Meta Platforms has overhauled its engineer performance evaluation framework, saying AI adoption dashboards and token usage will not be used to measure employee impact. An internal memo jointly issued by executives Maher Saba and Santosh Janardhan on September 2 said managers should instead assess output quality, work speed, and the complexity and scope of problems. The memo disclosed that AI agents assist with 93% of Meta’s code changes, but said adoption is not the goal in itself; better outcomes, higher-quality delivery, and faster iteration are. The policy removes references to specific AI tools and task-based usage directives from the previous framework. The change follows an earlier push to mandate AI use and track token consumption, which contributed to the so-called tokenmaxxing behavior, in which employees consumed AI tokens to demonstrate adoption. CTO Andrew Bosworth had warned in April that token usage was not a measure of impact. A Meta spokesperson declined to comment.