Even better, we can say that it's the actual hard prompt: this is real text written by real OpenAI employees. GPTs are well-known to easily quote verbatim from their context, and OpenAI trains theirs to do it by teaching them to break down word problems into pieces which are manipulated and regurgitated. This is clownshoes prompt engineering done by manager-first principles like "not knowing what we want" and "being able to quickly change the behavior of our products with millions of customers in unpredictable ways".