The difference (claimed by the comment above) is in the words

without loss of quality

So you can convert back and forth without the photo copy of a photo copy problem.

And you don't have to store the second copy of the file except for caching of frequently fetched files which I'm sure will just be an nginx rule.

And let’s not forget HEIF and JPEG-2000.

What they’re saying is that a web server can create a traditional jpeg file from a jpeg xl to send to a client as needed.

Other way around, you can convert a "web safe" JPEG file into a JXL one (and back again), but you can't turn any random JXL file into a JPEG file.

But yeah, something like Lemmy could recompress uploaded JPEG images as JXL on the server, serving them at JXL to updated clients, and converting back to JPEG as needed, saving server storage and bandwidth with no quality loss.