Coding to Understand

Every chapter in the series adds a practice to a codebase of your choice. Pick a project you care about, small enough to prototype in one session but complex enough to eventually need a spec and tests. You'll carry it through eighteen chapters.

Throughout the book, I'll walk through PlantAutopsy, a service I built to analyze photos of dead plants to explain what went wrong. Your exercises use your project. By the end, you'll have a codebase with a spec, decision log, context file, test suite, review history, and trace dissections.

Part 1: Prototyping and Intent

We start with speed. You will prototype by instinct, shipping fast without plans or tests (Chapter 1). Chapter 2 weighs the cost of that speed, a concept called cognitive debt. The rest of Part 1 offers a fix. You'll learn to plan so your intent survives implementation (Chapter 3), to break work into manageable decision-based chunks (Chapter 4), and to use AI as a tool for reasoning (Chapter 5).

Part 2: The Daily Craft

The best way to improve AI output is to improve what you give it. Chapter 6 introduces context engineering, defining constraints so the code belongs in your project. Chapter 7 reframes testing as a contract, focusing on verifying tests capture the right behavior. Chapter 8 introduces the decision log to record why you built it while reasoning is fresh. Chapter 9 teaches you to match review patterns to risk, teaching when to inspect every line and when to let the agent run.

Part 3: The Labor of Building

Chapter 10 covers debugging with AI by feeding AI everything to get ranked theories. This does not mean thoughtless copy and pasting of everything. Chapter 11 is “code archaeology”, the art of recovering intent from forgotten code, including your own prototypes from three months ago. In Chapter 12 we tackle refactoring. Your role is defining what good looks like; AI performs the labor while your test prove the system still works. To wrap up the section, Chapter 13 shows how to build your own tools. When creation cost drops to near zero, you can build exactly what you need.

Part 4: Personalization and Mastery

The final section is about you. Chapter 14 teaches you to review your AI sessions, find where communication failed, and build a personal playbook. Chapter 15 shows how to encode your standards into context files and style guides. Chapter 16 goes further, using hooks, commands, and skills to automate your workflow. Chapter 17 introduces parallel agents, which multiply throughput while managing the risk of cognitive debt.

The final chapter (Chapter 18) focuses on the human edge. The tool should be invisible. Quality comes from the person wielding it. These practices exists to make your expertise shine.