// Making AI "show its work" — why thinking out loud gives better answers
💡 The core idea: If you ask an AI "what is 27 × 33?" it might say "891" — wrong. But if you say "think step by step", it works through the multiplication and gets 891 correctly. Chain of Thought (CoT) is just prompting the AI to reason out loud before giving a final answer. Sounds simple. Works surprisingly well.
See the difference CoT makes
Problem: "If a shirt costs $15 and is 30% off, and you also have a $2 coupon, what is the final price?"
0 / 5 thoughts
Why does Chain of Thought work?
When an AI "thinks step by step", it's not actually thinking in a human sense. It's generating tokens that represent intermediate steps — and those tokens help the model access and use the right knowledge. It's like writing notes while solving a math problem: the act of writing helps you stay on track. Prompting "think step by step" gives AI a reasoning "scratch pad".