I’ve spent countless hours trying to perfect my prompts, only for the AI to get sidetracked or stop to ask for permission on something obvious. It’s super frustrating! Well, it looks like that’s about to change in a big way.
A new guide, supposedly based on official OpenAI docs, shows that GPT-5 uses a whole new system: “structured tags.” It’s like building a mini-configuration file directly into your prompt to control exactly how the AI thinks and acts. This is awesome!
⚙️ Here are the Core Tags:
🔬 <context_gathering>: This controls how the AI does its research. You can set it to be fast and efficient for quick tasks or tell it to do a comprehensive deep dive for complex analysis.
🚀 <persistence>: My absolute favorite. This tag sets the AI’s autonomy level. You can tell it to work completely on its own, making reasonable assumptions and only stopping when the job is 100% done. No more hand-holding!
✍️ <tool_preambles>: This shapes the AI’s communication style. You can ask for detailed, step-by-step progress reports or just a quick summary when it’s finished.
✨ Create Your Own Tags!
The real power is that you can define your own custom tags to enforce specific rules for any task. You can literally build a specialist agent on the fly.
📜 <code_quality_standards>: Make sure the AI follows your team’s exact coding style.
⚖️ <communication_style>: Define the precise tone you want, from formal reports to friendly chats.
This new method feels way more direct and powerful. The original post is packed with full, working examples for creating an autonomous coder, a research agent, and more.
Check out the full Reddit post for all the detailed code snippets and pro tips!
How to talk to GPt-5 (Based on OpenAI’s official GPT-5 Prompting Guide)
byu/carlosmpr in