The messing stuff up at scale is more because LLMs aren't capable, and people treat them as though they are.