but it seems that even when we re-run older experiments it's like it's lost a lot of skill in doing it