I built a "BS Detector" for supplements in 48 hours using Cursor + OpenAI (Roast my logic)
I got tired of buying "premium" supplements that were basically just expensive flavored water.
The industry runs on consumer confusion. They hide 10mg of "Pixie Dust" collagen behind a "PROPRIETARY BEAUTY BLEND" label and charge $50 for it. Existing scanner apps are useless because they rely on static databases. If a product is new or niche, they fail.
I wanted a cynical pharmacist in my pocket. Someone who checks the math, not just the ingredients.
So I built NutriDetector
Stack: I am a solo founder, so speed was everything. I used the Cursor + OpenAI stack to build this fast. Frontend: WordPress (Yes, really. It works.) Logic: OpenAI API.
The "Secret Sauce": A massive, heavily-tuned System Prompt that acts as a strict auditor.
The Hard Part (Where I need feedback):
The biggest challenge wasn't the code, it was the prompt engineering.
Early versions were too nice. They would see "Vitamin C" and give the product a Green Flag, ignoring the fact that it also contained Titanium Dioxide and Red #40.
I had to explicitly instruct the LLM to handle Battle Mode:
Trash vs Trash: If I compare two bad products, it doesn't just say "both suck." It forces a winner based on the "lesser of two evils."
Micro-dosing Detection: It calculates if the dosage is clinically effective. 20mg of Collagen? Flagged as a scam immediately.
What I need from you:
I’m looking for edge cases to break my logic: 1) paste a supplement label that you know is controversial 2) try “battle mode” ( compare mode) 3) tell me if the AI is too harsh or too lenient.🙏
Its free to use (no login required)
