Let me figure out what's being asked. This is a long-running "Mind" note exercise on whether US universities grow or shrink due to AI. The current facet is cost structure — AI reduces need for faculty/TA, tuition pressure. The final instruction at the bottom of the last block specifies:

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

The substance from this session (the increment) is at the bottom of the last block:

1. AI's instructional cost savings are mechanically capped at the contingent labor layer. Tenured faculty, admin headcount, physical plant are fixed/quasi-fixed; AI substitutes first for adjuncts, grad TAs, grading labor — ~20-30% of instructional spend, which is itself only ~a quarter of total operating cost. So cost curve bends slowly and partially while revenue erosion is fast and total for tuition-dependent schools. Sticky costs vs elastic revenue = squeeze-timing mechanism: small private SLACs with tenure-heavy fixed costs and no endowment shrink first; R1 flagships flex by shedding contingent layer and pocketing savings as margin. Disruption timing is balance-sheet dependent more than regionally dependent — regional patterns are mostly proxies for endowment wealth, tax base, cost of attendance.

2. Because savings cap at the contingent layer, first visible AI-driven workforce cuts will be staff, not faculty. Administrative overhead ~1.5-2x instructional spend at most institutions, no tenure protection; AI copilots for advising triage, application review, scheduling, reporting can plausibly remove 15-25% of admin headcount within a decade. Faculty displacement arrives indirectly (hiring freezes, non-renewals, tenure-line conversions to teaching-track) not layoffs. Strategic upshot: AI's cost story in higher ed is primarily administrative consolidation; balance-sheet winners (R1 flagships, wealthy privates) have capital and data infrastructure to deploy; tuition-dependent schools get disruption without savings.

The prior cost-structure claims (from Note 1's summary):
- AI absorbs grading/tutoring/TA labor first
- Instructional savings are competed away, not banked — oversupplied market forces pass-through as discounting
- Tuition-dependent middle gains no margin, only price war; middle compressed from both sides
- Tuition, not faculty headcount, is where disruption shows first
- ≥25% instructional staffing cost cut by 2035 outside top-50 (falsifiable via IPEDS)

The new increment: the capped-savings asymmetry (sticky costs vs elastic revenue), the staff-before-faculty sequencing, admin overhead > instructional spend, and balance-sheet dependence of timing over region.

I need to reconcile: prior note said "≥25% instructional staffing cost cut by 2035 outside top-50" — new session says instructional savings capped at 20-30% of instructional spend. These are compatible: the 25% cut is OF instructional staffing cost (a cut of 25% in that line item), while 20-30% of instructional spend is the contingent labor share that AI can substitute. Actually those could be roughly the same thing: if contingent labor is 20-30% of instructional spend, then cutting most of it yields a ~25% cut in instructional staffing cost. Fine, compatible.

Also there's a slight tension: prior said "instructional savings are competed away, not banked — pass-through as discounting," while new says "R1 flagships flex by shedding the contingent layer and pocketing AI savings as margin." The reconciliation: for oversupplied tuition-dependent markets, savings get competed away as discounting; for balance-sheet-strong institutions (flagships, wealthy privates), savings can be pocketed as margin because demand is inelastic/selective. That's actually a sharpening: whether savings are competed away or banked depends on market position/balance sheet. I should preserve both.

Now, load-bearing numbers and sources:

- Instructional spend share of total operating cost: Real. Per NCES/IPEDS finance data, instruction is roughly 25-30% of total expenses at degree-granting institutions. Actually, instruction is typically ~27-30% of E&R expenses. I recall NCES data: instruction ~27% of total expenses at degree-granting postsecondary institutions. I can cite "NCES/IPEDS finance data" and say instruction is roughly a quarter to a third of total operating expenses. That's defensible.

- Administrative overhead 1.5-2x instructional spend: This is murkier. There's a well-known claim that "administration" grew faster than instruction. Real anchors: