Software developer. Minimalist. Builder.
Field notes on homelabs, AI coding agents and terminal tooling: what I deploy, what breaks, what I'd do again. Not sure where to begin? Start with Ghost Protocol.
Minilab
Two houses, Proxmox nodes, Tailscale, agents that never sleep.
Engineering
Workflows, CI, debugging: trust nothing you haven't verified.
Terminal
Zsh tuned until the shell gets out of your way.
Productivity
Obsidian and a system that survives losing everything else.
Day Zero Protocol: Standing Up a Proxmox Homelab Node From Blank USB Stick to Tailscale Mesh
The origin of the homelab node behind Ghost Protocol and Pocket Protocol: Proxmox on a mini PC, one-line installs, a BIOS power-cut gotcha, and the Tailscale mesh.
Pocket Protocol: Driving Pi Coding Agents in Your Homelab From Your Phone
Drive Pi coding agents running in a homelab from your phone: a self-hosted Un Bien relay, pairing over a Tailscale mesh, answering agent prompts, and launching sessions remotely.
Ghost Protocol: Running AI Coding Agents 24/7 on an Always-On Homelab Dev Machine
Why I moved AI coding agents off my laptop and onto an always-on homelab dev machine: unattended agent pipelines, a three-layer Docker devcontainer, and remote access from any device via Claude Code and agent-deck.
Trust No Slicer: Proving the Printer Innocent with curl
Debugging two Orca Slicer failures with curl and raw FTPS logs instead of trusting the dialog box: a method for isolating what actually broke.
Everything Is Fine: The Comfortable Lie of Green CI Pipelines and 100% Test Coverage
Why a 100% green test suite doesn't mean your code is safe: the gap between passing CI and real confidence, and what actually closes it.