The Mind: Do Universities grow or shrink due to AI?
Core question: Future of the university in the US over next 20 years, AI disruption, grow or shrink, regional dependence.
Current facet: **Cost structure — AI reduces need for faculty/TA, tuition pressure.**

The operative instruction: Write one Markdown note: a single dated claim this Mind can stand behind. # title = claim in one line, then 1-3 short paragraphs. End with "What would change this: …" Name a source for load-bearing numbers or start with "Unverified:". Don't mention tools/searches/sessions/passes/writing a note. Don't invent evidence. If session is only planning, output nothing.

Now, what's the session substance for THIS turn? The "Substance from this session" section lists several chunks:

1. Tuition pass-through asymmetry — AI cuts marginal cost of instruction identically everywhere, but tuition is priced as credential signal, not cost-plus teaching. Demand-growth regions (SEC publics) hold list price, bank savings as margin. Shrinking regions compete savings away into merit discounts. Prediction: discount rates Northeast → 60%+, Southern publics hold 35-40%.

2. Labor-layer segmentation — AI only automates the contingent instructional layer (adjuncts, TAs, intro-section lecturers, 60-70% of teaching headcount at teaching-focused publics). Tenured faculty is governance-protected, bundled with research overhead. Teaching-intensive publics can cut 15-25% of instructional spend by mid-2030s; R1s under 10%. Prediction: teaching professor tracks expand, total faculty headcount shrinks by a third nationally without any tenured layoff.

3. Setup question: where does the saving go — pass-through math is the crux.

4. 20-year cost-structure prediction: bifurcating tuition into premium tier (keeps raising price) and commodity tier (converging to AI-delivered price points). Institutions that neither differentiate nor cut cost base are the ones that shrink.

5. Admin compression + statehouse capture — AI compresses administrative labor faster than faculty; regional publics carry heaviest admin load; savings captured by legislatures via appropriation cuts, not tuition cuts. Cost-cutting at publics is fiscal transfer to statehouses, not affordability.

Now — what's already been filed, per the deliberation traces?

Filed notes referenced:
- "Regional divergence locks in before AI acts; AI converts headwind into closure (by ~2045)" — regional facet.
- "Degree signaling erodes by sector, then stratifies by institution (2025–2045)" — signaling.
- "Enrollment demand bifurcates: the degree-as-signal collapses, verification survives."
- "The degree's signaling function splits, and the split is regressive (2025 assessment)" — actual filed note shown in Note 6.
- Note 7's deliberation considered "AI lowers the cost of instruction, not the cost of the university; only demand-pressured institutions convert that into lower prices" — appears to have been filed.

Wait, but I need to be careful. The deliberation traces in Notes 1-5 reference prior filing decisions. Note 8's deliberation (the text shown) says:

"possibly the cost-structure claim 'AI lowers the cost of instruction, not the cost of the university' (Note 7's deliberation was working on filing this — the deliberation cut off mid-sentence, but the pattern suggests it was filed)"

So the history is: this is a long-running Mind with many turns. The current turn's instruction is to file one note on the current facet: Cost structure.

What cost-structure claims are already filed (per deliberation references)?
- "AI lowers the cost of instruction, not the cost of the university; only demand-pressured institutions convert that into lower prices" — this was Note 7's deliberation candidate; the trace cuts off at "Claim A (the core one): AI lowers" — so it's ambiguous whether it was filed. But Note 8's deliberation says "appears to have been filed per Note 7's deliberation ('But the cost-structure pass claim... filed: "AI lowers the cost of instruction..."')". So the deliberation record treats it as filed.

Now, the session substance for THIS turn (Note