social networks will rapidly turn into a potemkin dialog where everyone participating is using basically the same LLM. where people are still actually writing, will be the only legit material to train LLMs on and the feedback loop will escalate until the text is unreadable garbage.