Prologue
This book was written with a machine.
It's a short book about the work that goes on before you build anything — the work that quietly settles whether a small thing flies — and how much of it you've already handed to a machine. I had the idea, the arguments, the stories. But a lot of the sentences came out of an AI, and I pushed them around until they sounded like me. It would be a strange book to hide that in. Half of what's coming is about what these tools do well and what they do badly. The sentences are a thing they do well. I'll get to the rest.
It costs about the price of a coffee. That's on purpose. I'm not sure yet whether this is a book or just the idea of one, so I'm shipping the cheap version and watching what happens. If enough people read it and tell me it was worth their time, there'll be a better one. If nobody does, I've lost a weekend, you've lost a couple of dollars and an hour, and we've both learned something.
That's the whole argument, done to myself before I ask you to do it to anything of yours. Build the small version. Put it in front of real people. Watch what they actually do, not what they say they'll do.
So do me a favour. When you finish — if you finish — tell me what landed and what didn't. The bits you nodded at, and the bits where you thought, no, that's not how it works.
I'm reading.
Charles B. Fry
Chapter One — The Cheap Back End
Two years ago, starting almost anything cost money you had to find first.
If you wanted to sell something you needed a website, and a website needed somebody who built websites, and that somebody wanted a few thousand dollars and six weeks. A logo was a person. Bookkeeping was a person. A contract was a person too, usually an expensive one who wrote in a language invented to make sure you'd need him again next year. Advertising copy, a person. A business plan, if you were the sort who wrote them, was either a long miserable weekend or another person. Every one of those was a small wall. Most people, standing in front of enough small walls, go back inside and watch the telly.
That's gone now.
All of it sits in a browser tab for about the price of a phone plan. The website builds itself while you make coffee. The logo turns up in four versions before you'd decided you wanted one. The contract, the bookkeeping, the plan, the personas, the market research, the customer emails — twenty dollars a month, and most of it good enough. Not all of it. Good enough.
So the walls came down. Which means a great many more people are about to start a great many more things than at any time I can remember, and I mean that plainly as a good thing. The window-cleaning round, the ceramics, the side thing that might become the main thing — the cost of finding out just collapsed, and a lot of people who'd have stayed inside watching the telly are going to come out.
Then there's the other half of it.
The same tab that builds the website will also tell you whether to build the thing at all. Ask it. It'll answer. It'll tell you the idea is strong and the timing is good and the market is ready. It'll tell you who your customer is, what to charge them, and how to know it's working. It does all of this instantly, in clean confident paragraphs, and it costs the machine nothing to be wrong.
Read that last part again.
It costs the machine nothing to be wrong.
When a builder quoted you six weeks, he was staking six weeks of his life. The advice in the tab has no skin in it. Nobody knows whether an unbuilt thing will work — not you, not me, not the machine. The difference is the machine doesn't know in flawless grammar.
It will be wrong, and the spelling will be perfect.
And here's the part the clean paragraphs won't tell you, because it isn't really in the bit of the internet they grew up on. Most things fail. They always have. Roughly eight in ten, depending whose numbers you read and what they're counting — across industries, across decades, funded and unfunded, boom and bust. The number barely moves. New cafés, new apps, new products coming off the line at companies with marketing departments bigger than my whole town. Eight in ten, near enough, end up in the ground.
And these aren't the amateurs. Quibi. Juicero. New Coke. Google Wave. Disney's pour-the-billions-in-and-watch-it-burn streaming years. The best people money could hire, research none of us will ever afford, and the cows still wouldn't use the robot.
So if the giants get it wrong eight times in ten with all that, the question for the rest of us — starting small, with our own money, in a browser tab — was never how do I get the machine to tell me I'm right. That's the easiest thing in the world to arrange. The machine will always tell you you're right. The question is how do I find out whether I'm right before it costs me the thing I can't afford to lose.
The whole answer comes down to what you decide to trust.
You can't trust what people say. They're kind, or they're polite, or they don't actually know what they'll do until the moment lands in front of them. Ask your mate if he'd buy it and he'll say yes.
Because he likes you.
You can't trust what the machine says. It's fluent and it's certain and it has never once paid for being wrong.
You can only trust what people do. Money changing hands. A name on a list that took some effort to get onto. Someone coming back a second time. What a person does when there's something real on the line — that's the one signal in this whole business that doesn't lie to you.
That's what the book is about. The work that happens before you build — the deciding work, the part that settles whether the thing flies before you've hammered a nail — has quietly been handed to a machine. Some of it the machine does well. Some of it the machine does badly, in a way that feels exactly like doing it well.
Telling the two apart is the job. And the first thing you handed over is the most basic one of the lot. You asked it whether your idea was any good.