Reading OpenAI's own docs beats secondhand Astra prompting guides
The author says most prompting courses for a new model are written before the documentation is out, while Astra's banned word list, testing-thoroughness quirk and mid-conversation effort switching are all published by OpenAI.
most "prompting masterclass" content for a brand new model is written before anyone's actually read the model's own documentation, and it shows. astra's real behavior, the banned word list, the testing-thoroughness quirk, mid-conversation effort switching, is all published directly by openai. reading that primary source carefully turns out to beat almost every secondhand guide circulating right now, including the ones with more views. full masterclass below.
This post quotes another post: x.com/1373665408193036293/status/210039675107295…