So let’s say you have a graph stream generated by the triangle-free process. What if you mix in new graph data? Could this be an easy test case for anomaly detection? hm.