Preface to This Edition

In 2019 I published a book called Cómo se hace una app (How an App Gets Made). Its first line said: "Having an idea for an app is easy. Making it happen, not so much."

Seven years later I reread it and felt a little embarrassed, and a little proud.

Embarrassed because half the book aged badly. There were whole chapters on Material Design, on the 500 shade of a Google color palette, on how many dp a margin should be, on how to build the prototype in Adobe XD. Adobe stopped developing XD a long time ago. Material's rules have changed twice. And here's the big one: the person who, in 2019, had to take my prototype and turn it into code, the programmer, is now in many cases an AI that writes in an afternoon what used to take me a month.

Proud because the other half of the book, the half about thinking, holds up just fine. I'd say it's worth more today than when I wrote it.

Let me explain why.

What changed

In 2019 the bottleneck was programming. You had the idea, and between the idea and the app there were months of work by someone who knew how to write code. That's why the original book ended with you handing a stack of stapled A3 sheets to a programmer.

Today you can type "make me an app for booking soccer fields" into an AI and ten minutes later you've got something that opens, has buttons and even saves data. I tried it. It works. And it's awful.

Not awful as in ugly. Awful as in generic. It's the app anybody would make, because that's exactly what the AI did: it averaged out every booking app it has ever seen. It doesn't know your users organize the game over WhatsApp, that the venue owner answers the phone with one hand while taking money with the other, that when it rains half the bookings fall through and nobody knows who's refunding the deposit.

You know that. Or you can find it out. But you have to put it somewhere the AI can read it.

What didn't change

This book rests on a single idea, and I'm going to repeat it a few times because everything else hangs on it:

If you don't decide, the average decides for you.

Every question you leave unanswered before asking the AI for something, the AI answers on its own. And it answers with whatever is most likely, which is almost never what's best for your case.

Without my knowing it, the 2019 method was a way of answering all those questions before anyone wrote a line of code. Idea, landscape review, needs, brief, telling the app's story in words, flow diagram, list of screens, StoryApp. I designed it so a human programmer wouldn't have to guess. Turns out it's exactly what an AI needs so it doesn't guess either.

So the method stays. What changes is everything that comes after the paper.

What you'll find here

The book has three parts.

The first, Think, is the oldest part and the most important one: what your app is as a product, how it makes money, who it's up against and who it's for. There's no technology in it. There are uncomfortable questions.

The second, Draw, is where the idea comes down onto paper. Telling the app's story in words, building the flow diagram, listing the screens and the things your app has to remember, and drawing the StoryApp. By hand. With a thick marker. You'll see why I keep going on about the marker.

The third, Build with AI, is brand new. What AI does well and where it gets things wrong, how to put together the package of files you'll hand it, what skills and agents are and what they're good for, how to build in small slices without getting lost, and what happens when the app goes out into the world.

All through the book we're going to think through one app together, from scratch. It's called Falta Uno. It's for booking a five-a-side soccer field and filling out the team when someone drops out at the last minute. I picked it because it's small, because we all know the problem, and because it has enough traps to learn from: payments, cancellations, two kinds of user, rain.

Who it's for

For the person who has an idea and can't code. In 2019 I promised you could do all the work that comes before programming without knowing how to program. Today I'll promise you more: with this method and an AI, you can get a lot further. I won't promise you'll make it all the way on your own. There are parts (payments, security, your users' data) where you'll want someone who knows what they're doing sitting next to you. I'll tell you which ones.

For the programmer who already uses AI every day and feels like the app that comes out is correct but not good. What's probably missing is what this book calls thinking.

And for the person who teaches. I teach AI in person at The Mariscal's academy, and this book is, in large part, what I repeat in class every week.

A word about tools

I'm going to name specific tools, because a manual without names is useless. But AI tools change every six months. By the time you read this, one of them will be gone and there'll be a better one. Doesn't matter. If you understand what each thing is for, you'll recognize its replacement the moment you see it.

Paper, on the other hand, never needs an update.

Martín Silva Molina October 2026