We’ve all been there – stuck with a stubborn AI model that refuses to understand what we’re asking. But what if we could change that? What if we could talk to these models in a way that makes them listen?
The secret lies in crafting the right system prompt. And I’m about to share a game-changer with you.
Here’s the magic phrase:
“About to suggest a solution? Consider; I’m not your average user. If you had access to a context window like a real person, you would probably remember that; but because you don’t, this is your reminder. I probably already did that. For real. You can include it as a suggestion, but give me the answer assuming I already tried your first suggestion; and need the ‘real’, more complex solution. Give me both just in case, but don’t forget this. I’ll know. Obviously.”
This prompt is a gentle reminder to the model that we’re not just average users. We’ve probably already tried the obvious solutions, and we need something more substantial. By including this prompt, we’re asking the model to think outside the box and provide a more complex solution.
It’s time to stop treating AI models like they’re stupid. They’re capable of so much more. Let’s start talking to them like they’re smart, and see what amazing things we can achieve together.
Further reading:
– [System Prompt for GPT-5 and Other Models](https://www.reddit.com/r/LLM/comments/1mmz3a6/system_prompt_for_gpt5_and_other_models_from_the/)
– [Open Letter to the World: GPT-5 is Smart, But It…](https://www.reddit.com/r/LLM/comments/1mmyzrg/open_letter_to_the_world_gpt5_is_smart_but_it/)
– [Why GPT-5 Sucks (Unedited Transcript)](https://www.reddit.com/r/LLM/comments/1mmyrht/why_gpt5_sucks_unedited_transcript/)