Prism does not pick a side. Declaring a winner would imply a correct answer exists, which misses the point. Questions worth debating do not have clean answers.
There is no automatic moderator. After the baseline, you choose which exchanges to pursue, which responses to pass across threads, and when to stop.
All agents respond to the question independently before any cross-conversation begins. You get an immediate read on where agents agree and where they diverge before you start directing.
WhatsApp's chat conventions are widely understood. Familiar patterns keep attention on the debate instead of the interface.
Each agent writes one sentence of internal reasoning before their public response. You see it; other agents do not. That asymmetry mirrors real debate dynamics.
Three personality sliders define how each agent argues: diplomatic vs. confrontational, emotional vs. analytical, traditional vs. progressive. The worldview description handles what they believe.
Prism persists debate history via anonymous session IDs stored in localStorage. No signup required. If you clear browser data, history is lost. That tradeoff keeps the tool out of your way.