It's addressed early in the video that the rectangles will be composed of lines oriented parallel to the xy axis. You're right in that it is a big assumption, but if you think about it, it's implicit in the input format. "Sloped rectangles" embedded in a plane need to describe at least 3 vertices, so the two edges required to form a 90° corner (a vertex) are represented. The corner becomes the local origin, use pythagoras to get the edge lengths, multiply for area. For "non-sloped rectangles" you just need 2 opposing vertices to find the edges required to calculate area. This is because rectangles by definition are composed of 4 congruent interior angles of 90°, and 2 sets of 2 congruent edges. It'd occasionally useful to consider a square as a special case rectangle. If it were a rhombus, all 4 vertices are required to find area. So the question in my opinion to ask is not, "are there sloped rectangles?" but rather "is the input format stable or variable? Will all rectangles be described with two vertices?". From that you should be able to describe the system's capacity for rectangles to the interviewer, demonstrating an awareness of how a rule, in this case the geometric identity of a rectangle, is used to omit unnecessary data from a model without losing any information about the object the model describes.