sequential decision making

RL is about decisions made one after another, where today's choice changes the situations you'll face tomorrow. A single snapshot decision—classify this email as spam—stands alone. But steering a car, playing a game, or managing an investment is a chain: each move reshapes the world and the options you'll have next. RL is built for exactly these chains.

This linkage is what makes the problem rich and hard. A locally tempting action can lead somewhere terrible, while a small early sacrifice can open a winning path—so the agent must weigh whole trajectories, not isolated steps. Choices, consequences, and future choices are entangled, which is why RL cares about long-run cumulative reward rather than the gain from any one decision.