Has anyone given the same coding task to the same agent/model two days in a row and then diff-ed the resuts? (I have not, so just wondering).