Friday, June 6, 2025

Building KubeSkippy: Learnings from a thought experiment

So, I got Claude Code Max and I thought of what would be the most ambitious thing I could try "vibe"? As my team looks after Kubernetes, and I know a bit about the challenges there, I figured - An AI-powered Kubernetes Operator that just fixes all the things! That should be challenging enough! 

The tl;dr is, this was not perfect, however it was actually very impressive what I was able to get Claude to do...

It wrote operators, integrated to local Ollama, deployed everything got it running, added in Prometheus and Grafana, build dashboards to try show everything working, build demo apps to simulate scenarios and actually triggered AI powered healing events.

Was it all easy? no
Was Claude sometimes frustrating not listening to config or commands? yes
Did everything go to plan and work? no
Would I trust any of the X lines of code running anywhere but a local test cluster? no
Was it impressive actually how much it achieved, given that I had not yet known to spend much more time on the planning and getting the pre Context and Prompts sorted? Yes

I mean look at Architecture Overview and AI-Driven Healing Explained as examples

Building KubeSkippy: Learnings from a thought experiment

So, I got Claude Code Max and I thought of what would be the most ambitious thing I could try "vibe"? As my team looks after Kuber...