"Claude currently covers just 33% of all tasks in the Computer & Math category" [33%]
The 33% figure directly quantifies the gap between the 94% theoretical capability (β measure) shown in the same figure and actual observed usage, establishing the core structural claim that deployment lags feasibility.
Reasoning from this article
The article systematically compares theoretical capability (Eloundou et al. β metric) to observed usage across occupational categories, showing consistent underdeployment. This gap persists even for tasks where Claude usage is highest (68% of Claude usage on β=1 tasks), indicating the constraint is not technical feasibility but adoption friction—legal, organizational, and workflow integration barriers. As these frictions resolve over time, the red area in Figure 2 will expand toward the blue, creating a predictable wave of labor market disruption.