Describe something you are about to do: a price change, a launch, a new policy, an announcement. A crowd of AI personas (customers, staff, a competitor, a journalist, a regulator) reacts to it round by round, replies to each other and changes their minds. Step in between rounds to respond, then read the room in a short report and question anyone one to one.
The model reads your decision and casts the people whose reactions decide whether it works, each with a motive, a voice, a starting stance and a reach.
Each round the crowd posts, replies, shares or stays quiet, reacting to you and to each other. Stances can shift one step a round, and only for a reason in the feed.
The report names who backs you, who blocks you and who could swing, the risks, your next moves, and cheap real-world tests to run before you commit.
Inspired by research on multi-agent "swarm" simulation, including the open-source MiroFish project. Built from scratch by Cocoon; every persona is fictional. Want your team to build tools like this? Talk to us.