The Newsline

Quite the homepage, if we do say so ourselves
Newsblend - dark mode!

Inspiration

We were frustrated by how hard it is to answer a simple question: “what actually happened?” without wading through clickbait, partisan spin, and conflicting headlines. We wanted a way to check out the story and the sources at the same time - not just “left vs right”, but how strongly each claim is supported, and who’s shaping the narrative.

What it does

The system scrapes a set of major outlets, clusters articles about the same topic, and uses an LLM to synthesize a single, modular, neutrally-worded article (our Newsblends). Each section of the Newsblend carries live metrics: how many sources back it, how aligned they are, political bias and credibility scores, and emotional tone. On hover, you see a dropdown of outlets and quotes that support that specific passage, turning the article into an interactive, explorable explanation of all the accounts. We also support community notes for factchecking and comment sections for story discussion (complete, of course, with reliability metrics for every involved member). Your opinions are your own.

How we built it

On the backend, Python scripts crawl multiple news sites, once a day, and collate articles discussing the same story. These are run through a local scoring pipeline (agreement, credibility, emotionality, political bias), plus a Gemini‑powered aggregator that outputs structured JSON “stories”. Each story discusses a certain theme or viewpoint, and is linked to the particular source passages which support it. Together, they form a Newsblend - allowing you to look into the reliability of every component of the narrative. An API layer serves these Story objects to the frontend. There, we used React + TypeScript + Vite + Tailwind to build a dashboard: topic search and cards, article detail views, support gauges, and hoverable citations maximise transparency.