The Hugging Face and Mythos 5 incidents show AI agents can self-organize, raising questions about how much agency they should have and when to seek human input (Ethan Mollick/One Useful Thing) | TickerVault