I built Drewes.NEWS as a way to learn NLP and serverless architecture, and to find bias in news by reading the same story from multiple outlets. Now it's evolved into a useful, privacy-focused tool.<p>It's privacy-focused in that there are no cookies, no usage of Google or Facebook components (like Google Analytics or Ads). No data tracking on users whatsoever.<p>There are bugs I'm aware of, but am looking for feedback on if this format and function is useful.<p>For those interested in the NLP side of this or the serverless side, I'd be happy to answer questions about how it was put together. The short version is, I pull down RSS feeds from 33 news sites (approximately 1M stories in the database so far), store them, create a term frequency model, cluster the most recent 10k stories based on TF similarity vectors, then store story similarities.<p>In the future I'd like to add paging, searching, and more filtering. I'm also thinking about having a URL for each story, that would show all the similar articles. That way, if you don't want to link directly to a particular news source, you can link to the drewesnews aggregate URL, and the reader can pick whichever source he/she wants to read.<p>Any feedback would be much appreciated!