for me the takeaway here isn't that sorting should use some smarter/better mechanism for inferring semantic intent from filenames, but rather that sorting should not try to infer semantic intent from filenames in the first place