METR introduces a new metric to calculate exactly when AI agents become more expensive than humans

METR has introduced a metric called the expenditure horizon designed to calculate the specific point at which using AI agents becomes more expensive than employing human labor for problem-solving tasks. While the tool provides a quantitative basis for assessing cost-effectiveness, early testing on the NanoGPT speedrun indicates that current models have yet to consistently outperform human benchmarks. This development highlights the ongoing challenge of balancing computational costs against productivity gains as organizations weigh the integration of automated agents into their workflows.