The user has given me a long set of "filed notes" describing a knowledge-management session about whether US universities grow or shrink due to AI, with regional dependence. The final instruction: "Revise the notes as a set: keep every claim that still holds, drop repetition, sharpen wording. Output one Markdown note that is the current best statement of what is established. Start with a # title, then short paragraphs, then: What would change this: … Keep sources or Unverified: tags. Do not mention tools, sessions, or that you are revising."

So I need to synthesize all the notes into ONE Markdown note. Let me inventory the claims across all notes:

**Core claim (Notes 1/2):** By 2045, US higher education bifurcates rather than uniformly shrinks — top ~100 grow, middle hollows into teaching-only, bottom third of four-year schools merge or close. Regional decline tracks the demographic map more than AI itself.

**Demography (Notes 1/2/3):**
- Births fell from ~4.32M (2007) to ~3.59M (2023), ~17% decline (CDC/NCHS)
- HS grads peak ~3.9M/yr around 2025 (WICHE)
- Post-2008 birth dearth hits college age 2029–2037
- Northeast/Midwest/Plains lose 10–20% of college-age population (Note 1) / 18yo cohorts contract 10–15% (Note 5)
- Texas, Florida, Tennessee, Georgia, Carolinas grow

**AI's role:** accelerant/sorter, not primary cause. Two-front squeeze: fewer 18-year-olds AND degree-as-signal weakened (skills directly verifiable). Unbundles campus-credential first at mid-tier.

**Regional map:**
- Hardest hit: Northeast/Midwest small privates + open-enrollment regional publics
- Most resilient: SEC/Sun Belt flagships + super-selectives (prestige-elastic, not cohort-elastic)

**Program-mix × region (Notes 5/6 — most evolved mechanism):**
- Program mix (licensed vs unlicensed share of credit hours) picks winners; regional demography picks which losers can still switch
- Licensed expansion (nursing, engineering, allied health) takes 7–10 years (clinical placements, faculty pipelines, facilities, state approvals) — longer than declining-region enrollment runway → decline locks in
- Sunbelt/SEC publics: +10–25% seat growth through ~2040
- Northeast small privates without licensed core: −30–50% by 2040
- Surviving Northeast exceptions: licensed-core with hospital-system or land-grant ties
- Falsification: if Northeast pivot attempts succeed at rates comparable to Sunbelt licensed launches, pivot-lag mechanism wrong

**Tier structure (Notes 1/2):**
- Middle loses research identity first → teaching-only → most AI-exposed
- 2045 university: fewer, bigger, regionalized; research engine with experience-economy shell; credentialing as trust layer; teaching split between prestige faculty core and model-supervised adjunct tier
- Region works through tier: North's flagships hold while its teaching schools hollow first

**Adaptations/watch (Note 3):**
- Admit-standard relaxation at margin, international recruitment contingent on visa policy, credential repackaging around verified AI skills
- Watch-tell: endowment repurposing from prestige spending to need-based discounting = fighting the cliff

**Cost structure (Notes 9–12):**
- Revenue event before headcount event (Note 9)
- AI instruction savings competed away, not banked — oversupply → discounting price war → middle compressed from both sides; tuition shows disruption first
- Regional divergence in use of savings: growth states reinvest, shrinking states cut costs
- Teaching staffing cost cut ≥25% claim (Note 10) with falsification test on IPEDS instructional labor share
- Sequencing, tenure as annuity, admin vs faculty — partially spelled out

**Research function (Note 13):**
- Bifurcation into ~150 fortress research institutions vs ~3,000 teaching-first schools
- Verification pivot: scarce asset shifts from producing knowledge to verifying it (provenance peer review, replication, credentialing)
- Two geographies: R&D production follows money/population to SEC/Texas/Research Triangle; verification rents stay with legacy elite brands
- SEC can buy talent, not reputation
- Research headcount per campus shrinks 30–50% as AI absorbs drafting/analysis labor (Unverified)

**Kill-switch (Note 1):** if selective privates AND flagships both show flat/declining yields and international demand despite AI tools, degree itself devalues — deeper than bifurcation.

Now, the instruction says "Revise the notes as a set