ReAct interleaves reasoning and acting: the model thinks, calls a tool, reads the result, thinks again. It adapts naturally to surprises because there is no plan to invalidate, and it drifts on long horizons for exactly the same reason.

ReAct

The agent first produces an explicit multi-step plan, then executes the steps, re-planning only on failure. Because the plan is written down it can be inspected, approved by a human, and used to check whether the run actually did what it set out to do.

Plan-then-Execute

A topic of the Building AI Agents course. Planned for 2.5 hours at Hard difficulty.

University of Michigan - Ann Arbor

A category of the Building AI Agents course, covering 4 topics.

Planning, Orchestration and Evaluation

Planning Strategies: ReAct and Plan-then-Execute

A topic of the Building AI Agents course. Planned for 2 hours at Hard difficulty.

Multi-Agent Orchestration Patterns

Evaluating Agent Trajectories

A topic of the Building AI Agents course. Planned for 1.5 hours at Medium difficulty.

Learn Before

Related

Learn After