WILD CORPUS · reddit
PQS 30 (F) - prompt from www.reddit.com
Source: www.reddit.com · Scraped 2026-05-04 · Scored 2026-05-04
Score
F30 / 80
gemma4:latest · local · pqs-v2.0 · canonical
Clarity8 / 10
Specificity4 / 10
Context7 / 10
Constraints3 / 10
Output format2 / 10
Role definition4 / 10
Examples1 / 10
CoT structure1 / 10
The prompt
Something I’ve noticed a lot when people try to write "expert-level" prompts: * Saying “you surpass any human expert” doesn’t actually make the AI smarter. If anything, it sets up false expectations and makes the model more likely to confidently BS you. * Real experts show their thinking, limitations, uncertainty, nuance, tradeoffs. The best prompts should encourage that, not gloss over it. * Vague success criteria like “give me highly strategic, top 0.1% insights” aren’t helpful unless you define what that *actually* means in your context. * You’re not telling the model what kind of output you want; should it give you a breakdown, a recommendation, a plan, or a debate? That matters. Here’s a stronger approach: You are an expert consultant in [TOPIC], with up-to-date knowledge of current research, best practices, and emerging trends. Give a well-reasoned, multi-perspective analysis. Be clear about assumptions, uncertainties, and where expert opinions may differ. This kind of structure gets you way better responses that are more grounded, useful, and realistic. I actually built a tool called [TeachMeToPrompt](http://teachmetoprompt.com) that helps you write prompts like this. It helps you position the AI correctly, define your goals, and set concrete expectations for what a “good” answer looks like. Super helpful if you’re tired of vague, fluffy outputs. Let me know what you think. I made it beginner-friendly too.
This prompt was scraped from a public source. The score reflects the input as written, not the quality of any output it produced. The AI input quality problem is the gap between what people type and what the model can act on.