Mhm, I just wonder how many samples they get and how much time they have to come to the conclusion.

Like a short example is easier to grade, but not in the same ballpark as a whole codebase.