Reading the world as of now, live.

How to read this. Every question is about an official US statistic's next first print (FRED/ALFRED): no markets, no odds. Upcoming forecasts are committed to an append-only, hash-chained ledger the first time they are made and never edited. Settled questions are scored against the first print. Cards marked replay were not forecast before their release; they were computed afterwards as of the release eve, which the as-of mask makes time-safe. Two forecasters are shown: Tev, and Kāla. Since 18 October 2026, Kāla's forecasts before a print come from Ask Kāla's four models (gpt-oss, Gemini, Qwen, Kimi), each given the series' first-print history, simple facts computed from it and recent dated news, with their probabilities averaged; earlier forecasts (committed on 17 October) came from an older system that returned about 57% for every series, and stay on the record unchanged. After a print, Kāla reads the published value. A call is marked confident only at 90% or more one way; otherwise the card shows which way it leans. Tev (R2) is Kāla's 0.6B precursor, until Kāla-1 is trained: it never leaks, but it does not yet read evidence well, and its scores show that. All numbers are measured, not curated; the board updates daily.

Data source. FRED® and ALFRED®, Federal Reserve Bank of St. Louis, from the original releases of the U.S. Bureau of Labor Statistics, Bureau of Economic Analysis, Census Bureau, Federal Reserve Board, Department of Labor and Energy Information Administration. Series carrying a third-party copyright are excluded. This board is not endorsed by the Federal Reserve Bank of St. Louis. Download the full ledger.