Market context, summed up in one number.

45+ feeds polled continuously (near-instant Telegram squawk, Reuters/AP/Bloomberg newswires, WSJ, FT, official Fed/ECB/BLS/EIA sources, geopolitics, energy/gold/Europe packs). Duplicates are merged: a story picked up by 30 outlets becomes ONE item corroborated 30 times. Everything is shown, but only relevant headlines count in the score: questions, opinions, lists, local prices and off-topic items stay visible in the feed, grayed out, with no weight.

impact = severity × surprise × reliability × freshness × corroboration. Severity ranks the event (FOMC/CPI = maximum, analyst noise = minimum). Reliability depends on the content as much as on the source: a verifiable fact (a figure, an official decision) counts in full, a reported event a little less, speculation (“could”, “warns”) barely more than half. Freshness decays exponentially, in market time: a weekend hour counts three times less than a session hour, so that a Saturday bombshell still weighs at the Sunday reopen (Gross-Klussmann & Hautsch, 2011).

The headline is projected onto macro axes (Fed policy, inflation, risk aversion, oil supply, dollar, tech, Europe) and then crossed with each asset's sensitivity matrix. “Oil soars”: bullish CL (supply) but bearish NQ (inflation → rates → tech). Only events give a direction, never a mere topic word: “war” or “Hormuz” alone say nothing, whereas “Iran offers to reopen Hormuz” is a resolution and reads the opposite way to an escalation. Negations (“denies”, “calls off”) and the conditional (“could”, “threatens”, at half weight) are taken into account. No asymmetry: good and bad news of the same importance weigh the same.

Items are grouped by theme: a theme's weight grows with the logarithm of the number of pick-ups, never linearly. 50 articles on the same CPI do not weigh 50 times more than the first one. The sum goes through a hyperbolic tangent that bounds the score between −10 and +10. That level is then compared with its recent norm (exponential average, 3-day half-life): an uninterrupted war news flow for weeks is already in the prices, only the deviation from that norm enters the score.

Upcoming events are shown, but no direction is drawn from the consensus: it is public and already in the prices. And in the 45 minutes before a major release, the score is deliberately flattened: a news bias just before a figure is dangerous noise.

Actual values (CPI, NFP, claims…) are fetched from official sources, FRED (Federal Reserve Bank of St. Louis) and the ECB, to compute the real surprise vs consensus, the one that moves prices.