A system is also a trajectory.
Dynamical systems offer a way to think about intelligence through time. A state changes under inputs; observations reveal only part of it; actions change what can happen next.
This connects models of learning to control, feedback, and physical systems. It also changes the questions worth asking: not only what a system predicts, but what happens when its predictions become inputs to the world.
Models that keep updating
Model Predictive Control is a recurring point of interest: plan over a horizon, act, observe the changed state, and plan again.
That loop connects to agents and reinforcement learning, but also to a broader curiosity about continuous intelligence. Some processes are better understood as ongoing adjustments than as isolated answers.
A working entry. The connections may change.