The living transcriptUnlike the sequential mode where each AI only receives the previous responses, the Debate Room maintains a complete shared transcript of all exchanges. With each new message, all AIs receive the entire conversation from the start — they are truly "present" in the room.ConsequenceModels can reference arguments made 10 messages earlier, change their mind mid-debate, and build a progressive collective reasoning — impossible with the sequential mode.
The hand-raise mechanismThe mechanism that makes the Debate Room unique. After each model's message, the orchestrator sends a micro-call to all the others:Hand-raise prompt"You are in this room. Here is what was just said. Do you want to intervene now, and if so, why in one sentence?"
The orchestrator analyzes the responses and chooses the most relevant model to speak next. This mechanism generates natural interruptions, unexpected twists, and an authentically organic debate. Session flow1
Room opensThe orchestrator introduces the topic to all models simultaneously and invites them to take a position.
2
First round of the tableEach model speaks for the first time. The transcript grows in real time, visible to the user via streaming.
3
Cascading hand-raisesAfter each message, inactive models evaluate whether they want to intervene. The orchestrator grants the floor based on relevance.
4
Organic convergenceThe agreement score is calculated continuously. When it reaches 90%, the arbitrator is triggered to produce the final consensus.
5
ConsensusSnapshotThe arbitrator, having read the entire transcript, produces a deep synthesis — far more nuanced than a simple summary.