stateful-input

The author argues that treating video as stateless input (i.e., without temporal context) is a limitation that holds back AI agents, and suggests that stateful video processing could improve performance.