Comment by xpl

LLMs feed on their own output during inference, so there is a loop on itself, isn't it obvious?