Understanding AI Coding Agents

00:00 Plenty of AI tools can help you write code, but not all of them are agents. So what makes the difference? Standard AI chatbots give you one-off answers. You ask, they respond, and that’s the end of it.

00:12 Coding agents are built for something different: autonomy. They operate through a continuous execution loop that keeps going until the task is done.

00:22 That execution loop has four steps: read, reason, act, and evaluate. During the read step, the agent reads relevant files from your codebase to build its context.

00:34 Then comes the reason step, where it determines the logical steps needed to achieve your goal. After that comes the act step, where it executes those steps, editing files, running terminal commands, or using external tools.

00:48 And finally, during the evaluate step, it checks the results to see if more work is needed. Then the loop starts again. And that’s what separates agents from simple autocomplete or one-off prompts: they bridge the gap between suggestion and execution, working autonomously through the development workflow.

01:08 There are different agent types, but the core loop stays the same across all of them. What changes is where the agent runs, and that shapes everything about how you interact with it.

01:19 In an editor, it works alongside you in real time. In a terminal, you guide it step by step. In pull requests, it reviews changes on its own schedule. And in the cloud, it works in a managed environment and reports back later.

01:34 Those environments define four primary workflow types. This diagram will be your roadmap for the rest of the course. You’ll explore each workflow one at a time, starting with IDE agents, terminal agents, pull request agents, and cloud agents, so you can understand where each one fits and when to use it.

01:57 See you in the next lesson, where you’ll learn more about IDE agents.