Agents, Environments & Rewards

Supervised learning hands the model a labeled dataset and asks it to copy the pattern. Reinforcement learning hands it nothing — no labels, no answer key. An agent observes the state of its environment, takes an action, and receives a reward: a single number saying how that went. Then it loops, millions of times.

The twist that defines RL: actions change what data you see next. Which move you make decides which board you face; which states you visit decides which experiences you learn from. The agent doesn't just learn from experience — it generates its own.