When did we decide your code was cooler relative to someone else's when you end up typing fewer characters per line for the same functionality? If anything, the added density per line makes reading code (yours or, worse, someone else's) more difficult.
I nticd ur nt tping lk ths. Sentences are much easier to understand if you use words, even if the abbreviated forms can be parsed out to the same meaning.
If your aim is to save yourself a couple characters per line, fine. But it does come at a cost.
The interesting thing for me is that the K programs are fewer lines...a lot fewer. 9 in K vs. 96 in Perl in one example. When one remembers that perl is historically the king of one-liners, this is pretty impressive. I'd probably be willing to learn how to interpret line noise if I only have to parse 9 lines for every 96 I'm parsing now (I'm primarily a Perl programmer). I, obviously, don't believe this is representative--a large number of programming tasks aren't reduce-able beyond a certain level. But for a 50% LoC reduction, I might consider it.
There's a lot of other considerations, and right now Haskell and Erlang are leading my "what'll my next back end projects use" list (after Perl 6...which I'm getting excited about after seeing Audrey Tang's talk about it from a few days ago, as it appears it will fix every single major problem with Perl, by my reckoning)...but I'm not going to rule out a language merely because my first glance at it makes me uncomfortable or confused (if that were the case, I would have never used Perl, Python, Ruby, and I would be avoiding messing with Erlang and Haskell...I might have even continued trying to figure out Java, since it seemed comfortably familiar based on my C experience, but I believe that would have been a very bad instinct).
"The individual tokens should be short as well. Perl and Common Lisp occupy opposite poles on this question. Perl programs can be almost cryptically dense, while the names of built-in Common Lisp operators are comically long. The designers of Common Lisp probably expected users to have text editors that would type these long names for them. But the cost of a long name is not just the cost of typing it. There is also the cost of reading it, and the cost of the space it takes up on your screen."
I'd guess that if you want to be better at Arc than all the Arc-philes who know Lisp, K would be worth studying.
Why would I want to be better at Arc than Lispniks? I'm a Haskeller. We're better than them anyways. ;-)
Seriously, this is one case where I think PG is wrong. Typing isn't the main productivity burden on programmers, thinking is. Any time you make someone decipher an identifier, you're slowing them down. Unpronounceable functions are particularly bad: I dunno about others, but I always think "F-N" when I see "fn", while I think "def" when I see "def".
I think Python and Scheme have it basically right, Haskell is okay but has an unfortunate tendency towards obscure vocabulary and cryptic conventions (nub? foldl? mapM_?), and Common Lisp and Java are way too verbose.
Maybe I'm missing something.