How this scenario works
Create bounded AI-driven scenarios whose assumptions, agent actions and outcomes can be inspected. For training, the model focuses on repeatable practice where performance and outcomes can be reviewed.
Start with explicit assumptions, change one or more conditions, preserve state history, and compare why trajectories diverge.