[ personal site / 2026 ]

I'm a product leader working out loud on that question — trusted data foundation, agent context, taste and evals.

IC → leader → builder. Years of scoping and setting direction, plus the last twelve months spent shipping it myself.

If you're building in context engineering, knowledge graph, or applied AI in complex workflows, I'd enjoy comparing notes.

01 / why

Why this site exists

Three reasons.

01
Show, don't tell.
The cheapest way to claim you build with AI is to write a LinkedIn post about it. The most expensive — and most honest — is to ship something. This site is the something. Every line of copy, every component, every deploy is a small bet on my own taste.
02
A craft challenge.
Anyone can have an LLM produce a site. Almost nobody can produce one that feels like them. That's the real test of the AI-native era — not whether you can prompt your way to output, but whether your judgment shows up in what you ship.
03
Compound notes.
I'm sharing the workflows, mental models, and mistakes from months of building. If you're earlier in the journey, you may find shortcuts. If you're further along, push back — I want to be wrong faster.

The five modules02 / map