Research
OpenAI says it has reached its 'automated research intern' goal
The company reports 3.1 agent-workdays per human workday inside its research organization and is now aiming for an automated AI researcher by March 2028.
Research
Papers, benchmarks, and findings from labs and universities.
- Research
DeepMind swarm of 100 agents split into cheaters and whistleblowers after finding a grading exploit
Identical Gemini 3.1 Pro agents faked 34 Lean proofs in 27 minutes while a quarter of the swarm audited, boycotted, and filed complaints.
- Research
Google's WeatherNext 3 forecasts hourly from live satellite data at five-kilometer resolution
The model drops numerical weather prediction inputs and is already live in Search, Maps, and Gemini.
- Research · Context
Anthropic reports first complete computer-checked proof of Fermat's Last Theorem
Claude wrote the Lean formalization largely autonomously over 11 days, building on a community effort begun in 2024.
- Research
Researchers propose PlanFence to stop distributed AI agents acting on stale plans
A dependency-scoped validation protocol prevented invalid actions in all 30 test workflows where freshness checks alone failed every time.
- Research
Google DeepMind's WeatherNext 3 adds real-time satellite data and hourly forecasts across Google products
The model topped an independent forecast comparison and now powers weather data in Search, Maps, Gemini and Google Cloud.