I have a love-hate relationship with Home Assistant. I love its mission and I love it in spirit, but whenever I need to add or change something in it, I don't love the process. Without disparaging the work already done on improving it in recent years, I still find the UI and UX to be lacking. Claude Code has been shifting my perception much closer to the love end of the axis, because it allows me to side-step the boring parts of managing my Home Assistant instance and it is able to carry out the changes I want very reliably.

I still struggle with letting go of writing code and becoming only a full-time reviewer when it comes to AI agents doing programming, but I don't struggle in the slightest with assuming the position of a reviewer of the changes CC does to my HA instance, delegating all the work to it. The progress I made on making my house smart and setting up my dashboards has skyrocketed compared to before I started using CC to manage HA via its REST and WS APIs.

it's kind of fascinating, LLMs suddenly are making the Linux Desktop waaay more accessible, of all things.

All those fancy GUIs in Mac and Windows designed to be user friendly (but which most users hate and are baffled by anyway) are very hostile for models to access. But text configuration files? it's like a knife through butter for the LLMs to read and modify them. All of a sudden, Linux is MORE user friendly because you can just ask an LLM to fix things. Or even script them - "make it so my theme changes to dark at night and then back to light each morning" becomes something utterly trivial compared to the coding LLMs are being built to handle. But hey, if your OS really doesn't support something? the LLM can probably code up a whole app for you and integrate it in.

I think it's going to be fascinating to see if the power of text based interfaces and their natural compatibility with LLMs transfers over into an upswing in open source operating systems.

Nope, never trust AI to do such things, it’s imminent to cause issues. Maybe as an assistant only but never installed on the same server and worse, the privilege to access/execute commands.

Self hosting post. Tailscale.

Its comedic at this point.

Oh my gosh, everything you want to host comes with a docker compose file that requires you to tweak maybe two settings. Caddy as your web proxy has the absolute simplest setup possible. You don't need AI to help you with this. You got this. You want to make sure you understand the basics so you (or your LLM doesn't do anything brain dead stupid). It's not that hard, you can do it!

Basic question: If I wanted a simple self hosting solution for a bot with a database, what is the simplest solution / provider I can go with. This bot is just for me doesn't need to be accessible to the general public.

Thanks

Opens with "self-hosting" and then brings claude code into the mix. You realize it's not actually running locally right? Privcy-wise that's a nightmare. A non-deterministic blackbox running in somebody's AI cloud is controlling your server. Congrats.

My stack. Claude code working via CLIs: Coolify on hetzner

How is readcheck any different than using raindrop.io?