NewsRank is a realtime news intelligence engine. It watches hundreds of RSS sources, strips out the duplication and noise, and surfaces the stories that actually matter — with entity extraction and AI summaries so you read the signal instead of scrolling the feed.
The problem
The news isn't hard to find; it's hard to rank. The same story breaks across fifty outlets within minutes, each with a slightly different headline. A raw feed buries the one update you care about under near-identical copies. NewsRank exists to collapse that noise into a single, ranked view.
How it works
- Continuous ingestion. A Go service polls hundreds of feeds and normalizes everything into a common shape as it arrives.
- Deduplication. Near-duplicate articles are clustered so a single event shows up once, not fifty times.
- Ranking. Stories are scored on signal — recency, source spread, and importance — so the feed leads with what matters.
- Entities & summaries. Named entities are extracted and each cluster gets an AI summary, so you grasp a story without opening it.
- Live updates. WebSockets push new and re-ranked stories to the client with sub-second latency.
Highlights
- Streaming pipeline across hundreds of sources
- Sub-second updates pushed straight to the browser
- Entity extraction + AI ranking on every story
Stack
A Go ingestion and ranking engine backed by Postgres, a SvelteKit frontend, and WebSockets for realtime delivery. Live at newsrank.ai.