Comment by throw-the-towel

But then you hit the same problem LLM slop exploits: the cost of refuting bullshit becomes way higher than the cost of generating it.