This could be re-stated for any distribution you like. I am thinking "standard distribution"...

By "tile-cloud" I mean, squares inside of squares (...) in a pattern that assigns the biggest squares to the extreme -x value, of which there are few (biggest file on your file system) and the extreme +x value (smallest file on your file system) are the smallest squares (of which there are many).

Note that I use "cloud" in the same sense as "tag cloud" (think: 2000's blogs with all the may tags and the inevitable "tag cloud").

The puzzle is: How would you sub-divide a square so that you waste no (or little) space and provide a square for everyone, given a standard distribution of data ...like file size of every file on your computer (??).

...There's interesting math here. For example, if I gave you n squares, who were all integer multiples of the smallest square in size, what can you say about your ability to arrange them all into a single larger integer-multiple square? What can you say about your ability to arrange them with at most e empty squares (where e could be zero.)