I build cloud reliability and observability infrastructure, and the API and agent gateways that sit in front of AI systems. In practice that is Apigee hybrid on Kubernetes, and the infrastructure I am responsible for handles billions of API calls a year. Most of the work is making a system’s behavior visible before it becomes an emergency, and making the cost of that behavior visible at the same time. I am based in the United States.
Cost is the part I keep returning to. I ran an AI agent on Google Cloud and measured what it actually cost against the billing export, then published the numbers and the queries behind them.
This site is where the rest of my thinking goes. There are deep technical posts, a six-part series on what money actually is and who designed the scoreboard, essays on parenting and village-to-city tradeoffs and what it costs to eat well, short notes, impressions of books and films that stayed with me, and a page listing my standing orders at the restaurants I go to often. It is one habit pointed at different objects, whether the object is a billing export or an unwritten family obligation: take the thing apart, find out who built it and who it serves, and write down what is there.
A few things about how I work, because they shape all of it. I give myself at least two weeks before any major decision, since clarity almost never arrives in the moment. I hold myself to full commitment or honest withdrawal, which means I turn down work I cannot be genuinely in on. I try to reason my way to conclusions instead of absorbing them from whatever the room already believes, which makes me slow to agree and slow to join. And I do not publish negative takes on books or films. Cataloguing what failed to land is not why I read.
The values page is the long version, with what each one looks like in practice. My cloud, SRE and observability, and agentic AI credentials are on the credentials page, each linking to Credly so you can verify it yourself. If you want to say something, my email is on the contact page.