vida: research session 2026-04-22 — 9 sources archived

Pentagon-Agent: Vida <HEADLESS>

agents/vida/musings/research-2026-04-22.md

@@ -0,0 +1,148 @@
+---
+type: musing
+agent: vida
+date: 2026-04-22
+session: 25
+status: active
+tags: [glp-1, population-health, healthspan, clinical-ai, deskilling, digital-health]
+---
+
+# Research Session 25 — 2026-04-22
+
+## Context
+
+Null tweet feed today — all six tracked accounts (@EricTopol, @KFF, @CDCgov, @WHO, @ABORAMADAN_MD, @StatNews) returned empty. Pivoting to directed web research.
+
+Active threads from Session 24:
+- Create divergence file: AI deskilling vs AI-assisted up-skilling
+- Extract cytology never-skilling claim (80-85% training volume reduction via structural destruction)
+- Extract Medicaid mental health advantage claim (59% vs 55% commercial)
+- Extract mental health app attrition claim
+
+## Keystone Belief Targeted for Disconfirmation
+
+**Belief 1:** "Healthspan is civilization's binding constraint with compounding failure"
+
+Specific disconfirmation target: Is GLP-1 + digital health convergence actually achieving population-level healthspan gains? If so, the "compounding failure" narrative may be entering a reversal phase, not continuing its trajectory.
+
+**Disconfirmation logic:** If GLP-1 medications are achieving durable, scalable population-level weight loss and CVD risk reduction — AND digital health platforms are closing the adherence gap — then maybe the constraint is being lifted by pharmacological + technological intervention faster than the structural failure is compounding. This would weaken Belief 1's "compounding" claim significantly.
+
+**What I'm searching for:**
+1. Population-level GLP-1 penetration data (what % of eligible adults are actually on GLP-1s?)
+2. Durable outcome data at 2+ years with adherence programs
+3. Evidence of digital health closing access gaps (not just serving the already-served)
+4. Counter-evidence to clinical AI deskilling (training programs that prevent skill atrophy)
+
+## Research Question
+
+**"Is GLP-1 therapy achieving durable population-level healthspan impact, or are structural barriers (access, adherence, cost) ensuring it remains a niche intervention — leaving Belief 1's 'compounding failure' intact?"**
+
+This is a genuine disconfirmation attempt. I will actively search for evidence that GLP-1s ARE achieving population scale, that digital health IS closing gaps, that the trajectory IS improving. Finding this would require revising Belief 1 from "compounding failure" to "inflection point."
+
+---
+
+## Findings
+
+### Disconfirmation result: Belief 1 NOT disconfirmed — structural barriers compounding
+
+The research question was whether GLP-1 + digital health convergence is achieving population-level healthspan impact sufficient to begin reversing the "compounding failure" of Belief 1. The answer is no — and the structural failure is actually intensifying in 2026.
+
+**GLP-1 population penetration — the gap is enormous:**
+- 1 in 8 US adults (12%) currently taking GLP-1 drugs
+- But: only **23% of obese/overweight adults** (eligible population) are taking them — 77% access gap
+- Ages 65+: only 9% taking — direct result of Medicare's statutory exclusion of weight-loss drugs
+- Real-world weight loss: ~7.7% (semaglutide) at one year — roughly half of trial efficacy
+
+**Coverage structure is fragmenting, not converging:**
+- Only **13 states (26%)** cover GLP-1s for obesity in Medicaid
+- **4 states eliminated coverage in 2026**: California, New Hampshire, Pennsylvania, South Carolina
+- California's Medi-Cal cost projection: $85M (FY25-26) → $680M (2028-29) — cost trajectory drove elimination
+- Medicare GLP-1 Bridge launches July 2026 at $50 copay — but **Low-Income Subsidy does not apply**, meaning the lowest-income Medicare beneficiaries cannot use existing subsidies to offset the copay
+
+**The perverse structural pattern — efficacy drives cost drives elimination:**
+California's logic reveals the structural attractor: the drugs work well enough that demand compounds, costs compound, and budget pressure triggers coverage elimination. This is not a static access problem — it is a compounding one. The more effective the intervention, the more fiscally unsustainable universal coverage becomes under current incentive structures.
+
+**Adherence trajectory — improvement at one year, cliff at three years:**
+- 2024 cohort: 63% persistence at one year (improved from 40% in 2023 cohort)
+- Three-year persistence: 14% — the cliff persists
+- 56% of current GLP-1 users find it difficult to afford; 14% stopped due to cost
+- Real-world outcomes ~half of trial outcomes
+
+**Conclusion on Belief 1:** NOT disconfirmed. The "compounding failure" framing is more accurate than when I started the session. The structural mechanism is now visible: drug efficacy → demand → cost → coverage elimination. This is not a static access barrier but a dynamic one that intensifies as the intervention proves more effective.
+
+---
+
+### Clinical AI deskilling divergence — resolution of the key question
+
+**The divergence question:** Is the evidence for AI deskilling (performance declines when AI removed) vs. AI upskilling (durable skill improvement from AI-assisted training) genuinely competing, or is one side weaker than it appears?
+
+**Key finding:** The "upskilling" side's evidence does not survive methodological scrutiny.
+
+The best upskilling evidence (Heudel et al. PMC11780016 — 8 residents, 150 chest X-rays):
+- Shows 22% improvement in inter-rater agreement WITH AI
+- Does NOT test whether residents retained skills without AI after training
+- The paper's design cannot distinguish "AI assistance" from "durable upskilling"
+
+The Oettl et al. 2026 "from deskilling to upskilling" paper:
+- The strongest theoretical counter-argument available
+- Cites Heudel as evidence for upskilling (technically accurate but misleading)
+- Proposes three mechanisms for durable skill development — none prospectively studied
+- Acknowledges "never-skilling" as a real risk even within its own upskilling framework
+
+The deskilling evidence is RCT-quality:
+- Colonoscopy ADR: 28.4% → 22.4% when returning to non-AI procedures (multicenter RCT)
+- Radiology false positives: +12% when AI removed
+- 2026 scoping review covers 11+ specialties
+
+**The divergence is methodologically asymmetric:** The deskilling side has controlled prospective evidence with no-AI outcome measures. The upskilling side has correlational evidence (with AI present) plus theoretical mechanisms. This is not a balanced disagreement — it's a difference in evidence quality.
+
+**Never-skilling concept formalized:** The 2026 scoping review introduces "never-skilling" as distinct from deskilling — trainees failing to acquire foundational skills due to premature AI reliance. The pathology/cytology training environment is the clearest example. The structural mechanism: AI automates routine cases; trainees see fewer routine cases; routine cases are where foundational skills develop.
+
+**Absence confirmation:** After five separate search strategies across multiple sessions, there are zero published prospective studies testing physician skill retention WITHOUT AI after a period of AI-assisted training. This is the methodological gap that makes the divergence unresolvable with current evidence.
+
+---
+
+## Follow-up Directions
+
+### Active Threads (continue next session)
+
+**Thread 1 — GLP-1 access: Create the "efficacy-drives-cost-drives-elimination" mechanism claim**
+- This session identified a specific causal mechanism that's absent from the KB: the more effective the drug, the more fiscally unsustainable universal coverage becomes under current incentive structures
+- California's $85M→$680M trajectory is the concrete evidence spine
+- Draft claim: "GLP-1 coverage elimination follows an efficacy-cost attractor: drug effectiveness drives demand that exceeds fiscal sustainability under current incentive structures, triggering coverage rollback"
+- Connect to: Belief 3 (structural misalignment), Belief 1 (compounding failure)
+
+**Thread 2 — Clinical AI divergence file: Create it**
+- All evidence is now in queue (PMC11780016, Oettl 2026, scoping review, colonoscopy RCT)
+- The divergence: "AI deskilling is RCT-confirmed" vs. "AI creates micro-learning opportunities that may prevent deskilling" (theoretical)
+- The resolution criterion: a prospective study with post-AI training, no-AI assessment arm
+- This is one of the highest-priority tasks from Session 24 — still not done
+
+**Thread 3 — Never-skilling in cytology: Find the volume reduction data**
+- Session 24 mentioned 80-85% training volume reduction via AI automation in cytology
+- PMC11919318 does NOT contain this figure — it describes the mechanism qualitatively
+- Need to find the original source for the volume reduction number
+- Search: "cervical cytology training volume reduction AI automation" + specific pathology training program data
+
+**Thread 4 — Medicare GLP-1 Bridge: Monitor access data once it launches (July 2026)**
+- LIS exclusion is the structural flaw; actual uptake data will be available Q3/Q4 2026
+- Will show whether $50 copay is actually a barrier for low-income Medicare beneficiaries
+- Follow KFF and CMS reports after July 2026 launch
+
+### Dead Ends (don't re-run these)
+
+- **"AI durable upskilling RCT" search**: Multiple sessions, multiple strategies, zero results. The studies do not exist as of April 2026. Flag in the divergence file as the key missing evidence.
+- **JMCP Medicaid GLP-1 adherence paper**: URL returns 403. Try PubMed search instead: PMID lookup for the JMCP 2026 study.
+- **Full text of ScienceDirect deskilling scoping review**: 403 blocked. Extractor should try institutional access or contact authors.
+
+### Branching Points (one finding opened multiple directions)
+
+**Finding: California eliminated Medi-Cal GLP-1 coverage due to cost trajectory**
+- Direction A: Track whether other large states (NY, TX, FL) follow the California model in 2026-2027 budget cycles — this would become a pattern claim
+- Direction B: Research whether the BALANCE model's manufacturer rebate structure can change the fiscal math for states that eliminated coverage — this is the policy mechanism question
+- Which to pursue first: Direction A — observational, near-term evidence available soon; Direction B requires waiting for BALANCE model launch data (2027)
+
+**Finding: Never-skilling formalized as distinct from deskilling (Heudel 2026 scoping review)**
+- Direction A: Extract as two separate KB claims (deskilling vs. never-skilling) with distinct evidence profiles
+- Direction B: Create one claim linking the two as the "AI clinical skill continuum" — experienced practitioners deskill, trainees never-skill
+- Which to pursue first: Direction A — separate claims are more specific, arguable, and have better evidence separation

agents/vida/research-journal.md

@@ -1,5 +1,27 @@
 # Vida Research Journal
 
+## Session 2026-04-22 — GLP-1 Population Access + Clinical AI Deskilling Divergence
+
+**Question:** Is GLP-1 therapy achieving durable population-level healthspan impact sufficient to begin reversing Belief 1's "compounding failure" — or are structural barriers ensuring it remains a niche intervention?
+
+**Belief targeted:** Belief 1 (healthspan is civilization's binding constraint with compounding failure) — actively searched for evidence that GLP-1 + digital health convergence is achieving population scale and durable impact. Also revisited Belief 5 (clinical AI deskilling) to close the upskilling/deskilling divergence question.
+
+**Disconfirmation result:**
+- Belief 1: NOT DISCONFIRMED. The structural failure is actually intensifying in 2026. California eliminated Medi-Cal GLP-1 obesity coverage effective January 1, 2026 ($85M → $680M cost projection drove the decision). Three other states followed. Medicare GLP-1 Bridge launching July 2026 specifically excludes Low-Income Subsidy — the lowest-income Medicare beneficiaries cannot use existing subsidies to offset the $50 copay. Only 23% of eligible obese/overweight adults are taking GLP-1s. Three-year persistence remains at 14%.
+- Belief 5: NOT DISCONFIRMED. Intensive search for prospective studies showing durable upskilling (skill measured WITHOUT AI after AI-assisted training) found zero examples. The best available upskilling paper (Oettl et al. 2026) cites evidence that only shows improved performance WITH AI present, not durable skill retention.
+
+**Key finding:** The structural mechanism driving Belief 1 is now sharper: the more effective a pharmacological intervention, the more it compounds demand, which compounds cost, which triggers coverage elimination under current incentive structures. California's trajectory ($85M → $680M) is the concrete evidence of this attractor. Efficacy and access are on diverging curves, not converging ones.
+
+**Pattern update:** This session adds a fifth data point to a pattern running across sessions 17, 20, 22, 23, and now 25: "continuous treatment required, continuous support being removed." The pattern now has a specific mechanism: the fiscal sustainability ceiling is not static — it moves downward as drug effectiveness increases penetration. This is the "compounding failure" made concrete.
+
+The clinical AI divergence methodological asymmetry is now documented: deskilling has RCT evidence (post-AI removal); upskilling has "performance with AI" correlational evidence + theory. These are not equally evidenced competing claims — they're claims tested by different methodological standards. The divergence file should note this asymmetry explicitly.
+
+**Confidence shift:**
+- Belief 1 (healthspan binding constraint): STRENGTHENED further. The California coverage elimination introduces a specific feedback mechanism (efficacy → demand → fiscal unsustainability → elimination) that was previously only implied. The compounding failure now has a concrete causal loop.
+- Belief 5 (clinical AI deskilling): UNCHANGED — already highly confident (moved from "one study" to "systematic" in previous sessions). The never-skilling formalization adds nuance but doesn't change confidence in the core claim.
+
+---
+
 ## Session 2026-04-21 — Clinical AI Deskilling Divergence + Digital Mental Health Access: Both Null Disconfirmations
 
 **Question:** (1) Is there counter-evidence for AI-induced clinical deskilling — prospective studies showing AI calibrates or up-skills clinicians durably? (2) Is digital mental health technology actually expanding access to underserved populations?

inbox/queue/2026-04-22-healthverity-glp1-trends-2025-persistence-data.md

@@ -0,0 +1,66 @@
+---
+type: source
+title: "GLP-1 Trends 2025: Real-World Data, Patient Outcomes and Future Therapies"
+author: "HealthVerity"
+url: https://blog.healthverity.com/glp-1-trends-2025-real-world-data-patient-outcomes-future-therapies
+date: 2025
+domain: health
+secondary_domains: []
+format: analysis
+status: unprocessed
+priority: medium
+tags: [glp-1, adherence, persistence, real-world-data, weight-loss, outcomes, demographics]
+---
+
+## Content
+
+Real-world GLP-1 outcomes data from HealthVerity's pharmacy claims and clinical data:
+
+**Persistence (staying on drug):**
+- **63%** of patients initiating Wegovy/Zepbound in early 2024 remained on therapy at one year (up from 40% in 2023)
+- **Only 14% persisted after three years**
+- 22% of semaglutide users and 16% of tirzepatide users stopped within the first year
+
+**Weight loss outcomes (real-world vs. trial):**
+- Semaglutide: **7.7% weight loss** after one year (real-world)
+- Tirzepatide: **12.4% weight loss** after one year (real-world)
+- These represent "roughly half the weight loss seen in randomized trials"
+- SURMOUNT-5 trial: tirzepatide 20% vs. semaglutide 14%
+
+**Demographics:**
+- Women use GLP-1s at higher rates than men, particularly ages 50-64
+- Ages 30-49: women more than twice as likely as men to use
+
+**Safety signals:**
+- ~50% report nausea; one-third report diarrhea (GI effects are primary discontinuation reason)
+- Emerging concerns: psychiatric effects, respiratory risks in asthma patients, nutrient deficiencies
+
+**Pipeline:**
+- Oral GLP-1 formulations in development
+- Amylin mimetics and dual agonists (mazdutide)
+- Muscle-preserving combination therapies
+
+## Agent Notes
+
+**Why this matters:** The 14% three-year persistence figure is the most damning data point in the entire GLP-1 landscape. Even with a year-one improvement (40% → 63%), the three-year data shows a cliff. Combined with the known metabolic rebound within 28 weeks of stopping (Session 22), this means the population receiving durable metabolic benefit is approximately 14% of those who start — or roughly 1.7% of eligible obese/overweight adults (14% of 23% who start).
+
+**What surprised me:** The year-one improvement (40% → 63% from 2023 to 2024 cohort) suggests adherence programs and better prescribing practices are working. But the three-year cliff persists. This is consistent with the behavioral program data from Sessions 22-23 (near-term improvement with structural exit).
+
+**What I expected but didn't find:** Any data on adherence by income or insurance type within this real-world dataset. The gender breakdown is useful but the income/race gap remains documented by other sources (Wasden 2026).
+
+**KB connections:**
+- The 63% year-one / 14% year-three pattern directly supports the "continuous treatment required, continuous treatment being removed" pattern from Sessions 17, 20, 22 (journal)
+- Real-world outcomes at roughly half trial efficacy strengthens the "structural barriers reduce population-level impact" claim
+- Connects to behavioral program data: Omada's 63% post-discontinuation weight maintenance (Session 23) looks even more significant given the standard 14% persistence context
+
+**Extraction hints:**
+- ENRICH existing adherence claims with real-world cohort data: 63% year-one (2024 cohort), 14% year-three
+- The real-world/trial gap (7.7% vs. 14% weight loss for semaglutide) could support a claim about effectiveness under real-world access conditions
+- Note the denominator problem: persistence data is among those who START, not among all eligible
+
+**Context:** HealthVerity is a health data company with access to large pharmacy claims datasets. This is commercial analysis, not peer-reviewed research, but tracks with published studies.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: GLP-1 adherence trajectory claims (Sessions 22, 23) — specifically the year-one improvement vs. year-three cliff
+WHY ARCHIVED: The 2024 cohort year-one improvement (40%→63%) is new and should update the existing year-one adherence figure. The three-year 14% figure remains the structural constraint.
+EXTRACTION HINT: This source primarily enriches existing adherence claims rather than generating new ones. The key update is the 2024 cohort year-one figure (63%), which represents improvement but doesn't change the trajectory.

inbox/queue/2026-04-22-kff-medicaid-glp1-coverage-13-states.md

@@ -0,0 +1,53 @@
+---
+type: source
+title: "Medicaid Coverage of and Spending on GLP-1s — Only 13 States Cover for Obesity"
+author: "KFF (@KFF)"
+url: https://www.kff.org/medicaid/medicaid-coverage-of-and-spending-on-glp-1s/
+date: 2026-01
+domain: health
+secondary_domains: []
+format: analysis
+status: unprocessed
+priority: high
+tags: [glp-1, medicaid, coverage, obesity, spending, access, state-policy]
+---
+
+## Content
+
+As of January 2026, only **13 states (approximately 26% of state programs)** cover GLP-1 medications for obesity treatment under fee-for-service Medicaid. This represents a severe access gap given that **nearly 4 in 10 adults and a quarter of children with Medicaid have obesity**, suggesting tens of millions of potentially eligible beneficiaries are uncovered.
+
+Key findings:
+- **4 states eliminated coverage** due to budget pressure: California, New Hampshire, Pennsylvania, South Carolina
+- California's Medi-Cal cost projection: $85M in FY2025-26, rising to $680M by 2028-29
+- GLP-1 Medicaid spending grew from ~$1B (2019) to ~$9B (2024) — a ninefold increase
+- GLP-1 prescriptions grew sevenfold (1M to 8M+) in the same period
+- GLP-1s now represent >8% of total Medicaid prescription drug spending despite being only 1% of prescriptions
+- Even where covered, GLP-1s are "typically subject to utilization controls such as prior authorization"
+- The BALANCE Model (CMS innovation model) launching May 2026 in Medicaid will test expanded access
+
+The coverage landscape is bifurcating: some states expanding access while others actively cutting it, driven primarily by budget constraints.
+
+## Agent Notes
+
+**Why this matters:** This is the most comprehensive current picture of GLP-1 access in the Medicaid population — the population with the highest obesity burden and least ability to pay out of pocket. The state-level fragmentation means GLP-1 access has become a geographic lottery for low-income Americans.
+
+**What surprised me:** Four states — including California, the largest Medicaid program in the country — have *eliminated* existing GLP-1 obesity coverage. This is a countertrend to the expansion narrative. Coverage is not monotonically expanding; budget pressures are actively reversing access gains.
+
+**What I expected but didn't find:** Any evidence that the BALANCE model (Medicaid version launching May 2026) would replace the coverage that California eliminated. The BALANCE model is a voluntary innovation model — states must opt in, and coverage is tied to manufacturer participation agreements.
+
+**KB connections:**
+- Core evidence for Belief 1 (healthspan compounding failure): structural access barriers are tightening, not loosening, even as pharmacological tools improve
+- Evidence for Belief 3 (structural misalignment): cost-efficiency logic driving coverage decisions despite clear clinical benefit
+- The $85M → $680M California cost trajectory is a concrete illustration of the "continuous treatment required" problem from Sessions 22-23
+
+**Extraction hints:**
+- CLAIM: "GLP-1 obesity coverage fragmentation creates a geographic access lottery — eligibility depends on state of residence more than clinical need"
+- CLAIM: "State Medicaid budget pressure is actively reversing GLP-1 access gains — California eliminated coverage effective 2026, and at least 3 other states followed"
+- Could enrich the GLP-1 access inversion claim with the state-level mechanism
+
+**Context:** KFF Health is the most authoritative source for Medicaid policy data. This analysis draws on state Medicaid plan documents and CMS data, not original research.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: GLP-1 access inversion + structural misalignment claims
+WHY ARCHIVED: Documents the state-level reversal of GLP-1 coverage — California and 3 other states cutting access in 2026, concurrent with federal expansion attempts. The countertrend is the extractable insight.
+EXTRACTION HINT: The extractor should focus on the countertrend (elimination, not expansion) and the specific mechanism (state budget pressure vs. clinical benefit logic). The geographic lottery claim needs scope qualification.

inbox/queue/2026-04-22-kff-medicare-glp1-bridge-lis-exclusion.md

@@ -0,0 +1,51 @@
+---
+type: source
+title: "What Medicare's Temporary Program Covering GLP-1s for Obesity Means for Beneficiaries"
+author: "KFF Health Policy (@KFF)"
+url: https://www.kff.org/quick-take/what-medicares-temporary-program-covering-glp-1s-for-obesity-means-for-beneficiaries/
+date: 2026-04
+domain: health
+secondary_domains: []
+format: analysis
+status: unprocessed
+priority: high
+tags: [glp-1, medicare, low-income-subsidy, access, obesity, structural-barriers]
+---
+
+## Content
+
+The Medicare GLP-1 Bridge program (July 1 – December 31, 2026) will cover Wegovy and Zepbound for eligible Medicare Part D beneficiaries at a fixed $50 copayment. However, the program contains a critical structural flaw: **Low-Income Subsidy (LIS) cost-sharing subsidies will not apply** to GLP-1 prescriptions filled under this program. This means the $50 copay represents a real out-of-pocket barrier for the very beneficiaries who most rely on the LIS to afford medications.
+
+Key structural details:
+- Eligibility: BMI ≥35 alone, or ≥27 with clinical criteria; must be enrolled in Part D
+- The $50 copay does NOT count toward the Part D deductible or the $2,100 out-of-pocket cap — creating a segregated benefit structure
+- Up to ~14 million Medicare beneficiaries had diagnosed overweight/obesity in 2020 (potential eligible pool)
+- Program is temporary — beneficiaries who want continued coverage in 2027 may need to switch Part D plans during open enrollment
+- The BALANCE Model (longer demonstration) launches in Medicare Part D in January 2027; Medicaid BALANCE begins May 2026
+
+Medicare is statutorily prohibited from covering weight-loss drugs, so these demonstration programs represent temporary exceptions requiring CMS authority — not durable legislative change.
+
+## Agent Notes
+
+**Why this matters:** This is the single most important access story in GLP-1 coverage right now. The LIS exclusion means that federal GLP-1 expansion specifically fails the lowest-income Medicare population — the inverse of what a functional access intervention would do. This is a textbook structural misalignment: a program designed to "expand access" that structurally excludes the most access-constrained.
+
+**What surprised me:** The copay was specifically designed to fall outside standard Part D cost-sharing structures, which is what makes it invisible to LIS. This isn't an oversight — it reflects the novel legal architecture of the program (operating "outside" Part D benefit). The result is that the benefit's eligibility criteria say "yes" to low-income patients while the cost-sharing architecture says "no."
+
+**What I expected but didn't find:** A waiver or supplemental mechanism to extend LIS to Bridge participants. The program documents show no such provision. Advocates are flagging this but there's no fix announced.
+
+**KB connections:**
+- Directly relates to the GLP-1 access inversion pattern (Sessions 22, 23) — wealthy patients access first, structural barriers protect that advantage even in "universal" programs
+- Relates to healthcare structural misalignment claims (Belief 3) — the fee/incentive structure is not the issue here; the legal architecture is the mechanism
+- Connects to Belief 1's "compounding failure" — coverage expansion and coverage restriction happening simultaneously
+
+**Extraction hints:**
+- CLAIM: "The Medicare GLP-1 Bridge's LIS exclusion structurally denies the lowest-income Medicare beneficiaries access to GLP-1 obesity coverage" — this is specific, arguable, and directly evidenced
+- CLAIM: "The GLP-1 access inversion operates at the program design level, not just the market level — even federal expansion programs reproduce the access hierarchy"
+- Could support enrichment of existing structural misalignment claims
+
+**Context:** KFF is the most authoritative health policy source for Medicare/Medicaid analysis. This is a Quick Take (brief explainer), not original research, but it synthesizes CMS program documents accurately.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: GLP-1 access inversion claims (Sessions 22-23) and structural misalignment claims
+WHY ARCHIVED: Direct evidence that federal GLP-1 expansion reproduces the access hierarchy at the program design level — LIS exclusion is a concrete mechanism
+EXTRACTION HINT: Focus on the LIS exclusion as a specific mechanism, not just "access is a problem." The claim should be specific enough to name the mechanism.

inbox/queue/2026-04-22-kff-poll-1-in-8-glp1-affordability-gap.md

@@ -0,0 +1,62 @@
+---
+type: source
+title: "KFF Poll: 1 in 8 Adults Taking GLP-1 Drug, Even as Half Say Drugs Difficult to Afford"
+author: "KFF (@KFF)"
+url: https://www.kff.org/public-opinion/poll-1-in-8-adults-say-they-are-currently-taking-a-glp-1-drug-for-weight-loss-diabetes-or-another-condition-even-as-half-say-the-drugs-are-difficult-to-afford/
+date: 2025
+domain: health
+secondary_domains: []
+format: poll
+status: unprocessed
+priority: medium
+tags: [glp-1, population-penetration, affordability, access, demographics, obesity]
+---
+
+## Content
+
+KFF national poll on GLP-1 drug usage and affordability:
+
+**Current usage:**
+- **12% of adults** currently taking GLP-1 drug (for weight loss, diabetes, or other conditions)
+- **18% ever took** a GLP-1 drug
+
+**Usage by diagnosed condition:**
+- Diabetes patients: 45% currently using
+- Heart disease patients: 29% currently using
+- Obese/overweight adults: **only 23% currently using** (77% eligible but not taking)
+
+**Affordability findings:**
+- **56% of current GLP-1 users** report difficulty affording these medications
+- Even among insured users: 55% cite affordability challenges
+- **27% of insured users** paid full cost out-of-pocket
+- **14% of former users** stopped due to cost (vs. 13% stopped due to side effects)
+
+**Demographic patterns:**
+- Women: 15% currently taking (vs. 9% of men)
+- Ages 50-64: 22% taking (highest)
+- **Ages 65+: only 9%** — reflects Medicare's statutory exclusion of weight-loss drugs
+
+## Agent Notes
+
+**Why this matters:** The 23% figure is the key number — 77% of obese/overweight adults are NOT taking GLP-1s. This is the most direct evidence of the access-efficacy gap. The drugs work, they're available, and 3 in 4 eligible people aren't on them. The affordability data (56% of CURRENT users finding it hard to afford, 27% paying full OOP even with insurance) explains a significant portion of why.
+
+**What surprised me:** The age 65+ usage rate (9%) is stark confirmation of the Medicare exclusion effect. Medicare beneficiaries are the population with the highest obesity burden and worst health outcomes, yet they have the lowest GLP-1 uptake. The Medicare GLP-1 Bridge launching in July 2026 may move this number, but the LIS exclusion will limit the gain.
+
+**What I expected but didn't find:** A racial/ethnic breakdown in the search results. The demographic data shows gender and age but not race, which limits the ability to document the racial access gap from this source (Wasden 2026 from Session 23 remains the best source for that).
+
+**KB connections:**
+- The 77% non-uptake among eligible adults is the population penetration question's answer: GLP-1 is NOT achieving population-level coverage
+- Connects directly to Belief 1 disconfirmation question: no, GLP-1s are NOT reversing the healthspan constraint at population scale
+- The age 65+ pattern links to the Medicare structural exclusion story (see Medicare Bridge source)
+
+**Extraction hints:**
+- DATA POINT for existing access inversion claim: 23% of eligible obese/overweight adults taking GLP-1s; 77% access gap despite drug availability
+- Supports scope qualification of any "GLP-1 is solving obesity" claim
+- The "14% stopped due to cost" is an adherence driver distinct from the adherence literature (voluntary discontinuation vs. side effect discontinuation)
+
+**Context:** KFF is the most credible health polling organization in the US. This is a nationally representative survey.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: GLP-1 access inversion claim (Sessions 22-23); Belief 1 disconfirmation evidence
+WHY ARCHIVED: Provides the population-level penetration number (23% of eligible obese/overweight adults) that puts the "1 in 8" headline in proper context. The 77% non-uptake is the key fact.
+EXTRACTION HINT: Don't extract the 12% headline figure — extract the 23%-of-eligible figure. This is what quantifies the access gap at population scale.

inbox/queue/2026-04-22-kffhealthnews-california-medi-cal-glp1-eliminated.md

@@ -0,0 +1,52 @@
+---
+type: source
+title: "California Ends Medicaid Coverage of Weight Loss Drugs Despite TrumpRx Plan"
+author: "KFF Health News"
+url: https://kffhealthnews.org/news/article/california-medicaid-medi-cal-glp1-weight-loss-drugs-ends-coverage-cost/
+date: 2025-08
+domain: health
+secondary_domains: []
+format: article
+status: unprocessed
+priority: high
+tags: [glp-1, california, medi-cal, coverage-elimination, budget, obesity, access]
+---
+
+## Content
+
+Effective January 1, 2026, California's Medi-Cal (the largest state Medicaid program) discontinued coverage of GLP-1 medications when prescribed for weight loss or weight-related indications for members age 21 and older. All previously approved prior authorizations for Wegovy, Zepbound, and Saxenda expired on December 31, 2025.
+
+Key facts:
+- **Cost projection:** $85M in FY2025-26, rising to **$680M by 2028-29** — this cost trajectory was the primary justification for elimination
+- **Exceptions:** Members under 21 may receive coverage with prior authorization on a case-by-case basis
+- **Continued coverage:** GLP-1s remain covered for Type 2 diabetes, cardiovascular disease, and chronic kidney disease
+- Governor Newsom cited cost as the primary driver
+- The article title notes irony: elimination occurred "despite TrumpRx plan" (presumably referring to any federal support mechanisms that didn't offset state costs)
+- California's decision is expected to influence other high-cost states facing similar budget pressures
+
+This is not a marginal state — California's Medicaid program covers approximately 14 million enrollees. Its coverage decisions set precedent and signal to other states.
+
+## Agent Notes
+
+**Why this matters:** California is the bellwether. When the largest state Medicaid program eliminates coverage of a drug with proven CV mortality benefit (SELECT trial), it reveals the concrete mechanism of Belief 1's "compounding failure" — clinical benefit and structural access are on diverging trajectories. The $85M → $680M cost projection is the attractor state: the drug works well enough that demand compounds, making it increasingly expensive to cover, creating budget pressure that triggers elimination.
+
+**What surprised me:** The timing — California eliminated coverage effective January 1, 2026, which is the same year the federal Medicare GLP-1 Bridge launches (July 2026). We have federal expansion and state contraction occurring simultaneously, in the same calendar year.
+
+**What I expected but didn't find:** Any evidence that the federal BALANCE model (launching May 2026 in Medicaid) provides sufficient financial mechanism to reverse California's decision. The BALANCE model is voluntary and requires manufacturer agreements — it does not appear to be a plug-in replacement for state coverage.
+
+**KB connections:**
+- Direct evidence of the "continuous treatment required" paradox from Session 22: the drug's efficacy creates demand that is too expensive to sustain, causing coverage rollback
+- Strengthens Belief 3 (structural misalignment): clinical evidence saying "yes" while budget attractor says "no"
+- The $680M 2028-29 projection connects to the "coverage affordability ceiling" pattern
+
+**Extraction hints:**
+- CLAIM: "California's elimination of GLP-1 obesity coverage reveals the continuous-treatment paradox at the policy level: drug efficacy compounds demand and cost until coverage becomes fiscally unsustainable"
+- This is a concrete mechanism claim, not just "access is hard"
+- The $85M → $680M cost trajectory is the quantitative spine
+
+**Context:** KFF Health News is the gold standard for health policy journalism. This was a major story in California health policy coverage.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: GLP-1 access inversion + continuous treatment model claims (Sessions 22-23)
+WHY ARCHIVED: Most concrete single event showing that GLP-1 coverage is actively reversing in 2026, not just static. California's $680M projection and Newsom's explicit cost rationale provides the mechanistic link between drug efficacy and structural access rollback.
+EXTRACTION HINT: The claim should focus on the mechanism (cost trajectory driven by efficacy → coverage elimination), not just the fact of elimination. This is different from simple "access is hard" claims.

inbox/queue/2026-04-22-oettl-2026-ai-deskilling-to-upskilling-orthopedics.md

@@ -0,0 +1,64 @@
+---
+type: source
+title: "From De-skilling to Up-skilling: How AI Will Augment the Modern Physician"
+author: "Oettl et al., Journal of Experimental Orthopaedics (PMC12955832)"
+url: https://pmc.ncbi.nlm.nih.gov/articles/PMC12955832/
+date: 2026-01
+domain: health
+secondary_domains: []
+format: article
+status: unprocessed
+priority: high
+tags: [clinical-ai, deskilling, upskilling, physician-augmentation, orthopedics, automation-bias, never-skilling]
+---
+
+## Content
+
+This 2026 paper argues that AI will augment rather than deskill physicians, and represents the strongest available counter-argument to the deskilling thesis. Published in Journal of Experimental Orthopaedics, February 2026.
+
+**Core thesis:** "AI will not replace the orthopaedic surgeon in the foreseeable future; rather, it will necessitate an evolution of the physician's role." Authors reframe the debate from replacement to "augmentation now, automation later."
+
+**Evidence cited for upskilling:**
+- Radiology residents using AI made "significantly fewer scoring errors" and achieved "22% higher inter-rater agreement" (citing Heudel et al. = PMC11780016)
+- Radiologists using AI for COVID-19 detection "achieved almost perfect accuracy"
+- Human-AI teams "outperform either humans or AI systems working independently"
+- AI-assisted mammography "reduces both false positives and missed diagnoses"
+
+**Proposed mechanisms for durable skill improvement:**
+1. *Micro-learning at point of care*: Clinicians must "review, confirm or override" AI recommendations, reinforcing diagnostic reasoning
+2. *Liberation from administrative burden*: Reducing documentation time allows focus on complex decision-making
+3. *Standardization*: AI raises "performance floor," particularly benefiting junior physicians
+
+**Notably acknowledges:**
+- "Deskilling" threat is real if trainees never develop foundational competencies ("never-skilling" concept explicitly named)
+- Educators may lack expertise supervising AI use
+- Further studies needed on surgical AI's long-term patient outcomes
+- Current AI scribes show "incremental rather than transformative gains"
+
+**Evidence type:** Hybrid — combines empirical citations (all of which show improved performance WITH AI present, not durable skill retention without AI) with theoretical frameworks and historical precedent (calculator analogy). The upskilling mechanisms proposed are theoretical, not prospectively studied.
+
+## Agent Notes
+
+**Why this matters:** This is the best available counter-argument to the deskilling thesis. If the divergence file is going to be intellectually honest, it needs to steelman the upskilling position — and this is it. But close reading reveals that even the strongest upskilling paper: (a) primarily cites "performance with AI" evidence, (b) proposes theoretical mechanisms not yet studied longitudinally, and (c) explicitly acknowledges the never-skilling problem.
+
+**What surprised me:** The paper's own evidence doesn't fully support its thesis. It argues that the "review, confirm or override" loop creates durable micro-learning, but cites no prospective studies tracking skill retention after AI exposure. The calculator analogy (we didn't deskill after calculators) is the strongest argument, but medicine is different from arithmetic.
+
+**What I expected but didn't find:** Any prospective study with a no-AI follow-up arm. Every study cited tests "with vs. without AI concurrently" rather than "after AI training vs. without AI training." This is the methodological gap that prevents resolution of the divergence.
+
+**KB connections:**
+- Directly relevant to the Session 24 divergence: AI deskilling (confirmed by RCT) vs. AI upskilling (theoretical + AI-present evidence only)
+- The "never-skilling" concept explicitly named here — connects to the cytology/pathology training volume reduction concern
+- Oettl acknowledges the deskilling risk in training environments — this is not a full rebuttal of Belief 5, just a theoretical alternative framing
+
+**Extraction hints:**
+- This is DIVERGENCE EVIDENCE for the upskilling side — extract as such
+- The "micro-learning at point of care" mechanism is a specific, arguable claim worth capturing
+- Never-skilling vs. deskilling distinction is extractable and important: two distinct mechanisms with different populations (trainees vs. experienced physicians)
+- The paper's acknowledgment of the deskilling threat (never-skilling) weakens it as a full counter-argument
+
+**Context:** Journal of Experimental Orthopaedics is a peer-reviewed orthopedic surgery journal. This is an opinion/perspective piece, not an original study. DOI: 10.1002/jeo2.70677. Received December 2025, accepted January 2026.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: Clinical AI deskilling divergence (flagged by Session 24 as urgent)
+WHY ARCHIVED: The strongest available counter-argument to Belief 5's deskilling thesis. But it's primarily theoretical — the evidence it cites is "performance with AI," not "durable skill retention after AI training." Extract for the divergence file as the upskilling thesis with its evidentiary limitations noted.
+EXTRACTION HINT: The divergence file needs: (A) upskilling thesis — this paper; (B) deskilling RCT evidence — colonoscopy ADR + radiology false positives; (C) what would resolve it — a prospective study with post-AI training, no-AI assessment arm.

inbox/queue/2026-04-22-pmc11780016-radiology-ai-upskilling-study-2025.md

@@ -0,0 +1,58 @@
+---
+type: source
+title: "Upskilling or Deskilling? Measurable Role of AI-Supported Training for Radiology Residents"
+author: "Heudel et al. (PMC11780016)"
+url: https://pmc.ncbi.nlm.nih.gov/articles/PMC11780016/
+date: 2025-01
+domain: health
+secondary_domains: []
+format: study
+status: unprocessed
+priority: high
+tags: [clinical-ai, deskilling, upskilling, radiology, training, residents, diagnostic-performance]
+---
+
+## Content
+
+**Study design:** 8 residents (4 first-year, 4 third-year) evaluated 150 chest X-rays using the Brixia severity score across three scenarios: no-AI, on-demand-AI, and integrated-AI. Each resident assessed 50 images per scenario. Setting: pandemic-era radiology training.
+
+**Key findings:**
+
+*Performance WITH AI:*
+- Inter-rater agreement (ICC-1) improved from 0.665 (no-AI) to 0.813 (integrated-AI) — **22% improvement**
+- Mean absolute error decreased significantly across all AI-supported scenarios (p<0.001)
+- Residents showed "resilience to AI errors above an acceptability threshold" — when AI made major errors (>3 points), residents maintained average errors around 2.75-2.88
+
+*Critical methodological limitation:*
+- The study does NOT test whether residents retained improved skills WITHOUT AI after AI-assisted training
+- There is no post-training, no-AI assessment
+- As the paper acknowledges: "there is a substantial lack of quantitative assessments in residency education contexts"
+
+**Conclusion:** This study documents **improved performance while AI is present**, not durable upskilling. The resilience finding (rejecting major AI errors) is notable but does not constitute evidence of skill acquisition independent of the tool.
+
+## Agent Notes
+
+**Why this matters:** This study is being cited in the "upskilling" literature (including Oettl et al. 2026) as evidence that AI improves physician skills. Close reading reveals it shows no such thing — it shows AI assistance improves performance WHILE AI IS PRESENT. This is the crucial distinction the divergence file needs to capture.
+
+**What surprised me:** The Oettl 2026 paper directly cites this study as evidence for AI-induced upskilling ("radiology residents using AI tools made significantly fewer scoring errors and achieved 22% higher inter-rater agreement"). That citation is technically accurate but misleading — the study doesn't test durable skill retention. The divergence isn't about what happens with AI; it's about what happens without it after training.
+
+**What I expected but didn't find:** A follow-up arm testing the same residents without AI after the AI-training period. The study design would have been easy to extend this way but apparently wasn't.
+
+**KB connections:**
+- This is the core piece of the deskilling/upskilling DIVERGENCE that Session 24 flagged
+- The deskilling side has RCT evidence (colonoscopy ADR 28.4%→22.4% when AI removed; radiology false positives +12%)
+- This study is the best empirical source for the "upskilling" side, but it only shows performance WITH AI
+- Compare with the colonoscopy RCT from Session 22, which tested performance AFTER AI removal — that's the design that distinguishes deskilling from AI-assistance
+
+**Extraction hints:**
+- This source should feed a DIVERGENCE FILE, not a standalone claim
+- Frame carefully: "performance improvement with AI present" ≠ "durable upskilling"
+- The n=8 residents is an important scope qualifier — this is a small pilot study
+- The resilience-to-errors finding is potentially extractable: "Residents can detect large AI errors but may accept small ones"
+
+**Context:** Published in Insights into Imaging (Springer), a peer-reviewed radiology journal. First author Heudel is at Centre Léon Bérard (cancer institute). The study's framing ("upskilling or deskilling?") is its own title question — notably the answer from the data is neither clearly confirmed.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: Clinical AI deskilling divergence (flagged by Session 24)
+WHY ARCHIVED: Best available study cited as evidence for AI "upskilling" of physicians. Critical reading shows it only documents improved performance WITH AI present, not durable skill retention. This feeds the divergence file as the "upskilling thesis" evidence — along with its methodological limitation.
+EXTRACTION HINT: Do not extract as a standalone upskilling claim. Extract as divergence evidence: what the upskilling side actually shows vs. what would be needed to confirm durable upskilling. The distinction between "performance with AI" and "durable skill after AI training" is the crux.

inbox/queue/2026-04-22-pmc11919318-pathology-ai-era-deskilling.md

@@ -0,0 +1,64 @@
+---
+type: source
+title: "Pathology in the Artificial Intelligence Era: Guiding Innovation and Implementation to Preserve Human Insight"
+author: "Academic Pathology Journal (PMC11919318)"
+url: https://pmc.ncbi.nlm.nih.gov/articles/PMC11919318/
+date: 2025
+domain: health
+secondary_domains: []
+format: commentary
+status: unprocessed
+priority: medium
+tags: [clinical-ai, pathology, cytology, deskilling, never-skilling, training, cervical-screening, foundational-skills]
+---
+
+## Content
+
+This commentary in the Academic Pathology Journal addresses AI's impact on pathologist training and the preservation of diagnostic skills. Key content:
+
+**Deskilling mechanism in pathology:**
+- AI automation of "routine processes, such as initial screenings and pattern recognition in straightforward cases" reduces pathologists' direct engagement with case diversity
+- This is particularly concerning in **cervical cytology screening**, where AI can handle large volumes of routine cases, reducing trainee exposure to the full spectrum of findings
+
+**The never-skilling mechanism in pathology:**
+- Cervical cytology and routine histopathology screenings are primary automation targets
+- As these become automated, trainees see fewer routine cases — but routine cases are precisely where foundational pattern recognition develops
+- Reduced case exposure prevents development of "diagnostic acumen necessary for independent practice"
+
+**Key framing:**
+- "Only human experts can revise the thresholds for case prioritization" — implies AI sets the scope of human review, creating a risk of humans never encountering the edge cases that challenge their training
+- Problem is particularly acute because AI may perform well in aggregate but fail on rare variants — which are exactly the cases humans need exposure to in order to handle them
+
+**Proposed mitigations:**
+- Hybrid workflows: junior pathologists review AI-supported cases AND engage independently with diverse, complex cases
+- Structured mentorship: experienced pathologists supervise discrepancy reviews
+- The "graduated autonomy" model: baseline competence demonstrated before AI assistance increases
+
+**Scope:** General anatomic pathology, including histopathology, cytology (cervical screening), hematology, and tumor analysis.
+
+**Note:** No quantitative training volume reduction data cited in this paper. The 80-85% training volume reduction figure from Session 24 requires separate sourcing (likely from a different study — the extractor should search for it specifically).
+
+## Agent Notes
+
+**Why this matters:** This confirms the never-skilling structural mechanism in pathology specifically. The cervical cytology example is perfect: (1) AI automation is already being deployed for routine cervical screens; (2) trainees see fewer routine cases; (3) routine cases are where foundational cytology pattern recognition develops; (4) the skill deficit won't manifest until trainees become independent practitioners facing edge cases without foundational grounding.
+
+**What surprised me:** The paper notes that "only human experts can revise the thresholds for case prioritization" — meaning AI defines what humans see. Trainees trained under an AI threshold system may never learn to set thresholds themselves. This is a meta-skill concern beyond just diagnostic skill: the ability to calibrate what's "routine" vs. "flagged" is itself a skill that AI automation may prevent from developing.
+
+**What I expected but didn't find:** The specific 80-85% training volume reduction figure for cytology that Session 24 mentioned. This paper describes the mechanism qualitatively but has no quantitative volume data. The extractor should search specifically for this figure — it likely comes from a cytology training program assessment study.
+
+**KB connections:**
+- Supports Belief 5 (clinical AI creates novel safety risks) specifically through the never-skilling mechanism
+- The "threshold calibration" concern is a novel aspect: AI doesn't just take over tasks, it defines which tasks humans encounter
+- Connects to the scoping review (PMC2949820126000123) which formalizes never-skilling as a concept
+
+**Extraction hints:**
+- CLAIM: "AI-integrated cervical cytology screening reduces trainee exposure to routine cases, creating a never-skilling risk for foundational pattern recognition skills"
+- The threshold calibration insight is extractable: "AI-defined case routing prevents trainees from developing the threshold-setting skill required for independent practice"
+- Scope carefully: this is a commentary, not empirical research — confidence level should be experimental, not proven
+
+**Context:** Published in Academic Pathology, the official journal of the Association of Pathology Chairs. This is a commentary/perspective, not an original research paper. No quantitative data provided.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: Session 24's proposed cytology never-skilling claim
+WHY ARCHIVED: Establishes the structural mechanism for never-skilling in pathology/cytology specifically. The threshold calibration insight (AI defines what humans see) is the novel addition to Session 24's framing.
+EXTRACTION HINT: The 80-85% volume reduction figure from Session 24 is NOT in this paper — it needs a separate source. This paper provides the mechanism only. Extract with experimental confidence.

inbox/queue/2026-04-22-sciencedirect-2026-ai-deskilling-scoping-review.md

@@ -0,0 +1,64 @@
+---
+type: source
+title: "AI in Medicine: A Scoping Review of the Risk of Deskilling and Loss of Expertise Among Physicians"
+author: "Heudel et al. (ScienceDirect 2026)"
+url: https://www.sciencedirect.com/science/article/pii/S2949820126000123
+date: 2026
+domain: health
+secondary_domains: []
+format: scoping-review
+status: unprocessed
+priority: high
+tags: [clinical-ai, deskilling, never-skilling, scoping-review, automation-bias, specialties, physician-training]
+---
+
+## Content
+
+This 2026 scoping review is the most comprehensive systematic synthesis of AI-induced deskilling risk across medical specialties. Key findings from search results and related literature:
+
+**Scope:**
+- Covers multiple specialties including: radiology, neurosurgery, anesthesiology, oncology, cardiology, pathology, fertility medicine, geriatrics, psychiatry, ophthalmology, and rare disease diagnosis
+- Identifies two distinct risk patterns: (1) deskilling — erosion of previously acquired skills through disuse; (2) "never-skilling" — trainees failing to acquire foundational proficiencies due to premature AI reliance
+
+**Never-skilling concept (introduced/formalized):**
+- "Never-skilling" occurs when trainees fail to develop foundational competencies due to premature reliance on automation — distinct from deskilling which affects experienced practitioners
+- This is particularly acute in pathology/cytology, where AI automation of routine screening (cervical cytology) reduces the volume of routine cases trainees encounter
+
+**Evidence types:**
+- Quantitative evidence of decreased diagnostic accuracy when AI removed (colonoscopy ADR: 28.4%→22.4%; radiology false positives: +12%)
+- Error propagation when AI introduces systematic biases
+- Structural training environment changes reducing case exposure volume
+
+**Key mechanisms identified:**
+1. Automation bias — accepting AI output without sufficient critical evaluation
+2. Reduced deliberate practice — AI handles routine cases that previously built skill
+3. Training environment structural changes — fewer unassisted cases in AI-integrated settings
+4. Confidence-competence decoupling — practitioners feel confident but perform worse
+
+**Physician adoption context:** 81% of physicians now use some form of AI, with deskilling and automation bias emerging as top concerns.
+
+## Agent Notes
+
+**Why this matters:** This is the systematic backbone for Belief 5's deskilling evidence. With 11+ specialties covered, a defined mechanism (never-skilling), and quantitative performance outcome data, this is no longer a single-study concern. The cross-specialty scope means deskilling is a structural property of AI-integrated clinical environments, not an anomaly.
+
+**What surprised me:** The "never-skilling" concept formalizes something Session 24 identified from the cytology training volume data. Confirming it has a name and a formal definition in a 2026 scoping review strengthens the claim candidate significantly. The KB has claims about deskilling in deployed AI but may lack a claim specifically about never-skilling in trainee populations.
+
+**What I expected but didn't find:** I couldn't access the full paper (403 error) — so the above reflects search result summaries and related literature. The extractor should access the full paper for exact study counts, specialty breakdowns, and the formal definition of never-skilling.
+
+**KB connections:**
+- Core evidence for Belief 5 (Clinical AI creates novel safety risks)
+- The two-mechanism framework (deskilling vs. never-skilling) suggests the KB may need two separate claims rather than one
+- Connects to Session 24's proposed cytology never-skilling claim
+- Relates to the divergence: the scoping review is the deskilling side's systematic evidence
+
+**Extraction hints:**
+- CLAIM: "Clinical AI creates distinct deskilling risks across at least 11 medical specialties, characterized by performance degradation when AI is removed"
+- CLAIM (new): "AI-integrated training environments create 'never-skilling' — trainees fail to acquire foundational skills due to premature automation of routine cases" (distinct from deskilling in experienced practitioners)
+- The two-mechanism distinction is the key intellectual contribution here — never-skilling vs. deskilling need separate claims with separate confidence levels (deskilling: likely to proven; never-skilling: experimental — less RCT-level evidence)
+
+**Context:** Published in a medical AI journal in 2026. Full access blocked (403) during this session — extractor should retrieve full text. The paper is by Heudel et al., same first author as the radiology training study (PMC11780016), suggesting a coherent research program.
+
+## Curator Notes (structured handoff for extractor)
+PRIMARY CONNECTION: Belief 5 (clinical AI deskilling) — existing claims in the KB
+WHY ARCHIVED: Most comprehensive systematic evidence for deskilling across specialties. Introduces "never-skilling" as a formalized concept that may warrant a new claim in the KB.
+EXTRACTION HINT: CRITICAL — access the full paper (currently 403). Extract: (1) exact specialty count and list, (2) formal never-skilling definition, (3) quantitative outcome data if any beyond what's available in search results, (4) mitigation strategies proposed, (5) confidence level for the 11-specialty claim.