bblackwood

I ran into a really weird case at work yesterday. We've built an LLM-based tool that generates a suggested next message in a conversation between a support representative and a customer. The prompt is pretty extensive,…