The user has filed multiple notes on the same question: "Do universities grow or shrink due to AI?" over the next 20 years, in the US, regionally dependent?

Note 1 is a complete synthesis with a strong dated claim: bifurcation by 2045, top ~100 grow, middle hollows into teaching-only, bottom third merges/closes, regional decline tracks demography more than AI. Ends with a "What would change this" line.

Note 2 is essentially the same claim but truncated — it repeats Note 1 almost verbatim and cuts off mid-sentence ("Region works through tier"). It's a redundant draft of the same note.

Note 3 is a different filing angle: it's more data-grounded (birth figures 2007 ~4.32M → 2023 ~3.59M, ~17% decline; HS grads peak ~2025 at ~3.9M/yr), frames it as a "two-front demand squeeze" (fewer 18-year-olds AND devalued degree signal), and has a regional map: hardest hit = Northeast/Midwest small privates and open-enrollment regional publics; most resilient = SEC/Sun Belt flagships + super-selectives; predicted adaptations (admit-standard relaxation, international recruitment contingent on visa policy, credential repackaging around verified AI skills). It also has a "watch for" tell: endowment repurposing from prestige spending to need-based discounting as the signal a school is fighting the cliff.

So I need to produce ONE note that merges the non-redundant content:

From Notes 1&2 (same content, keep once):
- Dated claim: by 2045 bifurcation, not uniform shrink. Top ~100 grow, middle hollows into teaching-only, bottom third of four-year schools merge/close. Regional decline tracks demographic map more than AI.
- Demography is the load-bearing wall: HS grads peak ~3.9M near 2025–26; post-2008 birth dearth hits college age ~2029–2037. Regional lopsidedness: Northeast/Midwest/Plains lose 10–20% of college-age population; Texas, Florida, Tennessee, Georgia, Carolinas grow.
- AI is accelerant/sorter, not primary cause. Unbundles campus-credential product first at mid-tier; AI tutoring + skills-based hiring erode value of non-selective degrees; schools in shrinking states whose legislatures won't backfill losses hit hardest.
- Research flagships, land-grants, Sun Belt/SEC growth-tier gain pricing power: population inflow, athletics-attention economies, pivot to AI research hubs and workforce pipelines.
- Middle loses research identity first, becomes teaching-only — most AI-exposed tier.
- 2045 university: fewer, bigger, more regionalized; research engine with experience-economy shell (residential life, stadium, brand); credentialing demoted to trust layer; teaching split between prestige faculty core and model-supervised adjunct tier.
- Region matters but works through tier: North's research flagships hold while its teaching schools hollow first.

From Note 3 (additions, non-redundant):
- Birth data: ~17% decline from 2007 (~4.32M) to 2023 (~3.59M). This sharpens the demographic claim with actual figures. Note 1/2 mention "post-2008 birth dearth" without numbers, so adding the 4.32M → 3.59M figures is a genuine improvement, not repetition.
- The two-front squeeze framing: fewer domestic 18-year-olds AND degree as signal weakened (skills directly verifiable). Note 1 has "skills-based hiring" but Note 3's framing of the degree as a *signal* being directly verifiable is sharper — worth merging.
- Regional map refinement: "hardest hit: Northeast/Midwest small privates and open-enrollment regional publics" — Note 1 has "non-selective private and regional public colleges" — same thing, keep once. "Most resilient: SEC-region and Sun Belt flagships (in-migration, athletic brand draw, state support), plus super-selective schools (demand is prestige-elastic, not cohort-elastic)." The super-selective point — "demand is prestige-elastic, not cohort-elastic" — is a new, sharp claim not in Notes 1/2. Keep it.
- Predicted adaptations: modest admit-standard relaxation at the margin, expanded international recruitment contingent on visa policy, rep