Growth Studio

Growth Studio watches what is trending across seven tech communities, ranks it for a team, and turns any item into publish-ready content in that team's own brand voice.

The problem#

Content teams working in tech have the opposite of an information shortage. HackerNews, Dev.to, GitHub Trending, Lobsters, Mastodon, Product Hunt and Stack Overflow all surface what people are talking about, but each is a separate tab, each ranks by its own logic, and the same story appears in four of them under four different headlines. Deciding what is actually worth writing about means reading all of it, every day — and the work of turning that decision into a post only starts afterwards.

The approach#

The platform collects from all seven sources on a schedule, then does the part that makes the collection useful: deduplicating across sources so one story is one entry, and ranking what remains by velocity and by relevance to the team's own brand rather than by raw popularity.

From there any item can be expanded into platform-specific drafts — a Facebook post, a reel script, a LinkedIn article, a YouTube idea, a caption — written in a configured brand voice rather than a generic one. A chat interface sits over the same trend pool, so the collection can be interrogated conversationally instead of only browsed.

The AI layer is deliberately provider-agnostic: OpenRouter, Anthropic and OpenAI sit behind one interface, so the model can change without the application changing. Keys are supplied in-app and stored AES-256-GCM encrypted, rather than being baked into the deployment environment.

Architecture#

A React 19 single-page frontend built with Vite, using TanStack Query for server state and Radix UI primitives for the interface. The backend is Express 5 on Node 22 with Mongoose over MongoDB Atlas, using Atlas Vector Search for semantic retrieval across the trend pool, Redis for caching and node-cron for the ingestion schedule. Requests are validated with Zod and logged with pino.

The frontend deploys to Vercel; the backend runs in Docker via Dokploy. The API is documented with Swagger at /api/v1/docs.

What it does#

• Scheduled ingestion from seven community sources, deduplicated into one pool
• Ranking by velocity and configured brand relevance
• Generation of Facebook, LinkedIn and YouTube drafts in a defined brand voice
• Conversational querying of the trend pool
• Vector search over collected trends via MongoDB Atlas
• JWT authentication, encrypted provider credentials, and a documented REST API