Free Lesson
AI Agent Architectures: ReAct, Reflection and Planning
30 min
Oct 28, 2026 11:00 AM
By continuing, you agree to Maven's Terms and Privacy Policy.
What you'll learn
Understand how ReAct, reflection and planning work
See the loop each pattern runs, what the model sees at each step, and which failure it is meant to fix.
Choose an architecture for a given task
Match the pattern to the task: tool use, output quality, or long multi-step work, and what each costs.
Recognize where each pattern fails
Learn the typical failures: tool loops that never stop, self-critique that agrees with itself, stale plans.
Why this topic matters
Most agents are built from three patterns: ReAct interleaves reasoning with tool calls, reflection has the model critique and revise its own output, and planning writes the steps before running them. Each fixes a different failure and adds cost and new ways to go wrong. This lesson shows how each works in code and how to choose between them for a given task.
