A chart measuring artificial intelligence progress created by METR, a 30-person nonprofit in Berkeley, has become an obsession across Silicon Valley and Wall Street. The so-called 'METR time-horizon' chart tracks how long AI agents can reliably complete tasks, with the length doubling every three to four months recently. This exponential growth has fueled a hundreds of billions of dollars race to build more powerful AI systems.

METR's chart has become a Rorschach test for AI's future, with optimists claiming it signals artificial general intelligence is near while safety advocates warn of potential catastrophe. Companies like OpenAI and Anthropic have competed fiercely to improve their time-horizon scores, driving massive investments in data centers and specialized chips. The chart's influence has grown so significant that some analysts say it's helping to support the global stock market.

While the chart doesn't measure job displacement or AI takeover risks directly, its implications are profound. METR researchers estimate a 1-10% chance of an 'intelligence explosion' occurring this year, where AI systems could recursively improve themselves beyond human capabilities. The nonprofit's work, funded by philanthropies and computing credits from major AI companies, has become the most influential metric in tracking AI's rapid advancement.