Sources. Every 15 minutes, thisminute scrapes ~100 RSS feeds (BBC, Reuters, ESPN, Variety, etc.) plus 13 structured data APIs (USGS earthquakes, NOAA weather, WHO disease alerts, and more). Stories are placed on the map by location.
Presets. Each preset filters stories by source type. Sports shows stories from sports feeds. Markets shows business feeds. Planet shows weather/earthquake APIs. The "All" preset is unfiltered.
Situations. An AI (Claude) groups related stories into events, then synthesizes ongoing situations per domain: news, sports, entertainment, positive, and curious. Each domain has its own prompt tuned for what a "situation" means in that context.
Bright Side. The AI scores every story on a "bright side" scale. The Bright Side preset shows stories that are genuinely positive — not "the silver lining of a war," but actually good news.
Curious. Stories are scored on quirkiness and surprise. High scores go to unusual, delightful, "wait really?" stories. War, tragedy, and routine celebrity news score low, even if they're dramatic.
Limitations. Categorization is imperfect. A story about an SNL sketch set in Iran might get geocoded to Iran instead of NYC. A mainstream entertainment story might sneak into Curious. We're continuously improving the scoring and welcome feedback.
Philosophy. Anti-curation. You decide what to see, not editors. The map is the filter. Every story gets a dot. Presets are lenses, not editorial choices.