The ‘got ours, now let’s fix this mess’ of ‘Poison Fountain’ seems rather performative. HTTP get twice and check the results are the same? Further there’s no indication of how the ‘poison’ data is generated. There were commits to Git for Grok some time back showing they were already having to filter data from social media that was excessively aligned with X’s userbase and discursive style from its own search ingress; this will be no different.via