Luka van Maren · UBC commerce student
I'd rather finish something and live with it.
I study commerce at UBC and spend my own time on the overlap between business judgement and what AI now lets a very small team build. Describing a product teaches me very little. Using one I made every day teaches me a lot.
What I'm trying to learn
How to find a problem that is actually worth solving, turn a pile of tools into a workflow I keep using, and tell the difference between something I like and something other people would choose. The software is one part of that. The product decisions and the honesty about evidence are the rest.
Commerce gives me the language for markets, costs and positioning. Building gives me the part that argues back: a decision that sounded clean in a note gets rejected the first morning I have to live with it.
How this gets built
I choose the problem, decide the tradeoffs, use the result daily, and review every change against the workflow it is supposed to serve. I am responsible for what I put in front of people, including this site.
AI coding tools do a large share of the implementation and testing. They research approaches, write and revise code, and run checks; I direct that work and reject what does not hold up in real use. I do not describe this as solo, unaided engineering, and I do not describe it as client work.
Problem choice · product decisions · tradeoffs · daily use · review and rejection · what ships
Research · implementation · refactoring · test and browser checks · drafting
Where to look next
Health OS is the most complete example right now: a tool I built for my own health routine and still open every day. It is private, owner-only, and the case study is as clear about what it does not prove as about what it does.