State of Technology is built around a simple frustration: most news products show articles, but readers are trying to understand events. One AI model launch, chip shortage, antitrust filing, or platform change can appear as a wall of repeated headlines. The useful unit is the underlying story and the sources confirming it.
The X News Inspiration
This project is partly inspired by the useful part of X's news experience: fast public conversation can make breaking stories legible before traditional homepages settle. The goal is not to recreate X's private ranking system. SoT has no access to it and does not claim to model it.
The narrower attempt is to capture a few visible behaviors in an open, inspectable way: group related coverage, expose source corroboration, detect momentum, and keep the reader close to the original reporting.
Ranking Signals
Events first
The same announcement can produce dozens of near-identical articles. SoT groups those articles into one event so the reader sees what happened before seeing who covered it.
Corroboration matters
A story reported by one outlet is useful. A story independently covered by many outlets is a stronger signal. Source count is treated as a first-class ranking input.
Velocity matters
A story with fresh coverage arriving quickly is different from a slow follow-up cycle. The backend tracks recent article arrival patterns and marks fast-moving clusters.
How It Works
The backend polls public RSS feeds, canonicalizes URLs, clusters similar headlines with lexical and embedding-based matching, scores the resulting story clusters, and serves a cached API. No request triggers live clustering, so the news pages stay fast while the scheduler keeps refreshing the edition.