IMF President Georgieva: Artificial intelligence growth may change labor market structure and raise questions about how governments generate tax revenue.

Heat

0–100 points (no unit)Source breadth 40 + best rank 35 + freshness 25. Breadth = min(sources ÷ 8, 1) × 40; rank = max(0, 1 − (best rank − 1) ÷ 30) × 35, so #1 on a trending list scores full and #31 or lower scores 0; freshness is 25 for a trending item seen only 1–2 times in the last 24 hours or an RSS article under 24 hours old, otherwise 12.5. All counts use the last 24 hours, not the calendar day. Templated market tickers are multiplied by 0.45; breaking keywords add up to 15, capped at 100.

Heat level

Surging ≥ 70, Hot ≥ 45, otherwise normal (circulation, not CVE-style severity)This grades how far a story has spread, not how risky it is.

Rising / falling

Heat differenceCompared with the previous crawl, not the previous page rebuild — one crawl rebuilds the pages several times. Crawls run every 4 hours. Differences under ±8 show as steady.

Tracking floor

Heat ≥ 45Only events at or above this line are written to the time series, so the rising / falling / net-change columns cover only them; events below it still count toward event totals and category heat.

Sources

countHow many different outlets carried the story (sections of one outlet, such as its newsflash and latest feeds, count once).

Mentions

headlinesEntity pages only: headlines that day matching the entity's keywords, counted from raw headlines — cross-posts are not de-duplicated and this is not a heat score. It measures share of coverage.

Evidence confidence

high / medium / lowHow many events this AI paragraph can be traced to (headline match weighted by word rarity, at least two distinct words). It measures traceability, not whether the analysis is right.

Time

UTC+8All dates and times on the page are UTC+8; "today" means the UTC+8 calendar day.

Heat measures how widely and how prominently a story is carried, not how serious, credible or market-moving it is. A major disaster covered by two outlets scores lower than entertainment news reposted everywhere — that is the definition, not a bug. This site has no model of severity or market impact and does not pretend to.