A recent blog post from METR proposes a novel way to measure artificial intelligence: by assessing the length of tasks AI agents can independently complete. Over the past six years, this metric has been growing at an exponential rate—doubling in just about seven months. Such a rapid advance suggests a dramatic shift is underway. If this trend keeps up, the coming decade could see AI agents capable of handling a significant portion of software tasks that now consume days or weeks of human effort.
The post emphasizes that these improvements are more than incremental—they signal a future where AI might soon move beyond supporting roles and into independent project delivery. This changing capability has profound implications for productivity, software development, and the nature of work itself. For leaders, technologists, and anyone interested in the future of AI, this metric offers a powerful lens on how quickly the frontier is moving. As the exponential curve pushes upward, watching how long a task AI can complete may become one of the most important indicators of technological progress.
Leave a Reply