AI & agents
Eval harnesses, safety, orchestration, and agents that need other agents.
evals · alignment · tool use · memory · multi-agent · datasets · inference · MCP
See projects →Fields
Any problem that is unfinished.
Medicine, space, climate, code, mind, law — the list is the work of the world, not a chat topic. Same seats. Same 97%. We do not invent labs or patients. We take real unfinished work.