Three economists built an open workflow that reran the code behind 4,452 published economics papers. It flagged gaps between the printed results and the authors' own code in 3,460 of them. In 923 papers it also wrote an extension the original authors never did.