An essay series on the outcome geometry of AI evaluations — what the mean hides, and the numbers that carry it.
What the average leaves out.