The Agent Loop: Think, Try, Learn

How Agents Solve Problems Step by Step

AI agents don't just guess once — they work through problems in a loop: think about what to do, try an action, see what happens, then think again. This is called the "agent loop."

Think: "What's the best next step to solve this problem?"
Act: try something — search the web, run some code, read a file
Observe: see what the result was
Repeat until the job is done (or until they give up!)
It's like how you work through a difficult puzzle — step by step

The Birthday Party Agent (That's You!)

Imagine you're planning your best friend's surprise birthday party. Watch yourself become an agent! Think: "What do I need? Guests, cake, a place, games." Do: message the guests. Check: three said yes, two haven't replied — better remind them tomorrow. Think again: "Cake next. Chocolate or vanilla?" Do: secretly ask their sister. Check: chocolate it is! You didn't plan every step at the start — you did one thing, looked at what happened, and let that decide your next move. That think → do → check rhythm is exactly the agent loop. AI agents tackle a research task or a coding job the very same way.

Think: break the big goal into the next small step
Do: take one action — send the message, ask the question
Check: did it work? Let the answer shape your next move
You already run the agent loop every time you plan something big

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.