WILD CORPUS · reddit
PQS 23 (F) - prompt from www.reddit.com
Source: www.reddit.com · Scraped 2026-05-04 · Scored 2026-05-04
Score
F23 / 80
gemma4:latest · local · pqs-v2.0 · canonical
Clarity7 / 10
Specificity4 / 10
Context6 / 10
Constraints2 / 10
Output format1 / 10
Role definition1 / 10
Examples1 / 10
CoT structure1 / 10
The prompt
not "act as an expert." everyone does that. stopped working the moment everyone started using it. this instead: *"you've seen a thousand people fail at this exact problem. tell me where they fail before you help me."* what came back wasn't the generic answer. it was the failure map. where people go wrong that nobody admits. worth more than any solution it could give directly. **the vantage points that actually work:** *"you've reviewed a thousand versions of this. what separates the top one percent."* stops giving average advice. starts giving edge. *"you've watched people spend months on this and get nowhere. what were they doing wrong that they couldn't see."* the blind spot answer. the thing you're probably doing right now. *"you built this from scratch and it failed. what did you miss."* post mortem energy without the actual failure. *"you tried the obvious solution. it didn't work. what did you try next."* skips the first layer. goes straight to the interesting part. the difference between role prompting and vantage point: "act as an expert" gives credentials. a vantage point gives a relationship to the problem. an expert knows the answer. someone who watched a thousand people fail knows where the answer breaks in practice. completely different kind of useful. what question have you been asking the same way for months that a different vantage point would break open?
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.