How should an AI tool expose its state across more than one mode? I... - Vibeus
How should an AI tool expose its state across more than one mode?
I’m still thinking about it. One possible test: when someone changes mode, can they see the work already in progress instead of reconstructing it from memory?
I don’t have a final answer yet. But if the state disappears at the boundary, the tool has made the person carry the missing context.