As a guy with a terrible memory, I think of it the opposite way. For Q (and more so K), the actual number of distinct items to remember is very small, as basically anything complex is created by composition of the simple functions/variables. This way you don't need to remember nearly as many libraries/variables/function names, and also the function itself becomes its own description.

That being said, for more production-heavy applications and interfacing with other languages, variable/function naming could be more descriptive :)

Agreed. The small(ish) vocabulary of operators makes K easier for me to use.

Would you rather write/read:

DIVIDE X BY 5 GIVING Y

or

y=x/5;

The first is COBOL (designed to make code easier for "normal" people to read. The second is C (which looks more like the math that we learn in grade school).

k/APL/J simple moves further in the direction of math.

As a far as descriptive names go, in k, names can always be aliased with longer names (assuming no conflicts). I do this when exposing a library k code as a web interface.