Why I'm building practical AI
A short note on what I mean by 'practical AI,' why I think it matters now, and the kinds of problems I'm choosing to work on.
There is a lot of noise around AI right now. New demos every week, new tools every month, new claims of imminent transformation every few days. Some of it is real. A lot of it isn't.
I'm less interested in spectacle and more interested in what holds up when you use it for a month.
What "practical AI" means to me
When I say practical AI, I mean three things at once:
- It solves a real problem someone actually has. Not a problem the technology is convenient for — a problem a real person is paying time, money, or attention to right now.
- It earns its keep. It saves time, reduces friction, improves a decision, or unlocks something that wasn't possible before. The user can feel the difference.
- It behaves responsibly. Especially in sensitive domains — health, finance, relationships — it knows what it's doing, what it isn't doing, and where its limits are.
That's a higher bar than "uses an LLM." Most AI products today fail it on the first criterion alone.
Why now
The interesting thing about the current moment isn't the models. It's that the cost of trying things has collapsed. Five years ago, building a useful AI feature meant weeks of training. Today it's an afternoon of prompting and a few days of iteration.
That's a gift, but it's also a trap. The cheaper experiments get, the more carefully you have to choose which ones to ship.
What I'm working on
The projects on this site share a common thread: they take a real, often unglamorous human problem and use AI as one ingredient — not the whole meal. Wellness Coach AI helps people stick with healthy habits. Decision Studio helps teams remember why they made a call. Focus Companion explores what local-first AI can do for daily work.
None of these need to be revolutionary. They just need to be useful.
What I'm publishing here
I'll be writing about the decisions behind the projects, the things I'm learning, and the things that aren't working yet. Honest notes from someone in the middle of the work, not retrospective polish.
If that sounds useful, I'm glad you're here.