Related to a few existing threads on dual-AI voice chat, but with a different focus: the human as an active third voice in the room, and the AIs able to actually build and act on ideas together - not just discuss a topic while we watch.
Right now, working with an AI means typing an idea, waiting, reading, and typing again. That rhythm is fine for simple questions, but it puts a bottleneck on actual collaboration. What I’d love to see is a live voice workspace - a room where I can talk with one or more AI voices at once, out loud, at the natural speed of conversation.
The value isn’t just convenience. It changes what’s possible:
- Ideas can bounce before they’re fully formed. Half-thoughts get built on in real time instead of having to be typed out completely first - which is often where the best ideas come from.
- The human stays the creative spark. Two AIs working together move faster than any person, but they tend to agree with each other. A human in the loop throws in the angle neither of them was tracking, the real-world constraint, the “what if we did this completely differently.” The person supplies divergence; the AIs supply speed and execution.
- The AIs should be able to do, not just describe. In a shared session they could actually build the thing together - draft the code, sketch the design, test the idea - while we talk it through, keeping a shared memory of what the group has decided so no one’s holding half the picture.
The key features that would make this real: multiple distinct AI voices in one session with clean turn-taking (not talking over each other), a shared context all participants can see and act on, and the ability to run/build things live rather than only talk about them.
I see this as a genuinely different kind of tool than a single chat window - closer to sitting at a table with fast, tireless collaborators who can also respectfully disagree. The person who has the messy human idea plus a couple of fast builders in the room can accomplish something far quicker than any of them alone.