"Both estimate that around half a million fewer coders are working today than would have been if pre-LLM-era employment trends had continued." [half a million]
The 500,000-coder figure represents agreement between Federal Reserve official statistics and Stanford payroll analysis—two independent methodologies converging on the same magnitude of displacement. This is the article's core evidence that AI job loss is no longer theoretical.
Reasoning from this article
The article emphasizes this is not absolute decline but 'marked slowdown in growth'—a distinction that matters for policy but confirms the displacement mechanism is active. The convergence of official and private data sources removes the excuse that earlier findings were artifacts of incomplete measurement. The article treats this as the moment when 'evidence on white-collar work displacement is beginning to match the theory.'