Meta drops token usage as AI performance metric

Meta's move to scrap token-based metrics signals a maturing view of AI productivity, where tool adoption is no longer treated as a proxy for engineering value. With 93% of code changes already AI-assisted, the focus shifts to output quality and outcomes, which could pressure other firms to recalibrate similar dashboards. The internal memo's emphasis on AI not being an end in itself suggests a broader industry reckoning with measuring human-AI collaboration. Whether this recalibration leads to more nuanced performance frameworks or just moves the goalposts is the open question.