In terms of usage of AI, I'm thinking "doing something a million people already know how to do" is probably on more secure footing than trying to go out and pioneer something new. When you're in the realm of copying and maybe remixing things for which there are lots of examples and lots of documentation (presumably in the training data), I'd bet large language models stay within a normal framework.

As the great Linus Torvald said:

It's why I strongly want this to be that "just a tool".

The problem I’ve seen is the lack of knowledge retention when AI feeds you stuff, buddy wouldn’t even bother to read nor memorize what it’s telling him and just copy-paste commands thinking it’ll fix whatever obscure issue he is encountering.

I’ve been using Debian for the last ~3ish years now relying on documentation from others so I’ve seen how fragile it can get.