Building the Machine, Day 6: The Brakes Worked Too Well

I built a checkout flow that takes a $3,500 commitment and walks a customer all the way to their dashboard. Then my own safety system locked me out of my own database.

Yesterday I built a business in a day. Almost.

Start to finish: a stranger lands on the site. They tell the machine what they want automated in three questions. The machine decides - right there - whether they’re a fit or whether they need to go somewhere else. If they’re a fit, it shows them exactly what they’re buying and what it costs. They pay a deposit. They get walked into creating an account. They set a password. They land in their own dashboard.

That whole path - cold visitor to paying customer with a login - worked by the end of the day. Real Stripe. Real database. Real account creation. I tested it three times and watched it work.

And then you hit the dashboard, and there’s nothing there.

Because the next piece - the part where the machine actually shows you your build and lets you approve it - isn’t ready yet. So right now the machine takes your money, learns your name, hands you keys to a room, and the room is empty. That’s the honest state of it. I’m not going to pretend otherwise. Building in public means showing you the empty room too.

But that’s not even the part that stopped me cold yesterday.

Two days ago I wrote about brakes before gas. About the gate I built to sit between my AI and anything dangerous - deploying, deleting, touching money. A checkpoint it can’t argue past.

Yesterday that gate turned on me.

The machine needed to add four columns to a table. Not delete anything. Not drop a table. Just add four empty slots so new customers had somewhere to put their name and phone number. The most harmless thing you can do to a database.

The gate slammed the door. Classified “add a column” as a destructive operation - same bucket as “wipe the whole thing.” Demanded my authorization, signed with a key it doesn’t have, before it would move.

So I sat there frozen. The machine I built to move fast stopped dead. By the safety system I built to keep it from moving too fast. I built a cage and locked myself in it.

Here’s the lesson, and it’s the whole game:

A safety system that blocks everything isn’t safe. It’s useless in a different direction.

The first version of any guardrail is always too aggressive. You’re scared, so you gate everything. And it works - nothing bad happens. But nothing good happens either. Instead of the machine breaking things, the machine does nothing, and you’re back doing it all by hand at midnight. You didn’t build a business. You built a very expensive intern who needs permission to breathe.

The skill was never building the brakes. Anybody can build brakes. The skill is calibration.

Below, I’m going to show you exactly how I built yesterday’s flow - the three-question bouncer, the deposit-to-account handoff, the one-card-at-a-time dashboard - and the exact rule I used to teach the gate the difference between “let it fly” and “stop and wait for a human.” This is the part nobody selling you the “AI runs your business while you sleep” dream will ever show you. It’s free for my paid subscribers because this is the stuff I’d have paid to know three weeks ago.