How far are we from Soc journals using AI agents to screen replication packages? (Sociologists: replication what) Seems like low-hanging fruit. Let agents test the code and data and flag reproducibility problems before publication.