How did Anthropic measure AI's "theoretical capabilities" in the job market?

Anthropic's new simulation claims to predict which jobs AI could theoretically replace with 85% accuracy, but critics argue theory and practice diverge sharply in the real labor market. The model analyzes tasks across 923 occupations, scoring each for AI feasibility. Yet the gap between simulation and reality undermines the approach: it fails to account for employer resistance, regulatory hurdles, and training costs that delay deployment. This limitation becomes stark when comparing Anthropic's method to #5-ranked studies that only capture historical impacts—both miss the messy dynamics of adoption. Without concrete validation against actual job displacement data, the simulation remains a provocative estimate rather than a reliable forecast.