I got tired of two things: news apps that are just an aggregator with no memory of what happened last week, and "AI news summarizer" apps where you can't tell what's a real signal vs. the model's mood that day. So for Horizon, I made a rule for myself early on: an LLM only gets to judge the one thing that's genuinely subjective (picking and summarizing the day's actual world/geopolitics events). Everything else — Markets, Cyber, Disasters, Sanctions, Space weather — had to come from a fixed, documented threshold against a real public API, never from asking a model "how bad is this." That constraint shaped most of the build: finding a free, reliable, no-key API for each domain (GDACS, NOAA SWPC, OFAC/UN sanctions lists, IODA, IMF PortWatch...), then a threshold I could defend rather than tune to look dramatic. The other thing that evolved along the way: I started treating the daily archive as a real feature, not a byproduct — dated pages, public RSS, nothing behind a login. It's free, no account, no tracking. Would love feedback, especially on what's missing or what feels off.

Horizon

Daily world news & markets, scored by fixed rules — not AI

Daily world news & markets, scored by fixed rules — not AI