I'm Mike McMahon — infrastructure/SRE engineer, AI practitioner, homelab builder. This site is where I document the work: what I'm building, how it broke, and what I learned along the way.
Use the sidebar to navigate. Start with How I Got Here if you're new, or jump straight into the Session Notes for technical content.
Subscribe to new posts: Email digest (daily) or RSS feed
A migration war story: a one-result retrieval bug led me to discover I'd built a whole new knowledge store and never cut over to it. Finishing that cutover — and the deploy pipeline that had been silently dead for weeks — became a lesson in why engineers need PM support, even when the engineer and the customer are the same person.
Phase 2 moved the last Technitium node, dns-1, off Docker. The cluster came up throwing TLS 'UntrustedRoot' — a certificate error that was really a four-hop-deep Kubernetes pod-IP-versus-VIP cascade in disguise. The hard part wasn't fixing it; it was not believing the error message.
Moving a Technitium DNS secondary from Docker to a Talos Kubernetes cluster. The YAML was the easy part; the hard parts were source-IP preservation for DNS NOTIFY, a self-signed cert missing its SAN, and a binary config that can't be pre-seeded.
A controlled experiment: two AI agents, one infrastructure deployment task, same Proxmox home lab. Hermes took 10 hours and $75+. Claude Code took 3 hours and $40. Here's what actually happened.