I am a perfectionist. I had to learn which corners to cut.

The moment it actually clicked

Early in my career I worked at a security company, and we shipped something I had argued hard against. I said no in the room. We shipped it anyway. It was not good, and it was the kind of not-good that matters in security.

For years I carried the obvious lesson from that: I was right, we should have slowed down, we should have been more careful. That lesson is not wrong, exactly, but it is lazy, and it is the same lazy lesson the whole industry reaches for. "Be more careful" sounds like wisdom and functions like guilt. It does not tell you what to do on Tuesday with the next thing on the board.

The real lesson took me a lot longer to see, and it is not "be more careful." It is that careful was only the right answer because of the surface we were on. We were on the one where being wrong cannot be taken back. If we had been arguing that hard about a settings page nobody would die over, I would have been the problem, the perfectionist holding up a launch over something a patch could fix next week. The thing that made my "no" correct was not that I have high standards. It was the surface.

The sort, in plain terms

Here is the whole thing, and it is almost embarrassingly simple once you see it. Every decision you ship sits on one of two surfaces, and you tell them apart with one question: if I am wrong, can I quietly fix this next week, or is the damage permanent the moment it ships?

Most of what I obsess over is reversible. Onboarding copy. A layout. A feature that is not in yet. The pace of a wizard. If I get those wrong, a real user shows me, and I fix it, and the cost was a bad week. My perfectionism on those is not diligence. It is waste, and worse, it is waste that steals time from the surface that actually deserves the paranoia.

And then there is the other surface. A security hole. Someone's data. A bug that moves real money. A kid's information in a product built for kids. There is no "ship it and iterate" there, because by the time you would iterate, the data is already out, the money is already gone, the trust is already broken. On that surface, the caution that looks like perfectionism is just the job, and I will hold a launch for it without apologizing.

The "ship fast and iterate in public" advice everyone absorbed in the last few years was real, but it was never permission for everything. The teams I admire shipped rough products on purpose, the reversible surface, while keeping the irreversible one locked down hard. They were not careless. They were sorted.

How I run the fleet on it now

Across the products my team and I build, this sort is the operating rule, and naming it out loud changed how we work.

On the reversible surfaces, I make myself ship. We put the working version in front of real users and let the evidence redraw it, instead of letting me polish it in private against a standard no user asked for. That has been the harder discipline for me personally, and it is the one that has bought us the most speed.

On the irreversible surfaces, we go all the way. When we touch security, the defaults fail closed: the unsafe configuration is the one that does not run, not the one that quietly ships. When we build for kids, we treat the data and the AI as their own irreversible surface and we review it against real safety standards before launch, not after, because "we will fix the kids' safety issue in the next release" is a sentence that cannot be allowed to exist. The high-stakes actions in any AI we deploy go through an approval queue, with the policy encoded in the tool instead of trusted to a prompt.

The honest part, the part I will not pretend I have beaten: I still catch myself polishing the wrong corner. Last month I lost real time to something reversible while telling myself it was important. The sort is a discipline, not a cure. What changed is that now I can catch it, name it, and move the energy to where being wrong is permanent.

What this article is not

This is not me claiming I have perfectionism solved. I have a rule that beats it more often than it used to. That is all.

This is not an argument for shipping carelessly. It is the opposite. It is an argument for spending your care where it cannot be recovered instead of spreading it thin across everything.

This is not a fixed checklist. Where the irreversible surface starts is different for a fintech, a kids' platform, and a marketing site. The question transfers. The exact line is a judgment you make per build, and it is most of what experience buys you.

One-sentence takeaway

You can iterate your way to a great product, but you cannot iterate your way back from a breach, so stop spreading your perfectionism evenly and spend it all on the surfaces where being wrong is permanent.