I'm not a haskell expert (obviously), but the time it takes me to parse things like this is time I would rather spend reading 5-10x the number of lines and getting the meaning right away.
let loeb x = fmap ($ loeb x) x in
loeb [ (!!5), const 3, liftM2 (+) (!!0) (!!1), (*2) . (!!2), length, const 17]
(btw, I have no idea what the hell this does. Something to do with spreadsheets, apparently. I found it on http://www.haskell.org/haskellwiki/Blow_your_mind, which has enough cleverness to make me want to cry)
I've seen cleaner and more readable code in production haskell, but this sort of thing happens enough that I'm very cautious.
I think the first paragraph of the page linked explains why that code is so impenetrable. You will not have to read or write Haskell like that, ever. Good find though! On a similar note, the Haskell community is amazing. You can learn absolutely everything between the Haskell wiki, freenode irc and hackage. How great is it that EVERY lib/package/framework is documented on hackage in exactly the same format? Very great. Coming back to JavaScript is a bummer :(
I don't think you'll ever find this kind of code in production.
With years of Haskell experience, I rarely encounter code that is hard for me to read. This is a good counter-example, and is not typical code.
I read 10 lines of Haskell code roughly as fast as I read 10 lines of Python code -- yet the 10 lines of Haskell code can pack much more useful information.
So Haskell is a great tool for more efficient communication between programmers, who can write shorter messages to each other to convey the same information.
I've seen cleaner and more readable code in production haskell, but this sort of thing happens enough that I'm very cautious.