Blog
Notes on cloud architecture, platform engineering and AI for DevOps, from things I've built, broken and fixed.
I work on several projects at once. This is the harness loop I built around my coding agents (worktrees, tickets as a dependency graph, and a cleanup step), and how it evolved from a shell alias into something I trust.
What I set up (and agree on with the team) before anyone installs Argo CD. Spoiler: most of it is conventions, not tools.
Everyone wants an AI agent in their pipeline. Here's where I've seen them pay off, where they cause trouble, and the setup I suggest to keep them out of production.
Every cloud has its own landing zone accelerator, and they all look different. Underneath, I keep coming back to the same five questions.