These four mental models will up-level your perception of AI. Use a Universal Prompt Pattern and more tips to improve prompting today.

Why Your Prompts Might Not Work

Images created with Gemini

Images created with Gemini

I believe the biggest reason there's confusion and ambiguity around AI is because our emotions and previous experience make us use outdated mental models to describe what AI really "is" and what it can do.

Factor #1 - LLM’s have a wide range of capabilities

Some see AI as a superhero that can do anything faster than a human. Others see it as a broken machine that occasionally produces decent results. The reality is in between, where you get more value by writing better prompts that maximize its strengths while limiting its weaknesses.

Factor #2 - Fear

Right now, the world (social media, news, company messaging) tells us AI will take our jobs and these feelings are too valid to easily dismiss. These fears trap us in outdated mental models, killing motivation and creating a self-fulfilling prophecy: weak prompts produce generic results, which decrease motivation and confirm our fault assumptions.

To break this cycle, I've distilled years of research, personal experience, and dozens of interviews into four mental models and practical tips for better AI results today.

Mental Model #1: AI is an Unreliable Assistant

image.png

How to Self-Identify: You approach AI with low expectations, treating it like a junior assistant/apprentice you have to micro-manage to get the result you want. You prompt it like a human and need 5-10+ replies to get generic results that still needs a significant manual refinement.

The Limitation: This begins the self-fulfilling prophecy where minimal effort leads to mediocre results, reinforcing the belief that "AI isn't useful" and only worthy of continued minimal effort.

When this Makes Sense: Quick, low-stakes tasks where you need a rough draft and don't care about quality.

How to Grow Beyond: Treat every prompt as your custom programming function. Use the universal prompt pattern and iteratively learn which words/context/constraints work best for your specific needs.

Example Prompts: