Stefan Jansen
Free Lesson

How to Build AI Agents with ReAct, Reflection and Planning

30 min
Jan 7, 2027 11:00 AM

By continuing, you agree to Maven's Terms and Privacy Policy.

What you'll learn

Trace ReAct, Reflection, and Planning in Code

Inspect what each loop gives the model, which actions it permits, and how it decides to continue or stop.

Choose a Pattern for a Defined Task

Match tool use, output revision, and multistep coordination to the task and its verification requirements.

Recognize Failures Before Adding More Agent Loops

Identify repeated tool calls, unhelpful self-critique, and stale plans, then choose checks for those failures.

Why this topic matters

ReAct interleaves model reasoning with tool calls, reflection critiques and revises an output, and planning specifies steps before execution. Each pattern addresses different failures and introduces costs of its own. Reuse worked agent loops to see what each pattern changes, where it fails, and how to choose an architecture for a defined task.

You'll learn from

Stefan Jansen

Stefan Jansen

Builds production ML and agent systems, then teaches the operating method.

See all products from Stefan
Get free access