Anthropic's economist breaks with the boss
Peter McCrory, the economist Anthropic hired to model AI's labor effects, says he does not expect US unemployment to be noticeably higher a year from now because of AI. That is the opposite of what his own CEO has spent a year predicting.
Dario Amodei told Axios in May 2025 that AI could erase half of entry-level white-collar jobs and drive unemployment to 10-20 percent. This week the person Anthropic pays to study exactly that published the rebuttal. In a long essay on X, Head of Economics Peter McCrory reports US unemployment at 4.2 percent as of June 2026, the level the Federal Reserve treats as full employment, and no sign that the jobs most exposed to Claude are faring worse than the rest.
"No occupation in the US Department of Labor's O*NET database has every task fully handled by Claude." — Peter McCrory
His method is what makes the claim land. He ranked occupations by how much of their work Claude can actually do, using the Labor Department's task-by-task catalog of what each job involves, then checked whether the high-exposure jobs (consultants, lawyers, financial analysts, technical writers, data entry, customer support) had shed workers faster than everyone else. They had not. His explanation: no job in that catalog has every one of its tasks fully handled by Claude, so human oversight stays load-bearing, and the leftover human-only work acts as a floor under employment.
McCrory does not paper over the one crack that matters. Hiring has weakened for younger workers in exposed fields, the same early signal Stanford's payroll research has flagged, and the honest tension in the whole debate is aggregate calm sitting on top of cohort-level erosion. The larger oddity is corporate: the same company now hosts the loudest forecast that AI will gut white-collar work and its sharpest data-driven rebuttal, both about its own product, both under named employees. When the doomer and the debunker share a payroll, the number is worth reading twice.
The lenses
The facts
Concepts
How this connects
Tap a node to open it