Twenty stories. One cup of coffee.
AI news doesn't have a supply problem — it has a filtering problem. Every morning, hundreds of headlines pile up across dozens of outlets: half of them are the same story rewritten five different ways, and finding the handful actually worth your time means scrolling through all of it. Solas Drop does that filtering for you and hands you twenty stories worth reading, ranked to your taste, before your coffee's gone cold.
The volume of AI coverage has outpaced anyone's ability to read it. Newsletters, RSS feeds, Hacker News, a dozen tech outlets all covering the same three announcements — the signal is in there, but it's buried. Most people cope by reading nothing, or by doomscrolling a feed that doesn't know or care what they actually care about. Neither is a real solution.
Every day, Solas Drop automatically pulls new stories from a curated set of RSS feeds and Hacker News, generates a vector embedding for each one, and collapses near-duplicate coverage of the same story into a single card — so you never see the same announcement from five outlets in a row. What's left is ranked using cosine similarity between each article's embedding and your personal preference vector, a running fingerprint of the kind of AI news you actually like, built automatically from your swipe history. The more you swipe, the sharper the ranking gets. No account setup, no topics to configure by hand — just like and skip, and the feed learns.
Stories are pulled on a schedule from RSS feeds and Hacker News — never manually entered, so the drop never goes stale and never depends on someone curating it by hand.
When five outlets cover the same announcement, embedding similarity collapses it to one card, so your twenty stories are twenty different stories.
Every swipe updates your preference vector. The feed ranks unseen stories by how closely they match what you've liked before — no settings, no sliders, just swiping.
Solas Drop never reproduces full articles. Each card gives you a headline, a one- or two-sentence summary, and a link straight to the original publisher, who gets your traffic and your attention.
Every past drop lives on at /archive — no account needed to browse what's already run.
Twenty stories. Refreshed every morning. Takes about as long as your coffee.