About this site

I'm Josh Hornsby. I've spent close to twenty years as a software engineer, most of it as a principal IC on systems where being wrong is expensive. Large-scale healthcare infrastructure, most recently, and agentic AI now.

The constraints are different. Non-determinism instead of consistency, latency tails instead of throughput, rubrics instead of SLAs. The shape of the job is the same: figure out what actually happens in production, measure it, decide what to do about it.

This blog is the field notes from that work. Usually a problem I hit, a small experiment, and what the numbers actually say.