@textfiles@mastodon.archive.org
Well, that does answer that. :-) Of course I can't know how many people used the feature, just know I did. :-)
I will agree there were many "hallucinations" and honestly some of the categories were a bit hilarious (if it was a picture of a person, there might be a tag for "arm" or "torso").
Discoverability (an interesting concept in itself) is (for me at least, I assume others) a problem in large data sets such as this (and the Archive as a whole). Obviously, this is one reason people like algorithmic feeds, it brings up "new, related stuff automatically." Which is a great thing is the feed is honest in some way. Yes, I can search for a specific thing and browse, but sometimes just don't know where to start.
The Public Domain Review has a really sweet random image stream
https://pdimagearchive.org which I find lovely to just let wash over me and occasionally click on some images to explore.