Imagine you’re a factory owner in Manchester in 1880. Orders are growing faster than you can fulfil them, and you need another loom.
Buying one means persuading a banker. Running it means hiring skilled workers. Paying for it means finding customers willing to buy what comes off it. Every step forces another conversation, and every conversation can stop you.
None of that was designed as a validation process. It was just that expensive capacity couldn’t be acquired alone, and the people you had to go through in order to get it were entirely capable of telling you the plan was wrong.
For most of business history, that was the shape of it. Reality interrupted the building process.
The software era ran on a version of the same physics. Let’s hire another engineer meant finding one, convincing them to leave a job that already paid, and then answering the question every good engineer asks in the first week, which is some version of what is this for, and who wants it? That is market validation wearing a hiring conversation’s clothes. You simply couldn’t get the resource without meeting people, and people ask things.
So the interruptions came early, before much had been built, when being wrong was cheap.
Three products by Monday
Now suppose a founder is unsure whether his product is for recruiters, sales teams or investors.
Two years ago, serving all three would have meant three sets of customer conversations, three sets of product decisions, and more engineers than he could realistically hire. The cost of not choosing was immediate and obvious. So he would have chosen badly, perhaps, but he’d have had to form a view, and forming it would have required going and asking somebody.
Now he can build all three over a weekend.
This looks like progress and sometimes it genuinely is. But on Monday he has the same unanswered question he had on Friday. He merely has three products attached to it.
This is what I think is actually happening, and it’s more interesting than AI removing friction. The interruptions haven’t disappeared. They’ve moved. Reality still shows up but it just shows up after the building now instead of during it, which means founders are meeting it at the point where being wrong is most expensive rather than least.
And there’s a belief underneath that shift which I’ve started hearing everywhere.
Founders now have effectively unlimited productive capacity, and with it an almost irresistible conviction: that every uncertainty is just another problem to compute through. Enough prompts. Enough prototypes. Surely, eventually, enough generated intelligence will make the decision obvious.
You can hear it in the language.
It was just me and Claude over the weekend.
That’s a sentence that barely existed two years ago. Now it’s quietly become normal.
The conviction isn’t unfounded. The instinct has been rewarded so consistently over the past two years that it’s hardened into a general method.
That works surprisingly well for engineering.
Markets are different. They’re discovery problems, not reasoning problems. The information you need doesn’t exist anywhere in your possession, however much of it you generate, because it lives inside other people and hasn’t been articulated yet; sometimes not even to themselves. No amount of thinking retrieves it. It has to be gone and got.
There’s an obvious objection here, and it’s a good one.
Cheap experimentation is exactly what startups should exploit. Building three landing pages to learn which customers respond is an experiment, and a founder who can run it in a weekend has a real advantage over one who couldn’t. The difference is what the building was for. An experiment has a question it was designed to answer, and a result that would change what you do next. Building three versions because choosing one felt premature, and then calling the resulting activity validation, isn’t an experiment. Brute force merely produces more things.
It’s easy to see why founders end up there, and I don’t think it’s a character flaw.
Building feels brave. It is visible, it produces something by evening, and the productive feeling is real. Talking to potential customers offers you considerably less: a conversation that ends ambiguously, or a rejection, or the lingering suspicion you asked the wrong person and the answer would have been different from someone else. One of those activities gives you a result. The other gives you doubt. When the productive-feeling action is also the emotionally easier one, and when it now costs almost nothing, of course it gets chosen. It would be strange if it didn’t.
What used to prevent that choice wasn’t wisdom. It was expense.
The discovery call
It’s something I’ve begun noticing with surprising consistency in first calls with founders.
I ask what the company is and get a tour instead of a definition. I ask who wants this and get a description of what’s been built for them. I ask what the next twelve months buy and get a feature list. For a long time I misread them as being a presentation problem. Now I’ve come to realise that this is what it looks like when someone reaches an investor having never been interrupted. When every question I’m asking is one that could have been asked twelve months earlier, cheaply, by an engineer or a buyer or anyone at all outside the building.
By the time it reaches me, it’s expensive, because a fundraise is the one conversation where the first answer often has to be the right one. I’m not a hard audience because I’m an investor. I’m a hard audience because I’m the first one.
So the thing worth keeping:
AI didn’t remove the interruptions. It moved them to the end and the whole value of an interruption is that it comes early.
Nothing enforces that sequence anymore. The founders who’ll do well aren’t the ones who build less. They’re the ones who go and get told no while it’s still cheap to change course.
Written from the other side of the table.
INVESTOR TRANSLATION #3
Founder says X / Investor hears Y.
“It was just me and Claude over the weekend.” → Nobody in this process could refuse me.
“We’re building for all three segments.” → Choosing felt premature, so I built instead.
“We’ll validate once it’s live.” → I’ve scheduled reality for the most expensive possible moment.
Worth Keeping
Albert E. N. Gray, The Common Denominator of Success (1940):
“Why are successful [people] able to do things they don’t like to do while failures are not? Because successful [people] have a purpose strong enough to make them form the habit of doing things they don’t like to do in order to accomplish the purpose they want to accomplish.”
Gray was writing in 1940, to insurance salesmen, about the men who wouldn’t pick up the phone. What’s changed isn’t the reluctance. It’s that avoidance now has something to show for itself.
ONE QUESTION
What’s the last thing you built that could have told you you were wrong and did it?
If nothing recent qualifies, that’s the finding. Send me yours; the replies to this one have been more revealing than anything else I’ve asked this year.
Clipped
Unrelated, and the best thing I saw this week: Nico Williams climbed into the stands after Sunday's final and put his World Cup medal around his mother's neck. She walked out of Ghana and across the Sahara in 1994, pregnant with his older brother.
He said it himself after the Euros: he runs very fast, but not as fast as his mother.