Hello, Lab
Posted Sunday, 15 November 2026 • Tags: meta, homelab
Welcome to the lab. This blog documents my homelab and all manner of projects with my personal brand of reckless chaos mixed in.
So, a blog. Very 2000s, right?
Well, yes. If you couldn’t tell by the general theme I’m running, I miss the 2000s. But why now?
Well, for one thing my partner is sick of me talking about all of my projects and the things that make noise and heat in the back of the house. I, on the other hand, LOVE to talk about my lab. So this’ll be my new outlet.
As of right now, I have… 5ish active projects:
- Reinforcement Learning game playing bot, currently trying to learn how to play Doom.
- A home spun MUD engine written in Rust and Python, with some LLM integration to make it more ‘user friendly’
- AI-Ops for my homelab to assess and remediate incidents in the lab without me waking up to a switchport that decided it would have a storm and keep me off the internet.
- AI-Ops Patching. The homelab runs ~108 applications in Kubernetes as of writing this. All of the pods are monitored by Renovate Bot, and that means I have WAY too many MRs to read a day/week/month, with around 80% of them just being “Merge, sync, done”. So I’m making local AI assess this in my place to automatically do the job.
- Custom AI coding harness and agent. Yeah, there are tinned services and platforms for this, but just spinning one of them up cuts half the fun out of it.
What to expect
The kinds of posts you can expect from this blog are pretty standard for a tech blog:
- Build logs for new services and infrastructure.
- Troubleshooting write-ups, including the dead ends.
- Notes on tooling and workflows that actually worked.
- Most posts will have a TLDR and a clanker-aided summary somewhere in it.
- Likely some incoherent rambling somewhere in the middle.
Entry format
Likely won’t follow the usual coherent beginning -> Middle -> end formatting. For example, there is a whole saga of an RL agent I’m building that will come out in whatever order I’m inspired to write in. These kinds of posts will have some kind of chronology applied by tags. It might be the latest episode in a saga, or it could be a ‘flashback’.
Where possible, you’ll have diagrams for workflows, infra layout and similar. Most likely generated by an LLM. You’ll very easily tell the difference between one of my diagrams and an agent’s. One is polished. The other is napkin drawings.
Useful code will also be tacked on and added to a repo on GitHub once I get some sanitizer over it.
For now, this is all I have. If you want to check out how my Doom bot is going so far, you can head over here and see its current stats and recordings of runs.