthand accounts from researchers using AI to advance scientific discovery.
openai.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.