The larger problem with software driven by machine learning and related things is that the behavior of the software is no longer predictable. You can see this in a lot of places even before LLMs (autocorrect, speech commands, etc). Software is generally easier to use when it is more predictable, so this trend away from behavior that is consistent is really annoying.
The larger LLM trend is the latest in a 30 to 40 year history of idiot tech managers getting FOMO over the next big thing even before the next big thing has actually been proven to be useful. The fact that it involves hugely inefficient use of computing resources at a time when cutting back on resource utiltization is important and the fact that except for certain interesting cases in software development (at least for me) it's been mostly useless is pretty upsetting. And that's even before we get to the fact that it's based on stolen IP.
This weekend the Pittsburgh Symphony played the Mahler 2.
I went twice. It was great.
The companion piece, by Boris Pigovat, called Yizkor, was also great both times, and a fitting prelude to the Mahler.
More on that here: https://pigovat.com/wordpress/yizkor/
I have a cabinet where I have put most of the single and double CDs that I've ripped in very thin plastic sleeves. As I rip them I give them a number and put them at the end of the drawer. So the drawer is in "when I ripped it" order. I actually spent the last year or so ripping and filing because I had done some, but not most, of what I owned before and then sort of gave up. Now pretty much everything is in that one cabinet except...
I have various boxed sets (anything with 4+ disks) scattered around the house. But I isolate the ones I have not ripped yet in one cabinet so I can usually find them.
I keep track of the serial numbers in iTunes/Apple Music, which stores all the ripped files.
Now ... LPs are a different story.