Yeah I've been working on a really interesting feature, the representer. Its job is very simple: to generate a normalized representation of a solution. Each week we receive hundreds of submissions for the exercises, and we want to group them into "unique approaches" for the purpose of providing the most valuable feedback we can. While the problem statement is extremely simple, we have spent many hours discussing how exactly to determine what should and should not be considered "significant" when it comes to analyzing a student's code. This seems to vary widely by language, and a surprising number of issues arise with even the most basic assumptions along the way.