Conversational AI

I have begun using conversational AI in my hobby Linux system administration projects. I've found it's much faster to arrive at a solution than querying the traditional resources: Wiki, IRC, forums, mailing lists, subreddits, etc. Not only does the agent have access to a lot more information, it can wait on your next prompt indefinitely, and won't usually disappear unless the service goes down, or you exhaust your allotted time/cycles for the day.

The key to using conversational AI is this: test and verify. Don't prompt once and expect the response to be perfect. We as humans give limited information, and the agent can only work with what you give it. If the agent presents something as fact, check another reputable source (instead of the one the agent cited).

The reason I call it "conversational AI" is that it is indeed a conversation, albeit with a machine. It's a dialogue. If you're doing it right, not only are you asking the agent questions, it asks questions and will seek to "understand" your prompts.

Most of my chats have been trying to fix some Linux system administration headache, or improve the system on my laptops, desktops, and servers. I would definitely not use conversational AI for any human problems, like with human relationships. I think the results would be disastrous, and I've heard horror stories confirming this.

I've listed the articles on this blog (discursions) that were written with the assistance of AI. The links below are paid content to the raw chat conversations, with dead ends, backtracks, and all. There are also a few hallucinations, at least one the agent admitted to itself without being called out on it!
NOTE: As of 2026-04-20, these links just bring the browser to the main page.

systemd-homed resize (out of space)
ThinkPad Hardening
Named Workspaces - Multi-Monitor
LUKS2+TPM2 Issue
Intermittent Fingerprint Reader Issue