Stuart_Armstrong comments on Why we should err in both directions