About
Building systems that still make sense under inspection.
I work on AI systems and software architecture, but I tend to approach both from the same direction: understand the structure first, then optimize. This site is where I write down the arguments, tradeoffs, and first-principles models that keep showing up.
Working model
My background sits somewhere between engineering practice and mathematical taste. I care about systems that are interpretable, robust, and boring in the places that should be boring. That usually means strong abstractions, measurable feedback loops, and less tolerance for hand-wavy claims than most hype cycles allow.
The writing here stays close to that. Expect essays on model evaluation, software design, typing, probability, optimization, and the occasional note on learning itself. If a topic cannot survive contact with production or first principles, it usually does not survive the draft either.
Defaults
- Prefer explicit systems over clever ones.
- Benchmarks matter less than deployment evidence.
- Mathematics is useful when it sharpens engineering judgment.
- Static sites, open feeds, and small tooling stacks age well.
Site stack
The blog is intentionally small: source posts live in content/posts,
drafts live in content/drafts, and ./build-site regenerates
the homepage, post pages, RSS feed, sitemap, and robots file. No framework runtime,
no CMS, no manual SEO drift.