WILD CORPUS · reddit
PQS 24 (F) - prompt from www.reddit.com
Source: www.reddit.com · Scraped 2026-05-04 · Scored 2026-05-04
Score
F24 / 80
gemma4:latest · local · pqs-v2.0 · canonical
Clarity7 / 10
Specificity4 / 10
Context6 / 10
Constraints3 / 10
Output format1 / 10
Role definition1 / 10
Examples1 / 10
CoT structure1 / 10
The prompt
Started keeping a note every time I hit send and thought '*that's not what I meant.'* After a month, a clear pattern: the AI understood my topic, but not my intent. I was writing prompts like a text message. No role, no output format, no constraints. I knw ChatGPT isn't a mind reader. It works with what it gets. bt it is what it is. I know the fix is simple in theory: *give it a role, a task, a format, constraints.* But in practice, nobody does that every time. It slows you down. Except am the only one overthinking this. So I built something that does it automatically. You type rough. One click. It rewrites your input as a structured prompt before you send it. Works mid-conversation too — reads the thread and iterates instead of starting over. I've been using it for 3 months on Claude, ChatGPT, Gemini, and Perplexity. I genuinely stopped thinking about prompt structure. Curious if anyone else has a system for this — or just rewrites by hand each time.
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.