Everything here is written by me, not generated. The colophon says which parts of the site aren't.
Writing
-
Writing myself out of the loop
I started my agentic development experiment from scratch, with a vanilla Claude Code: no skills, no plugins, no customization at all. My first two goals were quite clear: move them to the mini-PC and get agents to work nights. …
-
Plumbing in the dusty mini-PC
After scavenging for cables in a box full of different types of cables, I found all I needed to set up the mini-PC at my desk and installed Debian Trixie. Then I installed some basic tools for the setup I wanted (git, tailscale, ssh, docker) and created a bootstrap script to handle the provisioning of the mini-PC. Just a bunch of bash scripts in case I need to reprovision the mini-PC or move the workload to new hardware. It’s ugly but it works. …
-
The identities I left lying around
An agent working on one of my projects needed to open a pull request. The project was configured with a fine-grained personal access token that couldn’t see the repository, so the attempt failed. …
-
Hiring the dusty mini-PC to run my AI agents
I have caught myself staring at the screen waiting for the AI agent to finish or just peeking at what it is doing. Tapping the trackpad so there is some activity to prevent the laptop from sleeping. Walking towards the laptop just to see if the agent finished. As an anecdote it’s fine, as a routine for a professional, not so much. …