I imagine the actor model implemented on top of a stable core (eg. Erlang/Elixir on BEAM) might be a natural fit for a "society of mind" like zoo of agents bustling around, working asynchronously, communicating with each other other and with the user(s), etc. Personally, I'm also excited to see how Spritely Goblins shapes up, esp with because of its support for object capabilities -- which should hopefully be an excellent security model for agentic architectures.