TimeThe feed ingests configured RSS outlets and X accounts, extracts article text and usable media, deduplicates repeated URLs, embeds each story, and clusters similar coverage. Cluster summaries are generated from the grouped articles so the front page reflects the shared event, not just one outlet headline.
Kalshi markets are synced into the database with current prices and periodic snapshots. News clusters and markets are embedded, candidate markets are ranked semantically, and the matcher asks the model to choose the closest public Kalshi relationships. The strongest match is shown first; lower-confidence but still relevant candidates are grouped as related markets.
Reported probability is the Kalshi price captured when the story was matched or scored. Latest probability is the newest synced market price. Outlet rankings estimate how close each source's framing was to the market probability at scoring time, while optional OpenRouter commentary turns large price moves and nearby news into ticker-style market notes.
APIs are expensive and the vibedev ran out of money
reach out to freddy so he can top em up