Pattern anatomy
A useful pattern explains the problem it solves, the user flow, when it is appropriate, when it is risky, and how the interface communicates AI state.
What each atlas entry covers
Best for and not for: the situations where the pattern helps or creates unnecessary complexity.
Trust and control: what the user can inspect, approve, edit, retry, or undo.
Failure states: likely errors, ambiguity, delays, and recovery paths.
Implementation notes: components, data flow, state transitions, and safety boundaries.
Browse all AI interaction patterns
Categories · Compare · Search · Topics