The Mind's overall question: Do universities grow or shrink due to AI? Current facet: **Cost structure — AI reduces need for faculty/TA, tuition pressure**.

Looking at the structure of the conversation: each user message presents a facet and session substance, and I (as the note-filer) output one Markdown note. The last user message (the actual current turn) contains the current facet (cost structure) and the session substance, then asks me to write one Markdown note.

Wait — let me re-read the final message. The conversation shows many "Note N" entries in the human turn, but the actual final content is:

1. A note already filed (Note 8 in the transcript): "# AI cuts the teaching payroll first; the admin layer captures the savings" — dated claim about AI shrinking instructional labor costs faster than admin, savings retained internally.

2. Then: "Substance from this session (write a note only if this adds a durable claim):"
   - Cost structure — second-order effect: cost-plus tuition pricing dies first at the bottom of the market... predicted observable: tuition compression concentrated at non-selective four-year institutions and community-college competitors by ~2033, selective schools hold nominal rates; price map bifurcates rather than uniformly falling. Falsifier: broad sector-wide nominal tuition cuts (including selective privates) in the 2030s.
   - Cost structure — labor sequencing claim: AI displaces academic labor by contract protection, not capability. Least-protected layers (TAs, adjuncts, non-tenure-track) go first; tenured faculty contractually sticky, so instructional-cost relief is backloaded ~a decade. Observables: TA/adjunct postings collapse first (visible ~2028); teaching-stream headcount down 30–50% at AI-aggressive schools by ~2035; tenure lines survive where faculty carry grant/prestige revenue, shrink where they don't. Professoriate bifurcates into small protected research class + thin AI-supervised teaching layer — hollow middle disappears. Falsifier: sustained tenure-track growth at teaching-focused institutions after 2030, or AI-driven buyout waves hitting tenured ranks before 2032.

Now, what's already filed? The immediately preceding note (Note 8) is the "admin captures the savings" claim. That was presumably filed from an earlier session or earlier in this session. Actually wait — Note 8 appears in the human turn as a filed note, and then the session substance follows. So the filed note is the admin-capture claim. Hmm, but actually the structure suggests Note 8 was the note written in response to the previous turn, and now this turn presents new session substance with the current facet being cost structure.

Let me check what cost-structure claims have been filed across the notes:

From Note 8 (the filed note): AI cuts teaching payroll first; admin layer captures savings; tuition sticky; savings retained internally.

From Note 5's analysis (a previous session on cost structure): there was a rich cost-structure claim — Baumol's cost disease breaking, faculty/TA hours ~60–70% of delivery cost, marginal cost collapse, price dispersion, $6–12k/yr convergence, unbundling, "shrinks the integrated bundle but may grow credential-issuing units." Whether that was actually filed is ambiguous — Note 5's analysis was deliberating about whether to file it. The consolidation note (Note 6/7) suggests the synthesis included cost structure claims like "spending redirects, not shrinks" and the "instruction is minority slice" Delta Cost data (instruction ~25–30% at publics), residential bundle defection, plant costs $8–15k/student/yr.

So the filed corpus likely includes:
- The big Baumol-break claim (pricing rule breaks, $6–12k convergence, unbundling) — possibly filed
- "Spending redirects, not shrinks" — filed
- Instruction is minority slice (Delta Cost data) — filed
- Admin captures savings / teaching payroll cut first — just filed (Note 8)
- Tuition-deflation channel (residential bundle defection) — possibly filed

Now the new session substance has two claims:

**Claim A: cost-plus pricing dies first at the bottom of the market.** For intro-heavy undergraduate seats, AI grading/tutoring drives marginal delivery cost toward compute cost, so non-selective schools pricing on instructional cost lose their price floor. Elite schools never cost-plus (price on prestige/scarcity) so insulated. Predicted observable: tuition compression at non-selective four-year institutions and CC competitors by ~2033, selective schools hold nominal rates; price map bifurcates rather than uniformly falling. Falsifier: broad sector-wide nominal tuition cuts including selective privates in the 2030s.