
Raw updates from an open AI research lab. Early access to model drops on HuggingFace, training run results, and research findings before they go anywhere else. We document what works, what blows up, and what we learned either way, straight to your inbox.