The Problem Discovery Engine is built on the core insight that people don’t struggle with "ideas," they struggle with finding "real problems" worth solving.
Most idea generators provide generic, AI-invented concepts. This application reverses that process by identifying actual human complaints from online forums (currently Reddit), clustering them with AI, and suggesting valid, data-driven startup opportunities.
When you enter a keyword (e.g., "SaaS" or "Postgres"), the engine searches Reddit for posts containing markers of genuine frustration like "I hate it when," "Why is it so painful," or "Is there any tool for."
It feeds the raw Reddit text into Gemini 2.5 Flash. The AI performs a "market research" role—identifying repeating themes across disparate posts and grouping them into 5 distinct "Problem Clusters."
For every cluster, the engine calculates:
- Frequency: An estimate of the number of users affected.
- Severity: A 1-10 rating of how painful the specific problem is.
- User Pain Summary: A clear, concise breakdown of exactly what is frustrating people.
To help you move from "problem" to "prototype," the engine provides:
- Solution Idea: A specific product or feature to solve the cluster.
- Tech Stack: Recommended modern tools (React, Node.js, MongoDB, etc.) to build the MVP.
- Frontend: React (Vite) + Tailwind CSS v4 + Framer Motion
- Backend: Node.js + Express (Modular ES Module Architecture)
- AI: Google Generative AI (Gemini 2.5/2.0/Pro)
- Scraper: Reddit Search API (via public JSON endpoints)
- Environment: Add your
GEMINI_API_KEYtoserver/.env. - Backend:
cd server && npm run dev - Frontend:
cd client && npm run dev
“Stop building what nobody needs. Start building what people are already complaining about.” 🚀