AI writes a lot of my code, and AI-written code skips the boring security basics. I installed Cisco’s Project CodeGuard as a Claude skill, tested it on a small app I VIBE coded, and found a real vulnerability. Then I made it part of my AI harness.
For two months I built software mostly by describing problems to an AI in plain sentences. Out of that came a box inventory with QR labels, an offline-episode queue that replaces a feature Plex removed, an audiobook catalogue, an off-box “break glass” map of my network, a photo pipeline for a Leica, a slideshow maker, and a 3D-printed lens case that took ten tries. Then one ordinary reboot left half of it running blind for fifteen hours without a single alert. That night taught me more than the building did. This post covers the apps, the reasons I couldn’t just buy them, what broke, and the safety nets I built afterwards.
I picked up a Leica M EV1 and spent the first week walking Round Hill, Virginia with three manual-focus lenses: a Summilux-M 50 f/1.4, a Summarit-M 90 f/2.0, and a Voigtländer Color-Skopar 35 f/3.5. Nine frames from the first walks.
At the AFCEA San Diego Mission Systems Symposium, Gene Kim described how sailors had built eleven mission-critical applications themselves and were using them — work that once took “millions of dollars and two years” and now takes “tens of dollars in tokens and a couple of hours.” He was describing what I’d been doing in my homelab all year, and finally gave me the word for it: vibe coding. So after the session I showed a colleague, by building one. Twenty-three minutes later there was a live website with its own address, a real certificate, a public hostname, documentation and a change ledger. Without the AI harness behind it, that’s a month of evenings.
Working almost entirely by talking to an AI, I set out to stop losing track of homelab IP addresses — and accidentally rebuilt most of modern platform engineering. A running translation of what each ‘thing I made’ is actually called.
Managing a multi-server home lab environment with custom OPNsense routing and multiple unRAID boxes generates a mountain of documentation. Recently, after deploying a high-availability Kea DHCP server and a web-based subnet editor, I decided to have a local AI model write up the project notes for me. I set up a test on my macOS workstation using LM Studio, connecting […]
Introduction Every homelabber’s worst nightmare is the day their network goes dark because of a single point of failure. For me, that day arrived when my UDM Pro’s SFP+ port failed silently — taking down Plex, HDHomeRun, Home Assistant, and every other service dependent on DHCP. The device had become the bottleneck for my entire local network. That failure sparked […]
How I turned a ULANZI TC001 into a living dashboard for RSS feeds, system alerts, and home automation. The Hardware At the center of this project is a ULANZI TC001 Smart Pixel Clock running AWTRIX 3 firmware v0.98. This isn’t your average smart display—it’s a programmable LED matrix clock that supports custom apps via MQTT, HTTP APIs, and even sound […]
AI Prompts — Knowledge Wiki Context Claude Needed to Do This Right Before any implementation, Claude needed to read the memory files at: ~/.claude/projects/-Users-kcossabo-Documents-Coding-Container-Managment/memory/ Specifically: – project_container_pipeline.md — the Gitea/Komodo workflow, compose.yml conventions, unRAID labels, known landmines – project_komodo_containers.md — server inventory, IP ranges, bridge names, Komodo field gotchas Without this context, Claude would generate generic compose files that don’t […]
Lessons — Knowledge Wiki What Broke Git dubious ownership error on NFS mount git init on /Volumes/knowledge failed with “detected dubious ownership” because the NFS mount is owned by a different UID than the Mac user. Fix: git config –global –add safe.directory /Volumes/knowledge Run this before any git operation on an NFS-mounted share. Git initialized as master, not main git […]
Recent Comments