Immediately after the Jacob/Anthropic tweet thread went viral, when it was at around 2m impressions, I screen recorded the first subtweeters as the diffusion/ propagation of the visit was obviously sync’d. I was certain, and said so at the time that the tweets were very highly suspect and bs. To do some light forensics, these days, as X cut off a lot of the tools/API I knew I had to catch the data in real time. I have , thru a shop I’m an investor in, the full x firehose, but whatever. Anyhoo, with this stuff it’s called ‘Graph Neural Networks’ = GNN It’s a graph of nodes that operate on graphs where nodes are connected by edges. A. Node (or a vertice) is a a network of users, the subtweeters B. edges are the relationships/interactions between them This is ‘neighborhood aggregation’ What you’re looking for, in the way I simplistically did it sans tools, is to look thru thr assorted influencers and subtweeters and see if their engagement is abnormal / too high And then also look for those accounts use of bots and inauthentic behavior Anyhoo