The project to build out the homelab continues. The scope and scale has expanded and using an AI coding assistant to get this set up has been an invaluable tool to get me to the level of functionality that I’m striving for.
That said, in trying to troubleshoot an issue with little available time to do so, I had to rely on the AI tool to discover the root cause and fix the problem as I still have a day job. The rules for the lab still hold true – it is not load bearing, but it is still used, and systems down, I’m mostly annoyed and it adds something to my to-do list.
As I continue to build this out, the plan is to move from a hosted AI tool (currently Claude) and switch to something self-hosted. I have limited my budget for any new hardware acquisitions, so trying to find a GPU with sufficient memory to run a decent model, or buying a $10k+ Mac to do this work is way outside the realm of possibility right now.
Over the last few weeks, I’ve also had time to reflect on what I’m gaining by having the bulk of the work deployed by an AI agent –
- Experience working with AI tooling and learning some of the safety boundaries, along with understanding how my career trajectory is going to change as these tools continue to develop and become lower cost.
- The feature outcome, which is the majority of the point of this project, with a lower amount of time put into the environment than what I would have if I did it by hand.
- Exposure to new products and tools running in my own environment that I can safely test, destroy, and redeploy.
- A noticeably increased power bill.
What I’m giving up –
- The experience of deploying and directly supporting these tools.
- The joy and frustration of building something myself.
- Side note with this – The experience of deploying a piece of software that I can only see the results of is not as meaningful as seeing the server cabinet nicely cabled and clean. I’m not sure what to make of that yet.
The lab status page is updated a little less frequently now as my progress has slowed a bit, but you can check out the latest status at: https://lablog.cloudlabs.fun/status/