A million lines of code.
Most of it thrown away.

We built our own operating system. 85,000 lines of code. It manages our pipeline, builds software autonomously, handles our expenses, files our taxes with HMRC, and runs our billing. We watched it evolve from an idea on a whiteboard into something that genuinely runs our business.

It started with 34 microservices. They broke before they scaled....we could see it coming, to be honest. So we rebuilt everything into something that actually works. Large codebases became light skills. Rigid frameworks became self-healing fabrics. We dropped half a dozen standalone products by building one thing that replaced them all.

We built a system that builds systems. It monitors code, checks quality, and writes software autonomously. Then it checks its own work. The factory that builds things has its own watchdog. We remember thinking....that's either mad or genius. Turns out it was both, several times over.

We found every single way to make a brittle autobuild system. Version after version. Each one got closer but revealed new failure modes. Not the building....the knowing-when-it's-broken part. The recovery. The self-awareness. We almost lost count of the versions, but we didn't....we just kept going.

The dev community has a meme: “vibecode is shit.” Yes. That's the point. That's the starting material. The question is what you do with it.

Most people stop at the meme. They point at the shit and feel clever for noticing. We took it and put it under pressure. Again and again. A million lines of pressure. And what came out the other side is a diamond....not because we're smarter, but because we paid the cost of compression that nobody else has been willing to pay.

Anyone else starting this process has nearly 2 million lines of code to throw away before they match where we are today. We build every day, so by the time the market starts to try AI properly, it will take 4 million lines of code to catch up with us.

We don't burn wasted tokens. We form hard experience. And we are not afraid to iterate entire codebases....because in a traditional dev team it would take years to get where we do in weeks.