Physical AI runtime for embodied agents — governed action, verified experience, physical memory, and skill evolution.
↑ +1 today
Discover the best Skill Learning open source repositories on GitHub. Explore trending tools, live star counts, source code, and developer projects updated daily.
Physical AI runtime for embodied agents — governed action, verified experience, physical memory, and skill evolution.
Gradient descent, but the parameters are agents — a parallel, asynchronous framework for self-evolving agents (skills, prompts, harnesses). Diffs are the gradients; the aggregator is the optimizer.