17 NOV 2026 / Written by Lukas / 5 MIN READ
A home for our bits
Self-hosted services and pleasantly overbuilt ideas.
On this page
Shibabox is the domain behind our homelab and personal pages. It started in my apprenticeship back in 2018, and it had many many iterations since then.
How it started
First it was a raspberry pi, and alot of ambition. I tried things a couple times, it went from trying to use LAMP Server over to different tools that install and configure applications for you (Cloudron).
Everytime was a miss but I was making steady progress, especially after I was fresh out of apprenticeship and in my first real job with a real salary I could use. Eventually I started watching this guy called TechnoTimLive, first thing I ran into was Kubernetes and I was intrigued, however I needed a stepping stone first which of course was Docker.
This guy single-handedly taught me everything I knew about Linux, Containers and Orchastrators like Kubernetes and how to deploy with things like Helm and automate using Ansible. As the homelab started growing with shitty docker run commands I learned about portainer (it was free and good then) I started getting into hosting actual projects, first one being the same mail server I still run today :')
Some point in the future I decided to take my experience onto my CV and mentioned that I have only used these technologies in my homelab for learning and experimentation. However, it yet landed me my second job as a Junior DevOps Engineer. I got assigned a mentor and dove deep into the world of production ready Kubernetes in a CRITIS environment. I went from a naive young boy to a grown man that learned how to use the tools he was given for the greater good.
So I started offering more and more services, hosted blogs and card pages for friends. Set up gameservers and offered people cloud storage. Over the years I started having alot of fun and my home-lab started becoming something that I wasn’t just building to learn or for fun, but for actual personal use to replace SaaS solutions in my consumer life and get my data under control.
Where the journey went
Now I am a personal data conscious, open source supporting idiot that can’t shut up about how amazing Kubernetes is for scalability and how we can replace the current 20 nodes strong Proxmox Cluster full of Windows VMs with a K3S and Open Source solutions while saving the company a pretty dime. I am still driving change forward and bringing the knowledge my mentor gave to me to the table wherever I go.
Nowadays my homelab is 99% automated. I have my CI/CD pipelines for my own applications or configurations, stuff patches itself and I got security hardening done front and back. Even CVE triage is done by AI now and gives me a very quick way to react to ongoing security incidents.
I still lock myself out of my homelab from time to time, and yet I find new ways in - that remind me that no system ever is 100% proof as long as you have a synology NAS or a port forwarding going on.
The homelab now consists of 8 machines that handle various tasks. From mail-server, over media sharing (photos & videos) and even gameservers or a fully local DNS with wildcard SSL certificates.
The system “just works” and I am happy I can rely on it for most of the time.
The future?!
About those iterations I mentioned at the start, of course the homelab is far from done. Anything ever is… We are planning to move again in 2027 and I have grand plans. My plans involve changing all our networking gear to unifi, going all in on smart-home and into finding creative ways to use local AI.
Even now I am already building a personal voice assistant, ever so elegantly named “Jarvis”, a local “Qwen 4B: Instruct” model that works away tirelessly at my silly requests at an astonishing 30 tokens per minute, all interpreted by whisper and output using piper with the same elegant voice you might know from the movies. Okay well maybe not as elegant, it struggles a fair bit and I don’t have the hardware to run anything better as of right now. While Jarvis is slow, it helped me understand alot where local LLMs can be headed, even with older hardware they can still serve a purpose as a efficient placeholder for otherwise complex code.
I already am finding out so many applicable scenarios just using N8N for these things and having Ollama chirp away at my mailbox or giving me summaries. Ideally in the future I can get some stronger hardware for local models, because building my own “Jarvis AI” has been a dream for ever so long and it seems to finally slowly come true.
Outside of all that, my entire stack will stay in docker compose for now, Kubernetes is too hard to use for Mike and I don’t want to torture the little guy more than I already am with all the ansible automations and cronjobs and other IT shenanigans.
However I do still believe eventually it will reach its final state. As Alan Wake once said: