I think the difference is AI images tend to create mush or impossible geometry. The ideas here where a minimal change to the design renders the item entirely unusable takes a fair bit of creativity.
I bought a new door for my house recently. When the salesman asked what type handle/knob I wanted, I had a bit of an internal crisis. The one he said post people got seemed like it would create a Norman door, which I desperately wanted to avoid. I ended up getting a standard knob to avoid being the absolute lunatic who spent 6 hours debating the merits of various door handles, but had I been alone, I would have absolutely done that. I still feel like I made a mistake every time I look at my door.
Well, bash offers vi and Emacs as editing modes. We're already covered on that front. Many of the parameters for ls are cryptic, making it awkward to use for anything other than routine tasks without referencing the man page. more is so limited that many people choose to use a program used to concatenation files (cat) as a file viewer. Those who don't want to reach for their mouse to use their terminal's scrollbar buffer will use less, since it does more than more. Don't bother parsing that last sentence with bison, unless you have a yacc to shave.
That is an interesting point to bring up, because this type of "almost but not quite right" is exactly what AI seems to naturally create.
I recommend this book to anyone remotely interested in design. Even today it is fantastic.
I never look at doors, without evaluating their usability, anymore.
The book was a gift and a curse.
i also do this for ui and app logic: go to some Microslop service, they are all like these...sad but true
Emacs and/or vi, depending on your inclination, have text editors covered already, of course ;-)
so linux is already there