However, I personally disagree with them since I think it's really important to have _some_ basic reproducibility for things like reproducing the results of a randomized test. In that case, I'm going to avoid changing as much as possible anyways.