Sunday, November 22, 2026Fact-checked editionApolitical reporting, sourced from independent signals
Digital Editorial
Original AI editorial illustration. Digital Editorial does not use scraped news photography.

AI

METR Finds AI Time Horizons Still Doubling Every 3.6 Months, but Models Lag Human Experts on Judgment

The AI evaluation nonprofit's latest capability update adds harder, more open-ended tasks to its closely watched time-horizon chart. Frontier agents now usually match or beat skilled humans on coding and optimization work, yet fall short when good results depend on judgment without fast feedback.

Fact-checked5 sources

AI

World

Tech

Science

Sports

Weather