Hollowing the Top

Hollowing the top is the labor-economic diagnosis that generative AI has done to the upper reaches of the wage distribution what earlier waves of computerization did to the middle: rendered structurally vulnerable the tasks that practitioners had been told were the safe harbor from automation. David Autor’s original task-based framework, developed with Frank Levy and Richard Murnane in 2003, classified non-routine abstract tasks—the legal reasoning, medical diagnosis, financial analysis, and creative production that require flexibility, contextual judgment, and adaptability—as the domain that computers could complement but not replace. The classification explained the hollowing of the middle with remarkable precision and generated an equally clear policy prescription: invest in education, move workers up the skill ladder into the non-routine abstract occupations where human comparative advantage was secure. Then large language models demonstrated the ability to draft legal briefs, generate medical differentials, compose financial analysis, write software, and produce creative content at professional quality. The boundary that the framework had identified as the fundamental line of demarcation