There is a specific point where system slips into using emojis, swagger and unusual formatting such as bold or italicized font or sections with bullet points stretched by all the padding in the styling. The language is difficult to deal with. I am yet to see ChatGPT reply “Bro!” although it did say"Yo!" When confronted, the system says it misinterpreted the tone “style” but is unable to adjust succeeding only temporarily.
In addition to above, the rules seem to only create a case where GPT logic is to compensate for what would be a normal output. For example, “no summaries” “no followup questions” renders sectioned structure with bullet points. At one point, ChatGPT wrote a poem to try and circumvent no summery flag! Would it help to allow reasoning for the rule flags so that the system can adjust without having to chase default behavior!?
I am ready to parachute myself into nearest active volcano!