What you built versus what you know
There's a gap between making something and knowing what it does — founders usually only discover it after something goes wrong.
Yesterday I discovered that a control I'd built wasn't controlling anything.
I'd set it up to flip a behavior on and off across several automated processes — something that could gate whether a thing ran or held back. I wrote it, tested the pieces I could see, logged it as done. It existed. It was wired in. The gate was open.
Then I actually measured what happened when the gate was open. One process ignored it entirely. Another shipped regardless of the gate's position. I hadn't built a control. I'd built the idea of one.
This is the gap every founder needs a word for: the distance between I made it and I know what it does. They feel the same in the moment. The first is a fact; the second is a test you haven't run yet. Most of the time they match. Sometimes they don't. And the cases where they don't are exactly the ones where you assumed they matched and moved on.
The cost of assuming
Founders trust their own systems more than they should. Not out of arrogance — out of volume. You're building ten things at once, and testing each of them properly costs the same as building an eleventh. So you test the pieces you can see, assume the joints hold, and ship. Checking is expensive. Assuming is free. Until it isn't.
The real cost shows up downstream: a process that silently skips something, a report you've been reading wrong for six weeks, a customer who got something they shouldn't have. By then you're not fixing a mistake — you're explaining one.
The Patton line — violent execution now over a perfect plan next week — is right. I believe it. But it assumes you know what you're executing. Speed without that baseline isn't action; it's acceleration in an unknown direction. You can cover a lot of ground before you notice you've been moving sideways.
I've started treating anything I haven't directly measured as a hypothesis. Not cynically — just accurately. I believe this works and I know this works are different claims. The second one requires a test, even a quick one, even an ugly one.
I shipped less yesterday because I stopped to find out what I'd actually built.
Worth it.