I'm thinking about how to maximize the speed, bandwidth of collaboration with agents and teams to get to shared context as fast as possible. I think for the human, based on biology, its visual into to the human (out from agent) and voice out of the human (into the agent). Based on this, we are working on a local, agent-native workspace where you can collaborate with your coding agent visually in your sessions, markdown, mockups, code, tasks, etc... Called Nimbalyst. Would love feedback on it.