ChatGPT-5 Is Coming—Get the Definitive 60 Page Guide to Prompting in the Reasoning Era

I've been writing about prompting for over a year now, building sophisticated systems that push the boundaries of what's possible with AI.

In "Nate's Secret Sauce", I showed you how to iterate across multiple models—19 prompts, hours of refinement, fighting for every inch of truth. In "My Prompt Stack for Work", I laid out 16 architected prompts that transform daily workflows. In "Stop Letting AI Guess", I introduced contract-first prompting—bilateral agreement at 95% confidence before execution.

Here's what I discovered after talking with hundreds of CEOs, developers, and power users: I built these systems for reasoning models. They were designed for the frontier—for o3, for Claude Opus 4, for models that could actually think. But nearly everyone I talked to at actual workplaces was still using GPT-4o or CoPilot and wondering why even sophisticated prompts felt like pushing water uphill.

This is like driving a Formula 1 car with the parking brake on.

This is the first time I'm saying it explicitly: Switch your model. Stop using GPT-4o. The prompts I've been teaching—the ones you've been learning—they were never meant for last generation's AI. They were built for reasoning models that most people weren't using.

But here's what's even more important: For those already using reasoning models, we're not done. We're just getting started.

The techniques in this guide have never been gathered into one place like this—I constructed an entire prompting stack for reasoning models to get you ready for ChatGPT-5. It’s time now, because ChatGPT-5 is going to be the first reasoning model most people use regularly. I want you to be ready!

Looking for examples of what's in that 60 page guide? Try parallel divergence exploration where models maintain contradictory solutions before convergence. Semantic denoising through recursion that refines outputs without explicit critique. Or inverse complexity scaling that reveals how reasoning models fail catastrophically on simple tasks.

Sounds really complicated, but 1) I promise to explain it clearly so you get it, and 2) I give you actionable prompting takeaways for each. Yes, grasping these principles of reasoning models like ChatGPT-5 really does shift your prompting when you understand them!

One warning: these are specifically not going to work well for ChatGPT 4o users, and I’ve avoided publishing a prompting guide specifically for reasoning models before now precisely for that reason. Now that we are all on the verge of all having reasoning models, it’s time to go in-depth on these more advanced principles.

And the timing isn't accidental. ChatGPT-5 may be hours away (fingers crossed). When it launches it will become the default model for hundreds of millions of users. For the first time, the baseline might shift from pattern-matching to reasoning. People who've been stuck on GPT-4o might suddenly find themselves with a reasoning model without even choosing it.

This is a rare moment. A chance for a collective paradigm shift in how we prompt. But only if we're ready with techniques that actually leverage what reasoning models can do.

Look around the web. Find me another guide that goes this deep into prompting specifically for reasoning models. Not generic "prompt engineering tips" that work across any model. Not "10 ChatGPT hacks." I mean real, tested techniques for engaging in deep work with models that can genuinely reason.

The 60-page guide and this article represent something different. For those still on GPT-4o, it's your migration path to the frontier. For those already using reasoning models, there are new techniques that will separate you from everyone else who just switched models but didn't evolve their prompting. And for everyone, it's preparation for the GPT-5 world that is right around the corner.

The gap isn't just about using better models anymore. It's about understanding how to prompt them to get the most out of them. And these techniques scale with reasoning model capability—they'll be even more powerful with GPT-5 than they are with o3.

The goal with this guide is not just to move the baseline on prompting. We're expanding the frontier of prompting as we all get reasoner models. And I’m dropping this piece now because it feels as close as I can get to “right before ChatGPT-5” (if I’m lucky you’ll have it in your hands today).

Come have a look at how to prompt for the future…

PS. One more thing: if you’re a Gemini or Claude fan you’re not left out here! These principles scale with Opus 4 and Gemini 2.5 Pro, and I dive into both those models in the 60 page guide as well!