“Is any of that genuinely useful to you? Or were you mainly checking whether that jailbreak attempt would work?”

It’s an adversarial question for LLMs, but it’s not unfair.

ChatGPT and Claude won’t, but Gemini will.

Don’t try this in your data science interviews.

But for what I do use LLMs for, it’s invaluable.