A concrete example
I want Claude to sort emails. I give it 3 examples (input → category) in the prompt, and it keeps going on the rest.
Why it matters
A few well-chosen examples often replace a long explanation. It's the highest-payoff hack in prompt engineering.
Handy for locking in a stable format, classifying data, or copying a structure.
Don't mix it up with
Prompt engineering: Prompt engineering means phrasing a request with enough context, examples, and constraints to get a useful answer.
Chain of thought: Chain of thought is the step-by-step reasoning a model can work through before it answers.
Common mistakes
- Giving 50 examples: it clutters the context and quality drops.
- Picking unrepresentative examples (the model generalizes badly).
- Forgetting a reasoning model needs fewer examples than a classic model.
Quick checklist
- First I check whether the word names a concept, a tool, a risk, or a metric.
- I tie it to a concrete case: I want Claude to sort emails. I give it 3 examples (input → category) in the prompt, and it keeps going on the rest.
- I keep the main trap in mind: Giving 50 examples: it clutters the context and quality drops.
Quick questions
What is Few-shot / Zero-shot in AI?
Zero-shot means asking with no example. Few-shot means giving a few examples before your request.
Where will I run into Few-shot / Zero-shot?
Handy for locking in a stable format, classifying data, or copying a structure.
Which word should I read next?
Start with Prompt engineering, Chain of thought, Context engineering.