A Different Kind of Medicine

Nearly everything medicine knows about any given patient was measured during a scheduled appointment. A blood panel is a photograph of one fasted morning, usually taken because someone was already worried. Between appointments—a handful of hours out of the 8,760 in a year—the record is blank.

That blank is where disease happens.

Consider what the same patient looks like when it is filled in. A wrist sensor and a ring supply continuous heart rate, heart-rate variability, respiratory rate, skin temperature, and sleep architecture. A continuous glucose monitor supplies a metabolic trace that requires no fasting and depends on nobody's memory of what they ate. A quarterly multi-omics panel supplies transcriptomic, proteomic, and metabolomic state—not merely what the genome permits, but what the body is actually doing with it this quarter. Instrumented bathroom fixtures sample urinary and stool biomarkers during activities that happen anyway.

Individually, none of these streams is diagnostic. One night of depressed heart-rate variability means nothing. Mildly elevated overnight glucose means nothing. An inflammatory marker drifting toward the top of its reference range means nothing, and clinicians are right to ignore each of them in isolation.

Collectively, and across time, they mean a great deal. The signal lives not in any single measurement but in the covariance—several unremarkable trends moving together, in an order that population data associates with something specific. Autoimmune flares, viral infections, and metabolic decompensation all have prodromes that are legible in aggregate weeks before they are legible as symptoms. What has been missing is not sensors. It is any system capable of reading several million data points per patient per year and identifying the handful of moments that warrant a phone call.

That is the function AI performs here, and it is the only reason the rest of the stack is worth building.

The resulting change is categorical rather than incremental. Reactive medicine asks what is wrong, once. Continuous medicine asks what is changing, constantly. The intervention point moves from after symptoms to before them—which, for most of the diseases that kill people in wealthy countries, is the difference between management and cure.

The world is not there yet. Almost none of it is integrated, most of it is uninsured, and the evidence base for acting on early molecular signals remains thin. But every component exists, and the direction of travel is not in doubt.

2026 Snapshot — Where Prevention Stands Today

The shift toward preventive and predictive medicine has begun, but implementation remains fragmentary.

AI-assisted clinical decision support is increasingly deployed. Systems analyze medical images, flag abnormalities in lab results, suggest diagnoses, and recommend treatments. Most implementations are narrow—an AI that reads chest X-rays, another that analyzes ECGs—rather than integrated systems reasoning across all patient data.

Multi-omics testing is expanding. Whole genome sequencing is available for under $500 and falling. Transcriptomics (RNA expression), proteomics (proteins), metabolomics (metabolites), and other -omics are available but expensive and not routinely integrated. The technology exists; the clinical workflows and evidence base lag behind.

Liquid biopsies for cancer detection are in clinical use, with Grail's Galleri test and others screening for multiple cancer types from a blood draw. Sensitivity varies by cancer type and stage; the tests catch some cancers earlier but miss others. They are supplements to, not replacements for, traditional screening.

Wearables are ubiquitous. Hundreds of millions of people wear devices tracking heart rate, activity, and sleep. Apple Watch can detect atrial fibrillation and perform ECGs. Continuous glucose monitors are used by millions of diabetics and a growing number of non-diabetics interested in metabolic health.

Integration and cost remain the primary barriers. Individual technologies work; making them work together in clinical practice is hard. Multi-omics panels cost thousands of dollars and aren't covered by insurance. Data from wearables rarely reaches electronic health records. The vision of integrated, continuous, predictive health monitoring is achievable in boutique concierge practices; it's not yet available to ordinary patients.

The current state is one of promising components that haven't yet assembled into a coherent system.

Notable Players

Clinical Software and Healthcare IT

Epic Systems dominates the electronic health record market in the United States, with over 250 million patient records. Their integration of AI capabilities—clinical decision support, documentation assistance, predictive analytics—will shape how AI reaches most American patients.

Oracle Health (formerly Cerner) is Epic's primary competitor, with a similar trajectory toward AI integration.

Nuance Communications (acquired by Microsoft) pioneered clinical voice recognition and is now integrating large language models into clinical documentation through DAX Copilot.

Ambient AI startups—Abridge, Nabla, Suki, and others—are competing to automate clinical documentation using AI that listens to patient encounters and generates notes.

Precision Medicine and Diagnostics

Illumina remains the dominant sequencing platform, enabling the multi-omics revolution. Their push into clinical markets (through acquisitions like Grail) positions them at the center of genomic medicine.

Grail offers the Galleri multi-cancer early detection test, one of the most ambitious liquid biopsy products. Illumina acquired it, was forced to unwind the deal by antitrust regulators on both sides of the Atlantic, and spun it back out as an independent public company in mid-2024—an expensive demonstration that the sequencing layer and the diagnostics layer are not obviously allowed to be the same company.

Tempus combines clinical data with molecular profiling to enable precision medicine, particularly in oncology. Their database of genomic and clinical data is among the largest.

Color Health offers population-scale genetic testing and health programs, working with employers and health systems to make genetic screening more accessible.

23andMe pioneered consumer genomics and assembled one of the largest databases of linked genetic and phenotypic data—then filed for Chapter 11 bankruptcy in 2025, having never converted that database into a durable business. The wind-down triggered a serious public reckoning over what happens to fifteen million people's genomes when the company holding them becomes an asset in a bankruptcy proceeding. That question had been asked for years and answered casually; it is no longer hypothetical, and it is the single clearest warning attached to everything else described in this chapter.

Wearables and Continuous Monitoring

Apple leads consumer health sensing through the Apple Watch. Their research partnerships, health features (ECG, fall detection, AFib detection), and integration with HealthKit make them a major force in consumer health data.

Fitbit/Google combines wearable hardware with Google's AI capabilities and health platform ambitions.

Dexcom and Abbott dominate continuous glucose monitoring, with millions of users generating continuous metabolic data.

Oura, Whoop, and other wearable companies focus on recovery, sleep, and readiness tracking with increasingly sophisticated sensors and analysis.

Withings offers connected scales, blood pressure monitors, and other home health devices that feed into continuous monitoring ecosystems.

At-Home Testing

Everlywell, LetsGetChecked, and others have built platforms for at-home testing of various biomarkers, from hormone levels to food sensitivities to STIs.

Cue Health built a genuinely impressive home molecular testing platform on the back of COVID-19 demand, then ceased operations in 2024 when that demand evaporated—a reminder that consumer diagnostics businesses have historically been built on one acute use case and struggled to survive its passing.

The trend toward decentralized, at-home testing aligns with continuous monitoring: if data can be collected at home, it can be collected more frequently and with less friction. The commercial history of the category, however, is unkind. Theranos was fraud, but Cue was not, and both are gone. The technical case for home diagnostics has consistently been stronger than the business case.

From Reactive to Continuous: The Paradigm Shift

The fundamental change coming to medicine is a shift from reactive to continuous care.

Reactive medicine waits for problems. You feel sick, you see a doctor. You notice a lump, you get it checked. Symptoms trigger investigation. By the time you seek care, disease has often progressed—cancer has grown, cardiovascular damage has accumulated, neurodegeneration has advanced.

Continuous medicine monitors constantly. Sensors track biomarkers around the clock. AI analyzes patterns across data streams. Anomalies are detected before symptoms appear. Intervention happens early, when disease is easiest to treat—or prevented entirely.

This is not a small adjustment. It's a fundamental reconception of what healthcare is and when it happens.

The Technology Stack

Several technology layers must come together:

Continuous sensing provides the data. This means:

• Wearable sensors (wrist-worn, rings, patches) tracking cardiovascular, respiratory, and movement data
• Implantable sensors (still early) providing more precise, continuous measurement
• Home devices (smart scales, toilets, mirrors, toothbrushes) capturing data during daily activities
• Ambient sensors detecting behavioral patterns, air quality, and environmental factors

Periodic deep measurement provides precision. This means:

• Regular blood draws analyzed for hundreds of biomarkers
• Multi-omics profiling (genomics, transcriptomics, proteomics, metabolomics, microbiome)
• Imaging at appropriate intervals based on risk stratification

AI integration makes sense of it all. This means:

• Algorithms that detect patterns across data streams
• Models that predict disease onset from early signatures
• Systems that recommend interventions personalized to individual biology and context
• Interfaces that communicate insights to patients and providers in actionable ways

Clinical workflow integration translates insights into action. This means:

• Alerts that reach providers when intervention is needed
• Decision support that suggests appropriate responses
• Documentation that captures AI-assisted reasoning for accountability
• Payment systems that reimburse prevention, not just treatment

None of these layers is fully mature. But each is advancing rapidly.

Wearables to Implantables

The wearable market is enormous—hundreds of millions of smartwatches, fitness trackers, and specialty devices. But wearables have limitations: they sit on the skin surface, can only measure what's accessible there, and depend on users actually wearing them.

Current wearables can measure:

• Heart rate (optical sensors)
• Heart rhythm / ECG (electrical sensors)
• Blood oxygen saturation (pulse oximetry)
• Movement and activity (accelerometers)
• Sleep patterns (combined sensors)
• Skin temperature
• Some limited blood pressure estimation

Emerging wearable capabilities include:

• Non-invasive glucose monitoring (multiple companies attempting; none fully reliable yet)
• Continuous blood pressure (difficult but advancing)
• Hydration status
• Respiratory rate and patterns
• Stress indicators (heart rate variability, electrodermal activity)

Implantables could go further:

• Continuous glucose monitors are already implanted beneath the skin, lasting months per device
• Continuous cardiovascular monitors (like Abbott's Confirm Rx) detect arrhythmias over extended periods
• Future devices might measure a broader range of biomarkers from interstitial fluid or blood

The trajectory is toward more sensors, measuring more biomarkers, more continuously, with less friction. Each generation captures more of the body's internal state.

Home Diagnostics

The clinic visit is a bottleneck. You schedule an appointment, travel to a facility, wait, have a brief encounter, and leave. Data is collected at one point in time, in an artificial setting. If you need follow-up testing, you do it all again.

Home diagnostics bypass the bottleneck:

Smart toilets can analyze urine and stool for biomarkers of metabolic health, kidney function, and colorectal cancer. Products from companies like Withings and Toto are entering the market.¹ Toilet-based testing is frictionless—it happens during activities you do anyway, generating continuous data without any extra effort.

At-home blood testing through finger-prick collection and mail-in analysis is already available. The frontier is smaller sample sizes, faster turnaround, and more biomarkers. Companies are working toward home devices that provide lab-quality results from a drop of blood.

Breath analysis can detect markers of various conditions, from diabetes to liver disease to bacterial infections. The technology is earlier-stage but promising.

Saliva testing for hormones, genetics, and some infections is already routine in consumer products.

Skin analysis through smartphone cameras and AI can screen for concerning lesions, track skin conditions, and detect changes over time.

Each of these modalities contributes data streams that, integrated together, provide a far richer picture of health than any annual checkup could.

Multi-Omics: Reading the Body's Code

The genome was just the beginning. Understanding biology requires measuring not just DNA but everything downstream:

Genomics tells you what genes you have—your inherited risk factors, drug metabolism variants, and biological predispositions. It's stable over time; you only need to sequence once.

Transcriptomics tells you which genes are active now—which RNA is being produced in which tissues. This changes dynamically and reflects current physiological state.

Proteomics tells you which proteins are present and at what levels. Proteins are the functional molecules; their patterns reveal biological processes in action.

Metabolomics tells you which metabolites—small molecules like sugars, lipids, amino acids—are circulating. These reflect metabolism, nutrition, and environmental exposures.

Microbiome analysis tells you which microorganisms are living in your gut, on your skin, and elsewhere. The microbiome influences digestion, immunity, and even mental health.

Epigenomics tells you how your genes are regulated—which are silenced, which are activated—revealing how environment and experience have modified your biology.

Integrated together, these layers provide a comprehensive picture of biological state. Disease signatures often appear across multiple -omics layers before symptoms manifest.²

The challenge is cost, complexity, and interpretation. A full multi-omics workup can cost tens of thousands of dollars. Analyzing and interpreting the data requires sophisticated bioinformatics. Translating findings into clinical action requires evidence that often doesn't yet exist.

AI is essential for making multi-omics clinically useful. The data is too high-dimensional for human interpretation. Patterns that predict disease require machine learning to identify. Personalized recommendations require models that reason across an individual's complete biological profile.

Liquid Biopsy and Early Cancer Detection

Cancer is most treatable when caught early. But most cancers aren't caught early—they grow silently until symptoms appear or they're incidentally discovered. By then, they may have spread.

Liquid biopsy offers a different approach: detecting cancer from blood samples.

The biology: tumors shed cells, DNA fragments (ctDNA), and other molecules into the bloodstream. These can be detected with sensitive assays. Different cancers produce different signatures.

Grail's Galleri test screens for over 50 cancer types from a single blood draw.³ It analyzes methylation patterns—chemical modifications to DNA that differ between cancer types and normal tissue. A positive result indicates both the presence of cancer and its likely tissue of origin.

Performance varies dramatically by cancer type and stage. For cancers with good existing screening (breast, colorectal, cervical), the test adds less value—those cancers are often caught anyway. For cancers with no existing screening (pancreatic, ovarian, many others), the test could catch cancers that would otherwise be found only after spreading.

Current liquid biopsy tests are expensive ($1,000+), have meaningful false positive and negative rates, and aren't covered by most insurance. They're available to those who can pay and are beginning to enter clinical guidelines.

The trajectory: costs will fall, sensitivity will improve, and screening will expand. The goal is a routine blood test that catches most cancers at stage 1, when five-year survival often exceeds 90%, rather than stage 4, when it often falls below 20%.

AI as the Integration Layer

The technologies described above—continuous sensing, home diagnostics, multi-omics—generate enormous volumes of data. Without AI, this data is overwhelming and clinically useless. A patient with a smartwatch, a continuous glucose monitor, quarterly blood panels, genomic sequencing, and microbiome analysis generates millions of data points per year. No physician can review all of it. No patient can make sense of it.

AI makes the data useful:

Pattern recognition identifies anomalies that human observers would miss. A subtle change in heart rate variability, correlated with a shift in sleep patterns and an uptick in inflammatory markers, might be invisible in isolation but collectively signal early infection or disease.

Predictive modeling anticipates problems before they manifest. Machine learning models trained on population data can recognize signatures that precede disease onset by weeks, months, or years. The individual data gets interpreted through the lens of what those patterns have meant for similar people.

Personalization adapts recommendations to individual biology. The optimal diet, exercise routine, or medication dose varies between people based on genetics, microbiome, and physiological state. AI can learn individual responses and tailor recommendations accordingly.

Communication translates complex biological data into actionable guidance. Patients don't need to understand methylation patterns or proteomic signatures; they need to know what to do. AI systems can bridge the gap between data complexity and practical action.

Provider augmentation helps clinicians make sense of comprehensive patient profiles. Instead of reviewing thousands of data points, a physician sees a synthesized summary: what's changed, what's concerning, what's recommended, and why. The physician applies judgment; the AI handles data processing.

The risk is that AI becomes a black box—making recommendations that neither patients nor providers understand. Trust requires interpretability: the ability to explain why a particular alert was raised or recommendation made. Building systems that are both powerful and interpretable is an active area of research and engineering.

Second-Order Effects: What Changes When Prevention Succeeds

If the shift from reactive to continuous medicine succeeds at scale, the second-order effects will be profound.

The Economics of Healthcare Flip

Current healthcare economics reward treatment, not prevention. Hospitals profit from surgeries, not from healthy patients who don't need surgery. Drug companies profit from chronic disease management, not from preventing disease in the first place. Fee-for-service payment incentivizes volume of services, not value of outcomes.

Value-based care attempts to realign incentives—paying for outcomes rather than procedures. But progress has been slow, and the transition is difficult when the entire infrastructure is built around reactive care.

Continuous prevention, if it works, could force the economic question. If most diseases are caught and treated early—or prevented entirely—the volume of expensive late-stage treatment drops. This is good for patients and payers but potentially catastrophic for provider economics built on procedural revenue.

The health system will have to find new economic models: subscription-based continuous care, capitated payments that reward keeping patients healthy, or direct payment for prevention services. The transition will be contested and disruptive.

Insurance and Underwriting Transform

Health insurance depends on uncertainty: people pay premiums because they don't know if they'll get sick. If continuous monitoring makes health status transparent, insurance dynamics change.

On one hand, individuals might use their health data to shop for better rates—or to self-insure if their risk is low. On the other hand, insurers might demand data access for underwriting, potentially excluding high-risk individuals or charging them unaffordable premiums.

The policy implications are significant. Anti-discrimination laws (like GINA for genetic information) may need to extend to broader health data. Mandates for coverage regardless of pre-existing conditions become more important when pre-existing conditions are more visible. The social contract around health risk-pooling will need renegotiation.

Privacy Becomes Critical Infrastructure

Continuous health monitoring generates extraordinarily sensitive data. Your glucose patterns reveal when you eat and drink. Your heart rate variability reveals stress responses. Your location and activity reveal behavior. Your genomics reveals hereditary risks. Aggregated together, this data paints an intimate portrait of biological and behavioral reality.

Who can access this data? What can they do with it? How is it protected?

The stakes are high:

• Employers might discriminate based on health predictions
• Insurers might exclude or price-gouge based on risk profiles
• Governments might surveil citizens through health data
• Criminals might extort individuals using sensitive health information
• Commercial actors might manipulate behavior based on biological states

Strong privacy protections—technical (encryption, data minimization), legal (regulations with teeth), and social (norms around appropriate use)—become essential. Health as infrastructure requires privacy as infrastructure.

The Definition of "Health" Expands

When disease is defined by symptoms, the boundary between sick and healthy is relatively clear. When disease is defined by molecular signatures that precede symptoms by years, the boundary blurs.

Is someone with elevated inflammatory markers "sick"? What about someone with genetic variants that increase cancer risk? What about someone whose AI health assistant detects patterns associated with future depression?

This expansion creates challenges:

• Overdiagnosis: labeling people as sick who would never have experienced symptoms
• Overtreatment: intervening in conditions that would never have caused harm
• Anxiety: making healthy people worry about future risks they can't control
• Medicalization: expanding the domain of medical authority into previously normal aspects of life

The benefits of early detection are real—catching cancer at stage 1 rather than stage 4 saves lives. But the boundary must be drawn carefully, distinguishing between actionable risk (where intervention improves outcomes) and noise (where detection causes harm without benefit).

Health Becomes a Continuous Activity

In the reactive model, health is something you attend to when it fails. You're healthy until you're not; medicine matters when you're sick.

In the continuous model, health is something you optimize constantly. Every meal, every workout, every sleep choice feeds back into a system tracking your biological state and suggesting adjustments.

For some people, this is empowering—a sense of control over their own biology. For others, it's exhausting—a constant reminder of mortality and a never-ending optimization project.

The social dynamics are unpredictable. Will continuous health monitoring become a competitive status marker, like fitness tracking already is in some communities? Will it create new anxieties and obsessions? Will it widen the gap between the health-optimizing elite and those who opt out or can't access the technology?

These questions don't have obvious answers. The technology is coming regardless; how society adapts to it will be determined by choices not yet made.

Programmable Medicine: mRNA and Beyond

Prevention is one transformation; programmable treatment is another.

Traditional drugs are small molecules or proteins discovered through screening—you try many compounds and find ones that work. This process is slow, expensive, and constrained by chemistry: only certain molecules can be synthesized, delivered, and tolerated.

mRNA therapeutics offer a different approach: instead of delivering the drug, you deliver instructions. mRNA tells cells what proteins to produce. If you can design the right sequence, you can instruct cells to produce virtually any protein—antibodies, enzymes, receptors, or novel molecules.

The COVID-19 vaccines proved the concept at scale. Within weeks of receiving the viral sequence, researchers designed mRNA encoding the spike protein. Cells that take up the mRNA produce the spike protein, triggering an immune response. Billions of doses later, the technology is validated.⁴

But vaccines are just the beginning:

Cancer vaccines use mRNA to train the immune system against tumor-specific antigens. Because tumors are genetically unique, these vaccines can be personalized—sequencing an individual's tumor and designing mRNA to target the specific mutations present.⁵

Protein replacement could treat diseases caused by missing or defective proteins. Rather than manufacturing the protein externally and injecting it (expensive, complicated), mRNA could instruct the body to produce it endogenously.

Gene editing delivery increasingly uses mRNA to transiently express editing enzymes (like Cas9), reducing the risks associated with permanent genetic modification.

Therapeutic targets that were previously "undruggable" because no small molecule could affect them become accessible through mRNA-based approaches.

The constraints are delivery (getting mRNA into the right cells), stability (mRNA degrades quickly), and manufacturing (scaling production). All are areas of active development, with the COVID-19 experience providing an accelerated learning curve.

The broader principle is that biology is becoming programmable. If you understand the instructions, you can write new ones.

Risks and Guardrails

The transformation toward continuous, predictive, preventive medicine is not without risks.

Overdiagnosis and Overtreatment

More sensitive detection means more findings. Not all findings matter. Detecting a slow-growing prostate cancer in an 80-year-old that would never cause symptoms is not a benefit—it leads to anxiety, potentially unnecessary treatment, and harm without corresponding benefit.

The challenge is distinguishing between:

• True early detection: catching dangerous conditions when treatment is most effective
• Overdiagnosis: finding conditions that look like disease but would never cause harm
• False positives: apparent abnormalities that aren't really abnormal

This requires not just better detection but better understanding of natural history—which detected conditions progress and which don't. AI trained on longitudinal outcomes data can help, but the data must be collected and analyzed carefully.

Guidelines must evolve to specify when detection should trigger intervention versus watchful waiting. The default assumption that "finding it earlier is always better" must be tempered by evidence of actual benefit.

The Worried Well

Continuous health monitoring may create populations of "worried well"—people with no symptoms who are anxious about their health because of ambiguous data.

A spike in heart rate variability. A biomarker trending toward the edge of normal range. An AI alert about a pattern that "may" indicate future risk. For many people, these signals generate worry without actionable guidance.

The psychological burden of continuous monitoring may be substantial for some individuals. Systems must be designed to minimize false alarms, provide appropriate context for findings, and offer clear guidance rather than ambiguous warnings.

Some people may be better served by less monitoring rather than more—choosing not to know things they can't act on. Respecting autonomy means respecting the choice not to be continuously surveilled by one's own devices.

Inequality of Access

The technologies described in this chapter—multi-omics profiling, liquid biopsy screening, continuous monitoring with AI integration—are expensive. In their current form, they're available to the wealthy and those with exceptional insurance.

If these technologies significantly improve health outcomes, unequal access creates unequal health. The wealthy live longer and healthier lives; the poor die of conditions that could have been prevented.

This is not new—health disparities have always existed—but the magnitude could increase if breakthrough preventive technologies remain inaccessible to most.

Policy responses are needed: coverage mandates, public investment in access, regulatory pathways that accelerate cost reduction, and explicit attention to equity in healthcare AI development and deployment.

Data Security and Misuse

Comprehensive health data is extraordinarily valuable and extraordinarily dangerous.

Value: understanding individual biology enables personalized medicine, population health research, and drug development.

Danger: the same data enables discrimination, surveillance, manipulation, and exploitation.

The threats are varied:

• Hackers stealing health data for extortion
• Employers discriminating based on health predictions
• Insurers excluding high-risk individuals
• Governments tracking citizens through health infrastructure
• Commercial actors exploiting health data for advertising or manipulation

Technical protections (encryption, access controls, privacy-preserving computation) help but aren't sufficient. Legal protections (regulations specifying permitted uses and penalties for misuse) are essential. Social norms (expectations that health data remains private) must be cultivated.

The health system is building an infrastructure of intimate surveillance. Making that infrastructure safe requires deliberate attention that has not yet been adequately applied.

Algorithmic Bias and Validation

AI systems trained on biased data produce biased outputs. Healthcare data reflects existing health disparities: underrepresentation of minorities in clinical trials, differences in care quality across populations, historical biases in diagnosis and treatment.

AI trained on this data may perpetuate or amplify disparities. A system trained primarily on data from white patients may perform worse for Black patients. A system trained on data from academic medical centers may fail in community settings.

Validation must be comprehensive: testing AI systems across demographic groups, clinical settings, and use cases. Bias must be actively measured and mitigated. Deployment must be monitored for differential performance.

This is not just an ethical concern—it's a practical one. AI systems that work for some populations and not others will lose trust and face regulatory and legal consequences.

The Path Forward

The end of "disease" as society knows it does not mean the end of all illness. It means a transformation in how medicine works: from waiting for problems to continuous monitoring; from treating symptoms to addressing causes; from one-size-fits-all to personalized approaches; from episodic care to continuous optimization.

Near-term likely (5-7 years):

• AI-assisted clinical documentation becomes standard
• Liquid biopsy cancer screening enters routine care for high-risk populations
• Wearable-detected cardiac events trigger automatic clinical alerts
• Multi-omics panels become affordable for broader populations
• Predictive models for common chronic diseases reach clinical validation

Plausible (7-15 years):

• Comprehensive continuous monitoring integrates with clinical care for most patients
• Multi-cancer early detection catches a majority of cancers at stage 1
• AI health assistants provide personalized, continuous health guidance
• Preventive interventions based on molecular signatures become standard
• The economic model of healthcare shifts meaningfully toward value-based care

Wild (speculative):

• Most major diseases become rare through early interception
• Aging itself is treated as a preventable condition
• Health becomes a continuous optimization process rather than an occasional medical encounter
• The distinction between "healthy" and "sick" largely dissolves

The transformation is already underway. What remains undetermined is the pace, the completeness, and who ends up on which side of the access gap.

The following chapters explore specific conditions—cancer, heart disease, neurodegeneration—and ask what AI-accelerated medicine might achieve against the diseases that still kill people. The tools are being built. The question is what can be done with them.

Endnotes — Chapter 3

1. Smart toilet technology for health monitoring has been developed by companies including Toto, Withings, and various startups. Clinical validation for specific biomarkers varies. Products detecting colorectal cancer markers, urinary tract infections, and metabolic indicators are in development or early market.
2. Multi-omics integration for disease prediction is an active research area. Studies have shown that combining genomic, transcriptomic, proteomic, and metabolomic data improves predictive accuracy for various conditions. Clinical implementation remains limited by cost and complexity.
3. Grail's Galleri test received FDA breakthrough device designation and is commercially available. Performance data published in clinical journals shows varying sensitivity by cancer type and stage, with higher sensitivity for late-stage cancers. The PATHFINDER and other studies continue to generate real-world evidence.
4. mRNA vaccine development timeline and mechanism are well documented in the scientific literature. The Pfizer-BioNTech and Moderna COVID-19 vaccines demonstrated the platform's potential for rapid response to novel pathogens.
5. Personalized cancer vaccines using mRNA targeting tumor neoantigens are in clinical trials by multiple companies including BioNTech and Moderna. Early results in melanoma and other cancers have shown promise, though large-scale efficacy trials are ongoing.