A personalized news app that learns from what you actually do

Why interest checkboxes do not work

Most personalized news apps ask you to tick topics at signup and then treat that list as permanent. It goes stale within a week: people follow stories, not categories, and nobody goes back to update a settings page. Headlinne infers preference from behaviour instead — what you open, what you finish, what you skip in under a second.

Swiping is the interface, not a gimmick

A swipe is the cheapest possible way to express a preference, which is why it produces so much more signal than a like button nobody presses. Right to save, left to skip, up to open. Every one of those gestures feeds the ranking, so the model gets a usable picture of your taste in a single sitting.

• —Skips count as much as likes — a fast left swipe is real information
• —The feed rebalances between sessions rather than only at signup
• —A finite session with an end, instead of an infinite scroll

Personalized, without the echo chamber

The obvious failure mode of a personalized feed is narrowing until you only see one point of view. Headlinne works against that directly: every story carries a five-point political bias label, and the feed deliberately mixes the distribution rather than converging on whatever you clicked most. You can see the balance of what you have been reading and correct it.

You can see and reset what it learned

The preference profile is yours to inspect and to delete. Resetting it in Settings wipes every learned preference and starts the model over — useful when your interests shift, or when you simply want to see what the feed looks like cold.

Built for people who do not want a news habit

The point is not to maximise time in the app. Stories expire after 48 hours, sessions end, and there is no infinite scroll to fall into. A few minutes should be enough to know what happened — that is the whole design brief.