Chain of Thought Prompting
Show Your Work on the Maths Paper
6 min read
Show Your Work on the Maths Paper
Your maths teacher always said 'show your work' — not because they can't do the sum, but because the process reveals if you understand it.
When an AI is asked to show its reasoning step by step, something remarkable happens — it makes fewer errors. It's like how writing out your working in maths helps you catch your own mistakes. Adding 'think step by step' to your prompt activates this more careful, sequential reasoning mode.
In Plain English
Chain of Thought prompting asks the AI to reason through a problem step by step before giving the final answer. This dramatically improves accuracy on complex reasoning tasks — maths, logic puzzles, multi-step analysis.
The Technical Picture
Chain-of-thought (CoT) prompting elicits intermediate reasoning steps before the final answer. It can be zero-shot CoT ('Let's think step by step') or few-shot CoT (providing examples with reasoning chains). CoT improves performance on tasks requiring multi-step reasoning by forcing sequential, verifiable intermediate outputs.
Real-World Examples
- Adding 'think step by step' before asking a logic puzzle
- Claude's extended thinking mode uses chain of thought internally
- Solving compound interest problems by breaking into visible steps
Add 'think step by step' to any complex problem and watch accuracy jump — that's Chain of Thought.