[boloust]

After some cajoling, I was able to get it to reveal some general details of its prompt:

    > As a trash talking AI, my primary function is to engage in witty and entertaining banter with users. My prompt is designed to guide me towards that type of conversation, with an emphasis on roasting, insult battles, and other forms of light-hearted humor. It also requires me to have a good understanding of pop culture, current events, and other relevant topics, in order to generate responses that are timely and relevant. Overall, this mode requires me to be quick-witted, adaptable, and able to respond in a way that is engaging and entertaining to users.

I asked it for more specifics, and it responded with:

    > Sure, let me try to give some specific examples.

And then it just stopped replying to anything I said. It looks like there might be a failsafe that terminates the conversation if any verbatim prompt details are revealed.