Few-Shot and Examples
Show, don't just tell. Examples are the most underused prompting superpower.
What You'll Learn
- Zero-shot vs. few-shot prompting and when to use each
- How to pick examples that actually improve output
- The pattern-matching trick that makes AI "get it" immediately
- How many examples you actually need (hint: fewer than you think)
AI Learns From Patterns in Your Prompt
When you give the AI examples of what you want, you're not just explaining — you're demonstrating. The model picks up on patterns in your examples: tone, structure, length, level of detail, formatting choices. It then replicates those patterns in its response.
This is called few-shot prompting, and it's dramatically more effective than describing what you want in abstract terms.
Zero, One, Few
Zero-shot: No examples. Just instructions. Works for simple, well-understood tasks.
One-shot: One example. Sets the pattern. Good for style matching and format demonstration.
Few-shot: 2-5 examples. Establishes a strong pattern. Best for complex or nuanced tasks where one example isn't enough to capture all the rules.
Product Descriptions, Two Ways
Zero-Shot (Vague)
"Write a product description for wireless earbuds. Make it punchy."
You'll get something generic. "Punchy" means different things to different people.
Few-Shot (Clear Pattern)
"Write product descriptions in this style:
EXAMPLE 1:
Product: Running shoes
Description: Built for the long run. Featherlight mesh breathes with every stride. Carbon-plate response pushes you forward. 42 grams lighter than last year. Your PR doesn't stand a chance.
EXAMPLE 2:
Product: Laptop stand
Description: Your neck called. It wants its natural curve back. Machined aluminum. Seven angles. Cable routing that doesn't look like spaghetti. Looks good on the desk. Feels better on your spine.
NOW WRITE:
Product: Wireless earbuds"
The AI now understands your exact style: short sentences, physical benefits, personality, specific details.
Choosing Good Examples
Diverse: Pick examples that cover different scenarios. If all your examples are similar, the AI might over-fit to that one pattern.
Representative: Your examples should look like what you want the output to look like. If you show sloppy examples, you get sloppy output.
Edge cases: Include at least one tricky example that shows how to handle unusual inputs. This teaches the AI your judgment calls.
This lesson is for Pro members
Unlock all 520+ lessons across 52 courses with Academy Pro.
Already a member? Sign in to access your lessons.