A terminal for reading between the sources.
PRISM is a media-literacy tool. It doesn’t tell you what to think — it shows you how the same event is being told by different newsrooms, and lets you decide what matters.
We continuously pull articles from major publishers — Reuters, BBC, Bloomberg, CNBC, Guardian, NPR, Fox News, Al Jazeera, The Hindu and more — covering the same real-world events.
An embedding model groups articles that describe the same event, no matter how differently they are titled or framed.
For every article we score sentiment (VADER), detect editorial framing (LLM zero-shot), extract named entities and identify shared facts.
A neutral cross-article summary is generated by Claude Sonnet from up to eight publisher accounts — verifiable facts only, no editorial stance.
“The point is not to distrust every source, but to notice which parts of a story each source chooses to amplify — and which parts they leave on the cutting room floor.”
- · Live RSS ingestion every 60 minutes
- · Side-by-side four-way article compare with diff highlighting
- · Publisher bias profiles (30-day sentiment trend, frame usage)
- · Publication timeline scatter (who broke the story first)
- · Trending entities sidebar and entity drill-down pages