From thousands of headlines to a handful of good news

It all starts with a hand-picked list of trusted outlets, country by country: newspapers, radio stations, local newsrooms. We do not trawl the whole internet — we listen closely to sources we chose one by one.

The list is alive: sources that stop publishing with quality leave it, and good suggestions from readers join it after review.

Before any editorial judgement, automatic filters set aside what does not belong: clickbait headlines, paywall teasers, duplicates, empty press releases.

Nothing subjective happens at this stage — it only removes what could never become a story worth your time.

Each surviving story is read by an automatic classifier that scores it on quality, how genuinely uplifting it is, its real impact, and its own confidence in that judgement.

Stories with a clearly high score are published automatically; clearly low ones are set aside; the borderline cases go to a human shelf.

The stories the classifier is least sure about go to a human curation queue. A person reads them, promotes the pearls and corrects the misfires.

We do not promise every borderline story gets read — but everything that is published automatically can be unpublished by a human, and the criteria keep being tuned.

We publish only the headline and a short summary — never the full text. The visible link on every story takes you to the original source, where the credit and the traffic belong.

If you only remember one thing about how this site works, make it this one.

Every human decision — a promoted story, a corrected misfire, a source suggested by a reader — becomes a lesson the system learns from.

Over time, the criteria get sharper and the classifier makes fewer mistakes. It is a loop: better judgement brings better stories, which bring better judgement.