This entire thread is questioning why OpenAI themselves use repetitive negatives for various behaviors like “not outputting JSON”.

There is no magic prompting sauce and affirmative prompting is not a panacea.