Lesson 5 · 11 min
Chain-of-thought & reasoning
"Think step by step" — and when it actually helps.
The most overrated trick in the book
Chain-of-thought (CoT) means asking the model to reason out loud before answering. The classic phrasing: "Let's think step by step."
It sometimes helps a lot. It often does nothing. And on modern models with built-in reasoning (Claude Opus 4, GPT-o-series), it can hurt by interfering with their internal reasoning.