Open Kitchen
How Briffly works
under the hood.
Briffly started as a side project to learn how recommendation systems, vector embeddings, and content aggregation actually work. It turned into something people wanted. Here's how the pieces fit together.
Content Ingestion
We use RSS - the internet's original open standard.
Publishers already expose content through RSS feeds. We subscribe to 200+ of them. When a new article drops, we fetch it through the feed - the same way a podcast app fetches episodes. We extract clean article text using Readability, the technology behind Firefox's Reader View. No scraping, no crawling. Just the public feed the publisher chose to share.
Understanding Content
Every article becomes a point in space.
Each article gets a vector embedding - a numerical fingerprint of what it's about. Think of it as converting meaning into coordinates. Articles about similar topics land close together: a React Server Components piece sits near other React articles, far from one about Kubernetes networking. We store and search these embeddings with pgvector in PostgreSQL. The same approach powers similarity search at Spotify and Netflix - we use it for news.
Your Feed
Your interests shape your feed - automatically.
During onboarding, your topic and publication choices create an embedding of your interests. As you read, like, and save articles, that profile evolves. Your feed is a vector similarity search: find the articles closest to your interest profile. Same math that powers "you might also like" in music and movie apps, applied to tech news. No black-box engagement algorithm. Just: the articles most relevant to what you care about.
Built by a developer.
For developers.
I built Briffly to understand recommendation systems - not in theory, but by shipping one. Vector databases, embedding models, content extraction, real-time scoring: every piece was something I wanted to learn by doing. Turns out other people wanted what I built for myself: a way to stay current without drowning in noise. Curious about the details? I'd love to chat.
- Swapnil, Creator of Briffly