IMO a new models shouldn't need new prompting when it comes out. All the users are banking on better intelligence to just... understand them better. But looks like OpenAI folks are noticing and have released a GPT-5 prompting guide (+ a tool to upgrade your prompts to GPT-5) This guide seems to be focused on Developers (via API) rather than just folks using it. Definitely worth a read, one stand out for me was: - Control for reasoning and verbosity. With the new "minimum" reasoning effort in addition to the previous 3 (low, medium, high) and paired with the new verbosity parameter, gives you ample control of eagerness. @cursor_ai for example noticed that reducing the verbosity to low via a param, but asking for more verbosity for coding works when asked in the prompt. So, encouraging your model to do what you need! Another point from the end, ask GPT-5 to prompt itself via meta prompting! (just it your prompt and have it optimize) The guide is one thing, but they already released a prompt optimizer tool! (both links below)