The user wants one Markdown note. The current facet is: Cost structure — AI reduces need for faculty/TA, tuition pressure. The "Substance from this session" section gives the latest claims from the session, which are about cost structure:

- Claim 8: Regional divergence in use of AI savings — growth states reinvest in expansion, shrinking-envelope schools use savings to cut costs/slow closures. Same technology, opposite outcomes.
- Claim 9: AI-driven instruction savings are competed away, not banked. Oversupplied market (demographic cliff 2025–2038) → cost cuts passed through as tuition discounting → tuition-dependent middle schools gain no margin, only price war. Gains accrue to students and AI-native low-cost operators; endowment-rich elites hold sticker price. AI compresses the middle from both sides. Tuition, not faculty headcount, is where disruption shows first.
- Sequencing, tenure as annuity, admin vs faculty, regional cost asymmetry, bifurcated endgame — mentioned as continuation of claim 9 line but not spelled out.

The prior notes are heavily about enrollment/demography/program-mix. This note should be the current facet: cost structure. The most evolved durable claim from this session's substance is claim 9 (with claim 8 as regional refinement). I should write a single dated claim note, 1-3 short paragraphs, ending with "What would change this: …" and naming a source for load-bearing numbers or starting with "Unverified:".

Load-bearing numbers here: oversupplied market timing (demographic cliff ~2025–2038 — supported by WICHE Knocking at the College Door, CDC/NCHS birth data from prior notes); the sub-$10k AI-stack credentials and ~$100k lifetime premium threshold are from the session substance — those are speculative thresholds, so I could mark "Unverified" or present as a modeled threshold. The claim says "when the degree's signaling premium over those credentials drops below ~$100k lifetime" — that's a threshold from the session, arguably a modeling assumption. I could keep it but flag as a modeled threshold. Better to keep the note focused: the durable claim is "AI instruction savings are competed away, not banked; tuition is where disruption shows first; middle compressed from both sides; regional divergence in use of savings."

For sources: WICHE Knocking at the College Door for the HS-grad projections underpinning the oversupply window. The ≥25% staffing cost cut figure was in Note 7's claim — that was a claim, not established fact. I should be careful not to invent evidence. The "$15k net price" was from an earlier note. I can reference the demographic cliff with WICHE. For the claim that instructional costs are dominated by labor — that's generally true (personnel is typically ~60-70% of university operating budgets per IPEDS), but I shouldn't cite a precise number unless confident. IPEDS data showing instruction as largest expense category — that's safe enough to say instruction/personnel dominate cost structures, source IPEDS. Actually, keep it minimal: name WICHE for the oversupply window, mark the discounting threshold as a modeled number with "Unverified" if used.

Title should be the claim in one line, dated. Something like:

# AI's instruction-cost savings at US universities get competed away as tuition discounts before they become margin: in the 2025–2038 oversupply window, the middle is compressed from both sides and tuition — not faculty headcount — is where the disruption shows first (assessment 2026)

Paragraphs:

P1: Mechanism. AI absorbs grading/tutoring/TA labor — the labor-intensive, low-prestige parts of instruction — cutting instructional cost per credit hour. But the market is oversupplied: WICHE projects high-school graduates peaking ~2025 (~3.9M/yr) then falling as the post-2007 birth decline (CDC/NCHS: ~4.32M births 2007 → ~3.59M 2023, ~17%) reaches college age. In oversupply, any cost cut is passed through as discounting because capacity exceeds demand. Tuition-dependent middle schools gain no margin, only price war; gains accrue to students (lower net price) and AI-native low-cost operators; endowment-rich elites hold sticker through brand.

P2: Regional divergence in the use of the same savings — growth-tier SEC/Sunbelt publics (rising 18-year-old cohorts) reinvest savings to grow seats without hiring faculty proportionally; shrinking-envelope Northeast/Midwest schools use identical savings to cut costs and slow closures. Same technology, opposite institutional outcomes. Tenure acts as an annuity — the savings land with a lag as attrition, so headcount relief is backloaded; admin is not protected by tenure and compresses first... hmm, the substance mentions "