Glossary · Term

ReAct

← all terms

Definition

Plain language

A common pattern where an AI agent alternates between thinking out loud and taking actions.

As stated in the literature

A prompting pattern in which an agent interleaves chain-of-thought reasoning with tool calls and observations in a loop.

Why it matters: Interleaving reasoning with action lets a model both plan and react to real-world feedback, which is the basic ingredient of almost every modern agent.

For example, when asked about the weather, the agent writes "I should call the weather tool for Tokyo," calls it, sees the result, then writes "the temperature is 18°C, so I'll tell the user that."

Heard on the show

“The standard agent, called ReAct, takes them one at a time, cold.”
Episode 186 — How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining

Mentioned in 10 episodes

  1. 186
    How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining
  2. 168
    When Turning Experience Into Code Makes Your AI Agent Dumber
  3. 110
    How an Agent Got 44 Points Better by Mining Its Own Scratch Paper
  4. 083
    Training the Translator: How a Small Communication Model Lets Agent Teams Outperform Themselves
  5. 068
    The OS Trick That Makes Tree Search Practical for Coding Agents
  6. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
  7. 046
    When the AI Optimizer Edits the Grade Book: Why Harnessing Evolution Needs a Wall
  8. 030
    Why Your AI Agent Won't Stop Working — and Each Model Falls for a Different Trap
  9. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
  10. 005
    Why a Debugger Designed for Humans Is the Wrong Tool for an AI Agent

Related concepts

Related terms