I think it would be interesting if more programs were written with an aesthetic or artistic goal in mind, rather than functional. Like network code that has feelings, or fonts that get blurry the longer you look at them, or programs that, instead of crashing, simply find something else to do. Maybe a program written so the source is a poem about lost love, and results in bugs that randomly corrupts your memory on your lover's birthday.
You could write such a paper
Pearls in our code
Awk generates solutions to certain problems involving record and field delimited text whose succinctness cannot be approached by common scripting languages.
Cut provides massive bang for buck , just learning "-f -d" and "-c"
I recommend finding people with letter-related synesthesia and showing them your friend's poetry!
One of the earliest triumphs (from the scholarly point of view, although it's also super-scary for privacy) is the success of stylometry for authorship attribution of unknown books, letters, and manuscripts.
I had a high school teacher who (as an English grad student) used a concordance to make an argument about Mark Twain's vocabulary, showing either when a particular work was likely written or when Twain likely read some other work (I forgot which). But that kind of thing and even more impressive inferences would be pretty routine today using computer tools.
We have lots of systems which make software modular and composable, but they are not for the end user. At best they end up being build-blocks for programmers who make gigantic black boxes. More often, programmers of gigantic black boxes struggle with tools and libraries that are themselves gigantic black boxes.