The Case for Open-Book Testing