AI Productivity Gain, Skill Atrophy Trap

The lab findings are narrow: reduced persistence in specific task settings, a theoretical model with assumptions, questions about peer review expertise that haven't been tested at scale. None of that individually confirms an economy-wide skill atrophy crisis. What this week produced is something more limited and more precise: the first documented measurements of the cognitive cost side of the ledger, arriving at the same moment policy moved to accelerate adoption without a cost side. The structural driver is that productivity is easy to count and skill degradation is slow, diffuse, and shows up later. What did not change is the underlying mechanism — humans have always lost skills they stopped practicing. AI just industrializes the offloading.