Few-shot examples
Few-shot prompting means showing the model 2-5 examples of exactly the input/output pattern you want, instead of just describing it. Models are much better at copying a demonstrated pattern than following an abstract description - "show, don't tell" applied to AI.
This works because the model is a pattern-completer at its core - giving it several input→output pairs strongly conditions what "the next output" should look like, often more reliably than lengthy instructions. It's especially powerful for consistent formatting, tone-matching, or classification tasks where the "shape" of the output matters more than creative variation.
Ask the model to write a one-line product description with no examples first. Then give it 3 examples of the exact style/length you want and ask for a 4th. Compare consistency.
Why does showing a model 2-5 examples (few-shot) often work better than describing what you want?