AI's capability improvements haven't come from it getting less affordable

METR's frontier time horizons are doubling every few months, providing substantial evidence that AI will soon be able to automate many tasks or even jobs. But per-task inference costs have also risen sharply, and automation requires AI labor to be affordable, not just possible. Many people look at the rising compute bills behind frontier models and conclude that automation will soon become unaffordable. I think this misreads the data. The rise in inference cost reflects models completing longer tasks, not models becoming expensive relative to the human labor they replace.