NewsBlur is a news reader whose distinguishing feature is Intelligence Training: rather than showing every item from every feed, it learns what you actually want to read.
Training is explicit and under your control – you mark authors, tags, titles and phrases as liked or disliked, and the classifier filters accordingly, surfacing a Focus view of likely-relevant stories against the full unread firehose. That transparency is the interesting part. Where an opaque algorithmic feed decides for you, here the rules are yours, inspectable and adjustable, which is a meaningfully different relationship with a recommender.
The reader itself is mature: original-site view, text view, story styles, saved searches, saved stories, shared and widely-read stories, long reads, and a daily briefing. It is open source with a free tier and a paid premium level, and has been running for well over a decade – which for an RSS reader, a category littered with shutdowns, is the relevant credential.
