Exercises
The exercise is designed to practice specific techniques from the documentation sections. You'll alternate between writing prompts, reviewing AI output, running tests, and committing your work. Just like a real development workflow.
How the Exercises Work
Each exercise follows a progressive structure:
- You get a starting point — a problem description and some starter code
- You build the rest with AI — using the prompting, review, and workflow techniques described in the documentation
- Each task maps to a documentation section — so you can reference the concepts as you go
- Solutions are provided — including recommended prompts, expected code, and self-assessment criteria
The exercise tasks emphasize test-driven development: you write tests first with AI help, then implement code to make them pass. This mirrors how experienced developers use AI tools in practice.
How to Track Progress
For each task in an exercise, check:
- Time box - expected duration so you know when to move on
- Success checkpoint - the observable result that proves task completion
- Failure signals - common signs that your prompt or implementation needs revision
Prerequisites
- Python 3.10+ installed
- An AI coding tool set up and working (Claude Code, Codex, Gemini, GitHub Copilot, or similar)
- git installed and basic familiarity with version control
- uv package manager (install instructions) or your favourite Python package manager
Available Exercises
| Exercise | Difficulty | Topics Practiced |
|---|---|---|
| FizzBuzz ML exercise | Intermediate | TDD(test-driven development), prompt engineering, PyTorch, agentic workflows, project context |
Sources: Practices and Challenges of Using GitHub Copilot, Stack Overflow Developer Survey (AI tooling trends)