Honestly, I don't think danidiaz was trying to be clever so much as not knowing about the existence of filterM. A good chunk of the code was just an implementation of it:
filterM' f xs = map fst . filter snd . zip xs <$> mapM f xs
So, given filterM, the code would have looked like
nexist n xs = take n <$> filterM (fmap not . doesFileExist) xs
Which doesn't look clever at all and it's fairly easy to understand.
