To some, METR’s “time horizon plot” indicates that AI utopia—or apocalypse—is close at hand. The truth is more complicated.

Technology Review, published by MIT, is a leading source of news, analysis, and commentary on emerging technologies, scientific advancements, and their societal impacts. Covering topics such as artificial intelligence, biotechnology, renewable energy, and more, Technology Review provides  reporting and  articles on the intersection of technology and society. Readers can stay informed about the latest breakthroughs and trends shaping the future of technology.

MIT Technology Review

METR's "time horizon plot" shows AI models improving exponentially at completing tasks, but is widely misunderstood. The y-axis represents how long tasks take humans to complete, not how long AI can operate independently. The metric is based primarily on coding tasks and comes with substantial error bars. While Claude Opus 4.5 achieved a five-hour time horizon, this doesn't mean it can replace five hours of human work in real-world scenarios. The plot has limitations but remains a valuable scientific tool for quantifying AI progress, despite being misinterpreted by both AI doomsayers and boosters.

This is the most misunderstood graph in AI