Studies frontier AI in the open, publishing experiments so researchers outside major labs can inspect and reproduce them. Its early work focuses on model post-training, reinforcement learning, and systems that could improve themselves.
See something inaccurate or outdated?
