MC12: Intro to Chain of Thoughts and ReACT Prompting

MC12: Intro to Chain-of-Thought and ReAct Prompting
AI Residency Masterclass 12
Most people prompt an LLM and hope for a good answer. Builders shape how the model thinks and acts. This masterclass covers two foundational techniques behind reliable reasoning and today's AI agents: Chain-of-Thought (CoT) prompting and ReAct (Reason + Act) prompting.
We start with how Chain-of-Thought gets a model to reason step by step, and why that improves accuracy on multi-step problems, calculations and decisions. We then move to ReAct, the pattern that combines reasoning with action. The model thinks, calls a tool, observes the result and decides what to do next. This loop sits at the core of every modern agent framework.
What you will learn

• What Chain-of-Thought prompting is and when it improves model output
• Zero-shot CoT vs few-shot CoT, and how to write effective reasoning examples
• Self-consistency: sampling multiple reasoning paths to get more reliable answers
• The ReAct pattern: the Thought, Action, Observation loop
• How ReAct lets a model use tools, retrieve information and complete multi-step tasks
• Common failure modes: hallucinated reasoning, tool misuse, infinite loops, and how to control them
• Where CoT and ReAct fit in the Four-Layer Shift, as the bridge from Prompt Engineering to Harness Engineering

Hands-on
You will build a working ReAct agent step by step. You'll write the reasoning prompt, connect a simple tool and trace every Thought, Action and Observation the model produces, using DecoBot, our HR leave assistant, as the worked example. You will leave understanding exactly what happens inside an agent, not just how to call one.
Who this is for
AI Residency participants who want to go beyond basic prompting and understand the reasoning patterns that power AI agents and production AI applications.
Key takeaway
Agents are not magic. They are structured reasoning loops, and once you understand CoT and ReAct, you can design, debug and improve them with