One Page at a Time

A coding agent writes code faster than you can read it, and projects still run late. This book explains why and shows you how to change it.

One Page at a Time starts from zero and takes you to running whole projects with coding agents. It works whether you have never followed a process or have been coding for years. The core idea is simple: give the agent a written decision, a scope limit and a check before "done". Every delivery fits on one page, and you review that page before the agent writes any code.

What you will learn

- Why AI speed moved the bottleneck from writing to deciding and verifying, and how an agent sees your project.
- Spec-Driven Development as an idea.
- The engineering practices that make an agent's work reliable: simplicity (KISS, YAGNI, DRY), rules as pure functions, exceptions as values, code organized by feature rather than by layer, and one test for each piece.
- focus-kit as a method and a tool: the project documents, the queue and milestones, and the commands that define, build and close each delivery.
- Git from the essentials to advanced use, worktrees for running several agents in parallel, and the team's tools: pull requests, issues and boards.
- How to use the method on projects that are not software, such as proposals, analyses and documentation, including this book.
- What agents cost and where they pay off, how to bring the method to a team or a company, and how to get an agent to deliver more than its first answer.

How the book is written

Every chapter opens with what you will be able to do after reading it and shows, with evidence, what your team gains. The numbers are real and say where they come from. The code examples are in TypeScript, all built on the same lending library. The tool chapters explain the basics first, so readers who are not developers can follow along.

Free and open

The book is free and open source under the CC BY-SA 4.0 license, with editions in English and Portuguese. You can read it, share it and adapt it.