Very cool! I think the hype around “agents are so good that you never actually need to see the underlying commands they are running or interact with the terminal session that they’re running” misses out on a lot of very important use cases, particularly around long running processes that may be shared across multiple agents. This will be very cool to see how best practices evolve!

Same ! TUI are becoming more and more mainstream so there is a need for automated multiplexing , not used by humans but by programs (and agents)