Is psychohistory a thing now? They are just another actor, for sure they have relevant usage data but (1) their view are biased in multiple ways (at least due to perverse incentives), knowingly and unknowingly (2) they are experts in their field, but not necessarily in sociology, economy, and politics, which are arguably necessary to make serious, even if ineffective, predictions about what's going to happen with labor.

I'd rather see what institutions such as NBER have to say than to blindly trust frontier AI labs that won't even publish their methodologies most of the time.