Lesson 5 of 6~1 min read

Lesson 4 — Watching It Think

Run your agent on a few real questions and study what it does:

Lesson 4 — Watching It Think

Run your agent on a few real questions and study what it does:

Try these prompts:

"What are the best AI coding assistants available right now? Compare Cursor, GitHub Copilot, and Claude Code."

"Research how companies are using AI agents in customer service. What results are they seeing?"

"Find the top 5 Shopify apps for upselling, compare features and pricing."

What to watch for in the terminal output:

How many searches does it run before reading pages? (Usually 1–2 broad searches first)

Which URLs does it choose to read? (It's making judgment calls about relevance)

Does it read multiple pages or stop at one? (Good agents read 3–5 to triangulate)

How does the final report structure differ from a single-search answer?

**A key insight:** Notice that the agent often searches, finds something interesting, searches again based on what it learned, then reads specific pages. That adaptive behavior — changing strategy based on what it finds — is what separates an agent from a simple chatbot.

← PREVIOUSLesson 3 — Building the Agent Loop NEXT →Lesson 5 — Extending Your Agent