Startup IdeasDB

The Problem

Founders spend weeks manually scouring Reddit threads, Product Hunt launches, and Hacker News comments trying to identify validated problems worth solving. Most never move past ideation because the research phase is exhausting and unstructured.

Existing idea databases are either shallow listicles or community-driven forums with no quality filter. There was no research-grade platform that aggregated, validated, and enriched problem statements at scale, with technical depth a builder could actually act on.

Constraints

• Scraper had to survive API and layout changes across Reddit, Product Hunt, Hacker News, and Indie Hackers without manual fixes.

• Embedding 12,000+ items via OpenAI required careful batching to stay within rate limits and manage cost.

• Whitepaper generation had to be fast enough for on-demand use, multi-step LLM chains needed caching at each stage.

• Three payment providers (Razorpay, DodoPayments, PayPal) all had to update the same entitlement state reliably via webhooks.

• Search results had to feel research-grade, not SEO-stuffed, quality filtering on ingested content was a continuous problem.

How Nexolve Built It

StartupIdeasDB is a TypeScript monorepo, a public marketplace and an admin portal sharing components and types. Automated scrapers feed an AI validation pipeline, pgvector powers semantic search, and a multi-step LLM chain generates on-demand technical whitepapers.

1. 1

   pnpm monorepo architecture

   Public marketplace and admin portal share a TypeScript types package and Shadcn UI components via pnpm workspaces. Zod-validated API types ensure end-to-end type safety across both apps.

2. 2

   Automated scraper pipeline

   A Python scraper harvests Reddit, Product Hunt, Hacker News, and Indie Hackers daily. Raw posts are deduped, cleaned, and classified into problem-statement format before being fed into the embedding pipeline.

3. 3

   Semantic search with pgvector

   All 12,000+ problem statements are embedded via OpenAI text-embedding-3-large and stored in Supabase with pgvector. Founders search in natural language and get sub-50ms ranked results by cosine similarity.

4. 4

   AI whitepaper generation

   BuilderAI orchestrates multi-step LLM chains to produce 2,000-word technical whitepapers on demand, covering market sizing, competitor analysis, tech stack recommendations, and go-to-market strategy for any selected idea.

5. 5

   Multi-currency payments

   Razorpay handles INR, DodoPayments covers USD/EUR crypto-rails, and PayPal serves global buyers. All three providers update entitlement state via Supabase webhooks in real-time.

Tech Stack

Outcomes

• 12,000+ problem statements indexed and searchable via natural language, sub-50ms query latency.

• $13,852 in idea pack revenue from paying founders across 3 payment providers.

• 900+ registered users within the first months of launch.

• Automated daily scraper pipeline keeps the database growing without manual curation.

• On-demand 2,000-word whitepapers generated per idea, covering market sizing, competitors, and tech stack.