Behind the grapevine: building a listening layer for the web
By Grapevine Team
Always listening is a promise that's easy to make and hard to keep. Behind it sits an ingestion pipeline pulling from 180,000+ sources across 40+ countries, plus six AI models — and a hard requirement that a breaking mention reaches you in minutes, not hours.
Ingest everything, store what matters
We pull news, social, podcasts and broadcast continuously, normalize wildly different formats into a single mention shape, and run each through enrichment: language detection, entity extraction, sentiment and the narrative score. Cheap to ingest, expensive to keep — so we're deliberate about what we retain.
Fast by default
The app is served from the edge, with realtime collaboration and streaming summaries so the product feels instant even when the work behind it isn't. Background jobs handle the heavy lifting; the interface never makes you wait for it.
Speed is a feature. A story you hear about tomorrow is a story you missed.
There's a lot more to say about each layer — and we will, in future posts. For now: this is the machinery that turns the whole web's chatter into one clear signal.
First to know. First to move.
Your first topic takes two minutes. The mentions start landing right away.
Keep reading
Introducing AI visibility — see how assistants describe your brand
Reputations are increasingly shaped by how ChatGPT, Gemini and Claude answer questions about you. Today we're launching AI visibility tracking across the six models that matter most.
Read articleFrom firehose to signal: how we rank what actually matters
Most monitoring tools measure volume. We built a narrative score that weighs reach, sentiment and momentum so the one story you need surfaces first.
Read articleBroadcast is back: real-time TV and radio monitoring
Podcasts and broadcast were the blind spot in modern comms. Here's how we transcribe and index the airwaves the moment your brand is mentioned.
Read article