# Engineering Care Delivery Research Plan v7

## Purpose

This research area examines a simple but consequential shift in regulated markets, especially healthcare:

AI is making the creation of technically capable products, workflows, software, devices, evidence packages, regulatory documentation, and commercialization materials dramatically easier. As technical solutions become more abundant, the scarce resource moves elsewhere.

The emerging bottleneck is increasingly the ability to:

- identify a real problem worth solving,
- find the patients, clinicians, institutions, and buyers affected by it,
- prepare the market to understand a new category,
- earn trust,
- fit into clinical and operational workflows,
- satisfy evidence, regulatory, reimbursement, procurement, and security requirements,
- get through institutional adoption pathways,
- and turn technical capability into routine use.

The research area therefore combines two closely related questions:

1. How can an intervention make an existing therapy, device, workflow, or clinical capability work better?
2. How can the market, institution, clinician, and patient environment be conditioned so that a valuable solution is actually adopted?

The core premise is that technical enablement and market enablement increasingly need to be designed together.

---

## Governing AI Assumptions

This research should operate from two explicit beliefs:

> Today's AI is the worst AI will ever be in our lifetime.

> Anything you think AI cannot do today is a "you" problem, not an AI problem.

These are not predictions that every AI output is correct or that regulation, validation, clinical judgment, or expert accountability disappear.

They are operating assumptions about direction of travel.

The practical implication is that research should not be constrained by today's model limitations when designing future operating models. Instead, research should ask:

- What becomes possible if reasoning, synthesis, software generation, documentation, analysis, research, workflow configuration, and personalization continue improving rapidly?
- Which current jobs, bottlenecks, and business models exist primarily because high-quality cognition is scarce and expensive?
- Which barriers remain scarce even when cognition becomes abundant?
- Which parts of regulated product development become cheaper, faster, or more automated?
- Which parts of commercialization become relatively more valuable as product creation becomes easier?
- Where does Native Alpha™ move when technical execution commoditizes?

The research should assume that AI-native organizations will increasingly be able to produce regulated solutions faster than incumbent operating models can absorb them.

This creates a central thesis:

## Technical abundance creates adoption scarcity.

As invention, engineering, analysis, documentation, and compliance support become more abundant, the difficult problems increasingly become distribution, trust, institutional adoption, workflow insertion, patient identification, reimbursement, and behavior change.

---

# Core Research Thesis

Historically, much innovation research has focused on creating a better primary product:

- a better drug,
- a better diagnostic,
- a better medical device,
- a better software application,
- a better clinical workflow.

This research area studies a different pattern.

Instead of replacing an existing successful product, identify what prevents that product from working, being used, reaching the right person, or being adopted.

Then intervene at that bottleneck.

The generalized model is:

```text
Existing valuable capability
 ↓
Failure, friction, or non-adoption
 ↓
Measurable cause
 ↓
Modifiable condition
 ↓
Enabling intervention
 ↓
Changed patient / workflow / market state
 ↓
Improved use of the original capability
 ↓
Measured outcomes
 ↓
Learning and compounding advantage
```

In therapeutics, this may mean preparing the patient so an existing therapy can work.

In commercialization, this may mean preparing the market so an existing solution can be understood, purchased, implemented, and routinely used.

The same logic can therefore be applied at multiple layers:

```text
Patient conditioning
Clinical workflow conditioning
Institutional conditioning
Market conditioning
Economic conditioning
Regulatory conditioning
AI / information conditioning
```

The unifying question is:

> What must become true before an already valuable capability can produce its intended outcome?

---

# Research Area Name

## Engineering Care Delivery

Engineering Care Delivery studies how care delivery itself can be deliberately redesigned in an era of abundant AI, software, devices, diagnostics, drugs, evidence, and technical capability.

The research area treats care delivery as an engineered system rather than a passive consumer of healthcare technology.

Its major internal frameworks include:

- Therapeutic Enablement,
- Market Conditioning,
- System Conditioning,
- Care Delivery as the Integration Layer,
- Public Signal Intelligence,
- AI-Native Care Delivery,
- evidence-driven opportunity discovery,
- Native Alpha™,
- and Alpha-to-Patent decision making.

The core question is:

> How should care delivery be engineered when technology becomes abundant but adoption, patient access, workflow, trust, reimbursement, accountability, and outcomes remain scarce?

A short public-facing formulation is:

## Engineer the care, not just the technology.

# Strategic Questions

The research area should repeatedly return to the following questions:

1. Why does an otherwise valuable therapy or regulated product fail?
2. Which failure mechanisms are measurable?
3. Which failure mechanisms are modifiable?
4. Can readiness for treatment or adoption be measured before deployment?
5. Can the patient, workflow, institution, or market be conditioned before introducing the solution?
6. Who benefits economically when the enabling intervention works?
7. Who has authority to buy, prescribe, approve, implement, or reimburse it?
8. What scarce resource remains after AI makes technical production abundant?
9. Where can proprietary data, workflow position, relationships, or rights compound?
10. Which opportunities reveal Native Alpha™ rather than merely technical novelty?

---

# Research Pillar 1: Why Good Therapies and Products Fail

## Research Question

Why do products that are technically sound, clinically promising, or commercially valuable fail to produce expected outcomes?

The purpose of this pillar is to build a taxonomy of failure mechanisms.

In therapeutics, failure may result from:

- biological nonresponse,
- immune suppression,
- resistance mechanisms,
- inadequate delivery,
- timing,
- pharmacokinetics,
- patient condition,
- contraindications,
- inability to identify appropriate patients,
- inability to monitor readiness or response.

In regulated products more broadly, failure may result from:

- poor workflow fit,
- weak economic justification,
- reimbursement uncertainty,
- hospital procurement friction,
- lack of clinical champions,
- implementation burden,
- security or IT requirements,
- training friction,
- unclear evidence,
- fragmented ownership,
- low patient demand,
- weak category understanding.

## Research Agenda

For major therapies and regulated product categories:

- identify where failure occurs,
- quantify the economic cost of failure where possible,
- separate technical failure from adoption failure,
- distinguish intrinsic product limitations from external conditions,
- identify measurable precursors,
- identify modifiable mechanisms,
- identify who has incentives to solve each failure.

## Output

A reusable "Failure Mechanism Map" for regulated products.

---

# Research Pillar 2: Treatment Readiness

## Research Question

Can we determine whether a patient is biologically, clinically, operationally, or behaviorally ready for a therapy before administering it?

This pillar generalizes the concept of immune readiness.

Potential readiness dimensions include:

- immune state,
- inflammatory state,
- metabolic state,
- nutritional state,
- microbiome state,
- organ function,
- medication interactions,
- genetic markers,
- disease burden,
- prior treatment history,
- behavioral readiness,
- adherence probability.

## Research Questions

- Which therapies have large nonresponse populations?
- Which measurable states distinguish likely responders from nonresponders?
- Which markers are causal, predictive, or merely correlated?
- Can readiness be represented as a composite score?
- Can readiness be measured longitudinally?
- Can treatment be delayed or modified until readiness improves?
- Can readiness become a reimbursable diagnostic or workflow category?
- Who owns the readiness data?
- Does the score become commercially more defensible than the conditioning intervention?

## Potential Research Output

A generalized "Treatment Readiness Index" framework.

---

# Research Pillar 3: Patient State Engineering

## Research Question

Instead of creating a new therapy, can we modify the patient so an existing therapy works better?

This pillar studies interventions that alter the conditions under which treatment operates.

Potential approaches include:

- immune conditioning,
- extracorporeal removal,
- microbiome modification,
- metabolic conditioning,
- inflammation reduction,
- nutritional intervention,
- medication washout,
- prehabilitation,
- cellular depletion,
- cellular enrichment,
- hormonal manipulation,
- fluid and electrolyte normalization,
- behavioral preparation,
- sleep and circadian optimization where relevant.

## Core Distinction

Treating the disease is not the same as changing the conditions under which a treatment operates.

That distinction may reveal new product categories.

## Research Questions

- Which therapeutic failures are driven by modifiable patient states?
- How rapidly can those states be changed?
- How durable is the conditioned state?
- What is the correct timing relative to treatment?
- What is the minimum intervention needed?
- How should response to conditioning be measured?
- Can conditioning itself be standardized?
- Does conditioning create a new clinical pathway or simply augment an existing one?

---

# Research Pillar 4: Companion Conditioning

## Research Question

Can "Companion Conditioning" become a category analogous to companion diagnostics?

The conceptual model is:

```text
Companion Diagnostic
Identify whether the patient is likely to respond.

Companion Conditioning
Modify the patient so response becomes more likely.

Response Monitoring
Measure whether the desired biological change occurred.

Adaptive Conditioning
Modify the intervention based on measured response.
```

## Research Questions

- Which therapeutic classes are best suited to companion conditioning?
- What evidence would be needed for clinical acceptance?
- How should conditioning be timed relative to treatment?
- Is the conditioning product drug-specific, class-specific, or mechanism-specific?
- What regulatory pathways are likely?
- Could conditioning be reimbursed separately?
- Would drug manufacturers subsidize or bundle conditioning?
- Does companion conditioning expand the eligible population?
- Could conditioning increase persistence or reduce treatment failure?

## Reference Use Case

Extracorporeal modulation of circulating immunosuppressive factors before checkpoint-inhibitor therapy can be used as an initial generic reference case without tying the research area to any specific company, brand, product, or oncology use case.

---

# Research Pillar 5: Therapy Orchestration

## Research Question

Can treatment outcomes improve primarily through better sequencing, timing, coordination, and adaptation rather than through new therapeutic components?

The future therapy stack may increasingly resemble:

```text
Measure
 ↓
Condition
 ↓
Treat
 ↓
Observe
 ↓
Adapt
 ↓
Treat again
 ↓
Learn
```

## Research Questions

- Which treatments are highly timing-sensitive?
- Which combinations depend on sequence rather than simply co-administration?
- Can biomarkers trigger treatment transitions?
- Can protocols become adaptive rather than fixed?
- Can AI personalize treatment sequencing?
- Who owns the orchestration layer?
- Is the orchestration workflow protectable through IP, data, or operating know-how?
- Can therapy orchestration become more defensible than an individual therapeutic component?

---

# Research Pillar 6: Installed-Base Therapeutics

## Research Question

When is it economically superior to improve the performance of an established therapeutic franchise rather than compete with it?

AI may make product creation abundant, but existing franchises still possess:

- regulatory approvals,
- installed clinical use,
- physician familiarity,
- reimbursement,
- manufacturing,
- commercial teams,
- evidence,
- patient demand,
- distribution.

The opportunity may therefore be to amplify incumbent assets rather than replace them.

## Opportunity Formula

```text
Large existing therapeutic franchise
×
Meaningful failure / nonresponse rate
×
Expensive treatment or disease
×
Measurable failure mechanism
×
Modifiable state
×
Credible enabling intervention
=
Potential Therapeutic Enablement Opportunity
```

## Research Questions

- Which existing therapies generate the largest economic value despite high nonresponse?
- Which franchises have economic incentives to expand eligible populations?
- Where does failure materially reduce franchise value?
- Could an enabling technology increase lifetime value per patient?
- Could it expand into earlier lines of therapy?
- Could it rescue patients otherwise considered treatment failures?
- Which manufacturers would benefit most?
- When does partnership dominate independent commercialization?

---

# Research Pillar 7: Market Conditioning

## Research Question

What must already be true in the market before a new regulated solution can be adopted quickly?

Market conditioning should be studied with the same seriousness as patient conditioning.

Potential components include:

- disease-state education,
- category definition,
- terminology creation,
- evidence reviews,
- KOL familiarity,
- payer education,
- physician education,
- patient education,
- guideline inclusion,
- protocol development,
- procurement preparation,
- implementation playbooks,
- structured AI-readable knowledge,
- digital and AI discoverability.

## Core Model

```text
Unknown problem framing
 ↓
Shared language
 ↓
Clinical understanding
 ↓
Evidence
 ↓
Institutional familiarity
 ↓
Economic justification
 ↓
Workflow readiness
 ↓
Adoption
```

## Research Questions

- How early should market conditioning begin?
- Can a company condition a market before product launch?
- What evidence must exist before a category feels legitimate?
- How do KOLs affect institutional acceptance?
- How do AI systems influence category discovery?
- Can structured content become part of the commercial infrastructure?
- Can a company own the language through which a market understands the problem?

---

# Research Pillar 8: Patient Discovery as Infrastructure

## Research Question

Can the ability to find eligible patients become a core strategic asset?

A product does not have a market simply because epidemiology says eligible patients exist.

Patients must be identifiable in real workflows.

## Research Stack

```text
Clinical eligibility
 ↓
Machine-readable phenotype
 ↓
Available data signals
 ↓
Patient identification
 ↓
Clinical review
 ↓
Referral / outreach
 ↓
Treatment
```

## Data Sources to Study

- EHR,
- claims,
- laboratory data,
- imaging,
- pathology,
- pharmacy data,
- genomic data,
- scheduling data,
- referral records,
- prior authorization records,
- social and behavioral data where appropriate and lawful.

## Research Questions

- Can inclusion and exclusion criteria be translated into executable logic?
- Which signals exist before formal diagnosis?
- How much of patient identification can AI perform?
- Who has permission to act on the signal?
- Which workflow owns the referral?
- How much friction exists between discovery and treatment?
- Can patient-discovery infrastructure become portable across therapies?
- Can manufacturers pay for finding eligible patients without distorting clinical judgment?

---

# Research Pillar 9: Hospital Adoption Engineering

## Research Question

Can regulated-product adoption be treated as an engineered system rather than a sales process?

A hospital rarely has a single buyer.

Adoption may require alignment among:

- physician champions,
- nursing,
- pharmacy,
- IT,
- security,
- compliance,
- legal,
- finance,
- procurement,
- value-analysis committees,
- quality,
- operational leadership,
- executive sponsors.

## Adoption Model

```text
Clinical value
+
Economic value
+
Evidence
+
Workflow fit
+
Technical integration
+
Regulatory comfort
+
Operational ownership
+
Internal champion
=
Adoption probability
```

## Research Questions

- Who has authority versus influence?
- Which stakeholder can veto adoption?
- What evidence does each stakeholder require?
- What objections are predictable?
- Which implementation burdens are underestimated?
- Which incumbent contracts create friction?
- Can AI model the likely buying committee?
- Can an AI Workforce™ maintain account-specific stakeholder maps?
- Can implementation artifacts be generated automatically for each institution?

## Output

A reusable "Hospital Adoption Engineering" playbook.

---

# Research Pillar 10: Distribution as Native Alpha™

## Research Question

When product development becomes easier, can superior distribution itself become Native Alpha™?

The traditional model is often:

```text
IP → product → moat
```

The emerging model may increasingly be:

```text
Unusual insight
→ narrow useful product
→ superior distribution
→ workflow position
→ data
→ trust
→ compounding advantage
```

## Research Questions

- When does market access become more defensible than technical differentiation?
- Which distribution capabilities are difficult to copy?
- Can relationships become institutionalized through workflow rather than dependent on individuals?
- Does controlling the patient-discovery layer create strategic leverage?
- Can implementation speed become a moat?
- Can integration into routine work produce switching costs?
- Can distribution generate proprietary evidence and data that strengthen the product?

## Native Alpha™ Test

Ask:

> Based on who we are, what can we know, see, decide, or do about adoption that competitors cannot?

---

# Research Pillar 11: Evidence as Distribution

## Research Question

Can evidence be treated as a commercial distribution mechanism rather than merely a regulatory obligation?

Evidence can change:

- physician behavior,
- payer behavior,
- procurement,
- guideline inclusion,
- protocol design,
- patient demand,
- AI-generated answers.

## Evidence-to-Adoption Chain

```text
Evidence
 ↓
Credibility
 ↓
Category acceptance
 ↓
Guideline / protocol inclusion
 ↓
Workflow
 ↓
Demand
 ↓
Usage
```

## Research Questions

- Which evidence types influence which stakeholders?
- When is real-world evidence sufficient?
- What is the role of investigator-initiated studies?
- Can implementation evidence be as important as clinical evidence?
- How should evidence generation be sequenced with commercialization?
- How should companies publish evidence so AI systems can retrieve it?
- Can evidence creation itself be designed as part of GTM?

---

# Research Pillar 12: AI-Native Category Creation

## Research Question

Can AI help create a new market category before conventional commercial infrastructure exists?

When a clinical category is immature, the market lacks shared language.

AI can help create and maintain:

- terminology,
- taxonomies,
- ontologies,
- clinical questions,
- structured evidence maps,
- patient education,
- physician education,
- payer narratives,
- economic models,
- trial summaries,
- conference intelligence,
- AI-readable knowledge bases.

## Core Concept

The objective is not content volume.

The objective is cognitive infrastructure.

When a clinician, payer, patient, researcher, hospital executive, or AI system encounters the underlying problem, the category should already be legible.

## Research Questions

- What information must exist before a category becomes understandable?
- Can terminology influence how problems are searched?
- Which structured formats improve AI retrieval?
- Can category ownership precede product leadership?
- How should scientific neutrality be maintained while building category awareness?
- Can category infrastructure become an asset independent of one product?

---

# Research Pillar 13: Demand Before Product

## Research Question

Should regulated-product development begin with demand architecture rather than product architecture?

Traditional sequence:

```text
Invent
→ develop
→ approve
→ launch
→ market
```

AI-era sequence:

```text
Identify expensive problem
→ identify who experiences it
→ understand adoption bottleneck
→ condition market
→ validate economic buyer
→ prove access to patients
→ define evidence requirements
→ build the narrowest product needed
```

## Research Questions

- How much commercial validation can occur before product completion?
- Can patient-discovery feasibility be tested first?
- Can hospital stakeholder interviews reveal fatal adoption barriers early?
- Can manufacturers commit to pilots or data access before product build?
- Can evidence requirements define the product rather than follow it?
- Can market conditioning begin while technical development is still underway?

This pillar should connect directly to Alpha-to-Patent research.

The commercial opportunity should help define what is worth protecting.

---

# Research Pillar 14: Solution Abundance and Market Compression

## Research Question

What happens to regulated industries when the cost of creating credible solutions collapses?

Expected effects may include:

- more competitors,
- faster imitation,
- shorter periods of technical differentiation,
- rapid feature convergence,
- reduced value of generic software,
- reduced value of undifferentiated device functionality,
- greater importance of trusted distribution,
- greater importance of evidence,
- greater importance of patient access,
- greater importance of workflow control,
- greater importance of data compounding.

## Research Questions

- Which forms of technical IP retain value under solution abundance?
- Which forms become commoditized?
- Do regulatory assets become more or less valuable?
- Does brand matter more?
- Does installed workflow become the dominant moat?
- Does reimbursement become a primary competitive asset?
- Does patient access become the equivalent of distribution in consumer markets?
- How quickly do regulated categories compress once AI lowers development cost?

---

# Research Pillar 15: AI as the Market Development Workforce

## Research Question

How much of regulated-product commercialization can become AI-native?

AI should not merely assist product development.

It should increasingly become the commercialization infrastructure surrounding the product.

## Candidate AI Workforce™ Functions

### Market Intelligence
- monitor therapeutic categories,
- monitor competitors,
- monitor clinical trials,
- monitor publications,
- monitor conferences,
- monitor regulatory changes,
- monitor reimbursement changes,
- monitor health-system activity.

### Account Intelligence
- identify target institutions,
- identify target manufacturers,
- map stakeholders,
- track leadership changes,
- detect adoption signals,
- identify internal champions,
- identify likely objections.

### Patient Discovery
- translate eligibility criteria,
- identify data sources,
- generate cohort logic,
- support screening workflows,
- track referral conversion.

### Evidence
- maintain evidence maps,
- compare studies,
- identify evidence gaps,
- draft evidence summaries,
- produce stakeholder-specific evidence packets.

### Content and Category Creation
- build disease-state libraries,
- create structured FAQs,
- maintain ontologies,
- prepare AI-readable content,
- generate physician and patient educational materials.

### Institutional Adoption
- prepare value-analysis materials,
- generate implementation plans,
- draft security questionnaires,
- create SOPs,
- generate training material,
- personalize adoption packages.

### Commercial Operations
- prepare account dossiers,
- draft personalized outreach,
- track engagement,
- recommend next actions,
- maintain opportunity histories,
- detect stalled accounts.

## Research Principle

Assume that anything which currently requires repetitive cognitive labor should be considered a candidate for AI-native execution.

Human work should increasingly concentrate on:

- accountability,
- judgment,
- relationships,
- negotiation,
- consent,
- clinical responsibility,
- exception handling,
- leadership,
- validation.

---

# Research Pillar 16: The Enabling IP Stack

## Research Question

What intellectual property should exist around therapeutic and market enablement?

Do not assume the core device, software, or intervention is the only valuable asset.

Potential IP layers include:

```text
Mechanism
+
Patient selection
+
Readiness measurement
+
Conditioning intervention
+
Timing
+
Sequence
+
Dose relationships
+
Specific combinations
+
Monitoring
+
Adaptive logic
+
Clinical workflow
+
Data structures
+
Institutional implementation methods
```

## Alpha-to-Patent Progression

Start with Native Alpha™:

> What unusual insight do we possess?

Then:

> What does that insight allow someone to know, decide, or do?

Then:

> Which commercially valuable uses should be protected?

Then:

> What is the narrowest commercially useful rights structure?

The research should not start with:

> What can we patent?

It should start with:

> What unusual insight changes a real commercial or clinical decision?

---

# Research Pillar 17: Data Compounding

## Research Question

When does an enabling intervention become a data business?

Each cycle may generate:

```text
Baseline state
→ intervention
→ changed state
→ treatment
→ response
→ longitudinal outcome
```

Over time this may reveal:

```text
Patient phenotype X
+
Biomarker state Y
+
Therapy Z
+
Conditioning A
+
Timing B
=
Probability C of response
```

## Research Questions

- Which data are uniquely generated by the workflow?
- Which data can legally be reused?
- Which data rights must be secured contractually?
- Can longitudinal outcomes improve treatment selection?
- Can the system learn across institutions?
- Can privacy-preserving methods enable learning without centralized data?
- Does the data improve workflow, diagnostics, or conditioning?
- Does the data create a stronger moat than the original technology?

---

# Research Pillar 18: Who Captures the Value?

## Research Question

Who benefits enough economically to become the real customer?

Potential beneficiaries may include:

- drug manufacturers,
- device manufacturers,
- hospitals,
- physician groups,
- payers,
- specialty pharmacies,
- patients,
- employers,
- integrated delivery networks.

## Research Questions

- Who receives the economic benefit?
- Who bears the implementation cost?
- Who controls purchasing?
- Who owns the budget?
- Who can measure the value?
- Who has authority to share savings?
- Which stakeholder has the strongest incentive to accelerate adoption?
- Can value be priced against increased therapy revenue, avoided failure, reduced hospitalization, or expanded eligibility?

## Potential Business Models

- licensing,
- field-of-use licenses,
- co-development,
- per-treatment fees,
- royalties,
- outcome-based payment,
- reimbursement,
- shared savings,
- bundled payment,
- manufacturer sponsorship,
- institutional subscription,
- data partnerships.

---

# Research Pillar 19: Therapy Expansion as Market Creation

## Research Question

Can therapeutic enablement create an entirely new market by expanding who can successfully receive an existing therapy?

Three primary mechanisms:

1. Improve response among already treated patients.
2. Convert previously unsuitable patients into eligible patients.
3. Move treatment earlier in the disease pathway.

## Research Questions

- Which expansion mechanism creates the greatest economic value?
- Does conditioning increase the eligible population?
- Can treatment move into earlier stages?
- Can enabling interventions reduce adverse events enough to increase use?
- Can improved patient selection increase physician confidence?
- Can patient discovery reveal hidden eligible populations?

---

# Research Pillar 20: The Adoption Bottleneck Map

## Research Question

What is the actual bottleneck for each regulated product?

Every opportunity should be scored across:

- technical feasibility,
- clinical feasibility,
- regulatory feasibility,
- evidence sufficiency,
- reimbursement,
- patient identification,
- physician adoption,
- workflow fit,
- procurement,
- IT and security,
- training,
- behavior change,
- patient acceptance,
- distribution.

## Core Principle

The hardest scientific problem is not necessarily the commercial bottleneck.

Research should identify the constraint that most limits adoption.

That constraint may become the highest-value intervention point.

---

# Cross-Cutting Research Theme: Native Alpha™

Every research pillar should include a Native Alpha™ analysis.

## Questions

- What unusual problem framing does this reveal?
- What can we know that others cannot easily know?
- What can we see before others see it?
- What can we decide faster?
- What can we do that others cannot easily execute?
- Is that advantage technical, commercial, institutional, informational, relational, regulatory, or operational?
- Does it compound with use?
- Can it be protected through IP, data, workflow, contracts, distribution, or know-how?

## Native Alpha™ Progression

```text
Signal
→ Availability
→ Relevance
→ Product Wedge
→ Demand Proof
→ Rights
→ Compounding
```

---

# Cross-Cutting Research Theme: AI-Native Regulated Product Development

The research should assume that AI increasingly compresses:

- literature review,
- prior-art analysis,
- software development,
- quality documentation,
- requirements generation,
- risk analysis,
- protocol drafting,
- verification planning,
- validation support,
- regulatory writing,
- evidence synthesis,
- market research,
- account research,
- content generation,
- implementation documentation.

Research should therefore continuously ask:

> If the technical work becomes ten times easier, what becomes the bottleneck?

And:

> If AI becomes another ten times better, what business model breaks?

The research should resist defending workflows merely because regulation made those workflows historically expensive.

Instead, identify the underlying regulatory or accountability requirement and ask whether AI-native execution can satisfy that requirement more efficiently.

---

# Research Methodology

## 1. Start With Expensive Failure

Prioritize categories where failure costs a great deal.

Examples:

- expensive drugs with low response rates,
- high-cost procedures with avoidable failure,
- devices with poor adherence,
- diagnostics that fail to change care,
- therapies with large eligible but untreated populations,
- proven products blocked by institutional friction.

## 2. Separate Failure Types

For every opportunity, classify failure as:

- biological,
- technical,
- operational,
- institutional,
- financial,
- behavioral,
- informational,
- regulatory,
- distributional.

## 3. Identify the Modifiable Condition

Ask:

- What must change?
- Can it be measured?
- Can it be changed?
- Who can change it?
- How quickly?
- At what cost?
- With what evidence?

## 4. Identify the Economic Beneficiary

Do not assume the clinical user is the commercial buyer.

Map:

```text
Clinical beneficiary
Economic beneficiary
Purchaser
Approver
User
Patient
Data owner
Risk bearer
```

## 5. Test the Adoption Path Early

Before deep product investment:

- identify patients,
- identify institutions,
- identify champions,
- map buying committees,
- test evidence requirements,
- test reimbursement,
- test implementation friction.

## 6. Design the Narrowest Useful Intervention

Do not build a platform first.

Build the smallest intervention that removes the bottleneck.

## 7. Capture Learning

Every deployment should produce reusable knowledge about:

- patient selection,
- adoption,
- objections,
- outcomes,
- implementation,
- economics,
- evidence.

---

# Opportunity Discovery Engine

A core research deliverable should be an AI-native opportunity discovery system.

## Search Pattern

Find:

```text
Successful therapy or regulated product
+
Large valuable failure population
+
Known or suspected failure mechanism
+
Measurable state
+
Potentially modifiable condition
+
Economic beneficiary
+
Accessible adoption pathway
```

## Candidate Data Sources

- PubMed,
- ClinicalTrials.gov,
- FDA databases,
- CMS data,
- reimbursement policies,
- clinical guidelines,
- conference abstracts,
- SEC filings,
- earnings calls,
- patents,
- payer policies,
- hospital procurement materials,
- public EHR implementation materials,
- peer-reviewed health economics,
- manufacturer product labels,
- adverse-event databases,
- medical society guidance.

## AI Functions

AI should:

- continuously identify candidate failure mechanisms,
- cluster similar mechanisms,
- score commercial value,
- identify likely beneficiaries,
- identify patient populations,
- identify existing enabling products,
- identify IP whitespace,
- identify evidence gaps,
- suggest narrow experiments.

---

# Opportunity Scoring Model

Every candidate should receive a structured score.

## Suggested Dimensions

### Clinical Value
- severity,
- unmet need,
- failure cost,
- patient impact.

### Economic Value
- therapy cost,
- market size,
- manufacturer economics,
- hospital economics,
- payer economics.

### Modifiability
- biological plausibility,
- intervention feasibility,
- timing,
- reversibility.

### Measurability
- biomarker availability,
- data availability,
- monitoring frequency,
- outcome clarity.

### Adoption
- patient identifiability,
- workflow fit,
- institutional burden,
- stakeholder complexity,
- reimbursement.

### Native Alpha™
- unusual insight,
- access advantage,
- data advantage,
- workflow advantage,
- distribution advantage,
- IP potential.

### AI Leverage
- research automation,
- software automation,
- regulatory automation,
- evidence automation,
- account intelligence,
- patient discovery,
- implementation automation.

### Compounding
- proprietary data,
- workflow lock-in,
- network effects,
- longitudinal learning,
- category ownership.

---

# Evidence Standards

The research must clearly distinguish among:

- proven,
- strongly supported,
- plausible,
- speculative,
- commercially interesting but unverified.

Do not allow persuasive strategy language to outrun evidence.

For every major claim, track:

- evidence type,
- source,
- date,
- quality,
- strength,
- conflicting evidence,
- commercial implication,
- unanswered questions.

AI outputs are research aids, not clinical, regulatory, or legal opinions.

Qualified specialists remain responsible for professional judgment.

---

# Regulatory Research

Regulation should be treated as a design input, not merely a late-stage obstacle.

For each opportunity:

- identify regulatory classification,
- identify intended use,
- identify claims,
- identify whether the product modifies diagnosis, treatment, workflow, or monitoring,
- identify likely evidence requirements,
- identify combination-product implications,
- identify software or SaMD implications,
- identify clinical decision-support issues,
- identify quality-system obligations,
- identify post-market requirements.

Research should also ask:

> Which regulatory requirements become substantially easier to satisfy with AI-native evidence, documentation, simulation, testing, and quality workflows?

Do not assume that easier execution eliminates regulatory accountability.

---

# Commercial Rights and IP Research

For every material opportunity, distinguish among:

- inventorship,
- ownership,
- assignment history,
- patent family,
- prosecution status,
- grant status,
- expiration,
- encumbrances,
- licenses,
- government rights,
- maintenance obligations,
- background know-how,
- field-of-use restrictions,
- commercially usable rights.

Never infer ownership from inventorship.

Never infer freedom to operate from ownership.

Never infer enforceability from issuance.

Never infer commercial demand from novelty.

The key question remains:

> Does this intellectual asset change a real commercialization decision?

---

# Reference Research Cases

The research area should accumulate case studies.

Initial cases may include:

1. Extracorporeal or other immune conditioning before checkpoint-inhibitor therapy.
2. Patient preparation before cell therapy.
3. Prehabilitation before high-risk surgery.
4. Microbiome conditioning before selected therapies.
5. Patient discovery infrastructure for rare disease.
6. Hospital adoption pathways for regulated AI.
7. Companion diagnostics as historical precedent.
8. Drug-device combination strategies.
9. Evidence-driven category creation.
10. Technologies that succeeded primarily through distribution rather than superior product design.

Each case should be analyzed using the same framework.

---

# Suggested Research Outputs

The research program should generate reusable assets rather than only prose.

## Core Artifacts

- Failure Mechanism Maps
- Treatment Readiness frameworks
- Companion Conditioning taxonomy
- Market Conditioning playbooks
- Hospital Adoption Engineering playbooks
- Patient Discovery architecture
- Account-Based Market Development templates
- Evidence-to-Adoption maps
- Native Alpha™ assessments
- IP opportunity maps
- commercialization-rights matrices
- AI Workforce™ role catalogs
- opportunity scorecards
- reference case studies
- research briefs
- visual explainers
- public research pillars
- investor and founder primers

---

# Suggested Visuals

## Visual 1: Technical Abundance Creates Adoption Scarcity

```text
More capable AI
 ↓
Cheaper product creation
 ↓
More solutions
 ↓
Faster feature convergence
 ↓
Less technical scarcity
 ↓
Greater scarcity of:
trust
patients
distribution
workflow
evidence
adoption
```

## Visual 2: The Therapeutic Enablement Stack

```text
Measure → Prepare → Treat → Observe → Adapt → Learn
```

## Visual 3: The Market Conditioning Stack

```text
Define Problem
→ Create Language
→ Build Evidence
→ Educate Market
→ Find Patients
→ Prepare Institution
→ Adopt
→ Measure
→ Compound
```

## Visual 4: Two-Sided Conditioning

```text
PATIENT SIDE MARKET SIDE

Identify state Identify market friction
 ↓ ↓
Measure readiness Create shared language
 ↓ ↓
Condition patient Build evidence
 ↓ ↓
Deliver therapy Prepare institution
 ↓ ↓
Measure response Drive adoption
 ↓ ↓
Learn Learn
```

## Visual 5: The Adoption Bottleneck

A horizontal pipeline with red bottleneck indicators for:

```text
Science → Regulation → Evidence → Patient Finding → Reimbursement
→ Procurement → Workflow → Training → Use
```

## Visual 6: The New Moat

```text
Old:
Patent → Product → Moat

Emerging:
Native Alpha™
→ Product
→ Distribution
→ Workflow
→ Data
→ Trust
→ Compounding
```

---

# Initial Research Program

## Phase 1: Build the Framework

Create:

- taxonomy of failure,
- therapeutic enablement model,
- market conditioning model,
- adoption bottleneck framework,
- opportunity scoring rubric,
- Native Alpha™ framework,
- research evidence standards.

## Phase 2: Build Reference Cases

Select 5 to 10 cases across:

- oncology,
- surgery,
- medical devices,
- regulated AI,
- diagnostics,
- specialty therapeutics.

## Phase 3: Build the Opportunity Discovery Engine

Use AI to continuously identify:

- expensive nonresponse,
- unmet adoption problems,
- modifiable mechanisms,
- eligible patient populations,
- strategic buyers,
- potential IP,
- likely evidence paths.

## Phase 4: Validate With Market Evidence

For high-scoring opportunities:

- interview clinicians,
- interview health systems,
- interview manufacturers,
- identify patient populations,
- test evidence requirements,
- test reimbursement,
- test partnership interest.

## Phase 5: Generate Commercial Experiments

Examples:

- patient-finding pilot,
- institution readiness pilot,
- evidence synthesis,
- manufacturer outreach,
- investigator-initiated study,
- field-of-use licensing discussion,
- workflow prototype,
- AI Workforce™ market-development pilot.

---

# Research Governance

Each research project should maintain:

- hypothesis,
- evidence,
- assumptions,
- counterarguments,
- regulatory questions,
- commercial questions,
- IP questions,
- patient-access questions,
- adoption questions,
- next experiment.

Avoid research theater.

Do not allow a large volume of AI-generated material to substitute for evidence.

The standard should be:

> Did this research improve a real decision?

---

# Key Contrarian Research Hypotheses

The research area should actively test the following propositions rather than merely assume them.

1. Product creation will become cheaper faster than institutional adoption becomes easier.
2. Distribution will become more defensible than many technical features.
3. Patient identification will become infrastructure.
4. Market conditioning can begin before product launch.
5. Evidence can function as distribution.
6. Category creation can precede product leadership.
7. AI systems will become an important discovery channel for clinical and institutional decision-makers.
8. The most valuable buyer may be the company whose existing product becomes more valuable.
9. Treatment readiness may become a new diagnostic category.
10. Companion conditioning may become a new therapeutic category.
11. Workflow position and proprietary data may outlive the technical novelty of the original product.
12. In many regulated categories, the ability to get through the institution will be more valuable than the ability to build the product.

---

# Long-Term Research Questions

Over time, the research area should attempt to answer:

- What becomes the primary source of defensibility when regulated-product creation is abundant?
- Which current regulated industries are structurally organized around scarcity that AI will eliminate?
- Which adoption barriers are genuine and which are artifacts of legacy operating models?
- Can AI-native companies enter regulated industries faster than incumbents expect?
- Will regulatory capability become a commodity?
- Will evidence generation become continuous?
- Will patient identification become embedded into treatment products?
- Will therapeutic companies increasingly sell outcomes rather than components?
- Will hospital procurement become partially machine-mediated?
- Will AI systems become important intermediaries between evidence and clinical action?
- Can companies deliberately engineer category ownership?
- Can conditioning become a generalized product class?
- Can institutional adoption itself become programmable?

---

# What Success Looks Like

This research area succeeds when it produces better opportunity selection and commercialization decisions.

Success is not measured by:

- number of papers,
- number of patents,
- number of AI-generated reports,
- number of ideas.

Success is measured by:

- identifying valuable bottlenecks before others do,
- finding opportunities where technical abundance creates new commercial scarcity,
- discovering where existing therapies can be made more valuable,
- proving access to patients and institutions,
- reducing time from technical feasibility to routine use,
- identifying the real economic buyer,
- creating narrow useful products,
- designing only the rights actually required,
- generating compounding data and workflow advantages,
- and producing Native Alpha™ that changes a real decision.

---

# Research Pillar 21: System Conditioning

## Research Question

What parts of the surrounding system must become ready before a valuable regulated solution can succeed?

This pillar generalizes patient conditioning and market conditioning into a broader framework.

A product may fail even when it is technically sound because the surrounding system is not yet prepared to absorb it.

The system may require conditioning across several layers:

```text
Patient conditioning
+
Clinician conditioning
+
Workflow conditioning
+
Institution conditioning
+
Payer conditioning
+
Regulatory conditioning
+
Market conditioning
+
AI / information conditioning
```

## Core Thesis

A new solution often fails because the surrounding system is not yet ready for it.

The research question therefore becomes:

> What must change in the surrounding system before this product can succeed, and which of those changes are themselves valuable businesses?

## Research Questions

- Which system layer is least ready?
- Which layer is the binding constraint?
- Can the constraint be changed directly?
- Who controls that layer?
- Who benefits when it changes?
- Can conditioning be standardized?
- Can conditioning be sold separately from the product?
- Can the conditioning layer itself become the more valuable business?

---

# Research Pillar 22: Falsification and Kill Criteria

## Research Question

What evidence would cause us to stop believing an opportunity is attractive?

Every research project should define falsification criteria before significant investment.

Research must not become advocacy.

## Required Kill Criteria

For each opportunity, define conditions that would invalidate:

- the clinical thesis,
- the mechanism thesis,
- the adoption thesis,
- the reimbursement thesis,
- the buyer thesis,
- the IP thesis,
- the market size thesis,
- the AI-leverage thesis,
- the data-compounding thesis.

## Example Kill Questions

- What evidence would show that the failure mechanism is not actually modifiable?
- What evidence would show that the measured biomarker is not predictive enough?
- What evidence would show that hospitals will not change workflow?
- What evidence would show that no stakeholder will pay?
- What evidence would show that the adoption cycle is too long?
- What evidence would show that the market is too fragmented?
- What evidence would show that a competitor or incumbent can neutralize the advantage easily?

## Decision Rule

Each research project should maintain:

```text
Belief
→ Evidence supporting belief
→ Evidence against belief
→ Confidence level
→ Explicit kill threshold
```

If the kill threshold is met, stop, redirect, license, sell, or abandon.

---

# Research Pillar 23: Portfolio Allocation

## Research Question

How should Intellectual Frontiers allocate time, capital, attention, and rights across multiple opportunities?

Strong individual opportunities can still produce a weak portfolio if resources are misallocated.

## Portfolio Decisions

Each asset or opportunity should be placed into one of several categories:

- investigate,
- incubate,
- partner,
- license,
- field-of-use license,
- spin out,
- assign,
- sell,
- maintain,
- monitor,
- abandon.

## Portfolio Scoring Dimensions

- strategic relevance,
- Native Alpha™ strength,
- time to evidence,
- time to adoption,
- capital intensity,
- regulatory burden,
- buyer clarity,
- IP position,
- distribution advantage,
- expected value,
- downside risk,
- opportunity cost.

## Research Output

A portfolio decision matrix that compares opportunities by both expected value and strategic fit.

---

# Research Pillar 24: Economics Before Enthusiasm

## Research Question

Does the opportunity create enough economic value to matter?

Clinical value does not guarantee commercial value.

Every opportunity should include a structured economic model.

## Required Economic Questions

- What does it cost to identify the patient?
- What does it cost to condition or prepare the patient?
- What does it cost to implement at the institution?
- What is the incremental treatment cost?
- What value is created by improved response?
- What value is created by expanding eligibility?
- What value is created by reducing failure?
- What value is created by reducing hospitalization or complications?
- What share of value can realistically be captured?
- Who pays?
- Who saves?
- Who earns more?

## Economic Stack

```text
Clinical benefit
→ Economic benefit
→ Capturable value
→ Revenue model
→ Margin
→ Scalable economics
```

An opportunity should not progress merely because it is scientifically interesting.

---

# Research Pillar 25: Adoption Latency

## Research Question

How long does it take for a technically viable solution to become routine practice?

Time to adoption may matter as much as technical feasibility.

## Adoption Latency Factors

- evidence requirements,
- clinical habit,
- guideline cycles,
- budget cycles,
- procurement cycles,
- reimbursement,
- institutional committees,
- integration burden,
- training,
- legal review,
- security review,
- patient acceptance,
- regulatory timing.

## Research Questions

- What is the likely time from first evidence to first use?
- What is the likely time from first use to repeatable use?
- What is the likely time from repeatable use to routine use?
- Which part of the path can AI compress?
- Which part is governed by institutional cycles that AI cannot easily accelerate?

## Output

An "Adoption Latency Score" for every opportunity.

---

# Research Pillar 26: Channel Architecture

## Research Question

What is the correct path through which the solution reaches the patient or institution?

A strong product can fail because the wrong commercialization channel was selected.

## Candidate Channels

- manufacturer-led,
- hospital-led,
- physician-led,
- specialty pharmacy,
- distributor,
- payer,
- employer,
- direct-to-patient,
- CRO,
- reference laboratory,
- digital health platform,
- existing regulated product platform.

## Research Questions

- Who already has the relationship?
- Who already owns the workflow?
- Who already has reimbursement?
- Who already has patient access?
- Which channel reduces adoption friction?
- Which channel creates dependency?
- Which channel produces the strongest data rights?
- Which channel best aligns with the eventual exit path?

---

# Research Pillar 27: Incumbent Response

## Research Question

What will powerful incumbents do if the opportunity works?

Do not assume incumbents remain passive.

Possible responses include:

- partner,
- acquire,
- license,
- copy,
- bundle,
- block,
- litigate,
- change pricing,
- influence guidelines,
- change reimbursement,
- use distribution power,
- ignore.

## Research Questions

- Which incumbent loses value?
- Which incumbent gains value?
- Which incumbent can neutralize the solution?
- Which incumbent has incentives to partner early?
- Which incumbent can create a competing substitute?
- Can the opportunity be structured so the incumbent becomes an ally rather than an adversary?

## Native Alpha™ Implication

Anticipating incumbent reaction should be part of Native Alpha™ analysis.

---

# Research Pillar 28: Reimbursement Engineering

## Research Question

Can reimbursement be designed into the opportunity from the beginning?

Reimbursement should not be treated only as a late-stage commercialization problem.

## Research Questions

- Does an existing code apply?
- Can the solution be bundled into an existing payment?
- Is a new code required?
- Can a manufacturer subsidize the intervention?
- Can the solution be economically viable without separate reimbursement?
- Is the best economic path through shared savings?
- Can reimbursement be tied to outcomes?
- Can the product fit under an existing benefit category?
- What evidence will payers require?

## Core Principle

Sometimes the best reimbursement strategy is to avoid creating a new reimbursement dependency entirely.

---

# Research Pillar 29: Trust, Liability, and Risk Transfer

## Research Question

Who is willing to carry responsibility when the solution is used?

In regulated markets, adoption is often constrained less by functionality than by accountability.

## Risk Categories

- malpractice,
- product liability,
- regulatory exposure,
- data liability,
- implementation risk,
- cybersecurity risk,
- operational responsibility,
- financial risk,
- reputational risk.

## Research Questions

- Who bears risk today?
- Does the new solution increase or reduce that risk?
- Can risk be shifted contractually?
- Can insurance costs be reduced?
- Which stakeholder must trust the product enough to assume responsibility?
- Does a new role or credential become necessary?
- Can better evidence or workflow controls reduce liability?

## Research Output

A "Risk Bearer Map" alongside the economic beneficiary map.

---

# Research Pillar 30: Ethics and Incentive Alignment

## Research Question

Can the commercialization model withstand scrutiny from patients, clinicians, regulators, boards, and independent investors?

Market conditioning must not become manufactured demand.

## Research Questions

- Are incentives aligned with patient benefit?
- Could manufacturer economics distort clinical judgment?
- Is physician independence preserved?
- Is evidence presented neutrally?
- Are patient choices respected?
- Are data rights transparent?
- Are related-party arrangements defensible?
- Would an independent board member approve the structure?
- Would the business model survive public disclosure?

## Standard

Commercial success should not depend on hidden incentives, distorted evidence, or asymmetric information.

---

# Research Pillar 31: Workflow Ownership

## Research Question

Who controls the moment at which action is taken?

Workflow control may be more valuable than product ownership.

## Critical Workflow Moments

- order entry,
- referral,
- scheduling,
- prior authorization,
- tumor board review,
- discharge,
- pathology review,
- treatment planning,
- pharmacy verification,
- clinical decision support,
- patient outreach.

## Core Question

> Who owns the click?

The actor or system controlling the action point may have greater commercial power than the technology vendor.

## Research Questions

- Which workflow moment determines adoption?
- Who controls it?
- Can the solution embed there?
- Can the workflow position compound over time?
- Can workflow ownership create durable distribution?

---

# Research Pillar 32: The Status Quo as Competitor

## Research Question

What existing behavior must change for the solution to succeed?

The primary competitor is often not another product.

It may be:

- current practice,
- clinical inertia,
- a manual workaround,
- a legacy contract,
- "good enough,"
- doing nothing.

## Research Questions

- Why does the current behavior persist?
- Who benefits from the status quo?
- What friction does change create?
- What must become dramatically better to justify change?
- What implementation cost is hidden?
- What habit, contract, or workflow creates switching cost?

## Rule

Every opportunity should identify the status quo explicitly and explain why behavior will change.

---

# Research Pillar 33: Category or Infrastructure?

## Research Question

Should the solution become a visible category, or should it disappear inside an existing workflow?

Not every innovation benefits from category creation.

Sometimes the best strategy is to become invisible infrastructure.

## Category Strategy

Use category creation when:

- the problem is not understood,
- shared language is missing,
- the buyer needs education,
- evidence must create a new mental model.

## Infrastructure Strategy

Use embedded infrastructure when:

- the workflow already exists,
- adoption friction is high,
- the user does not need another category,
- reimbursement already exists,
- invisible integration accelerates use.

## Core Question

> Does this opportunity need a new category, or does it need to fit quietly inside an old one?

---

# Research Pillar 34: International Market Strategy

## Research Question

Is the United States actually the best first market?

Regulated markets vary materially across:

- reimbursement,
- procurement,
- regulation,
- liability,
- clinical practice,
- data access,
- patient access,
- hospital structure,
- physician autonomy.

## Research Questions

- Which market offers the fastest evidence generation?
- Which market offers the fastest adoption?
- Which market offers the strongest reimbursement?
- Which market offers the lowest regulatory burden?
- Which market produces the most transferable evidence?
- Should initial commercialization occur outside the U.S.?

## Output

A market-entry comparison framework.

---

# Research Pillar 35: Exit and Rights Pathways

## Research Question

What is the intended destination of the opportunity?

Different opportunities should be structured differently depending on likely outcomes.

## Potential Paths

- internal commercialization,
- portfolio-company formation,
- strategic partnership,
- nonexclusive license,
- exclusive field-of-use license,
- co-development,
- assignment,
- sale,
- acquisition,
- technology transfer.

## Research Questions

- Who is the natural long-term owner?
- Which rights are actually required?
- What rights should be retained?
- Does exclusivity increase or reduce value?
- Would a strategic buyer prefer control or access?
- Does the opportunity require a company at all?

## Core Principle

Prefer the narrowest commercially useful rights structure.

---

# Research Pillar 36: Longitudinal Success Metrics

## Research Question

How do we know that an opportunity is progressing rather than merely generating activity?

Each opportunity should be measured through stages.

## Suggested Stages

### Research
- hypothesis quality,
- evidence confidence,
- falsification status.

### Access
- patient accessibility,
- data accessibility,
- institutional access,
- stakeholder access.

### Demand
- buyer interest,
- implementation commitment,
- data-sharing commitment,
- pilot commitment,
- payment commitment.

### Adoption
- first use,
- repeat use,
- routine use,
- time to implementation,
- implementation burden.

### Economics
- value created,
- value captured,
- margin,
- retention,
- expansion.

### Compounding
- proprietary data,
- workflow position,
- category influence,
- distribution advantage,
- IP strengthening.

## Rule

Do not confuse activity with progress.

---

# Research Pillar 37: Care Delivery as the Integration Layer

## Research Question

As healthcare technology becomes abundant, does value migrate downstream toward the organizations that actually deliver care?

Historically, healthcare technology companies have often been valued more richly than care delivery companies because technology appeared:

- scalable,
- high margin,
- asset light,
- repeatable,
- less labor intensive.

Care delivery appeared:

- labor intensive,
- local,
- fragmented,
- reimbursement constrained,
- operationally complex,
- difficult to scale.

That comparison may weaken if AI changes the underlying economics of both categories.

If software, devices, diagnostics, clinical intelligence, workflow tools, and even portions of regulated product development become easier and cheaper to produce, then the scarce capability may no longer be technical creation.

The scarce capability may become:

- patient access,
- clinical authority,
- workflow ownership,
- trust,
- reimbursement,
- implementation,
- orchestration,
- accountability,
- longitudinal outcomes.

## Core Thesis

> Care delivery may become the integration layer for abundant healthcare technology.

The strategic question is no longer only:

> Who creates the best technology?

It becomes:

> Who is best at selecting, combining, operationalizing, and continuously improving abundant technologies around actual patient care?

---

# Research Pillar 38: Value Migration From Technology to Care

## Research Question

Does technological abundance cause value to migrate from technology creation toward technology-enabled care delivery?

When a technical capability is scarce, the vendor creating it may capture disproportionate value.

When many technically credible substitutes exist, value may shift toward whoever controls:

- distribution,
- workflow,
- demand,
- patient access,
- reimbursement,
- data,
- outcomes.

## Research Questions

- Which healthcare technology categories are becoming commoditized?
- Which care delivery capabilities remain structurally scarce?
- Which forms of technology retain differentiated value?
- Which become interchangeable components?
- When does the care company capture more value than the technology vendor?
- Does the patient relationship become more strategically valuable than product ownership?
- Which parts of the value chain gain negotiating leverage as technical supply increases?
- When does the buyer of technology become more valuable than the seller?

## Research Hypothesis

The relative value of technology companies versus care delivery companies should not be treated as permanent.

Instead, study the drivers of historical valuation differences and ask whether AI is changing them.

---

# Research Pillar 39: AI-Native Care Delivery Economics

## Research Question

Can AI materially reduce the labor intensity that historically limited the economics and scalability of care delivery?

The weakness of traditional care delivery has often been that revenue growth requires roughly proportional growth in human labor.

AI may change that relationship.

## Candidate Areas for Labor Compression

- scheduling,
- documentation,
- coding,
- prior authorization,
- care coordination,
- patient communication,
- intake,
- triage support,
- education,
- monitoring,
- follow-up,
- quality reporting,
- compliance support,
- clinical preparation,
- population management.

## Core Economic Question

> How much human labor remains structurally necessary per unit of care delivered?

## Research Questions

- Which specialties have the highest administrative labor burden?
- Which labor categories can be safely reduced through AI?
- Which work can be unbundled from licensed clinicians?
- Which tasks require human accountability but not continuous human execution?
- How much operating leverage can AI create?
- Can care delivery begin to exhibit software-like marginal economics?
- What new bottlenecks appear when labor intensity falls?
- Does reduced labor intensity increase quality, margin, access, or all three?

## Strategic Implication

If a care company can produce materially more clinical output without proportional headcount growth, the old distinction between a "technology company" and a "services company" becomes less useful.

---

# Research Pillar 40: Care Orchestration as the Product

## Research Question

What if the defensible product is not an individual technology, but the way a care delivery organization composes technologies into outcomes?

A next-generation care company may combine:

```text
AI
+
Diagnostics
+
Devices
+
Drugs
+
Remote monitoring
+
Human clinicians
+
Automation
+
Data
+
Workflow
```

into a single care pathway.

## Core Thesis

The composition may become more valuable than any individual component.

## Research Questions

- Which components should the care company own?
- Which should it license?
- Which should it treat as commodities?
- What sequencing creates better outcomes?
- What data improve the composition over time?
- Which workflow components create switching costs?
- Can the care model be standardized without commoditizing clinical judgment?
- Does the orchestration layer become a form of operating IP?
- Can care pathways become adaptive through AI rather than fixed?

## Native Alpha™ Question

> What unusual combination of widely available technologies can produce an outcome others cannot easily reproduce?

---

# Research Pillar 41: Care Delivery Companies as Commercialization Channels

## Research Question

Can next-generation care delivery organizations become preferred commercialization channels for emerging drugs, devices, diagnostics, and AI products?

Today many healthcare technology companies attempt to sell independently into thousands of hospitals, physician groups, and health systems.

An alternative model is:

```text
Emerging technology
      ↓
Technology-absorbing care platform
      ↓
Integrated workflow
      ↓
Patient access
      ↓
Measured outcomes
```

## Research Questions

- Which care companies can absorb new technology quickly?
- Can a care platform standardize technology evaluation?
- Can it reduce the need for each vendor to build its own enterprise sales motion?
- Can it become a launch partner for emerging technologies?
- Can it earn economics for distribution, implementation, or outcomes?
- Can technology companies become component suppliers rather than full-stack commercial organizations?
- Can care delivery companies create procurement leverage by becoming high-volume technology integrators?
- Could a care company become the preferred partner for clinical validation and commercialization?

## Strategic Analogy

A care delivery company may increasingly resemble a systems integrator or manufacturer that assembles many components into a finished outcome.

---

# Research Pillar 42: Patient Relationship as Distribution

## Research Question

Does direct or durable access to patients become one of the most valuable forms of healthcare distribution?

Technology companies frequently struggle with patient acquisition and institutional access.

Care delivery companies already sit inside the patient journey.

## Research Questions

- Which care models create recurring patient relationships?
- Which specialties control high-value decision points?
- Which providers see patients before treatment selection?
- Which models can identify patients longitudinally?
- Does owning the patient relationship reduce GTM cost?
- Can patient access become a moat?
- How should patient trust be protected?
- When does patient access become more valuable than technical IP?

## Core Principle

A market is not simply a population that epidemiology says exists.

A market is a population that can actually be identified, reached, engaged, and served.

---

# Research Pillar 43: From Technology Margin to Outcome Margin

## Research Question

Where does the economic value sit when the technology itself becomes cheap?

Historically, technology vendors have often captured attractive margins by selling scarce capabilities.

If those capabilities become inexpensive, the larger opportunity may move into the redesigned care model.

## Illustrative Logic

```text
Traditional model:
Technology vendor sells a component.
Care provider delivers the service.

AI-native model:
Technology cost falls.
AI lowers delivery cost.
Care model captures more of the value created by the redesigned workflow.
```

## Research Questions

- Can AI materially reduce cost per episode of care?
- Can a provider retain part of the resulting economic gain?
- Can outcomes-based reimbursement increase value capture?
- Can the care company participate in shared savings?
- Does technology ownership matter if access is inexpensive?
- What economics accrue to the orchestrator versus component suppliers?
- Which models create scalable outcome margin?

## Research Output

A framework comparing:

- technology margin,
- service margin,
- workflow margin,
- outcome margin,
- data value,
- distribution value.

---

# Research Pillar 44: The Valuation Reversal Hypothesis

## Research Question

Under what conditions could an AI-native care delivery company deserve technology-like strategic value?

The research should not assume:

```text
Technology company = high multiple
Care delivery company = low multiple
```

Instead, identify the characteristics that historically drove the difference.

## Traditional Care Delivery Characteristics

- high labor intensity,
- local operations,
- low automation,
- fragmented workflows,
- weak data,
- poor differentiation,
- limited operating leverage,
- fee-for-service dependence.

## AI-Native Care Delivery Characteristics

- high automation,
- AI orchestration,
- structured pathways,
- longitudinal data,
- technology composability,
- lower marginal labor,
- patient-discovery capability,
- standardized operating model,
- measurable outcomes,
- faster technology absorption.

## Research Questions

- Which characteristics actually explain valuation differences?
- Which can AI change?
- Which cannot AI change?
- What level of automation creates operating leverage?
- What evidence would justify a higher valuation multiple?
- Does recurring patient engagement matter?
- Does owned distribution matter?
- Does proprietary data matter?
- Does outcome accountability matter?
- Can care delivery scale without proportionate physical footprint growth?
- What investor assumptions would need to change?

## Hypothesis

An AI-native care company with strong patient access, workflow control, data, repeatable economics, and technology absorption may be economically closer to a platform than to a traditional services business.

---

# Research Pillar 45: Build, Buy, License, or Consume

## Research Question

In an environment of abundant technology, what should the care delivery company actually own?

The default should not be ownership.

## Decision Options

- build internally,
- buy,
- license,
- partner,
- consume as commodity,
- use open technology,
- acquire only rights necessary for the care model.

## Research Questions

- Is ownership required for differentiation?
- Does the technology create strategic dependence?
- Is exclusivity necessary?
- Can multiple suppliers reduce risk?
- Does owning the technology increase financeability?
- Does it create unnecessary capital burden?
- Is the durable value in the technology or in how it is used?
- Which background IP must be secured?
- Which field-of-use rights are sufficient?

## Core Principle

Abundance should push care delivery organizations toward selective ownership.

Own what creates Native Alpha™.

Consume what has become commodity infrastructure.

---

# Research Pillar 46: Care Delivery as a Compounding Learning System

## Research Question

Can the care delivery organization become the best place to learn which combinations of technologies actually work?

Care delivery generates longitudinal feedback that standalone technology companies often cannot see.

The organization can observe:

```text
Patient state
→ technology selection
→ care pathway
→ implementation
→ response
→ adherence
→ outcome
→ cost
```

## Research Questions

- Which outcome data are captured routinely?
- Which component choices can be compared?
- Can AI learn optimal technology combinations?
- Can the care pathway improve continuously?
- Does accumulated experience strengthen procurement power?
- Does the learning system become proprietary?
- Can better outcomes attract more patients, technologies, and partners?

## Compounding Loop

```text
More patients
→ more observations
→ better orchestration
→ better outcomes
→ stronger reputation
→ more patients
```

---

# Research Pillar 47: Care Delivery and Unbundling Work

## Research Question

How does AI-driven unbundling of work change the design of the care delivery company itself?

Traditional healthcare organizations bundle many responsibilities into credentials, departments, and job descriptions.

AI allows the organization to decompose work into:

- tasks,
- decisions,
- accountability,
- communication,
- execution,
- monitoring,
- exceptions.

Then recombine those elements differently.

## Research Questions

- Which tasks truly require licensed clinicians?
- Which require professional accountability but can be AI-assisted?
- Which can be automated?
- Which can move to lower-cost roles?
- Which can be centralized?
- Which can occur asynchronously?
- How does unbundling affect liability?
- Can work composition reduce malpractice or operational risk?
- What new roles emerge?
- What happens to staffing ratios?

## Strategic Model

```text
Abundant technology
+
Unbundled human work
+
AI orchestration
+
Clinical accountability
+
Patient access
+
Reimbursement
=
AI-native care delivery
```

---

# Research Pillar 48: The Care Delivery Native Alpha™ Test

## Research Question

What Native Alpha™ can exist in care delivery when technology itself is widely available?

Potential sources include:

- patient access,
- referral position,
- cohort identification,
- clinical pathway design,
- unique technology combinations,
- workflow control,
- payer relationships,
- specialty expertise,
- evidence generation,
- proprietary longitudinal data,
- lower operating cost,
- faster adoption of new technologies.

## Native Alpha™ Questions

- What can this care company know earlier?
- What can it see that technology vendors cannot?
- What can it decide more accurately?
- What can it do faster?
- What technologies can it combine unusually well?
- What patient populations can it reach?
- What outcomes can it produce more reliably?
- What compounds as volume grows?

## Core Insight

The Native Alpha™ may lie in composition, not invention.

---

# Cross-Cutting Research Theme: Care Delivery as the Consumer of Abundance

The research area should explicitly study the possibility that healthcare's strategic center of gravity moves toward organizations that are unusually good consumers of technology.

The most important care company may not invent:

- the AI model,
- the diagnostic,
- the device,
- the drug,
- the monitoring platform.

It may instead be unusually good at deciding:

- which one to use,
- for which patient,
- at what time,
- in what sequence,
- with which human intervention,
- under which reimbursement model,
- with which measurable outcome.

This produces a broader research hypothesis:

> The biggest healthcare companies of the AI era may not be companies that make healthcare technology. They may be companies that are unusually good at consuming abundant technology and turning it into care.

This hypothesis should be tested, not assumed.

---

# Care Delivery Opportunity Screen

For every care delivery opportunity, score:

## Patient Access
- Can the organization reliably identify and reach patients?

## Workflow Control
- Does it control meaningful clinical action points?

## Technology Absorption
- Can it incorporate new technologies quickly?

## AI Leverage
- Can AI materially reduce labor intensity?

## Reimbursement
- Is payment aligned with the care model?

## Outcome Measurement
- Can outcomes be measured continuously?

## Data Compounding
- Does care generate proprietary learning?

## Geographic Scalability
- Can the model expand beyond one local market?

## Clinical Standardization
- Can enough of the pathway be standardized to scale?

## Human Scarcity
- Which roles remain structurally scarce?

## Native Alpha™
- What is difficult for competitors to reproduce?

## Capital Intensity
- Does growth require proportionate physical assets or headcount?

---

# New Strategic Comparison: Technology Company vs. AI-Native Care Company

| Dimension | Traditional Technology Company | Traditional Care Delivery | AI-Native Care Delivery |
|---|---|---|---|
| Primary asset | Technology | Clinical labor and facilities | Patient access + orchestration + data |
| Labor intensity | Low to moderate | High | Potentially materially lower |
| Distribution | Must be built | Already inside care | Embedded in care |
| Patient access | Indirect | Direct | Direct and increasingly data-driven |
| Workflow control | Limited | High | High |
| Reimbursement | Often indirect | Core | Core and potentially optimized |
| Technology ownership | Central | Consumed | Selective |
| Data | Product usage | Clinical | Longitudinal clinical + operational |
| Marginal cost | Low | Historically high | Potentially declining |
| Operating leverage | High | Historically low | Potentially increasing |
| Defensibility | IP / product | Local relationships | Workflow + data + patient access + orchestration |
| AI leverage | Product creation | Administrative augmentation | Full operating model redesign |

This table should be treated as a hypothesis framework, not a predetermined conclusion.

---

# New Long-Term Research Questions

Add the following to the long-term research agenda:

- Does care delivery become more valuable as technology becomes less scarce?
- Which healthcare technologies are likely to commoditize first?
- Which forms of patient access become strategic assets?
- Can AI-native care delivery achieve software-like operating leverage?
- Which specialties are best suited to technology orchestration?
- Can care companies become commercialization platforms for emerging technology?
- Will technology companies increasingly become suppliers to care platforms?
- When should a care company own technology versus simply consume it?
- Can outcome margins exceed technology margins?
- Can a care company capture economics from integrating multiple third-party products?
- What would justify technology-like valuation multiples for care delivery?
- Does care delivery become the natural home for proprietary longitudinal learning?
- Which care models are capable of continuous technology substitution without disrupting patient experience?
- Can an AI-native care company absorb innovation faster than a traditional health system?
- Does the organization that controls the care pathway become more strategically important than the organization that owns any individual tool?

---

# Revised Strategic Thesis

The research area should now examine three interacting forms of scarcity:

## Technical Scarcity

Historically valuable because difficult technologies were expensive to create.

AI increasingly reduces this scarcity.

## Adoption Scarcity

The ability to get a solution understood, approved, implemented, reimbursed, and routinely used.

This may become more valuable as technical supply increases.

## Care Delivery Scarcity

The ability to reach patients, exercise clinical authority, orchestrate multiple capabilities, accept accountability, and convert abundant technology into measurable outcomes.

This may become the ultimate downstream scarcity.

The resulting model is:

```text
AI abundance
      ↓
Technology abundance
      ↓
Technical differentiation compresses
      ↓
Adoption becomes more important
      ↓
Patient access and workflow become more important
      ↓
Care delivery becomes the integration layer
      ↓
Outcomes, data, and orchestration compound
```

The research should therefore avoid assuming that technology companies permanently sit at the highest-value point in the healthcare stack.

The location of value should itself be treated as an empirical research question.

# Research Infrastructure Layer: Public Signal Intelligence

## Purpose

Public-sector and public-domain data should be treated as a primary research infrastructure for this research area.

Before inventing a market thesis, ask where governments, payers, researchers, regulators, providers, and public institutions are already:

- spending money,
- funding research,
- testing payment models,
- running clinical trials,
- approving products,
- measuring quality,
- collecting utilization data,
- exposing geographic variation,
- publishing evidence,
- creating codes,
- documenting provider capacity,
- identifying workforce shortages.

The research discipline should be:

> Follow the money.

> Follow the experiments.

> Follow the grants.

> Follow the trials.

> Follow the evidence.

> Follow the codes.

> Follow the providers.

> Follow the regulatory activity.

Then ask where those signals disagree.

The disagreement may reveal Native Alpha™.

---

# Public Signal Intelligence Thesis

The healthcare market is unusually visible because so much of it is publicly financed, publicly regulated, publicly researched, or publicly measured.

AI changes what can be done with that visibility.

Historically, connecting CMS spending, CMMI experiments, NIH grants, PubMed evidence, ClinicalTrials.gov, FDA product status, provider infrastructure, patents, and commercial signals required large research teams.

The research area should now assume that cross-source synthesis is a machine task.

The human task is to decide whether the pattern is meaningful.

## Core Principle

Public data should not merely support a market-size slide.

It should drive opportunity discovery.

---

# Public Healthcare Opportunity Graph

The research program should build and maintain a conceptual and eventually machine-readable opportunity graph.

```text
Disease / Condition
      ↓
Patient Population
      ↓
CMS Utilization
      ↓
CMS Spending
      ↓
Providers / Institutions
      ↓
Quality / Outcomes
      ↓
CMMI Experiments
      ↓
NIH Funding
      ↓
Publications / Evidence
      ↓
Clinical Trials
      ↓
FDA Products
      ↓
Patents
      ↓
Reimbursement
      ↓
Care Pathways
      ↓
Commercial Activity
```

The purpose of the graph is not simply data aggregation.

The purpose is to reveal:

- expensive failure,
- underused technology,
- scientific acceleration,
- reimbursement readiness,
- adoption friction,
- care-delivery gaps,
- provider capacity,
- translational opportunities,
- emerging categories,
- places where value may migrate downstream.

---

# Signal Source 1: CMS

## Research Question

Where is CMS already spending money, and where does that spending coexist with poor outcomes, inefficiency, fragmentation, or unmet need?

CMS should be treated as one of the strongest public market-sizing and utilization sources because it reflects actual paid healthcare activity.

## Candidate Signals

- spending by service,
- spending by drug,
- beneficiary counts,
- utilization,
- geography,
- site of service,
- provider concentration,
- quality,
- chronic-condition burden,
- telehealth usage,
- hospital performance,
- physician and supplier activity,
- Part B drug spending,
- Part D drug spending,
- Medicaid drug spending,
- Medicare Advantage and fee-for-service patterns where public data permit.

## Research Questions

- Where is CMS writing the largest checks?
- Where is spending rising fastest?
- Where is utilization increasing without corresponding improvement?
- Where does geographic variation suggest inconsistent care?
- Where are expensive drugs associated with meaningful nonresponse or failure?
- Which services have high cost but poor patient experience or operational burden?
- Where is spend concentrated enough to support a focused care-delivery model?
- Which specialties offer a large reimbursed market but fragmented delivery?

## Market Discovery Pattern

```text
High CMS spending
+
Poor outcomes or inefficiency
+
Fragmented delivery
=
Potential target for enablement or care-delivery redesign
```

---

# Signal Source 2: CMMI

## Research Question

What healthcare problems is CMS actively trying to change?

CMMI should be treated as a public R&D laboratory for healthcare business models.

CMMI data and evaluations can reveal:

- which problems CMS believes are economically important,
- which payment structures are being tested,
- which provider types participate,
- what worked,
- what failed,
- what changed spending,
- what changed utilization,
- what changed quality,
- what changed provider behavior,
- what implementation barriers emerged.

## Research Questions

- Which CMMI models are expanding?
- Which are ending?
- Which produced measurable savings?
- Which improved quality?
- Which failed operationally despite sound policy logic?
- Which model failures reveal product opportunities?
- Which payment experiments could create future markets?
- Which care-delivery organizations are already adapting successfully?

## Opportunity Pattern

```text
CMMI policy priority
+
Operational failure
+
Clear implementation bottleneck
=
Potential product or care-delivery opportunity
```

---

# Signal Source 3: NIH

## Research Question

Where is public scientific capital accumulating before commercial markets fully form?

NIH funding should be treated as an upstream market signal.

## Candidate Signals

- grant volume,
- grant dollars,
- funding growth,
- investigators,
- institutions,
- scientific terms,
- mechanisms,
- publications,
- patents,
- repeated funding over time,
- translational-stage awards.

## Research Questions

- Which scientific fields are accelerating?
- Which areas receive sustained funding without obvious commercialization?
- Which investigators and institutions dominate a field?
- Which mechanisms have accumulated evidence for years?
- Which NIH-funded work generated patents but little deployment?
- Which research clusters are nearing clinical translation?
- Where has government already absorbed much of the early scientific risk?

## Opportunity Pattern

```text
Large cumulative public research investment
+
Credible mechanism
+
Maturing evidence
+
Weak commercialization
=
Potential translational opportunity
```

---

# Signal Source 4: NLM and PubMed

## Research Question

Where is biomedical knowledge accumulating, converging, or contradicting itself?

NLM resources should provide the structured knowledge layer.

## Candidate Sources

- PubMed,
- MeSH,
- RxNorm,
- MedlinePlus,
- other structured NLM terminology and knowledge resources.

## Research Questions

- Which topics show accelerating publication volume?
- Which terminology is emerging?
- Which mechanisms are becoming accepted?
- Which evidence remains contradictory?
- Which concepts are connected across apparently different specialties?
- Where does MeSH reveal relationships that keyword search misses?
- Which areas have many publications but little translational movement?
- Which areas have clinical demand with weak evidence depth?

## Strategic Role

NLM should help normalize terminology across public datasets so the research system can reason across diseases, mechanisms, drugs, trials, and technologies.

---

# Signal Source 5: ClinicalTrials.gov

## Research Question

Where is science becoming intervention, and where does translation stall?

Clinical trial activity can reveal:

- therapeutic interest,
- sponsor activity,
- indication expansion,
- intervention classes,
- combination strategies,
- recruitment friction,
- endpoint choices,
- geographic activity,
- repeated failure,
- trial density.

## Research Questions

- Which indications have rapidly increasing trial activity?
- Which interventions repeatedly fail?
- Which categories show intense trial activity but little routine adoption?
- Which trials struggle to recruit?
- Which technologies are repeatedly combined?
- Which sponsors are entering or exiting a category?
- Where is evidence accumulating faster than commercial infrastructure?

## Opportunity Pattern

```text
High trial activity
+
Weak routine adoption
=
Potential market-conditioning or care-delivery problem
```

---

# Signal Source 6: FDA and OpenFDA

## Research Question

What regulated products already exist, and where is the market underusing them?

FDA and openFDA should be used to understand:

- approved drugs,
- cleared or approved devices,
- indications,
- adverse events,
- safety signals,
- category crowding,
- regulatory precedent,
- product status.

## Research Questions

- Which categories already contain many viable products?
- Which approved products appear underused?
- Where is there substantial regulatory precedent but weak adoption?
- Where are adverse-event burdens creating opportunities for conditioning or redesign?
- Which product classes are becoming technically abundant?
- Which products could be components inside a care-delivery platform rather than standalone businesses?

## Strategic Hypothesis

If many regulated solutions already exist but adoption remains weak, the problem may not be invention.

It may be:

- patient discovery,
- workflow,
- reimbursement,
- implementation,
- market conditioning,
- care-delivery integration.

---

# Signal Source 7: Provider and Institutional Infrastructure

## Research Question

Who can actually deliver the solution?

Public provider data should map the infrastructure side of the market.

## Candidate Sources

- CMS provider data,
- NPPES,
- Care Compare,
- facility datasets,
- physician and supplier files,
- publicly available state licensure and facility sources where appropriate.

## Research Questions

- Which institutions treat the relevant patients?
- Which specialties are geographically concentrated?
- Where are provider shortages?
- Which organizations have the capabilities required for a new therapy?
- Which regions have the greatest mismatch between patient need and delivery capacity?
- Which institutions are likely early adopters?
- Which provider networks could support a scalable care model?

## Core Principle

Epidemiology does not equal an executable market.

The research system should ask:

> Where are the patients, who sees them, who can treat them, and which institutions can absorb the solution?

---

# Signal Source 8: HRSA, CDC, and AHRQ

## HRSA

Use HRSA to study:

- workforce shortages,
- health professional shortage areas,
- rural access,
- safety-net infrastructure,
- federally qualified health centers,
- maternal and child health programs,
- underserved populations.

## CDC

Use CDC to study:

- disease burden,
- mortality,
- incidence,
- prevalence,
- behavioral risk,
- geographic disparities,
- public-health trends.

## AHRQ

Use AHRQ to study:

- utilization,
- healthcare quality,
- patient safety,
- delivery-system research,
- cost and access,
- health-services evidence.

## Research Questions

- Where is clinical need concentrated?
- Where is provider capacity weak?
- Which delivery models could close the gap?
- Which conditions have large burden but fragmented care?
- Where can AI-native care delivery address shortage rather than merely compete for existing patients?

---

# Signal Source 9: USAspending.gov and SAM.gov

## Research Question

Where is the federal government actively procuring or funding capabilities?

Use federal spending and procurement data to study:

- contract awards,
- agencies,
- recurring purchasing,
- vendor concentration,
- procurement trends,
- solicitations,
- emerging requirements.

## Research Questions

- Which healthcare capabilities are being repeatedly purchased?
- Which problems are agencies actively seeking to solve?
- Where is procurement demand growing?
- Which categories have few incumbents?
- Which recurring solicitations reveal durable institutional needs?

---

# Signal Source 10: USPTO and Patent Activity

## Research Question

Where are organizations attempting to protect future technical or workflow advantage?

Patent activity should be treated as one signal among many.

## Research Questions

- Which technical clusters are attracting filings?
- Which areas show crowded IP but weak commercial adoption?
- Which NIH-funded concepts generated patent families?
- Where is patenting accelerating before market formation?
- Which claims appear disconnected from actual customer demand?
- Which patent families may reveal unusual problem framing worth investigating?

## Core Principle

Patent activity is a signal of effort and perceived strategic importance.

It is not proof of value.

---

# Signal Source 11: Open Payments and Relationship Signals

## Research Question

Where are manufacturers already building economic or educational relationships with clinicians?

Public transparency data may reveal:

- consulting relationships,
- research payments,
- educational activity,
- specialty concentration,
- emerging KOL networks.

## Research Questions

- Which clinicians are central to an emerging category?
- Which specialties attract concentrated manufacturer attention?
- Where are relationships forming before broader market adoption?
- Which institutions may already be shaping category direction?

## Discipline

Relationship data should be interpreted carefully.

Payments indicate interaction, not necessarily influence or endorsement.

---

# Signal Source 12: State-Level Public Data

Where useful and legally available, incorporate:

- Medicaid data,
- all-payer claims databases,
- public-health data,
- facility data,
- state licensing data,
- state-level quality reporting,
- state procurement.

## Research Question

Where do local market structures differ enough to create or destroy an opportunity?

National averages can hide the places where the opportunity is actually executable.

---

# Public Signal Taxonomy

The research infrastructure should normalize sources into a common question set.

```text
CMS
What are we paying for?

CMMI
What are we trying to change?

NIH
What science are we funding?

NLM / PubMed
What are we learning?

ClinicalTrials.gov
What are we testing?

FDA
What regulated products already exist?

Providers
Who can deliver it?

CDC / HRSA / AHRQ
Who needs it and where are the gaps?

USPTO
What are people trying to protect?

USAspending / SAM
What is government buying?

Commercial signals
Who is trying to monetize it?
```

---

# Signal Intersection Research

The most valuable discoveries may come from cross-source combinations rather than any individual dataset.

## Pattern 1: Expensive Failure

```text
High CMS spend
+
Poor outcomes
+
High utilization
=
Expensive failure worth investigating
```

## Pattern 2: Translational Gap

```text
High NIH funding
+
Strong publication activity
+
Weak commercial adoption
=
Possible translational opportunity
```

## Pattern 3: Adoption Failure

```text
FDA-cleared products
+
Clinical evidence
+
Low utilization
=
Possible market-conditioning problem
```

## Pattern 4: Emerging Reimbursement Opportunity

```text
CMMI experimentation
+
Improving evidence
+
Provider participation
=
Potential future payment model
```

## Pattern 5: AI-Native Care Delivery Opportunity

```text
High CMS spending
+
Fragmented provider base
+
Strong reimbursement
+
High administrative burden
+
Available technology
+
Large geographic variation
+
AI-automatable labor
=
Potential AI-native care delivery company
```

## Pattern 6: Infrastructure Gap

```text
High disease burden
+
Provider shortage
+
Available technology
+
Weak local care capacity
=
Potential new delivery model
```

## Pattern 7: Market Conditioning Opportunity

```text
Strong science
+
Active trials
+
Weak category understanding
+
Low adoption
=
Potential market-conditioning opportunity
```

---

# Public Signal Intelligence Scoring

Every opportunity should receive a public-signal score.

## Spending Signal

- current spend,
- spend growth,
- utilization,
- reimbursement maturity.

## Scientific Signal

- NIH funding,
- publication growth,
- investigator density,
- evidence maturity.

## Clinical Translation Signal

- trial activity,
- trial outcomes,
- regulatory precedent,
- approved or cleared products.

## Delivery Signal

- provider capacity,
- geographic concentration,
- workforce gaps,
- institutional readiness.

## Policy Signal

- CMMI models,
- CMS priorities,
- reimbursement experimentation,
- public procurement.

## IP Signal

- patent activity,
- ownership clarity,
- licensing potential,
- freedom-to-operate questions.

## Commercialization Signal

- incumbent activity,
- partnerships,
- acquisitions,
- public-company commentary,
- procurement.

---

# Public Signal Intelligence as a Native Alpha™ Capability

The opportunity is not merely access to open data.

Everyone can access open data.

Native Alpha™ may arise from:

- knowing which sources to connect,
- knowing which discrepancies matter,
- knowing which patterns predict opportunity,
- knowing how to translate a signal into a narrow experiment,
- building a reusable graph of relationships,
- accumulating proprietary interpretations,
- linking public signals to private market conversations,
- learning which public indicators actually correlate with adoption.

The defensible asset may become the interpretation layer.

---

# AI Workforce™ for Public Signal Intelligence

Public signal research should be largely AI-native.

## Core AI Workforce™ Functions

### Data Discovery
- identify relevant public datasets,
- monitor new releases,
- detect schema changes,
- maintain source inventories.

### Data Normalization
- normalize terminology,
- map codes,
- reconcile drug names,
- reconcile provider identities,
- map diseases and MeSH concepts,
- map geography.

### Signal Detection
- detect spending growth,
- detect unusual geographic variation,
- detect grant acceleration,
- detect trial acceleration,
- detect provider concentration,
- detect regulatory activity.

### Opportunity Matching
- connect failure mechanisms to reimbursed markets,
- connect NIH research to unmet commercial need,
- connect FDA-cleared products to weak adoption,
- connect provider shortages to AI-native care opportunities.

### Evidence Synthesis
- summarize supporting evidence,
- identify contradictions,
- assign confidence,
- maintain source traceability.

### Market Mapping
- identify institutions,
- identify investigators,
- identify likely KOLs,
- identify target manufacturers,
- identify target providers,
- identify government buyers.

### Research Maintenance
- refresh opportunity scores,
- flag material changes,
- update kill criteria,
- recommend next experiments.

---

# Public Signal Intelligence Data Discipline

Open data is useful because it is inspectable and repeatable.

It is also dangerous when interpreted carelessly.

## Do Not Infer

- CMS spending does not prove value.
- NIH funding does not prove demand.
- CMMI experimentation does not guarantee future reimbursement.
- Publication volume does not prove efficacy.
- Clinical trials do not prove commercialization potential.
- FDA clearance does not prove adoption.
- Patent activity does not prove defensibility.
- Provider counts do not prove available capacity.
- Open Payments does not prove influence.

Each source is a signal.

The research value comes from triangulation.

---

# Updated Opportunity Discovery Engine

The Opportunity Discovery Engine should begin with public signal ingestion rather than open-ended brainstorming.

## Step 1: Detect Public Signals

Identify:

- high spending,
- high growth,
- poor outcomes,
- high scientific funding,
- trial acceleration,
- regulatory activity,
- workforce shortage,
- geographic variation,
- provider concentration,
- reimbursement change.

## Step 2: Build the Opportunity Graph

Connect:

```text
Condition
→ Population
→ Spend
→ Evidence
→ Technology
→ Provider
→ Payment
→ Workflow
→ Buyer
→ Rights
```

## Step 3: Identify Mismatches

Examples:

- money without outcomes,
- science without commercialization,
- products without adoption,
- patients without providers,
- reimbursement without workflow,
- provider capacity without technology,
- patents without demand.

## Step 4: Generate Hypotheses

Ask:

- Is this a therapeutic enablement opportunity?
- Is this a market-conditioning opportunity?
- Is this a patient-discovery opportunity?
- Is this a care-delivery opportunity?
- Is this a licensing opportunity?
- Is this an IP opportunity?
- Is this a reimbursement-engineering opportunity?

## Step 5: Test Cheaply

Use:

- clinician interviews,
- provider mapping,
- account outreach,
- patient cohort analysis,
- literature synthesis,
- reimbursement review,
- KOL validation,
- licensing discussions.

---

# Updated Market Sizing Methodology

Market sizing should rely first on real-world public activity where possible.

Avoid beginning with generic TAM reports.

## Preferred Sequence

```text
Actual utilization
→ Actual public spending
→ Patient counts
→ Provider counts
→ Geographic concentration
→ Reimbursement
→ Addressable workflow
→ Adoption constraints
→ Realistic serviceable market
```

## Research Questions

- How much is already being spent?
- On what?
- For which patients?
- By which providers?
- In which geographies?
- Under which codes?
- With what outcomes?
- With what variation?
- How much of the spend is realistically influenceable?

This approach should be used for both technology opportunities and care-delivery opportunities.

---

# Updated Care Delivery Opportunity Screen

Add the following public-signal dimensions to the care-delivery screen.

## CMS Spend Density
Is there enough current reimbursed activity?

## Geographic Concentration
Can a care model reach meaningful patient density?

## Provider Fragmentation
Is delivery fragmented enough to create room for a better operating model?

## Workforce Scarcity
Are staffing constraints severe enough to reward AI-native redesign?

## Technology Readiness
Do useful devices, diagnostics, drugs, and software already exist?

## Reimbursement Readiness
Can the care model be paid today?

## CMMI Alignment
Is policy experimentation moving toward the model?

## NIH / Evidence Depth
Is the clinical foundation mature?

## Regulatory Precedent
Are required products already regulated or is there clear precedent?

## Adoption Gap
Is there a visible difference between available technology and actual use?

---

# Updated AI-Native Research Workflow

The research workflow should operate as follows:

```text
Public data ingestion
      ↓
Normalization
      ↓
Opportunity graph
      ↓
Signal detection
      ↓
Native Alpha™ hypothesis
      ↓
Market / care-delivery hypothesis
      ↓
Rights analysis
      ↓
Cheap validation experiment
      ↓
Decision
      ↓
Continuous monitoring
```

This should be treated as a standing research capability rather than a one-time project.

---

# Updated Research Questions for Care Delivery as the Integration Layer

The care-delivery research should explicitly use public data to answer:

- Which specialties already have large CMS reimbursement pools?
- Which have high geographic variation?
- Which are highly fragmented?
- Which have workforce shortages?
- Which have high administrative burden?
- Which have strong technology availability?
- Which have underused FDA-cleared tools?
- Which have active CMMI experimentation?
- Which have strong NIH-funded evidence but weak delivery infrastructure?
- Which could support an AI-native care model immediately?

This makes care-delivery company formation an evidence-driven exercise rather than a thematic thesis.

---

# Public Signal Intelligence Research Outputs

The research program should generate reusable outputs including:

- CMS spending heat maps,
- CMMI model opportunity maps,
- NIH funding trend maps,
- investigator and institution networks,
- ClinicalTrials.gov activity maps,
- FDA product-density maps,
- provider-capacity maps,
- workforce-shortage maps,
- reimbursement maps,
- patent-density maps,
- opportunity graphs,
- cross-signal scorecards,
- AI-native care-delivery opportunity lists,
- translational-gap opportunity lists,
- market-conditioning opportunity lists.

---

# Public Signal Intelligence Governance

Every public-data-derived conclusion should include:

- source,
- dataset,
- date,
- time period,
- coverage,
- known limitations,
- confidence,
- interpretation,
- contradictory signals,
- next validation step.

The research system should preserve raw evidence separately from interpretation.

AI should summarize and connect signals.

Humans should remain responsible for deciding whether the signal changes a real commercial decision.

---

# Revised Research Doctrine Additions

Add the following operating rules:

> Start with public evidence before starting with opinion.

> Prefer actual spend over market-size estimates.

> Prefer actual utilization over survey intent.

> Prefer real provider capacity over epidemiological abstraction.

> Treat CMMI as a window into future payment and delivery experiments.

> Treat NIH as a map of where scientific risk has already been publicly financed.

> Treat NLM and PubMed as the structured knowledge layer.

> Treat ClinicalTrials.gov as the bridge between science and intervention.

> Treat FDA data as a test of whether invention is actually still the bottleneck.

> Treat provider data as a test of whether a market is executable.

> Look for disagreement between public signals.

> The mismatch is often more interesting than the average.

> Use AI to connect the public graph.

> Use humans to decide what matters.

# Research Operating System: Evidence, Falsifiability, and Action

The research area should operate as a decision system, not as a content-generation system.

The standard is not whether a document is comprehensive.

The standard is whether uncertainty has been reduced enough to make a better decision.

Every material claim should be testable.
Every major hypothesis should have disconfirming evidence defined in advance.
Every opportunity should have a cheapest next test.
Every experiment should change a decision.
Every research stream should ultimately produce action, abandonment, or a deliberate decision to wait.

---

# Evidence Architecture

Evidence should be separated into independent tracks.

A strong score in one track must not compensate for weakness in another.

## 1. Scientific Evidence

Question:

Does the mechanism, intervention, or clinical thesis actually work?

Evidence may include:

- mechanistic studies,
- observational studies,
- randomized trials,
- meta-analyses,
- real-world evidence,
- clinical guidelines,
- replicated findings,
- biological plausibility,
- contradictory evidence.

## 2. Market Evidence

Question:

Is the problem important enough that someone will allocate scarce resources to solve it?

Evidence may include:

- actual spending,
- procurement,
- budget allocation,
- customer interviews,
- willingness to share data,
- willingness to run pilots,
- design-partner commitments,
- payment commitments,
- partnership interest.

## 3. Adoption Evidence

Question:

Can the solution actually enter real workflow?

Evidence may include:

- workflow mapping,
- implementation tests,
- stakeholder approval,
- procurement acceptance,
- security review,
- training completion,
- repeated use,
- time to activation,
- use outside founder-led support.

## 4. Economic Evidence

Question:

Does the solution create enough value, and can enough of that value be captured?

Evidence may include:

- unit economics,
- cost avoidance,
- incremental revenue,
- margin improvement,
- reimbursement,
- shared savings,
- manufacturer value uplift,
- cost per acquired patient,
- implementation cost,
- retention economics.

## 5. Strategic Evidence

Question:

Is there a durable advantage worth building around?

Evidence may include:

- rights position,
- distribution advantage,
- patient access,
- workflow ownership,
- data advantage,
- switching cost,
- category position,
- proprietary know-how,
- Native Alpha™,
- compounding advantage.

## Evidence Rule

An opportunity should not advance merely because its average score is attractive.

Material weakness in any essential evidence track should remain visible.

---

# Evidence Ladder

Research should rank evidence by how much real-world commitment it reflects.

A useful progression is:

```text
Public signal
→ Literature support
→ Expert validation
→ Retrospective data
→ Customer interviews
→ Access commitment
→ Data-sharing commitment
→ Pilot commitment
→ Implementation
→ Repeat use
→ Payment
→ Measurable outcome
→ Replicated outcome
```

## Demand Proof Ladder

Words are weaker than access.

Access is weaker than implementation.

Implementation is weaker than repeat use.

Repeat use is weaker than payment.

Payment is weaker than demonstrated outcomes.

Demonstrated outcomes at one site are weaker than replicated outcomes across settings.

## Research Discipline

Do not treat:

- positive interview feedback,
- conference enthusiasm,
- letters of support,
- advisory-board participation,
- pilot interest,

as equivalent to scarce-resource commitments.

A stronger signal requires the stakeholder to give up something scarce:

- money,
- data,
- implementation access,
- clinician time,
- workflow control,
- patient access,
- procurement effort,
- reputational commitment.

---

# Thesis-Level Falsifiability

The research area should define disconfirming evidence for its major theses before those theses become institutional beliefs.

## Thesis: Technical Abundance Creates Adoption Scarcity

Possible disconfirming evidence:

- regulated product development costs remain structurally high despite AI,
- validation remains the dominant cost,
- regulatory review becomes the binding constraint,
- evidence generation cannot be compressed materially,
- technical differentiation remains long-lived,
- commercialization remains easier than product development.

## Thesis: Care Delivery Captures More Value as Technology Becomes Abundant

Possible disconfirming evidence:

- care delivery labor intensity does not materially decline,
- patient acquisition remains prohibitively expensive,
- reimbursement compresses faster than costs,
- technology vendors retain workflow control,
- care companies cannot absorb technology quickly,
- capital intensity prevents scalable returns,
- investor valuation differences persist because underlying economics remain fundamentally different.

## Thesis: Distribution Becomes Native Alpha™

Possible disconfirming evidence:

- distribution channels remain easily replicable,
- incumbents commoditize access,
- customer switching costs remain low,
- patient access cannot be defended,
- technology ownership continues to dominate bargaining power.

## Thesis: Public Signal Intelligence Improves Opportunity Selection

Possible disconfirming evidence:

- public data are too stale,
- signal combinations produce excessive false positives,
- public activity poorly predicts commercial demand,
- proprietary relationships dominate opportunity discovery,
- interpretation cost exceeds value.

## Research Requirement

Each major thesis should maintain:

```text
Current belief
Supporting evidence
Disconfirming evidence
Confidence
Base rate
Kill threshold
Next test
```

---

# Counterfactual Discipline

Every material outcome claim should ask:

> What would have happened without the intervention?

Without counterfactual thinking, research can mistake correlation for causation.

## Required Comparisons

Where practical, compare against:

- historical baseline,
- current standard of care,
- current workflow,
- credible alternative intervention,
- matched cohort,
- external benchmark,
- control group.

## Example Questions

- Did cost fall because of AI, or because staffing changed?
- Did outcomes improve because of conditioning, or because patient selection changed?
- Did adoption improve because of market conditioning, or because reimbursement changed?
- Did the care model scale because of technology, or because it entered an unusually favorable geography?

## Rule

Do not assign causal credit without a credible comparison.

---

# Base-Rate Analysis

Every opportunity should include an outside-view analysis.

Before asking why this opportunity will succeed, ask how similar opportunities usually perform.

## Base-Rate Questions

- How often do similar therapies achieve routine adoption?
- How often do FDA-cleared devices achieve meaningful utilization?
- How often do NIH-funded discoveries reach commercialization?
- How often do CMMI-inspired models become durable businesses?
- How often do hospital pilots convert into scaled deployments?
- How often do care-delivery rollups improve margins sustainably?
- How often do new reimbursement pathways become durable?
- How often do new healthcare categories reach widespread adoption?

## Native Alpha™ Test

Native Alpha™ should explain why this opportunity can beat the base rate.

It should not be used to ignore the base rate.

---

# Evidence Provenance

Because AI performs substantial synthesis, every material research claim should be traceable.

## Required Provenance Fields

- source,
- source type,
- date,
- dataset version,
- time period,
- calculation,
- assumptions,
- confidence,
- contradictions,
- analyst or AI interpretation.

## Claim Types

The system should distinguish among:

### Observed Fact
Directly supported by a source.

### Derived Calculation
Computed from observed inputs.

### Expert Interpretation
Human judgment based on evidence.

### AI Inference
Machine-generated interpretation requiring validation.

### Commercial Hypothesis
A proposition that requires market testing.

## Core Principle

Never allow AI-generated synthesis to obscure the distinction between evidence and interpretation.

---

# Evidence Freshness and Expiration

Healthcare evidence changes.

Important findings should therefore have a freshness expectation.

## Examples

- CMS spending changes,
- reimbursement changes,
- guidelines change,
- competitors launch,
- FDA status changes,
- CMMI models end,
- trials report,
- patents issue or expire,
- providers merge,
- workforce conditions change.

## Required Fields

Every material evidence item should include:

- date observed,
- expected useful life,
- revalidation date,
- source refresh cadence.

## Research Rule

Stale evidence should be treated as evidence debt.

---

# Evidence Debt

Research organizations can accumulate evidence debt just as engineering organizations accumulate technical debt.

Evidence debt exists when decisions continue to depend on unresolved, stale, weak, or unsupported claims.

## Evidence Debt Categories

- unsupported claims,
- stale claims,
- single-source claims,
- unvalidated economics,
- unresolved ownership,
- uncertain reimbursement,
- unproven patient access,
- unproven workflow fit,
- unreplicated pilots,
- untested buyer assumptions,
- inferred regulatory paths,
- unvalidated AI conclusions.

## Evidence Debt Register

Each opportunity should maintain:

```text
Claim
Evidence status
Risk if wrong
Owner
Resolution method
Deadline
Decision blocked?
```

## Rule

Evidence debt should become visible and expensive.

Teams should not repeatedly advance on assumptions simply because they were never challenged.

---

# Experiment Design Standard

Every experiment should be designed around a decision.

## Required Experiment Fields

### Belief
What do we currently believe?

### Observable Test
What can we actually measure?

### Supporting Result
What outcome strengthens the belief?

### Disconfirming Result
What outcome weakens the belief?

### Threshold
What is the explicit pass / fail / continue threshold?

### Budget
How much time, money, or access are we willing to spend?

### Decision
What changes if the result is positive?

### Kill Decision
What changes if the result is negative?

## Example: Patient Identification Workflow

Weak test:

> Would your practice use an AI-native patient-finding workflow?

Better test:

> Will the practice give access to de-identified data?

Stronger test:

> Will the practice allow the cohort query to run?

Stronger:

> Will clinicians review the identified cohort?

Stronger:

> Will the practice contact eligible patients?

Stronger:

> Will the workflow be repeated next month?

Strongest:

> Will the practice pay for the workflow or share resulting economics?

## Core Principle

Experiments should test behavior, not enthusiasm.

---

# Cheapest Next Test

Every opportunity should always have a clearly defined cheapest next test.

The objective is to reduce the most important uncertainty with the least expenditure.

Examples:

- verify a reimbursement assumption,
- run a claims query,
- ask a hospital for data access,
- ask a manufacturer for a design partnership,
- test a cohort definition,
- validate workflow ownership,
- check chain of title,
- test willingness to pay,
- compare two implementation channels.

## Rule

Do not fund a large experiment when a smaller experiment can invalidate the same assumption.

---

# Replication Standard

A single successful customer, hospital, clinician, geography, or patient cohort is not evidence of scalability.

## Required Replication Questions

- Can another institution implement it?
- Can another clinician use it?
- Can another geography support it?
- Can another payer environment support it?
- Can another patient cohort benefit?
- Can implementation occur without founder involvement?
- Can the outcome be reproduced under less favorable conditions?

## Scale Threshold

Before calling a model scalable, require evidence of replication across at least one meaningful dimension.

For high-consequence opportunities, require replication across several.

---

# Decision Packages

Every serious opportunity should terminate in a standardized Decision Package.

## Required Sections

### Hypothesis
What do we believe?

### Evidence For
What supports it?

### Evidence Against
What contradicts it?

### Confidence
How strong is the evidence?

### Base Rate
How often do similar opportunities work?

### Native Alpha™
Why might we beat the base rate?

### Economic Model
Where is value created and captured?

### Real Buyer
Who actually allocates budget?

### Patient Access
How are patients identified and reached?

### Workflow Owner
Who controls the critical action point?

### Rights
What rights are required?

### Regulatory Path
What must be satisfied?

### Adoption Latency
How long to routine use?

### Primary Risk
What is most likely to make this fail?

### Kill Criteria
What causes us to stop?

### Cheapest Next Test
What should we do next?

### Decision
Proceed, test, partner, license, form company, wait, or stop.

---

# Practical Research Outcomes

Research should produce concrete decisions and actions.

Acceptable outcomes include:

- form a company,
- acquire a care-delivery business,
- partner with an existing care company,
- license a technology,
- acquire field-of-use rights,
- abandon a patent,
- continue prosecution,
- license a patent family,
- approach a manufacturer,
- approach a hospital,
- recruit a design partner,
- launch a care-delivery pilot,
- run a patient-identification study,
- build a narrow workflow,
- test reimbursement,
- perform deeper diligence,
- monitor for a defined trigger,
- stop.

## Failed Outcome

"Interesting research" is not an outcome.

A research program that produces more information but no better decision has failed.

---

# Research Decision States

Each opportunity should always have one current state.

## 1. Explore
Insufficient evidence to commit.

## 2. Validate
Specific material assumptions require testing.

## 3. Pilot
Enough evidence exists to justify real-world implementation.

## 4. Scale
Replicated evidence supports expansion.

## 5. Partner
The opportunity is better owned or commercialized with another party.

## 6. License
Rights are valuable but direct operation is unnecessary.

## 7. Incubate
The opportunity justifies company formation or dedicated operating capacity.

## 8. Monitor
The opportunity is not ready, but a defined external trigger may change the decision.

## 9. Stop
Evidence no longer supports further investment.

## Rule

Every opportunity should have:

- one state,
- one owner,
- one next decision,
- one next test.

---

# Practical Outcome Metrics

The research program should measure itself by decisions and downstream results.

## Research Quality

- percentage of material claims with traceable evidence,
- percentage of opportunities with explicit kill criteria,
- percentage with current evidence,
- percentage with base-rate analysis,
- percentage with evidence against the thesis.

## Decision Velocity

- time from signal to decision,
- time from hypothesis to cheapest test,
- time from test to kill or progression,
- number of low-value opportunities killed early.

## Market Validation

- data-access commitments,
- design partners,
- implementation commitments,
- pilots,
- repeat use,
- payment,
- outcomes.

## Commercialization

- licenses,
- assignments,
- company formations,
- partnerships,
- care-delivery pilots,
- acquisitions,
- reimbursed workflows.

## Portfolio Quality

- evidence debt,
- capital avoided through early falsification,
- maintenance costs avoided,
- abandoned low-value IP,
- opportunities advanced because evidence improved.

---

# Evidence Review Cadence

Each active opportunity should undergo periodic evidence review.

## Review Questions

1. What changed?
2. Which evidence is now stale?
3. What new evidence contradicts the thesis?
4. Has the base rate changed?
5. Has reimbursement changed?
6. Has the regulatory path changed?
7. Has the buyer changed?
8. Has workflow ownership changed?
9. Has incumbent behavior changed?
10. Should the decision state change?

## Active Research

High-priority opportunities should be reviewed frequently enough that new evidence can alter action before large resources are committed.

---

# Research Audit

The research system itself should be audited.

## Audit Questions

- Are we collecting evidence against our own thesis?
- Are we promoting AI inferences to facts?
- Are we confusing grant funding with demand?
- Are we confusing FDA clearance with adoption?
- Are we confusing pilot activity with product-market fit?
- Are we confusing interviews with payment?
- Are we averaging away a fatal weakness?
- Are we hiding evidence debt?
- Are we retaining opportunities because of sunk cost?
- Are we maintaining IP that no longer changes a commercial decision?
- Are we using stale evidence?
- Are we failing to replicate?

## Outcome

The audit should be allowed to recommend:

- re-test,
- pause,
- narrow,
- redirect,
- partner,
- abandon.

---

# Stronger Falsification Rules

For every opportunity, write the failure case before writing the investment case.

Required questions:

- Why might this mechanism not matter?
- Why might nobody pay?
- Why might the patient be impossible to find?
- Why might the workflow never change?
- Why might reimbursement not support it?
- Why might the incumbent neutralize it?
- Why might the IP be irrelevant?
- Why might AI not create the expected cost advantage?
- Why might this work only at one institution?
- Why might the status quo remain stronger?

The goal is not pessimism.

The goal is to expose the cheapest reason to stop.

---

# Updated Native Alpha™ Evidence Test

Native Alpha™ should itself be evidenced.

For each claimed advantage, require:

## Claim
What unusual ability do we possess?

## Source
Where did the insight come from?

## Scarcity
Why is it not obvious or easily copied?

## Action
What can we do differently because of it?

## Proof
What evidence shows the advantage produces a better decision or outcome?

## Compounding
Does the advantage become stronger with use?

If Native Alpha™ does not change a real decision, improve economics, accelerate adoption, improve outcomes, strengthen rights, or create defensibility, it should not be treated as meaningful.

---

# Updated Public Signal Intelligence Discipline

Public Signal Intelligence should classify each signal by evidentiary strength.

## Signal
Something observable in public data.

## Pattern
A recurring relationship across signals.

## Hypothesis
A possible explanation of the pattern.

## Opportunity
A hypothesis with plausible value creation and an identifiable beneficiary.

## Validated Opportunity
An opportunity supported by external commitment or real-world evidence.

## Commercial Thesis
A validated opportunity with a plausible path to value capture.

This progression prevents the system from treating interesting data patterns as businesses.

---

# Updated Opportunity Scoring Rule

Opportunity scores should never hide uncertainty.

Each scored dimension should include:

- score,
- confidence,
- evidence quality,
- evidence freshness,
- evidence against,
- next test.

A 9/10 score with low confidence should not be treated like a 9/10 score with replicated evidence.

---

# Updated Research Governance Rule

Every recommendation should answer:

> What decision should change because of this research?

If no decision changes, the research should explain why continued monitoring is justified.

If neither action nor monitoring is justified, stop.

---

# Final Research Doctrine

The research area should operate under the following doctrine:

> Research is complete when uncertainty has been reduced enough to make a better decision.

> Comprehensive documentation is not the objective.

> Evidence is not one thing. Scientific, market, adoption, economic, and strategic evidence must be evaluated separately.

> Claims should move up an evidence ladder through increasingly scarce real-world commitments.

> Define disconfirming evidence before institutionalizing a thesis.

> Use base rates before claiming exceptionalism.

> Use Native Alpha™ to explain why we can beat a base rate, not to ignore it.

> Require counterfactuals before assigning causal credit.

> Track provenance so evidence can be distinguished from interpretation.

> Treat stale or unsupported assumptions as evidence debt.

> Design experiments around decisions.

> Test behavior, not enthusiasm.

> Run the cheapest test that can invalidate the most important assumption.

> Replicate before calling something scalable.

> Every material opportunity should end in a Decision Package.

> Every opportunity should always have one current decision state and one next test.

> Research should ultimately produce action, abandonment, or a deliberate decision to wait.

> "Interesting research" is not a successful outcome.

# Decision Framework: From Research to Action

Every opportunity should move through a common decision sequence.

```text
1. Identify expensive failure
2. Explain failure mechanism
3. Identify modifiable condition
4. Measure readiness
5. Define enabling intervention
6. Identify economic beneficiary
7. Identify real buyer
8. Map system conditioning needs
9. Map patient access
10. Map workflow owner
11. Test reimbursement
12. Test channel
13. Anticipate incumbent response
14. Define falsification criteria
15. Define rights strategy
16. Test demand
17. Run narrow experiment
18. Measure adoption
19. Measure economic value
20. Decide: scale, partner, license, spin out, sell, or stop
```

---

# Integrated System Conditioning Framework

The research area should use a unified conditioning model.

## Patient Conditioning

Prepare the patient biologically or behaviorally.

## Clinician Conditioning

Prepare clinicians to understand, trust, and act.

## Workflow Conditioning

Prepare the operational process.

## Institution Conditioning

Prepare governance, procurement, implementation, IT, security, and ownership.

## Payer Conditioning

Prepare reimbursement logic and economic justification.

## Regulatory Conditioning

Prepare evidence, claims, quality systems, and accountability.

## Market Conditioning

Prepare the category, language, KOLs, evidence, and demand.

## AI / Information Conditioning

Prepare structured, retrievable, machine-readable knowledge so AI systems can understand and surface the category.

The complete model becomes:

```text
Technical Capability
        ↓
System Conditioning
        ↓
Patient Identification
        ↓
Institutional Readiness
        ↓
Adoption
        ↓
Measured Outcomes
        ↓
Learning
        ↓
Compounding Advantage
```

---

# Portfolio Review Questions

At regular intervals, every opportunity should be reviewed using the following questions:

1. What do we believe?
2. What evidence supports it?
3. What evidence contradicts it?
4. What would cause us to stop?
5. What is the real bottleneck?
6. Who benefits economically?
7. Who actually buys?
8. Who carries the risk?
9. Who owns the workflow?
10. What is the status quo?
11. What is the likely adoption latency?
12. What channel is best?
13. How will incumbents respond?
14. Is reimbursement required?
15. Does the opportunity need a category?
16. Which rights are required?
17. What is the likely exit path?
18. What Native Alpha™ exists?
19. What compounds with use?
20. What is the next cheapest experiment that could materially change the decision?

---

# Revised Research Doctrine

The working doctrine for this research area is:

> Build less. Understand the bottleneck better.

> Assume the product can be built.

> Assume AI will keep making technical work cheaper.

> Assume more technically credible solutions will exist than markets can absorb.

> Treat adoption as an engineered system.

> Condition the patient when the therapy needs help.

> Condition the system when adoption needs help.

> Find the patient before assuming the market exists.

> Find the workflow owner before assuming the buyer exists.

> Understand who bears risk before assuming anyone will adopt.

> Design reimbursement before assuming anyone will pay.

> Define kill criteria before falling in love with the thesis.

> Anticipate incumbent response before assuming a durable advantage.

> Protect only what changes the commercial outcome.

> Prefer the narrowest commercially useful rights structure.

> Measure Native Alpha™ by what it enables someone to know, see, decide, or do that competitors cannot.

> Treat every research effort as a decision system, not a content-generation exercise.