I didn’t start in software. I started behind a mixing desk, running sound for live television — X-Factor, The Voice, and a few hundred nights where a single missed cue was broadcast to everyone watching. Live production teaches you something no tutorial does: the show does not stop for you. You build a system that holds, or you get found out in real time.
That instinct carried straight into software. I came in through testing — the part of the industry everyone treats as a stepping stone — and found it was the best seat in the building. Testers see where things actually break. I spent years at Applause, Ubertesters and Smule learning exactly how production software fails, and then got tired of finding the same failures by hand.
So I automated it. Playwright suites, CI pipelines, visual feedback hooks that told you what broke before you asked. Somewhere in there the job changed shape: I wasn’t testing the product any more, I was building the machinery that tested it. That’s the moment the rest of my career turned on — the realization that the highest-leverage thing you can build is a system that does the work for you.
I’ve been chasing that idea ever since. First Vision One, a technical recruiting practice I ran through the pandemic — two years of matching engineers to teams by hand. Then Jobrain, which was me getting tired of doing that by hand and building the ML platform to do it instead. Same lesson twice: the bottleneck is almost never the code. It’s the number of things one person can hold in their head at once.
Which is what codmir is for. It’s an AI development platform — 70+ services spanning agent orchestration, cloud workstations, real-time collaboration and an IDE — built on one premise: a single engineer with a clear enough vision should be able to direct a fleet of autonomous agents the way a director runs a production. Not autocomplete. Agents that plan, write, review and ship, and keep working after you close the laptop.
I build systems that build themselves. That’s the whole thing, really.