About
How Distilled works
Every hour Distilled collects the most discussed things in tech and AI from five open sources and merges them into a single feed without duplicates. The author's notes are published alongside the trends.
Sources
- Hacker News: the front page plus stories with 40+ points from the last three days.
- GitHub: the most starred repositories created in the past week.
- Hugging Face: models and Spaces with the highest trending score.
- LessWrong: a selection of popular posts on rationality and AI safety.
- Lobsters: hot links from the programmers' community.
How the ranking works
Scores from different sites are not comparable: a GitHub star and a Hacker News upvote carry different weight. So each item is compared only with items from its own source (a percentile within the source), and only then is everything merged into one feed.
- Now: what is on the sources' pages right now, adjusted for freshness: the older the item, the lower it sits.
- Day, week, month, year: the best items by peak score among those that appeared during the period.
- A link discussed in several places is shown as a single row marked "also on".
- To keep one source from taking over the feed, consecutive items from the same source get a small penalty.
Updates and archive
The data is refreshed every hour. Everything that made the feed is kept in the day-by-day archive: each date has its own page with the best items.
Data for your own feeds
- Trends RSS and notes RSS.
- JSON:
/api/feed?sort=now(alsoday,week,month,year). - A description of the site for AI assistants: llms.txt.
Author
The site is run by Rom Burn. Trend titles and links belong to their authors and platforms; Distilled only collects and ranks them, always linking to the original source.