Hello,
I'm Phil.

I started where most of us start: writing code badly, then slowly writing it less badly. Developer, then senior, then architect — and then one day I looked up and I was managing the people writing the code instead of writing it myself. I love the code and building, and I love the people side more.

Then AI showed up, and I hated it. The output was bad, it took more effort to prompt the thing to do something than it did for me to just do it myself. Uninspiring. It couldn't read my mind.

So I did the only thing I know how to do with something I don't trust: I took it apart. Everyone else was so excited about LLMs and chatting, and I didn't get it at all. So I figured etiher everyone else was wrong (it happens) or I was. And it turned out it definitely was a "me" problem. I had been vague and expecting telepathy.

So I deep-dived (it's a word), got some API accounts and went to town, burned 30 million tokens in the first 3 months (which in today's agentic era is laughably peanuts).

I haven't really come up for air since.

I have a hard time leaving stones unturned. If I can't take something apart and explain how it works, I don't really believe I understand it — and if I don't understand it, I haven't earned the right to dismiss it.

I wanted to be absolutely sure AI sucked because it sucked — not because I sucked at it.

That's the whole thing, really. I went in to confirm a complaint and came out with a craft. It turns out there's no bottom to this one: every layer I open has another layer under it, and the better I get at the telling, the more the thing gives back. That's what keeps the lights on late -- I'm lucky if I ever get to bed before 2:00 AM.