Lesson 3.1: Chain of Thought (CoT) & Extended Thinking
Standard prompting often asks for an answer immediately. Chain of Thought (CoT) forces the AI to “think out loud” before giving that answer. This simple shift can improve accuracy on complex tasks by 30–50%.
1. What is Chain of Thought?
CoT is a technique where the AI breaks a complex problem into smaller, logical steps. Instead of jumping to a conclusion—which can lead to hallucinations—the model explores each piece of the puzzle sequentially.
-
Standard Prompt: “Calculate the ROI for this 12-month project.” (Risk of error)
-
CoT Prompt: “Calculate the ROI for this project. Think through the steps one by one before providing the final answer.“
2. Using “Extended Thinking” (2026 Feature)
In 2026, newer models like Claude 4.6 and 4.7 have a built-in “Extended Thinking” mode.
-
How it works: When enabled, Claude generates internal “thinking blocks” where it reasons, explores different paths, and self-corrects before you ever see the final text.
-
Benefit: This is essential for PhD-level research, complex coding refactors, or high-stakes financial analysis.
3. The “Let’s Think Step-by-Step” Trigger
The most famous “hack” in prompt engineering is still incredibly effective in 2026. By simply adding the phrase “Let’s think step by step” to the end of any prompt, you trigger a reasoning process that makes the AI more reliable and transparent.
4. Prompt Chaining
While CoT happens in one prompt, Prompt Chaining breaks a massive task into multiple separate prompts.
-
Prompt 1: “Identify the 3 biggest pain points in this customer survey.”
-
Prompt 2: “For each pain point identified in the previous step, suggest a specific product improvement.”
-
Why use it: It prevents the AI from getting “lost” in a 1,000-word prompt with too many instructions.
📝 Practical Activity
Task: Give Claude a difficult logic puzzle or a complex math problem.
First, ask for the answer directly.
Then, start a new chat and ask: “Solve this logic puzzle. First, explain your reasoning steps, and then give me the final answer.”
Observe: Compare the accuracy and the clarity of the second response. You will likely see the “thinking” process led to a more solid conclusion.
Next Lesson:
In Lesson 3.2, we will dive into Tree of Thoughts (ToT) and Self-Correction, techniques used when there isn’t just one right answer, but many possible paths to explore.