Amazon tells engineers to stop ‘tokenmaxxing’ to look productive

Amazon has issued new engineering guidelines instructing developers and managers to stop tokenmaxxing. The company asserts that AI usage metrics are deeply flawed unless they are directly linked to long-term outcomes. This directive warns employees against using AI simply to appear productive, suggesting that manual methods are sometimes faster. Similar concerns regarding AI spending have recently been reported at Uber and Accenture, prompting Amazon to prioritize actual results over the volume of AI tokens used.