A decade of watching what works — and what doesn't.
I came up inside agencies, watching good money build the wrong things.
I spent years in the agency and services world — close to founders, close to the work, close enough to see where it went wrong. And it went wrong in the same place almost every time.
Not in the code. The teams could build. It went wrong earlier — in the decision about what to build. A scope too big to ship. A stack chosen for the wrong reasons. AI bolted onto something that never needed it. By the time anyone noticed, the budget was half gone and the product still didn't work.
I kept seeing the same expensive mistake, dressed up in different clothes. So I started paying attention to the part everyone else skipped.
Most AI products fail before a line of code is written.
That's the conviction the years gave me. The build isn't usually the hard part — the thinking before the build is. What it actually needs to do. Where AI earns its place and where it's just expensive theatre. What to ship first, and what to leave for later.
"Get those decisions right and an ordinary team ships something good. Get them wrong and the best engineers in the world just build the wrong thing faster."
