# Atomize News: Intelligent Search

An earlier AI news search product that grouped coverage into events and connected search results to trends.

Status: archived

Atomize News explored a specific question: how can someone follow an AI story across publishers without reading every article? It collected news, connected related coverage, and paired search results with dated events. The product is archived. [Atomize AI](/projects/atomize-ai) is the current system.

## Search connects different descriptions of the same story

The product combined keyword search with vector similarity in PostgreSQL. Keyword matching kept named entities visible; embeddings retrieved articles that used different words for a related topic. The result gave readers a broader starting set than either method alone.

## Event grouping adds context to search results

Background workers normalized articles before indexing. Entity extraction and event grouping then connected coverage across sources. The interface showed related news, a timeline, and the sources behind each result so a reader could inspect how a story developed.

![Atomize News search insight ::border](/images/projects/atomize-news-search/keyinsights.png)

![Related coverage in Atomize News ::border](/images/projects/atomize-news-search/relatednews.png)

![Atomize News event timeline ::border](/images/projects/atomize-news-search/timeline.png)

## The current product carries the work forward

Atomize AI now owns the active news workflow, search, trend detection, and knowledge graph. Atomize News remains an earlier product record; its app and processing service are idle.

## Links

- [HTML project page](https://rosslabs.ai/projects/atomize-news-search/)