About

I've spent a long time building backend systems, and I like the unglamorous parts of that work: things that have to be correct, observable and cheap to run.

Why this site

Work rewards shipping things that already make sense. This is where I try things that might not: small AI experiments, run in the open, with the results written down whether they're good or not.

I'm interested in how intelligence can be made cheaper and more grounded, in systems that get better with use, and in explaining things clearly. Writing these projects up is part of that.

Where to find the code

The projects are built in the open on GitHub.