Most feeds are built to keep you scrolling. We wanted the opposite: twice a day, the handful of stories worth your time, in brief, and then it ends. No account, no tracking, no feed.
An edition at morning and one at evening. Each has six sections: World, Economy, Politics, Sport, Tech and Science. Every section shows three stories, plus a line of other things that happened, and "Curious for more?" if you want a little extra.
The summaries are written by an AI model in its own words. Each one links to the original article at the outlet that published it, so you can read the full piece there. Some outlets block automated readers or show only a short opening, so a summary can rest on a brief excerpt; the link always goes to the full article.
Every section opens with the three stories that reached the widest audience. We add up the weight of each outlet that covered a story, and extra outlets from the same country count less.
A story we could read in full counts more than one we only saw as a snippet, and a story loses half its weight every 24 hours.
Scores within five percent of each other count as a tie. A tie goes to the story covered in more countries, then to the longer text.
No two of the three top stories lead with the same outlet. A second story from the same outlet moves down to "Curious for more?".
Behind that link wait up to five more: the next best stories and, when we have them, one or two pieces that only one country's outlets reported, so you also see something from beyond the biggest outlets.
If you follow a place, stories from there move up on your device. Nothing about you leaves your browser for this.
Nobody at twice.day pins or boosts a story by hand. The rules sit in one short configuration file.
What you follow is stored in your browser, on your device. We never see it. To follow something new, type a phrase on your first visit or after pressing "change". Words we already know are matched in your browser. Only a phrase we cannot match locally is sent to a small service that maps it to our topic list; the service does not keep your text or your address.
The share button makes a link that carries your settings inside the address after the # sign. That part is never sent to any server. Open the link on another device and the settings come with it.
The switch at the top is on by default. If you leave it on, we count which edition was opened, which links were clicked (with the story's rank, its place in the section's full order, and its pick: one of the three top stories, one from "Curious for more?", or a one-country piece), and which topics found no stories, so we can see which sections and sources are useful. Turn the switch off and nothing is sent at all, immediately.
We use no cookies, there are no accounts and no tracking profile, and your IP address is not stored. The step that understands a new phrase you follow uses a one-way scrambled daily code made from your connection, never the address itself, and it changes every day, only to limit abuse. Usage counts are processed by an analytics service. What you type to follow a topic stays in your browser. The one exception: a phrase we cannot match on your device is sent once so it can be understood, and it is never logged.
More about the maker at alexratman.com.