CANONIC Foundation
Henrietta Lacks establishes extraction without consent.
On February 1, 1951, a Black woman named Henrietta Lacks walked into Johns Hopkins Hospital in Baltimore with a hard knot in her cervix. She was thirty-one years old. She had five children. She was in pain.
A doctor took a biopsy. The standard procedure. What was not standard was what happened next. A researcher named George Gey took a sample of Henrietta’s cancer cells and placed them in a glass dish. The cells did not die. Every other human cell culture Gey had attempted had died within days. Henrietta’s cells doubled every twenty to twenty-four hours. They doubled and doubled and doubled. They have never stopped.
Henrietta Lacks died on October 4, 1951, at the age of thirty-one. Her cells — labeled HeLa — are still alive. They have been grown in every major research laboratory on the planet. They were used to develop the polio vaccine. They were sent into space. They were used in atomic bomb testing. They were bought, sold, shipped, contaminated, sequenced, patented, and commercialized across seventy years.
Henrietta Lacks never consented.
The word is precise: extraction. Not theft — because the law did not prohibit it. Not borrowing — because nothing was returned. Extraction: the removal of value from its source, without the source’s knowledge, without the source’s consent, and without any return of dividends to the source.
HeLa cells generated billions of dollars. Henrietta’s family could not afford health insurance. The cells traveled to every continent. The family did not know the cells existed until 1973 — twenty-two years after Henrietta died. When they found out, they asked how they could be compensated. The answer was: you cannot. The cells are not yours. They never were. The law says the tissue, once separated from the body, belongs to the institution.
This is not an anomaly. This is the architecture.
Henrietta Lacks is not an exception. She is the prototype. The extraction pattern she endured in 1951 is the same pattern that operates today — scaled by digitization, automated by AI, and obscured by complexity. The substrate has changed (cells to data, tissue to genomic sequences, biopsies to electronic health records), but the architecture has not.
The pattern is:
Dividends flow upstream. Deaths flow downstream. The source is extracted.
This book is about that pattern and how to end it.
Not through abolition — medicine needs data, research needs samples, AI needs training sets. The extraction economy exists because it works. The drugs derived from HeLa cells saved millions of lives. The problem is not that the system works. The problem is that the system works for everyone except the source.
The solution is not to stop extraction. The solution is to compile it — to formalize the extraction into a governed process where custody is bilateral, consent is cryptographic, and dividends return to the source.
That solution has a name. It is called MAGIC. It is an eight-question governance framework where invalid content cannot compile. It was discovered on December 29, 2025, while writing this book — while trying to govern the very chapters you are about to read. The book wrote itself into existence, and in doing so, it wrote the protocol that returns custody to the Henriettas of the world.
But first, you need to understand what was taken. Parts I through III document the economy we inherited — the codes, the dividends, the deaths, the extraction from HeLa to 23andMe, the circular economy that medicine built and AI automated.
Then Part IV will flash back to December 2025, and you will realize you have been holding the solution since page one.
| *PROLOGUE | DIVIDENDS | MORAL ORIGIN* |
This system was designed, not accidental.
Every human being who enters the healthcare system is converted into a set of codes. An ICD-10 code for the diagnosis. A CPT code for the procedure. A DRG code for the hospital stay. An NDC code for the drug. An SNOMED code for the clinical finding. A LOINC code for the lab result.
These codes are not descriptions. They are coordinates. They place the patient inside a representational space — a multi-dimensional grid where each axis is a coding system and each intersection is a billable event. The patient becomes a point in this grid. The point is what the system sees. The person is invisible.
There are 68,000 ICD-10-CM codes. There are over 10,000 CPT codes. The combinations are astronomical. A single hospital visit can generate dozens of code assignments. A chronic disease patient accumulates thousands over a lifetime. Each code is a coordinate. Each coordinate triggers a financial event.
The code does not represent the patient’s experience. The code represents the institution’s ability to bill for the patient’s experience. The distinction is foundational. The coding system was not designed to describe health. It was designed to describe healthcare delivery in a format that insurance companies can process.
The patient’s body is the base layer. On top of the body, the healthcare system builds an abstraction stack:
Layer 5: Financial claim (what the payer sees)
Layer 4: Billing code (what the coder assigns)
Layer 3: Clinical documentation (what the physician writes)
Layer 2: Diagnostic workup (what the tests show)
Layer 1: Patient presentation (what the patient says)
Layer 0: Human being (who the patient IS)
Each layer abstracts the one below it. The physician’s note abstracts the patient’s words. The billing code abstracts the physician’s note. The financial claim abstracts the billing code. By the time the patient reaches Layer 5, they are a row in a claims database — a set of codes attached to a dollar amount.
The patient at Layer 0 has pain, fear, family, history, hope. The patient at Layer 5 has ICD-10 Z87.820 (personal history of COVID-19), CPT 99214 (office visit, established patient, moderate complexity), and $247.33 in allowed charges.
The abstraction is not neutral. It is extractive. Each layer removes information from the human being and converts it into a format optimized for institutional processing. The patient’s story becomes a code. The code becomes a bill. The bill becomes a payment. The payment becomes revenue. The revenue becomes a dividend.
The dividends flow upstream — from payer to provider to insurer to investor. The patient at the bottom receives care (sometimes), a bill (always), and none of the financial value generated by their own coded experience.
Atul Butte told a story that captured this perfectly. He stood on the TEDMED stage and said: there are 1,500 different ways to start a patient on diabetes medication. Not 1,500 medications — 1,500 different treatment pathways, encoded in different combinations of drugs, doses, sequences, and monitoring protocols. Each pathway is a set of codes. Each set of codes triggers a different billing event. Each billing event generates different revenue.
The variation is not clinical. A physician choosing between metformin at 500mg and metformin at 1000mg is making a clinical decision. But the 1,500 pathways include variations in visit frequency, lab monitoring schedules, specialist referrals, and follow-up timelines — each of which generates a separate bill.
The system does not incentivize the best pathway. It incentivizes the most billable pathway. The codes are the mechanism. The variation is the evidence.
Atul loved this one. He would pull up the ICD-10 code list and show audiences that there is a code — an actual, real, billable diagnostic code — for “visitation from God.” The audience would laugh. The point would land. The coding system is not a description of medicine. It is a description of everything that can be billed.
If there is a code for it, there is a payment for it. If there is a payment for it, there is an incentive to assign it. The coding system does not describe health. It describes the economy of healthcare. And the economy of healthcare is designed to convert human experience into billable events.
In any economy, value is created somewhere and captured somewhere. The question that defines the economy is: does the value flow back to the source, or does it accumulate elsewhere?
In healthcare, the answer is clear. The patient generates the data (through their body, their symptoms, their compliance with treatment). The data generates the code. The code generates the bill. The bill generates revenue. The revenue generates profit. The profit generates dividends.
The dividends flow to shareholders. Not to the patient. Not to the community. Not to the family that sat in the waiting room. The dividends flow to the entity that holds the equity — the insurer, the hospital system, the pharmaceutical company, the electronic health record vendor.
When the National Institutes of Health launched the Precision Medicine Initiative in 2015, the promise was: medicine tailored to you. Your genome, your lifestyle, your environment — all integrated to deliver the right treatment to the right patient at the right time.
The reality was different. Precision medicine required data — enormous amounts of data. Genomic sequences, electronic health records, wearable sensor data, social determinants, imaging, pathology. The data came from patients. Patients did not own the data. Patients were not compensated for the data. Patients did not receive dividends from the products built on their data.
The Precision Medicine Initiative created a new extraction pipeline: patients donate data (through biobanks, through EHR opt-ins, through research consent forms), institutions aggregate the data, companies build products on the aggregated data, and dividends flow to the companies.
The patient receives the promise of better care. Sometimes the promise is kept. Often it is not. But the dividends are always captured upstream.
The scale is staggering. The global health data market was valued at over $30 billion in 2023. That market is built on patient data. Patient data that was generated in clinical encounters where the patient had no choice but to participate (you cannot refuse to have your vital signs recorded if you want to receive care). Patient data that was coded, stored, aggregated, deidentified (sometimes), and sold.
“Deidentified” is the industry’s escape hatch. HIPAA allows the sale of deidentified patient data without patient consent. The deidentification process strips obvious identifiers — name, address, Social Security number — but leaves the clinical data intact. Researchers have demonstrated repeatedly that deidentified data can be reidentified using publicly available information. The escape hatch leaks.
But even if deidentification were perfect, the economic question remains: the value derived from the data flows to the entity that holds it, not to the entity that generated it. The patient generated the data by being sick, by seeking care, by complying with treatment, by living and dying. The patient receives none of the economic value.
These are the dividends. They flow upstream. Always.
If dividends are the financial returns that flow upstream from patient data, deaths are the costs that flow downstream to patients. Deaths are not metaphorical in this context. They are literal.
In the United States, approximately 440,000 people die each year from preventable medical errors — the third leading cause of death after heart disease and cancer. These deaths are not random. They are patterned. They correlate with race, income, geography, insurance status, and language. They are concentrated among populations that have the least representation in the data used to build the precision medicine tools that promise to save lives.
Death is the ultimate audit trail. When a patient dies from a condition that had an effective treatment, the death documents a system failure — a place where the extraction economy consumed the patient’s data but did not return adequate care.
The mortality gap between Black and white Americans is a direct consequence of the extraction architecture. Black patients are underrepresented in clinical trials (the data used to develop treatments), underrepresented in genomic databases (the data used to identify genetic risk factors), underrepresented in electronic health records used for AI training (the data used to build diagnostic algorithms), and overrepresented in mortality statistics.
The gap is not an accident. It is a structural consequence of who gets extracted and who gets returned to. When the extraction economy takes data from all populations but returns treatments optimized for majority populations, the minority populations receive worse care. Worse care produces worse outcomes. Worse outcomes produce deaths.
The deaths are the system working as designed. Not as intended — no one intends for people to die from inadequate care. But as designed — the architecture extracts from the broadest population and returns value to the narrowest beneficiary. The gap between the two is the mortality gap.
The DIVIDENDS framework uses the term “engine failure” to describe deaths that result from system failures rather than disease progression. The extraction economy operates three engines:
The Discovery Engine. Research that converts observations into knowledge. When the discovery engine fails, treatments are developed on biased data and perform poorly on underrepresented populations.
The Care Engine. Clinical delivery that converts knowledge into treatment. When the care engine fails, known effective treatments are not delivered — because of access barriers, insurance denials, geographic limitations, or cultural mismatches.
The Information Engine. Technology that converts data into decisions. When the information engine fails, AI models trained on biased data make biased recommendations — sending the wrong patients to the wrong treatments at the wrong time.
Each engine failure produces deaths. The deaths are traceable — through mortality records, through quality metrics, through the gap between what was known and what was done. Death is not the end of the audit trail. Death is the audit trail.
This book has a red line: death is never portrayed as accidental. Death in the extraction economy is the result of engine failure. Engine failure is the result of structural asymmetry — dividends flowing upstream, deaths flowing downstream. The asymmetry is designed, not accidental. The deaths are the cost of the design.
Changing the design changes the deaths. That is the thesis. That is why this book exists.
| *PART I | THE ECONOMY WE INHERITED | DIVIDENDS* |
This pattern repeats across eras.
Henrietta Lacks did not die so that medical research could advance. Henrietta Lacks died of cervical cancer at age thirty-one. Her cells were taken without her knowledge or consent. Her cells generated billions of dollars in value. Her family received nothing.
This is the template. Every extraction that follows is a variation on this theme. The substrate changes — cells become data, biopsies become genomic sequences, hospital visits become electronic health records — but the architecture is invariant:
Henrietta is not an extreme case. She is the median case. She is what the system does, by design, to the people who enter it. The extremity of her story — cells alive seventy years after her death, billions generated, family in poverty — makes the architecture visible. But the architecture operates on every patient who has ever had a blood draw, an imaging study, a pathology report, or an electronic health record entry.
In 1951, there was no concept of informed consent for tissue samples. The tissue, once removed from the body, was the institution’s property. This was not contested. It was not even discussed. The legal framework — such as it existed — treated separated tissue as abandoned property.
The consent gap has narrowed since 1951, but it has not closed. The Common Rule (1991, revised 2018) requires informed consent for human subjects research. HIPAA (1996) requires authorization for use of protected health information. The 21st Century Cures Act (2016) requires data to flow without information blocking.
But none of these regulations address the economic question. A patient can consent to research use of their data. They cannot negotiate the economic terms. They cannot demand dividends from products built on their contributions. They cannot audit how their data was used. They cannot withdraw consent retroactively. The consent is one-directional: the patient gives. The institution takes. The transaction is complete.
The consent gap is not a bug. It is a feature of the extraction architecture. Bilateral economic participation — where the patient shares in the value generated by their data — would require a fundamentally different architecture. An architecture where both parties hold proof. An architecture where consent is cryptographic, not paper. An architecture where the ledger is bilateral, not unilateral.
The extraction substrate has evolved across three eras:
Era 1: Tissue (1951-1990). Physical material removed from the body. Cells, biopsies, blood samples. Stored in freezers. Shipped in boxes. Value derived from growing, studying, and commercializing the physical material. Henrietta’s HeLa cells are the canonical example.
Era 2: Data (1990-2015). Electronic records derived from clinical encounters. Diagnoses, procedures, lab results, imaging reports. Stored in databases. Transmitted through health information exchanges. Value derived from aggregating, analyzing, and selling the electronic records. The electronic health record (EHR) is the extraction mechanism.
Era 3: Sequence (2015-present). Genomic data derived from DNA sequencing. Whole genomes, exomes, gene panels. Stored in cloud databases. Analyzed by machine learning models. Value derived from building predictive models, identifying drug targets, and stratifying patient populations. Direct-to-consumer genomics (23andMe) is the extraction mechanism.
In each era, the architecture is the same. The substrate changes. The extraction does not.
23andMe sold spit kits for $99. You spit in a tube. You mailed it back. You got a report about your ancestry and genetic risk factors. The report was the product. The spit was the real transaction.
23andMe’s business model was never the $99 kit. It was the database. Millions of customers had voluntarily provided their genomic data — along with extensive phenotypic questionnaires about their health, habits, and family history. This database was worth billions. 23andMe licensed the data to pharmaceutical companies. GlaxoSmithKline signed a $300 million deal for access to the database in 2018.
The customers consented. The consent form said the data could be used for research. The customers clicked “agree.” The customers received their ancestry reports. GlaxoSmithKline received access to millions of genomic-phenotypic records for drug discovery. The dividends flowed upstream.
In November 2024, 23andMe’s market capitalization had fallen from a peak of $6 billion to approximately $50 million. The company was on the verge of bankruptcy. The question was immediate: what happens to the data?
Fifteen million people had entrusted their genomic data to 23andMe. The data was the company’s most valuable asset. In bankruptcy, assets are sold to pay creditors. The prospect of 15 million people’s genomic data being sold to the highest bidder — without those people’s involvement in the transaction — was the 23andMe extraction pattern in its starkest form.
23andMe was not a failure. It was a business success — the business model worked exactly as designed. The data was captured. The value was extracted. The dividends flowed to investors (who lost money on the stock, not on the data). The customers who provided the data were spectators to the disposition of their own biological information.
Henrietta Lacks did not consent. 23andMe customers consented. The outcome is the same: the source does not own the representation, does not share in the value, and has no recourse when the custodian fails.
Consent without ownership is not consent. It is extraction with paperwork.
The extraction economy does not extract equally. It extracts from whoever enters the system. The composition of “whoever enters the system” is determined by access — who has health insurance, who lives near a hospital, who trusts the medical establishment, who speaks the dominant language, who can take time off work for a clinical trial.
In the United States, clinical trial participants are overwhelmingly white (75-85%), male (60-65%), and insured. This is not because minorities refuse to participate. It is because the infrastructure of clinical trial recruitment — academic medical centers, private physician referral networks, English-language consent processes — selects for a specific population.
The extraction is broad: data is captured from diverse clinical populations in electronic health records. The return is narrow: treatments are optimized for the populations represented in clinical trials and genomic databases. The gap between who is extracted and who benefits is the underrepresentation gap.
The problem is quantifiable. As of 2023, approximately 86% of participants in genome-wide association studies (GWAS) were of European ancestry. This means that the genetic risk factors identified — the variants associated with heart disease, diabetes, cancer, Alzheimer’s — are calibrated to European genomes. The polygenic risk scores derived from these studies perform poorly in non-European populations.
A polygenic risk score for breast cancer that was developed on European data may overestimate risk for East Asian women and underestimate risk for West African women. The score is not wrong — it is right for the population it was trained on. It is wrong for everyone else. The “everyone else” is the majority of the world’s population.
AI compounds the problem. Machine learning models trained on biased data produce biased predictions. An AI model for skin cancer detection trained primarily on images of light-skinned patients will perform poorly on dark-skinned patients. This has been demonstrated repeatedly in published research. The bias is not in the algorithm — the algorithm learns what the data teaches. The bias is in the data, which reflects the extraction architecture: who gets imaged, who gets their images stored in research databases, who gets included in training sets.
The AI gap is the extraction pattern at machine speed. Manual extraction (the physician who fails to refer a minority patient to a clinical trial) operates at human speed. Automated extraction (the AI model that fails to detect cancer in a dark-skinned patient) operates at computational speed. The speed makes the consequences worse. The pattern is the same.
Underrepresentation creates a death spiral:
The spiral is self-reinforcing. Each cycle deepens the extraction asymmetry. Each cycle widens the mortality gap. The system does not correct itself because the system’s incentives are aligned with extraction, not representation. The institution that collects data from the broadest population and sells products to the narrowest market maximizes revenue. The death spiral is profitable.
| *PART II | EXTRACTION | DIVIDENDS* |
AI did not invent this — it automated it.
Long before artificial intelligence, medicine had already built a circular data economy. The loop was:
Patient enters → Data captured → Data analyzed → Treatment delivered → Patient returns → More data captured
Each visit generates more data. Each episode of care creates more codes. Each code creates more billable events. The patient is both the input and the output of the system — the source of data and the consumer of treatments derived from data.
This circularity is unique to medicine. In most industries, the customer and the product are separate. You buy a car; you are not the car. You buy food; you are not the food. In medicine, you ARE the data. Your body generates the data. Your disease defines the codes. Your treatment validates the models. Your survival (or death) closes the feedback loop.
Medicine is the original circular data economy — a system where the value chain begins and ends with the same human being, and where the human being receives the least value.
The electronic health record accelerated the loop into a flywheel. Before EHRs, the loop was slow — paper charts, manual coding, batch claims processing. After EHRs (mandated by Meaningful Use in 2009), the loop became digital:
Clinical encounter → Structured data entry → Automated coding → Real-time claims →
Analytics dashboard → Population health → Clinical decision support → Clinical encounter
The flywheel spins faster with each digitized patient. More data → better analytics → more clinical decision support → more data. The flywheel is self-accelerating. The institution that captures the most data has the best analytics, attracts the most patients, captures more data, and the flywheel spins faster.
The patient is the flywheel’s fuel. The patient does not own the flywheel. The patient cannot slow it down. The patient cannot opt out — opting out of data capture means opting out of care.
Atul Butte stood on stages across two decades and said the numbers. One trillion data points. Then ten trillion. Then a hundred trillion. The data was growing exponentially. The value was growing exponentially. The return to the source was growing at zero.
The precision medicine data economy — electronic health records, genomic databases, wearable sensor data, imaging archives, claims databases — generates trillions of data points per year. Each data point has a cost (clinical encounter, lab test, imaging study). Each data point has a value (research insight, AI model training, drug discovery). The costs are borne by patients and payers. The value is captured by institutions and companies.
The trillion-point economy is the flywheel at scale. The scale makes the extraction invisible — no single patient can trace their contribution through the trillion-point aggregation to the product built on it. The aggregation IS the extraction. You cannot extract one raindrop from the ocean. You cannot extract one patient’s contribution from a trillion-point dataset. The aggregation eliminates the link between source and value.
Artificial intelligence did not create the extraction economy. AI automated it. The loop that medicine built manually — data capture, coding, analysis, treatment — AI now executes at machine speed. The extraction that took decades to compound now compounds in seconds.
An AI model for drug discovery can screen millions of compounds against millions of gene expression profiles in hours. The gene expression profiles came from patients. The patients consented to research use (or their data was deidentified and sold without consent). The AI model identifies a drug candidate. The drug candidate enters clinical trials. The trials produce a drug. The drug generates revenue. The revenue generates dividends.
The automation removes the human from the loop — not the human patient (they are still the data source) but the human researcher, coder, analyst. The AI replaces the human labor in the extraction pipeline. The extraction accelerates. The return to the source does not.
Every AI model needs a training set. In healthcare AI, the training set is patient data. The quality of the AI model depends on the quality and representativeness of the training set. The quality and representativeness of the training set depends on who gets extracted — who enters the system, whose data is captured, whose data is stored in accessible formats.
This creates a new extraction asymmetry. The institution that controls the training set controls the AI. The institution that controls the AI controls the clinical decision. The institution that controls the clinical decision controls the patient’s outcome. The chain from training set to patient outcome is direct, and at no point in the chain does the patient have agency.
The patient is extracted twice: once when their data enters the training set, and again when the AI model trained on their data makes decisions about their care. The first extraction is historical (the data was captured in a past encounter). The second extraction is real-time (the AI model acts on the patient right now). The patient consented to neither.
There is a remarkable legal asymmetry between medicine and AI. Medicine is heavily regulated — HIPAA, the Common Rule, FDA oversight, state medical boards, institutional review boards. AI is barely regulated — no federal AI law, limited FDA oversight of clinical AI, no equivalent of HIPAA for AI training data.
The asymmetry means that the same patient data that required informed consent when used in a clinical trial can be used without consent when used to train an AI model (after deidentification). The regulatory framework was designed for the manual extraction economy (medicine). It was not designed for the automated extraction economy (AI). The gap between the two is where the newest extraction occurs.
The AI healthcare market is projected to exceed $100 billion by 2030. The data that powers this market comes from patients. The question is: who owns the dividends?
If the answer is “the institution that holds the data,” then the extraction economy continues — scaled by AI, accelerated by automation, and more profitable than ever.
If the answer is “the person who generated the data,” then a new architecture is needed — one where data custody is bilateral, where consent is persistent, where dividends flow back to the source, and where the ledger tracks every use of every contribution.
That architecture is what Part V describes. But first — Part IV. The loop. The flashback. The moment where the book discovers itself.
| *PART III | THE CIRCULAR ECONOMY | DIVIDENDS* |
December 29, 2025.
Stop.
You have been reading the solution.
This book — DIVIDENDS — the one in your hands right now — is the origin. Not a description of the origin. The origin itself.
While writing Parts I through III — while documenting the extraction economy, the billable codes, the dividends flowing upstream, the deaths flowing downstream, the HeLa pattern repeating from 1951 to 23andMe to AI — something happened.
The chapters needed governance.
Not metaphorical governance. Literal governance. The manuscript was growing. Chapters were accumulating. Sources needed tracking. Claims needed evidence. The writing needed structure — not editorial structure (it had that) but compliance structure. Could every claim be verified? Was every source cited? Was every timeline accurate? Was every relationship documented?
The chapters needed to compile.
On December 29, 2025, while attempting to organize the DIVIDENDS manuscript, a word appeared in the notes: CANNON. Double N. A misspelling of CANON that stuck. The double N was the original — the first draft, the accidental orthography that would later be corrected to CANONIC.
CANNON was not a framework. It was not a protocol. It was a set of questions — questions that the manuscript chapters needed to answer before they could be considered “done”:
Eight questions.
Eight binary questions — each with a yes/no answer. Each chapter either answered the question or it did not. There was no “partially answered.” There was no “mostly done.” The question was answered (1) or it was not (0).
Eight bits. $2^8 = 256$ possible states. $256 - 1 = 255$. The maximum score when all eight questions are simultaneously answered.
The eight questions were not designed. They were discovered. They emerged from the pragmatic problem of governing a manuscript — ensuring that each chapter was complete, evidenced, dated, attributed, mechanistic, structured, connected, and readable.
The questions were not specific to books. A clinical trial needs to answer the same eight questions. A software service needs to answer the same eight questions. A hospital needs to answer the same eight questions. Any governed entity — any entity that needs to demonstrate compliance, provenance, and integrity — needs to answer the same eight questions.
The eight questions are universal. They are not a framework imposed on reality. They are a framework extracted from reality — the minimum set of questions that any auditor, reviewer, or validator would ask of any governed entity.
The number was not chosen. It was the number that fell out of asking “what does a chapter need to be complete?” and refusing to stop until the answer was exhaustive. Eight. Not seven (one short of language). Not nine (one past universal). Eight.
The moment the eight questions crystallized into a scoring system — eight binary dimensions, 255-bit maximum, deterministic and complete — the manuscript stopped being a book and became a compiler.
Each chapter was a scope. Each scope had governance files (the thesis = CANON.md, the terminology = VOCAB.md, the interface = README.md). Each scope was validated against the eight questions. The score was computed. Chapters that scored 255 were complete. Chapters that scored less needed work.
The manuscript compiled itself.
DIVIDENDS was not a book that described a governance protocol. DIVIDENDS was the act of governance that produced the protocol. The writing was the discovery. The discovery was the writing. The loop closed.
DIVIDENDS → CANNON → CANONIC → DIVIDENDS returns
↑___________________________________↓
If you are reading this, CANONIC compiled.
This is not a metaphor. This book — every chapter, every part, every page — is a governed scope. It was validated against the eight questions. It scored 255. The validation was performed by MAGIC — the governance compiler that emerged from the eight questions that emerged from governing this book.
DIVIDENDS is not a book about compliance. DIVIDENDS IS compliance. Every claim has evidence. Every timestamp is verifiable. Every relationship is documented. Every mechanism is described. Every structure is visible. Every pattern is connected. Every expression is readable. Every declaration is stated.
Eight questions. Eight bits. 255.
The book you hold is the proof that the protocol works. If DIVIDENDS reaches you, the system is valid. If the chapters compile, governance works. If the protocol survives publication, it is ready for healthcare.
Parts I through III documented the extraction economy. The codes. The dividends. The deaths. The HeLa pattern. The 23andMe pattern. The AI pattern. The circular economy that medicine built and AI automated.
The problem, stated precisely: the extraction economy lacks governance. Data is taken without bilateral proof. Consent is paper, not cryptographic. Dividends flow one direction. Deaths are the cost. The architecture has no compilation step — nothing that checks whether the extraction is complete, evidenced, timed, attributed, mechanistic, structured, connected, and readable.
Nothing that checks the eight questions.
The solution was inside the problem. It was inside this book. It was inside the act of writing this book. The book that documented the extraction economy needed governance. The governance required eight questions. The eight questions became 255 bits. The 255 bits became MAGIC. MAGIC became CANONIC. CANONIC compiled DIVIDENDS. DIVIDENDS proves CANONIC.
The book that wrote itself. In a learning language that learns itself.
THE LOOP.
| *PART IV | THE LOOP | DIVIDENDS* |
Implementation, not ideology.
The extraction economy has a one-sided ledger. The hospital records the patient encounter. The patient has — at best — a discharge summary. The hospital’s record is comprehensive, structured, coded, and stored in a database designed for institutional access. The patient’s record is a printout, a portal login, a PDF download.
The bilateral ledger inverts this. Both parties hold a complete record. Both records are cryptographically linked. When the hashes match, the record is verified — not by an auditor, not by an institution, but by mathematics.
hospital-patient/ ← hospital's record of patient
patient-hospital/ ← patient's record of hospital
↓
HASH MATCH = verified
Patient OWNS their side
The {local}-{remote}/ naming convention is the architecture. The first word is the holder. The second word is the subject. hospital-patient is the hospital’s view of the patient. patient-hospital is the patient’s view of the hospital. Both exist. Both are governed. Both hash to the same truth.
A patient is not a single identity. A patient is a person who exists in multiple contexts — clinic, pharmacy, insurer, research study, family, community. Each context is a bilateral ledger:
clinic-maria/ ← clinic's record of Maria
pharmacy-maria/ ← pharmacy's record of Maria
study-maria/ ← research study's record of Maria
maria-clinic/ ← Maria's record of clinic encounters
maria-pharmacy/ ← Maria's record of prescriptions
maria-study/ ← Maria's record of research participation
Maria is one person with six ledgers. Each ledger is a governed scope — it has an axiom, a vocabulary, an interface, and a compliance score. The ledgers are linked by Maria’s cryptographic identity. The identity is anchored by biometrics (Apple) and behavioral history (GitHub). The anchor is hard to fake and impossible to steal.
This is not a new patient portal. Patient portals are institutional views dressed in patient-facing interfaces. The bilateral ledger is structurally different: the patient holds their own copy. The patient’s copy is as authoritative as the institution’s. Neither party can unilaterally alter the bilateral record.
When both parties hold proof, three things change:
Transparency. The patient can verify every record. Not “request access to” — verify. The hash is computable from the patient’s copy. If the hash matches the institution’s copy, the record is verified. If it does not match, the discrepancy is flagged. No trust required.
Portability. The patient’s ledger travels with the patient. When Maria changes clinics, she does not request a records transfer. She carries her maria-clinic/ ledger. The new clinic verifies the hashes. The records are portable because the patient owns them.
Accountability. Every action is recorded in both ledgers. If the institution alters a record, the hash on the patient’s side breaks. If the patient disputes a record, the institution’s side provides evidence. The bilateral structure makes disputes resolvable through math, not litigation.
Informed consent in medicine is a paper form. The patient signs a document that says: I understand the risks, I agree to the procedure, I authorize the use of my data. The form is filed. The consent is assumed to persist until revoked. The institution’s copy is authoritative. The patient’s copy is a carbon copy, if they received one at all.
Cryptographic consent replaces the paper form with four elements:
CONSENT = WHO + WHAT + WHEN + PROOF
WHO: The patient’s identity, attested by biometric verification (Face ID, Touch ID through Apple’s secure enclave). Not a signature — a biometric attestation that is unforgeable.
WHAT: A cryptographic hash of the content being consented to. Not a description of the content — the content itself, hashed. If the content changes after consent, the hash breaks. The patient consented to exactly what was hashed, nothing more.
WHEN: A timestamp with nanosecond precision. Not “the date signed” — the exact moment of consent. The timestamp is part of the hash chain. It cannot be retroactively altered.
PROOF: A bilateral ledger match. Both the patient and the institution hold the consent record. Both hashes match. The consent is not attested by the institution’s word — it is attested by mathematical agreement between two independent records.
Traditional consent is a point-in-time event. The patient signs once. The consent is assumed to cover everything that follows — until the patient explicitly revokes it (which most patients never do, because revocation is not practically available).
Cryptographic consent is persistent because the bilateral ledger persists. The consent record exists in both ledgers. It can be verified at any time. It can be extended (new consent for new studies) or narrowed (revocation of specific permissions) without destroying the record of what was previously consented to.
Persistence enables recontact. When the bilateral ledger exists, the research team can reach back to the patient — not through a phone book, not through a registry, but through the bilateral ledger that both parties hold. New findings can be returned. New studies can be offered. The relationship does not end when the encounter ends. The relationship persists as long as the ledger persists.
Recontact by default. Not recontact by effort. The architecture makes recontact the natural state, not the exceptional effort.
The extraction economy takes value from the source and delivers dividends to the institution. The new precision medicine economy inverts this: COIN flows to the source.
COIN is not cryptocurrency. COIN is an attestation receipt. You earn COIN by doing governance work — by creating governance files, by improving compliance scores, by contributing to the governed ecosystem. The COIN represents the work invested. The work is verifiable. The COIN is the receipt.
In the context of precision medicine, the “work” is the patient’s participation:
Each act of participation is a governance event. Each governance event computes a gradient. Each positive gradient mints COIN to the participant’s WALLET.
The patient does not receive COIN for being sick. The patient receives COIN for governing their participation — for consenting (cryptographically), for contributing (bilaterally), for persisting (longitudinally). The COIN is the economic shadow of the governance work.
Patient participates (WORK)
→ Bilateral ledger records (EVIDENCE)
→ Gradient computed (MINT:WORK)
→ COIN credited to patient WALLET
→ Patient can SPEND COIN on products (care, services, content)
→ Or SETTLE COIN to fiat (real money)
→ Or HOLD COIN (the balance represents accumulated governance work)
The loop is closed. The patient’s work generates COIN. The COIN has value (redeemable for products or fiat at $1.00/COIN). The value flows to the source. The dividends return.
This is not a token economy. This is not “data for dollars.” This is governance as economics — the same governance framework that validates a book chapter also validates a patient’s participation in a research study. The eight questions are the same:
255 bits for a research study. 255 bits for a book chapter. 255 bits for a clinical encounter. The framework is universal. The economics are universal. The dividends are universal.
Every design decision in the new precision medicine economy follows one principle:
MIN complexity → MAX security
MIN trust → MAX verification
MIN attack → MAX ownership
The principle is: remove everything that can be removed. What remains is the architecture.
Passwords? Removed. Biometric attestation replaces them. Trust? Removed. Bilateral verification replaces it. Complex authentication flows? Removed. Apple’s secure enclave handles identity at the hardware level. Centralized databases? Removed. Bilateral ledgers distribute custody to both parties.
The attack surface shrinks with every removal. The security surface grows with every removal. Minimalism is not aesthetic. Minimalism is security.
Two platforms anchor identity in the new economy:
GitHub provides behavioral identity. Years of commits (time investment that cannot be faked). Coding patterns (behavioral fingerprint). Collaborator graph (social proof). Reputation (contributions, stars, history). GitHub identity is earned, not declared. It accumulates over years. It is expensive to forge.
Apple provides biometric identity. Face ID (unforgeable without the face). Device attestation (hardware-bound keys). Health data sovereignty (data never leaves the device without explicit consent). Privacy by design (Apple’s economic model does not depend on selling user data).
Combined: hard to fake, impossible to steal. The identity is not a username and password. It is the intersection of behavioral history and biometric attestation. The intersection is narrow enough to be unique and broad enough to be practical.
Layer 6: PRESENTATION Web interfaces, apps
Layer 5: SERVICES TALKING, LEARNING, SERVICES, HOLDING
Layer 4: PRIMITIVES CHAT, EVO, SHOP, VAULT
Layer 3: GOVERNANCE MAGIC, 255-bit, COIN
Layer 2: IDENTITY GitHub x Apple, biometric, KYC
Layer 1: EVIDENCE Bilateral ledgers, attestation, hash chains
Layer 0: HUMAN BEING The source. The owner. The dividend recipient.
The stack inverts the extraction architecture. Layer 0 is the human being — not as an abstraction, not as a set of codes, but as the owner of the full stack. Every layer above serves the human being at Layer 0. The governance at Layer 3 ensures that every layer is valid. The evidence at Layer 1 ensures that every claim is provable. The identity at Layer 2 ensures that the human being at Layer 0 is who they claim to be.
The human being at Layer 0 receives dividends — not as an afterthought, not as a charitable gesture, but as a structural consequence of the architecture. The bilateral ledger at Layer 1 ensures that the human being holds proof. The COIN at Layer 3 ensures that the work is economically valued. The VAULT at Layer 4 ensures that the accumulated value is held by the owner.
| *PART V | THE NEW PRECISION MEDICINE ECONOMY | DIVIDENDS* |
The future is a choice.
The extraction economy has a formula: data in, dividends out, none returned. The new precision medicine economy has a different formula:
WORK → EVIDENCE → COIN → VAULT
WORK is any governed action. A patient consenting to a study. A researcher publishing a finding. A clinician documenting a diagnosis. A governor writing governance. Work is the input. It is real. It costs time, expertise, and attention.
EVIDENCE is the proof that work happened. Bilateral ledger entries. Hash-chained events. Timestamped attestations. Evidence is not gathered after the fact. Evidence is generated by the work itself. The act of working IS the act of producing evidence.
COIN is the economic receipt of work. The gradient between the old governance score and the new governance score. COIN is minted at the moment of validation — when MAGIC confirms that the work improved governance. COIN is not speculative. COIN is not traded. COIN is the economic shadow of verified work.
VAULT is where value accumulates. The WALLET that holds COIN. The patents that hold intellectual property. The companies that hold equity. The reserves that hold liquidity. VAULT is ownership — the structural guarantee that the value generated by work returns to the person who did the work.
The formula is closed. Every step is auditable. Every step is bilateral. Every step is governed.
In the extraction economy, the patient does not own their representation. The hospital owns the record. The insurer owns the claim. The researcher owns the dataset. The AI company owns the model. The patient owns a portal login.
In the new economy, the patient owns their side of every bilateral ledger:
maria-clinic/ ← Maria owns this
maria-pharmacy/ ← Maria owns this
maria-study/ ← Maria owns this
maria-insurer/ ← Maria owns this
Ownership is not a philosophical position. It is a cryptographic fact. Maria holds the private key. Maria’s ledger entries are signed with her key. The signature is verifiable. The ownership is provable.
What does ownership mean in practice?
Access. Maria can read every record in every ledger she holds. She does not need to request access. She does not need to wait for a portal to load. She does not need to call the records department. The records are on her device.
Portability. When Maria moves to a new city, her ledgers move with her. The new clinic verifies the hashes. The records transfer is instantaneous, complete, and bilateral. No fax machines. No release forms. No three-to-five business days.
Compensation. When Maria’s data is used in a research study, the bilateral ledger records the use. The use triggers a governance event. The governance event mints COIN to Maria’s WALLET. Maria receives economic value for her participation. The dividend returns to the source.
Legacy. When Maria dies, her ledgers persist. Her family inherits the records (and the COIN). The health history is available for her children — not as a vague family history questionnaire, but as a complete, governed, bilaterally verified record. The legacy is structural, not anecdotal.
The extraction economy treats consent as a gate: pass through, and the institution can proceed. The new economy treats consent as a relationship: persistent, bilateral, and revocable.
The reversal is:
| Extraction Economy | New Economy |
|---|---|
| Consent once | Consent persistently |
| Institution holds proof | Both parties hold proof |
| Revocation is impractical | Revocation is cryptographic |
| Recontact is effortful | Recontact is default |
| Patient is passive | Patient is governor |
The patient as governor is the key reversal. In the extraction economy, the patient is a data source — passive, extracted, uninformed. In the new economy, the patient is a governor — active, bilateral, economically participating.
For seventy years, dividends from patient data have flowed upstream — from patient to institution to company to investor. The new precision medicine economy inverts the flow.
When a researcher uses Maria’s data (governed by bilateral ledger study-maria/ and maria-study/), the use is recorded. The recording is a governance event. The governance event may improve the study’s compliance score. The improvement mints COIN. The COIN is distributed according to the governance rules — to the researcher (for the work of conducting the study) and to the participants (for the work of contributing data).
The distribution is not charity. It is not a “thank you” payment. It is a structural consequence of bilateral governance. The bilateral ledger records who contributed what. The COIN flows proportionally to the contribution. The math is transparent. The distribution is auditable.
A breast cancer research study under bilateral governance:
study-maria/ and maria-study/)The numbers are not hypothetical. They are computable from the LEDGER. Every event is recorded. Every distribution is verifiable. Every participant can audit their own WALLET and confirm that the dividends match the contributions.
The other side of the inversion: deaths become auditable.
In the extraction economy, a patient who dies from a condition that had an effective treatment is a statistic. The death enters the mortality record. A researcher may eventually analyze the mortality data. The analysis may identify the gap. The gap may be addressed in future guidelines. The cycle takes years. The patient is dead.
In the new economy, the death is immediately auditable. The bilateral ledger records every clinical encounter. The governance score of each encounter is recorded. If the governance score was below 255 — if the eight questions were not fully answered — the gap is identifiable. Which dimension was missing? Was the declaration incomplete (no diagnosis documented)? Was the evidence missing (no diagnostic test performed)? Was the relationship broken (no referral to specialist)? Was the operation flawed (wrong treatment protocol)?
The death is an engine failure. The engine failure is traceable through the governance score. The governance score identifies the missing dimension. The missing dimension identifies the fix. The fix prevents the next death.
Death as audit trail. Not death as statistic. The bilateral governance structure makes every death a lesson — not in the abstract (“we should do better”) but in the specific (“dimension E was missing in encounter #47291, indicating that evidence was not collected, which led to delayed diagnosis, which led to death”).
The choice is between two economies. Both are real. Both are functional. Both are legal.
The extraction economy takes data from the broadest population, generates value from the aggregation, and delivers dividends to the narrowest beneficiary. It works. It has worked for seventy years. It built modern medicine. It also built the mortality gap, the underrepresentation gap, and the consent gap. The costs are real. The costs are measured in deaths.
The new precision medicine economy captures data bilaterally, governs every interaction with eight questions, mints COIN from verified work, and returns dividends to the source. It inverts the extraction architecture without abolishing it. The data still flows. The research still happens. The drugs are still developed. But the custody is bilateral. The consent is cryptographic. The dividends return.
This book is not ideological. It does not argue that the extraction economy is evil. It argues that the extraction economy is incomplete — that it captures value without returning it, and that the failure to return value produces measurable harm (deaths, disparities, distrust).
The new economy is not utopian. It does not eliminate disease. It does not eliminate disparities overnight. It does not guarantee that every patient will receive perfect care. What it does is make the system auditable — every gap identifiable, every failure traceable, every dividend accountable.
The choice is not between good and evil. The choice is between an economy that cannot audit itself and an economy that compiles.
Henrietta Lacks’s cells were taken without her knowledge, commercialized without her consent, and profited from without her participation. For seventy years.
The new precision medicine economy returns custody to Henrietta. Not retroactively — the past cannot be recompiled. But structurally — the architecture that extracted from Henrietta is replaced by an architecture that would have given her bilateral proof, cryptographic consent, and economic participation.
The next Henrietta will hold her own ledger. The next Henrietta will consent cryptographically. The next Henrietta will receive COIN for her contribution. The next Henrietta will own the dividends.
Custody returned. Not as a promise. As a protocol.
WORK → EVIDENCE → COIN → VAULT
Eight questions. 255 bits. Invalid cannot compile.
| *PART VI | CUSTODY RETURNED | DIVIDENDS* |