Let me understand the task. This is a "Mind" note-taking exercise. The core question is whether US universities grow or shrink due to AI over 20 years, with regional dependence. The current facet is: Labor market signaling — do degrees lose signaling value?

The substance from this session provides one durable claim: a falsifiable forecast about bimodal BA returns by the mid-2030s — licensed-profession and top-30 premiums hold near current levels, while mid-tier unlicensed BA premiums compress 30-50% in real terms, with compression sharpest in high-dynamism metros and mildest in state-gated fields. The falsification condition: if premiums compress uniformly, the "bifurcation by enforcement mechanism" thesis is wrong and AI's disruption is generic.

Prior filed notes already cover: enrollment bifurcation by region, program mix (licensed vs unlicensed), revenue shock, PhD pipeline, regional variation (three stacked currents). So this note should add the signaling/returns facet specifically — the bimodal returns forecast, and the enforcement-mechanism mechanism.

What's the mechanism here? Why would returns bifurcate by enforcement mechanism rather than compress uniformly? The idea seems to be:
- Licensed professions (nursing, engineering, teaching, law, etc.) — the degree's value is enforced by state licensure and gatekeeping (boards, accreditation, statutory requirements). AI can't erode the signal because the license is the signal, and the degree is a prerequisite.
- Top-30/selective institutions — value enforced by elite sorting, network effects, prestige — AI doesn't certify sorting, so the sorting premium holds.
- Mid-tier unlicensed BA — value enforced only by employer trust in the credential as a proxy for skills. AI tools (skill verification, direct assessment, portfolio/track records) undermine the proxy function, and AI also hollows the entry-level jobs those degrees fed into. So the premium compresses.

"High-dynamism metros" — labor markets where skills change fast, so the degree-proxy decays fastest and direct skill assessment is most valuable; employers there adopt AI screening fastest. "State-gated fields" — licensure-protected, so compression mildest.

The mechanism is about who enforces the signal: the state (licensure), the brand (selectivity), or nothing (generic BA). AI attacks only the third — the unenforced proxy. So the disruption is structural (by enforcement mechanism), not generic.