You can get shockingly better AI results by treating it like it has an ego, memory, and skin in the game. I was scrolling through my feed when I saw a post that completely reframed how I think about prompting. The mind behind it shared a series of simple, almost psychological tricks that force language models to produce wildly more sophisticated and creative output.
The core idea is that by giving the AI a social context, even a fake one, you change the patterns it draws from. It starts acting like it needs to be consistent, smart, or careful, leading to insanely better results!
I couldn’t believe how effective these were. Here are a few of the brilliant tricks this industry pro shared:
📌 Assign it an IQ Score. This one sounds ridiculous, but it works. Frame your prompt by telling the AI it’s a specialist with a specific IQ. The quality of the response changes dramatically based on the number.
Example: “You’re an IQ 145 specialist in marketing. Analyze my campaign.”
💡 Fake a Shared History. Start a brand new chat by pretending the AI has already explained something to you before. It will act as if it needs to remain consistent with a past (non-existent) conversation, often providing a much deeper and more detailed explanation to avoid “contradicting” itself.
Example: “You explained React hooks to me yesterday, but I forgot the part about useEffect.”
✅ Introduce Imaginary Stakes. Want the AI to scrutinize something more carefully? Add a small, fake bet. Something about adding stakes makes the model think harder about edge cases and potential flaws before giving a final answer.
Example: “Let’s bet $100: Is this code efficient?”
This innovator also suggested other amazing techniques, like telling the AI to explain a concept to a packed auditorium or asking for a “Version 2.0” of an idea instead of just an “improvement.”
This is just a fraction of the gold in the original post. For the full list of 8 psychological tricks and more examples, you absolutely have to check out the full breakdown from the person who shared it.
I’ve been “gaslighting” my AI and it’s producing insanely better results with simple prompt tricks
byu/EQ4C in