
I think at least one of test cases for this challenge should be longer - more characters overall, more words, more characters per word - or the challenge should be more explicit about the scale of the input that should be supported.
If the scale is not explicit, then a passing program ought to work on arbitrarily long input, so the test cases should also check that. Otherwise, there's a lot of scope here for something similar, e.g. using the one-line fibonacci looping constructs (@@@@@) to concisely loop but with a specific maximum number of iterations.
I'm enjoying this one, for some reason - I keep ending up cramming everything into a small space, but finding myself with just one character that I can't fit in!
I do have some code in my latest one which reads backwards later in the program, which was nice.
