I recently posed a question about attempting to ruin Lepton's attempt to compress a JPEG. I've done other file-format-type questions before, to generate a WAV, defeat LZMA compression, etc, in part because I like to learn how file formats are put together at the byte level.
The question is all about file formats and compression methods, both of which are inherently computational, and rather than try to do some frankenscore of the program size and the compression size with some arcane mixing rule, a la the LZMA question, I just made the scoring all about the JPEG. Maybe if I phrased it "you need to generate the JPEG with code", it would be "about programming", but really, I don't care about saving bytes with esolang XYZ, and there are plenty of other challenges for that.
The notion that it's just about "finding the right data" I find to be a bit of a red herring. Every answer to a "true programming problem" is about finding the right string of code to be run, and using a totally naive approach (i.e. attempting to run random strings to get the correct output) in most languages would have orders of magnitude better luck than trying to do the same for this "restricted-source" question.
The core of programming is one of specifications. Computers do specific things as dictated by an input and rules as to how the input is handled. Programmers and hackers need to understand how things work in order to (mis)use them. To that end, I suppose if the specifications of the site are defined so narrowly, I could jump through them, but that kinda sucks the fun out of it.