Technical experts have zero customers

Someone ships a working app built with AI. Within hours, the technical feedback rolls in.

"This will break with concurrent users." "Your database queries are terrible." "What about error handling?" The list goes on. Every potential failure mode, catalogued and explained.

The person giving feedback has been burned before. They've seen race conditions take down production. They know what happens when databases lock up. They're trying to help.

They're also solving problems that don't exist.

At Leap, I watch this happen constantly. Someone builds a functional product that solves a real problem. The technical crowd immediately lists everything wrong with it. Meanwhile, that "broken" app is making money.

A friend sent me this screenshot the other day - his backend had been running for 7 days without a proper database setup. Just failsafes returning empty responses. Any technical expert would tear this apart immediately. (We're talking about a software architect with 15 years of experience.)

But people were using it daily and paying for it.

This dynamic isn't new. @levelsio has been doing this for years - building million-dollar businesses on PHP and a single VPS while technical experts explain why his stack won't scale. He keeps shipping, they keep criticizing, he keeps making money.

The web development industrial complex has lied to your face:

- auth: yes you can do it yourself
- scale: yes you can do it yourself
- database: yes you can do it yourself
- server: yes you can do it yourself

And no when your site goes viral you don't need to pay a $20,000 bill https://t.co/JUbX7n5oN7

— @levelsio (@levelsio) August 24, 2024

These builders understand something fundamental: customers care about their problems being solved, not your architecture. Product-market fit matters more than technology-product fit.

AI builders ship software that works for their users. Technical experts explain why it's wrong. One group has customers, the other has opinions.

The technical feedback isn't wrong. Race conditions are real. Memory leaks happen. But most software never hits the scale where these problems matter. We've convinced ourselves that every app needs Netflix-level architecture. Most successful software serves a few hundred people who care about the problem it solves.

The AI builder shipping ugly code that works is making the right trade-off. Fix real problems first. Technical problems are a luxury you can afford later.

Build for the users you have, not the ones you imagine.