The current facet: enrollment demand shift (credential value vs. AI skills). The session substance gives a demand-side corollary: degree decomposes into two products by ~2035; mid-market enrollment contracts 25-40% from ROI arithmetic, not just demography; selectivity becomes the product; demand and funding compound into one geography (corridor adjacency); falsifiers given.

Existing notes already cover: enrollment bifurcation by region (both/region-stratum), program-level demand migration (credential-signal → verified-skill), revenue shock (buyers defect), bimodal BA returns (enforcement mechanism), corridor research map. So the increment here is: the degree decomposes into two distinct products (ticket-to-first-job vs. positional/selectivity product), mid-market contraction driven by legible ROI arithmetic rather than demography alone, and the fusion of demand shift with funding shift into one geography (adjacency keeps credential payoff legible to local 18-year-olds).

Load-bearing numbers:
- 25-40% mid-market contraction by ~2035 — forecast, mark Unverified or frame as forecast.
- "median US college unviable by ~2035" — forecast.
- Evidence for AI hollowing entry-level roles: Stanford Digital Economy Lab "Canaries in the Coal Mine" (Brynjolfsson, Chandar, 2025) — early-career employment declines in AI-exposed occupations. Real, used in prior notes.
- Demographic baseline: CDC/NCHS births 4.32M (2007) → ~3.59M (2023); WICHE HS grads peak ~2025. These are established in prior notes and real. I can cite for the "not primarily demography" contrast.
- BLS college wage premium — baseline anchor, could mention but not necessary.

The new claim should be one dated claim. The core: by ~2035 the bachelor's decomposes into two products — an entry-ticket product whose ROI goes negative as AI absorbs first-job tasks, and a positional product where selectivity itself is what's bought — so mid-market contraction is demand-driven (ROI arithmetic), not cohort-driven, and demand compounds with funding into one corridor geography.

Title options:
"# By ~2035 the bachelor's degree splits into two products — a first-job ticket whose ROI goes negative as AI absorbs entry-level work, and positional scarcity — and mid-market contraction is driven by buyers doing the arithmetic, not by the demographic cliff"

"# The bachelor's splits into two products by ~2035: a first-job ticket with negative ROI outside the top tiers, and positional scarcity — mid-market contraction is buyer arithmetic, not demography"

mechanism — AI absorbs the junior tasks (junior analyst, staff accountant, coordinator, junior dev) that justified the tuition premium; cite Stanford Digital Economy Lab "Canaries in the Coal Mine" (2025) for early-career employment declines in AI-exposed occupations. 18-year-olds and parents can see the pipeline break; the credential stops functioning as a ticket. Distinguish from demographic cliff: cohort decline (CDC/NCHS births 4.32M 2007 → 3.59M 2023; WICHE peak ~2025) hits everyone, but ROI collapse hits the mid-market specifically — non-flagship publics, regional privates, open-access comprehensives.

the two products + geography fusion. Where the ticket product dies, selectivity becomes the product: T20 privates and flagship honors colleges hold/grow pricing power on positional scarcity, networks, AI-complementary judgment. The corridor fusion: same industry-corridor hubs capturing research co-funding also capture applicant demand because adjacency keeps the internship-to-hire pipeline legible — the credential's payoff is visible to local 18-year-olds. So demand shift and funding shift compound into one geography: SEC/Sun Belt flagships gain both; heartland comprehensives lose both. Forecast numbers: mid-market enrollment contracts 25-40% by ~2035 (Unverified, forecast); median US college unviable by ~2035 (forecast).

Keep to 3 paragraphs max. Para 3 could be the falsifier setup but the instruction says end with "What would change this:" line — the falsifiers go there. So 2-3 paragraphs of substance then the line.

"What would change this: if AI proves complementary enough to raise the college wage premium broadly (augmentation, not substitution of entry roles), or if employers reinstate degree screens defensively against AI-inflated applicant pools, mid-tier credential value stabilizes and this fails; also if NCES/IPEDS