I build thingsthat runthemselves.

Twelve-plus years of running companies, six of them my own, put me on the paying side of software long before I shipped any. So when I build, I already know where the money comes from, which module sits closest to revenue, and what a wrong build costs, because it has been my budget. That is also why I would rather tell you what not to build than deliver a perfect version of the wrong thing.

AI coding agents write the code. I own the part that decides whether it survives contact with customers: the spec, the architecture call, the review, the deploy, and the phone when it breaks. The proof is systems taking real orders — MAS Group runs a 13-stage quote-to-order pipeline live in the field as a mobile app, Flyt quotes cross-border VAT with deposit and refund logic, QuickFix went from nothing to a trilingual EN/PL/IS site and funnel in 72 hours — and I operate these businesses myself, so a bug in production is my problem before it is anyone's ticket.

Every case study on this site names the alternatives I rejected and where the system still falls short, because the expensive mistakes are the ones nobody says out loud early. I came to Iceland in 2019 with no network and learned the market from the retail floor. What stuck is the habit of taking something genuinely hard and turning it into steps someone else can run, which is why every build ends with your team trained and me stepping away.