Ten advances in mathematics and theoretical computer science
OpenAI released new AI-generated solutions to long‑standing math problems, demonstrating models can tackle geometry, cryptography and complexity.
Also covered bySimon Willison
Catch-up issue
OpenAI unveils Astra and expands free ChatGPT access as Google rolls out Gemini Managed Agents
Daily edition
49 of 50 feeds responded
326 distinct stories
from the last 12 days
36 previously issued, held back
12 worth your time
Collection and selection
326 candidate events; 40 shortlisted stories; 35 with article text available; 12 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
Active filters: none. Sort: Ranked.
All 12 stories, in ranked orderMore filters
Gemini API Managed Agents: 3.6 Flash, hooks, and more
Google's Gemini API now lets a single call coordinate reasoning, code execution and web retrieval inside an isolated cloud sandbox.
Our position on open-weights models
Anthropic clarified its stance amid US discussions on banning Chinese open‑weights models, countering claims it seeks to restrict them.
OpenAI announces its "next major model" Astra by dropping ten previously unsolved math solutions
OpenAI's upcoming Astra model family will coordinate multiple agents for long‑running, complex tasks and will undergo U.S. government review.
AMD Releases Instella-MoE-16B-A3B: A Fully Open Mixture-of-Experts LLM With 2.8B Active Parameters Trained On Instinct GPUs
AMD released Instella‑MoE‑16B‑A3B, a 16B‑parameter Mixture‑of‑Experts LLM with 2.8B active parameters and open weights.
Codex Security
OpenAI's Codex Security CLI and TypeScript SDK help locate and fix code vulnerabilities, requiring Node 22+ or Python 3.10+.
Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling
Moonshot Labs' Kimi K3 and Alibaba's Qwen 3.8 launch publicly, challenging top‑tier models by matching Anthropic's Fable 5 performance.
AI coding agents can modernize research software but can't judge if the science is right
OpenAI field report shows AI coding agents can speed up legacy research software up to 60×, but they cannot verify scientific correctness.
Supabase Releases Evals: an Open Source Benchmark That Scores Claude Code, Codex and OpenCode on Real Supabase Tasks
Supabase open‑sourced Evals, a benchmark that runs Claude Code, Codex and OpenCode on real Supabase tasks, with a public leaderboard.
Scientific computing in the age of agentic AI
A new field report details how scientists use AI coding agents to modernize scientific software, accelerating development in genomics and data‑rich fields.
Accelerating scientific discovery with ChatGPT for Academic Researchers
OpenAI offers 100,000 researchers free access to its most advanced ChatGPT models to boost scientific discovery.
Introducing the ChatGPT for small business program
OpenAI's ChatGPT for Small Businesses program equips entrepreneurs with AI tools to automate tasks and expand capabilities.
No stories match. Active filters.