It remains wild to me just how bad AI generated anything looks when you're a subject matter expert in that thing.
Similarly, it remains really sad to me when I see Claude "generate" code, and then find the project from 8 years ago that it completely word for word stole from.
I assume LLMs are absolute magic to people who don't know anything about a given subject matter, and as someone who is currently going through wisdom of the ancients for exim configurations, I can almost understand the appeal, but like, the worst possible outcome imo here is that I use a chatbot, and it seems to work, but then I offer this as a service to real customers, and it breaks, and I literally don't understand the product well enough to help, so then I need to keep visiting chat bots over and over again to ever do my job, tacking on more and more expense and ruining my ability to troubleshoot anything, which is what I pride myself on: because I'm an engineer.
My whole profession and skill set is built around my ability to build complex systems, troubleshoot them, scale them, and optimize them for reliability and availability in worst case scenarios. There's an art to this work and the meticulous documentation to do it correctly such that it will run long after you die.
I know a lot of people will say "but I just use LLMs for the non-creative boring stuff" and like, as someone whose skillset is that "non-creative boring stuff": That fucking hurts.