$ cd ./blog
Hello from the basement
Why this blog exists, what it runs on, and what to expect: architecture deep-dives, incident write-ups, and honest notes on AI-assisted engineering.
Welcome to ProbablyFine, a blog about running real infrastructure at home, breaking it in interesting ways, and writing down what I learned.
What this is
I run a production-grade Kubernetes cluster in my basement: nine Talos Linux nodes, GitOps via ArgoCD, Ceph storage, GPU scheduling across three different NVIDIA architectures, and a firewall doing BGP for health-checked service VIPs. It’s over-engineered on purpose. The whole point is to practice the patterns real platform teams use, at a scale where I own every layer and every outage is mine to fix.
This blog is where the interesting parts get written down.
What to expect
- Architecture write-ups: how the cluster is put together and why, including the trade-offs that don’t make it into the diagrams.
- Incident post-mortems: the honest version, where a “healthy” service was quietly broken for two months and the fix was three faults deep.
- Tooling and opinion, including AI-assisted engineering, which I use heavily and refuse to be coy about.
On AI-assisted engineering
Some of these posts, including this one, are written with an AI agent in the loop. That’s
not vibe coding. I design the systems, make the technology calls, review the output, and
own the debugging judgment; the agent helps me implement faster than I could by hand. Posts
written that way are tagged ai-assisted. Owning that openly is the point.
It’s probably fine. Let’s find out.