The future is unknown as developers start taking advice from machines rather than peers. How will we keep the LLMs honest?