Sometimes in chat, the AI returns examples of how to use its suggestions. Sometimes, when the response is finalized, these end up with code-y looking bits of letters & symbols--html stuff. Not in the main response, but in the grey boxes that have examples of how to deploy the suggestion.
When the chat is still mid-stream in returning the response, it looks fine. It's only once it's finalized that it turns wonky.
This is a custom prompt, copied from the GP prompt. Is there I might have messed up in the prompt, for this to happen? The prompt does include the phrase "Always write your answer in Markdown format, don't use any HTML or XML tags."
I was using gpt-4o this time, but I think it's happened with other models.
Image attached. fwiw...this isn't a big deal. I can easily cut those wonky buts out if I use this (or just type it myself!) Was just wondering if there's something I'm doing wrong to create this.