Atomize News: Intelligent Search
An earlier AI news search product that grouped coverage into events and connected search results to trends.
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 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.
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.