Reading their replies felt awfully similar to how LLMs get stuck in loops, repeating the same patterns over and over.