-
-
No log-in, welcome page
Inspiration
Seeing older relatives fall for dangerous WhatsApp forwards and fake medical advice. I wanted to build a low-friction tool that doesn't require users to be tech-savvy to verify claims.
What it does
Rumor-Stop is a mobile-optimized web app where users can paste forwarded messages. It uses NLP heuristics to evaluate claims and outputs a strict contract: a Traffic Light Verdict Badge (🔴 RED / 🟡 YELLOW / 🟢 GREEN), a plain-English 2-sentence summary, and a verified citation.
How we built it
We used the Devpost Learn Skill Pack's "Plan-First" methodology to generate our scope.md, prd.md, and spec.md before coding. The app is built with Python and Streamlit. For the AI logic, it uses Groq as the primary engine with a fallback routing to Google Gemini to guarantee 100% uptime.
Challenges we ran into
We hit a 404 error mid-build when our primary free-tier LLM was deprecated. Thanks to our planned architecture, we easily swapped models. We also had to refine our prompt engineering to prevent "Prompt Overfitting" (where the AI demanded medical evidence for non-medical claims).
What we learned
The absolute value of planning before coding. Writing the planning docs first prevented scope creep and helped us confidently push "Voice Note Input" to our v2 roadmap instead of breaking our MVP.
Log in or sign up for Devpost to join the conversation.