Prompt Exercises

Introduction

Practice cements theory. The exercises below repeat across interviews and real jobs — do all of them, save your answers.

Beginner analogy: Like piano scales — boring alone, transformative together.

Understanding the topic

Core concepts to understand:

• Write a zero-shot prompt to classify movie reviews as positive/negative.
• Convert it to few-shot with 3 examples — compare accuracy.
• Add CoT and measure delta.
• Add JSON output + validation.
• Add refusal rule for empty input.

Syntax reference

Visual workflow / architecture:

Exercise loop: Write → Test → Measure → Improve

Real-world use

These are the exact exercises new prompt engineers do during onboarding at top AI startups.

Best practices

• Do every exercise yourself, not just read.
• Save your prompt + result for review.
• Repeat after Phase 4 — you'll see what changed.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. Walk through your zero-shot vs few-shot results.
• Q2. How does CoT affect this task?
• Q3. JSON validation strategy?
• Q4. Refusal rule design?