The Methodology
The carried problem, the North Star, the specification, the AI partner, the build loop, failure, and the work required to move from prototype to product.
Complete 25-chapter edition on Leanpub
The Wall Has a Door
Build with AI
AI can produce software faster than most people can decide whether it is right. The harder work is holding onto the problem, making your standards explicit, and refusing to confuse a working screen with a trustworthy system.
The Wall Has a Door follows one non-developer from a retirement question to working products and an agent system. It explains the method that emerged, where it failed, and what changed once the work had to survive beyond a single conversation.
The complete expanded edition is available now. Professional copy editing is underway, and Leanpub identifies the current editorial status on the book page.
This book is for you if...
You know exactly what needs to exist. You just never thought you could build it.
You're an operator, not a developer
You understand your problem better than anyone. You've been waiting for someone technical to care enough to build the right solution. This book gives you a way to build it yourself.
AI lets you move fast, sometimes too fast
You have watched polished output arrive before the thinking was finished. The book shows how context, specifications, review, and evidence keep speed from becoming slop.
You want the work to survive the conversation
You are ready to move beyond a single chat into durable context, memory, review, ownership, and an agent system you can supervise.
The expanded edition
Four parts. From a first build to an agent system.
The original edition explained how the first methodology emerged. The complete edition follows what happened next: speed created new failure modes, conversations became systems, and ownership became more important as the agents became more capable.
Working With AI
The velocity trap, the slop problem, context, precise prompting, choosing the right executor, and discovering what the tools can actually do.
Building an Agent System
Memory, orchestration, supervision, feedback, long-term drift, ownership, and the move from one capable conversation to an agentic team.
What This Unlocks
What changes when domain experience can direct execution, where the limits still matter, and how to choose a first bounded build that reality can test.
One method, two ways to use it
Build it yourself, or bring the problem and work through it with me.
Writing the book forced me to turn years of operational judgment into a method another person could inspect and use. The same sequence guides my work with restaurant operators: begin with the real problem, define what better means, keep the first promise small, test it against reality, and leave behind something the team can understand.
The book teaches you to direct the work. Hospitality operators can bring me a recurring problem instead. I use the same method to help them understand it, determine the next practical step, and carry out the work we agree on.
The complete edition is available now
Begin with the problem you already know.
Twenty-five chapters, four parts, and a glossary that keeps AI partner, executor, agent, and orchestrator distinct.
Leanpub carries the current price and editorial-status details.