Episode 46 · The UpStream Life · Vishal Krishna in conversation with Anirudh

The lab report before the diagnosis — and a clean slate for Indian biomarkers.

Anirudh was twenty-six during the pandemic when his bloodwork came back with a marker called anti-CCP elevated above the reference line, and a single phrase at the bottom of the report: rheumatoid arthritis. A telemedicine doctor confirmed it calmly, prescribed medicines, and queued more tests. A second laboratory returned the same elevated number. A rheumatologist called it early onset and said the prognosis was two to three years of declining pain followed by a sharply truncated lifespan. Four months later, after he had cleaned up his diet, joined his parents in Lucknow because he could no longer stand up without pain, and entered the literature looking for an operating mechanism that explained any of it, a third lab returned a clean panel. The marker, which doctors do not retest once it goes up, had never been elevated in the first place. Beyond is the company he is building because the lab is a system, not a number, and the system is upstream of the disease.

Guest Anirudh · Founder, Beyond (getbeyondhealth.com)· Host Vishal Krishna· Length 46 min· Recorded 2023
Video thumbnail
The journey of Beyond — using data and AI to live longer and better by understanding your body chemistry
Opens YouTube

In sixty seconds.

Most of the health data an urban Indian generates today — lab tests once a year, wearable step counts every minute, sleep scores nightly, food log entries when guilt strikes — sits in silos that do not speak to each other. The trend is invisible because the year-over-year HbA1c lives on a paper printout, the cortisol curve lives in a screenshot, the cholesterol number lives in an email attachment. Anirudh's diagnosis: the data exists, the integration does not, and the reference ranges underneath the data were never calibrated for the body actually being measured.

Beyond is a preventative health company that draws blood at home through an NABL-certified lab, ships an FDA-approved continuous glucose monitor for the two-week window it takes to learn what your food is actually doing, pulls your wearable's sleep and stress and activity data into the same chart, and presents one evolving picture of your body across the years it can afford to track. The thesis underneath the product: almost every preventative protocol on the internet was peer-reviewed on fifty to two hundred Caucasians, and the reference ranges your local lab uses are downstream of that same data, which means an Indian body is being scored against a chemistry it was never sampled inside.

The conversation moves through Anirudh's own anxiety-as-first-job experience in 2015, the cortisol-HbA1c-cholesterol vicious cycle nobody warned him about, the four-month false-positive rheumatoid-arthritis episode that became the company's origin story, and the structural argument that most health-tech in India is a Swiggy patched onto a wearable. Beyond's bet is that the durable startup in this category is built slowly, on a thousand clean longitudinal records before a hundred thousand noisy ones, with the protocol engine learning from a population that finally looks like its own users. The cost is real. Beyond is expensive, on a waitlist, and unapologetic about both.

Where to land in the conversation.

Each chapter opens the YouTube video at that timestamp in a new tab.

Six ideas to carry into your own work.

Mental models lifted from the conversation that travel beyond preventative health. Each one is the kind of thing you could quote in a clinical meeting on Tuesday or repurpose in any category that runs on data, trust, and slow compounding.

01

The trend is the diagnosis, not the number.

Anirudh's core editorial rule: an HbA1c of 5.7 today carries almost no information; a six-month move from 5.4 to 5.7 and a six-month move from 5.9 to 5.7 carry opposite information about the same person. Beyond's data model is built around delta-on-baseline rather than single-point thresholds. The corollary is that any first test is a calibration, not a verdict, and the lab that does not retain the second and third tests has thrown away the only signal that mattered.

A point estimate is a snapshot; a trajectory is a story. The trajectory is almost always the actionable one.
02

Optimal is not undiseased.

The diagnostic standard a doctor uses is binary — medicine, or no medicine. A 5.5 HbA1c is not prescribed against because there is no medicine to write at 5.5, but a 5.5 is also not optimal for a thirty-year-old with active glucose machinery and decades of compounding ahead. Beyond's editorial line is that the reference range underneath the lab report was designed to catch disease, not to define health, and the gap between the two is where preventative work lives.

A measurement system built to detect failure cannot, by construction, optimise for performance. The two require different scales.
03

The clean-slate-data bet.

Most preventative protocols circulating on Indian wellness platforms are downstream of peer-reviewed studies run on fifty to two hundred Caucasian subjects. Anirudh's argument is structural: the recommendation engine cannot be better than the population the underlying trials sampled. Beyond is building what Anirudh calls a remote clinical trial — collecting clean longitudinal data on Indian bodies first, validating Western protocols against that data, and updating the recommendations as the population responds. The bet is slow data wins.

A category whose default truths were established on someone else's population is one good cohort away from being repriced.
04

The pre-analytic chain is the lab.

Anirudh names the failures that turn a clean blood test into a noisy one before the analyser ever sees it: needle gauge, time of draw, the hours the sample sits in the sun before reaching the centrifuge. A NABL certification is a floor for the analytical phase. The pre-analytic phase — the part between the puncture and the lab door — is unmonitored at most home-collection vendors, and it is where the false positives and the spurious trends are born.

In any measurement chain with a tightly regulated last mile, the unregulated first mile is where the systematic error compounds.
05

You cannot build a Swiggy of health.

The most-quoted line of the episode. Anirudh's diagnosis of Indian health-tech: most companies took the consumer-internet playbook — light infrastructure, deep marketing, fast scaling — and tried to retrofit it onto a category where the customer's body is on the line. The structural answer Beyond is testing: build thoughtfully, charge real prices, accept the slower curve, treat the first thousand users as collaborators, and price the product as a quality signal rather than an acquisition lever.

When the cost of being wrong is borne by the user's body, the e-commerce growth model is the wrong shape; the long-form services company is the right one.
06

Protocol engine plus doctor-in-the-loop.

Beyond's stack is not a black-box recommender. The protocol engine surfaces a candidate intervention from the data; a human clinician verifies it; the verified set then trains the engine's next iteration. The architecture treats the algorithm as a draft and the doctor as the editorial gate. Over years, the gate's interventions become training data for a system that does not need the gate as often. The discipline rules out any product where the algorithm ships directly to the user.

The high-stakes machine-learning product is built with the expert in the loop until the loss function knows what the expert knows.

Fifteen things to walk away with.

Each one carries the timestamps where the moment lives and a transferable note for work that isn't preventative health. The order is the order of escalation — from the anxiety that begins the personal story to the company that begins the structural one.

01

Anxiety as a first-job inheritance.

Anirudh's anxiety did not start in college; it started the year he joined his first job, 2015. Before that, no episodes. After that, peaks and troughs that he treated as something to power through with mind alone — new framework, change-your-thinking, get on the podcast and don't look weak. The pattern repeated for a decade. It is a familiar shape for anyone who entered a high-output workplace in their early twenties and assumed the discomfort was character-building rather than a signal that the body's regulatory loop had shifted.

The reframe came when he started measuring. The anxiety was the visible end of a metabolic loop — cortisol rising, HbA1c rising, cholesterol rising, each one feeding the others — and the cognitive intervention had been operating on the wrong system. The body was sending the signal; the mind was misreading it as a software bug. Anirudh is careful to say he sat down for the podcast that day not anxious, because the loop had been broken upstream, not because he had finally learned to manage it.

Beyond clinics. When a chronic state arrives at a known life transition, the temptation is to treat it as a stage of growth. The cheaper test is to measure the underlying chemistry first; the mental intervention can wait until the metabolic loop is ruled out.
02

The COVID lab test that started the company.

Locked in at home, waking up with stiffness in his fingers and legs, Anirudh did what most urban Indians do because lab access is uniquely easy here — he ordered a panel from home. The report flagged anti-CCP elevated above the cutoff, with a single phrase at the bottom: rheumatoid arthritis. A telemedicine consultation confirmed the reading calmly and queued medicines. He did not trust it. A second lab from a different provider returned the same elevated marker. A rheumatologist looked at the file and said early onset, watch and wait, no medication yet, prognosis bad.

The episode is the company's origin story for a reason. The lab system that gave him three confident answers was not three independent measurements; it was one questionable pre-analytic chain replicated. Four months later, after he had moved home to Lucknow because he could no longer stand without pain, a third lab returned a clean result. The marker had never been elevated. The body had been responding to the diagnosis, not the disease. Beyond's first product principle is that a lab number you cannot verify against an integrated picture is a number you have to defend against, not act on.

Beyond clinics. A single-source signal that triggers a high-stakes intervention deserves at least one independent confirmation in a different chain. Replication inside the same vendor is correlation, not corroboration.
03

Twenty-six, two to three years, what now.

The rheumatologist's note in plain English: two to three years of less pain, then a steady progression as cartilage went, then a sharply shortened lifespan. Anirudh was twenty-six. He describes the moment as the one where you either give up or go very deep, and the deep version came with what he calls hare-brained thoughts — if I can stall this long enough, medical science might catch up. He spent weeks reading the operating mechanism of the disease, watching how it propagated, looking for the lever that might bend the trajectory.

The reframe is structural rather than emotional. The diagnosis was wrong, but the response to the wrong diagnosis was the right response to a real one. He cleaned up diet, audited every input, ran research models on the disease's biology. Months of work that he later describes as solving the wrong problem — but the discipline laid down in those months is the discipline the company runs on now. The founder who reads the mechanism rather than accepting the label is the same founder who insists Beyond is a data company, not a wellness brand.

Beyond clinics. A wrong diagnosis received at the right age can be the most useful event in a founder's career, because the time spent solving an imagined problem builds the muscle to solve the real one. Do not waste the dry run.
04

The Caucasian-data critique.

Anirudh names the problem squarely. Almost every preventative protocol circulating in Indian wellness apps is downstream of a peer-reviewed paper that ran on fifty to two hundred Caucasian subjects, often in a North American or Western European centre, often with a randomised control trial design that does not survive past phase two of preventative work because nobody funds the long arm of prevention. The protocol then arrives in India as if it were a universal law of human biochemistry. It is not. The reference ranges your local laboratory uses are themselves downstream of the same data, so the score you are graded against was calibrated on a chemistry that was never sampled on your continent.

The remediation Beyond proposes is what Anirudh calls a remote clinical trial — collecting clean longitudinal data on Indian users, validating each Western protocol against that data, and progressively updating the recommendations as the cohort grows. The premise is that you do not need a hundred thousand Indians to do this; you need a few thousand thoughtfully studied with high-quality instrumentation, which is what the waitlist and the price point allow him to assemble. The data is the long-term moat. The protocols are the proximate product.

Beyond clinics. Any recommendation system built on a foreign population's calibration data has a precision ceiling that no amount of UX can raise. The remediation is collecting the right data on the right cohort, not adding more inputs to the same one.
05

Trends, not single points.

The single most carrying-power editorial move in the conversation: an HbA1c reading of 5.7 has almost no information in it on its own. A six-month move from 5.4 to 5.7 means one person; a six-month move from 5.9 to 5.7 means the opposite person. The reading is identical. The trajectory is opposite. Anirudh's product rule is that the first test is calibration, not result, and that any health platform that returns single-point thresholds without baselining against the user's history is leaving the only actionable signal on the floor.

The operational consequence is that Beyond digitises every prior lab report a user can supply on intake, sets the baseline before the first home draw, and then keeps the trajectory live. The dashboard shows direction, not just position. Anirudh's example of the user who looks fine on this year's panel but moved by half a unit in the wrong direction is the user the doctor will not see until the panel crosses the threshold, which is three to five years too late for prevention.

Beyond clinics. Any metric that drives high-stakes decisions has to be reported with its trajectory. Position-only dashboards systematically miss the cohort that is moving fastest in the wrong direction while still inside the green band.
06

Optimal is not the same as not diseased.

The reference range on a standard Indian lab report is a disease threshold. The 5.7 HbA1c sits inside the green band for a clinician because there is no medicine to prescribe at 5.7. Anirudh's editorial line: not diseased is not good enough. A thirty-year-old at the upper edge of the green band has two to three decades of compounding glucose exposure ahead, and the gap between 5.7 and 5.2 is the gap between a body that will and a body that will not develop pre-diabetes by forty-five. The doctor cannot help inside that gap because the doctor's tool is medication. Beyond's tool is lifestyle, and lifestyle works in the gap.

The implication for the product is that Beyond defines its own optimal ranges by sex, age and activity level — not the lab's clinical cutoffs — and reports against those. The doctor's report says fine. The Beyond report says fine for not being sick, and here is the work to be optimal. The editorial split between those two registers is the company's central editorial decision; it is why a clinical lab cannot become Beyond by adding a wearable.

Beyond clinics. The standard of measurement embedded in any incumbent diagnostic is the standard of failure detection. A category that wants to optimise the same variable has to introduce a parallel scoring system rather than fight the incumbent's threshold.
07

The pre-analytic chain is what fails.

Anirudh walks through the failure modes that turn a clean blood test into a noisy one before it ever reaches the analyser. Needle gauge affects the cell counts. Time of draw affects the hormone panel. The window between the puncture and the centrifuge matters disproportionately for the more delicate markers; a tube that sits in a phlebotomist's bag for hours in the sun before reaching the laboratory door has chemistries that no analyser can recover. A NABL certification regulates the analytical phase well; it says almost nothing about the pre-analytic phase, which is where the home-collection vendors compete on price and shave on quality.

Beyond's operational answer is to control the pre-analytic chain end-to-end. The phlebotomist is trained on the company's protocol. The collection window is constrained. The sample reaches the lab inside the time-and-temperature envelope the parameters demand. The cost is real and the company is unapologetic about it. The alternative is the kind of false-positive Anirudh himself lived through, which is what happens when the first analytic phase is good and the chain in front of it is unmonitored.

Beyond clinics. Any quality system that audits only the regulated last mile of a measurement chain will be defeated by the unregulated first mile. The end-to-end audit is the only one that meaningfully predicts the output.
08

The CGM is the Abbott Freestyle inside everyone's patch.

Most Indian continuous-glucose-monitor products are doing one of two things. They are taking an Abbott Freestyle CGM — the third-generation device that is FDA-approved internationally and clinically validated for diabetes care — and they are writing a software layer on top of it. The patch you see in the marketing photograph is the same patch any other vendor is shipping. The differentiation is the app. Anirudh is unsentimental about this; Beyond uses the same FDA-approved underlying device because there is no clinical reason to reinvent the hardware. The point is to do the right thing with the data the hardware produces, not to claim a hardware moat that does not exist.

The deeper observation is about the gap between marketing claim and engineering reality in Indian health-tech. The companies advertising a proprietary CGM are not, by and large, building one. The companies claiming an end-to-end product are usually shipping a software wrapper. Anirudh's editorial position is that the customer is better served by the honest version — FDA device underneath, NABL lab in the middle, Beyond's protocol engine on top — than by the marketing version that obscures the supply chain.

Beyond clinics. Honesty about the parts you did not invent is a market signal. The product whose marketing matches its bill of materials earns trust the louder product cannot match.
09

The city is engineered against your metabolism.

Anirudh's diagnosis of urban Indian life is environmental. Jobs are sedentary by default. Food arrives at the door via Swiggy at a price point that makes cooking economically irrational. The walk to the restaurant, which used to be the baseline minimum of daily movement for a city worker, is gone. Add fifteen-minute grocery delivery and the door is the perimeter of a closed system. Movement, the variable with the largest long-run impact on metabolic health, has been engineered out of the day by a stack of conveniences that each looks like a small win.

The remediation he names is not anti-convenience; it is sustainable habit. Beyond's protocol for any new user includes hard-thing prescriptions a few days a week — fasting, a long walk, a session that the user resists initially because the framing of fasting in particular scares most people — alongside changes to the default diet that the user does not have to think about. The point is to interrupt the closed system rather than replace it. The hard things are the spike; the diet adjustments are the floor.

Beyond clinics. A category whose users have been infrastructurally optimised away from a desired behaviour cannot be fixed with willpower campaigns. The remediation has to introduce environmental friction the platform itself supplies.
10

You cannot build a Swiggy of health.

The conversation's sharpest commercial line. Anirudh's diagnosis: most Indian health-tech startups have run the consumer-internet playbook — light infrastructure, deep marketing, an e-commerce model, fast scaling — and the category is structurally hostile to that shape. The user's body is on the line. The mistakes show up in lab results months later. The unit economics of moving fast and breaking things are catastrophic when the thing being broken is the customer's metabolic floor. Anirudh's read is that none of those companies are actually working; they look like e-commerce because they are e-commerce.

Beyond's structural answer is the opposite shape. Charge real prices because price is a quality signal at the top end of the health market. Move slowly because the protocol engine needs clean data more than it needs scale. Treat the first thousand users as collaborators, not acquisition targets. Refuse the categories of growth that would compromise data hygiene. The bet is that the durable Indian preventative-health company looks like a thoughtful services business with a software backbone, not a software company with a services line.

Beyond clinics. Any category whose failure modes are felt by the user's body has to be priced and paced like services, not like software. The growth model has to fit the cost-of-being-wrong, not the other way round.
11

The doctor-in-the-loop protocol engine.

Beyond's stack does not ship a black-box recommendation to the user. The protocol engine surfaces a candidate intervention from the integrated data — blood biomarkers, wearable signals, CGM curve, family history. A clinician verifies the candidate. The verified intervention then goes back into the training data. Over time the engine learns the verifier's judgement. The architecture treats the algorithm as a draft and the doctor as the editorial gate. This is the structural answer to the question of how a small startup with a few thousand users competes with a category that wants to be Cure.Fit at a million.

The discipline rules out a class of products that ship the model directly to the user. Beyond's claim is that the high-stakes machine-learning product belongs to the species of system that keeps the expert in the loop until the loss function has internalised what the expert knows — and that this is decades of work in preventative health, not months. The interim product is a tightly orchestrated combination of clinician time and algorithm output; the long-term product is an engine that needs the clinician less often. The order of operations is the editorial choice.

Beyond clinics. The expert-in-the-loop machine-learning product is not a transitional architecture; it is the durable one in any category where the cost of an unaudited recommendation is paid by the user's body.
12

Price is a quality signal in health.

Anirudh is direct about Beyond's price point. The company is expensive, on a waitlist, and not apologising for either. The reasoning has two parts. The first is operational — the end-to-end pre-analytic control, the FDA-approved CGM, the NABL laboratory, the clinician time, the engineering of the integrated dashboard. The price reflects the bill of materials. The second is editorial. Especially in health, especially in a market full of two-thousand-rupee panels of dubious provenance, the price tag is itself a signal about what the user is buying. A cheap full-body test is almost always wrong somewhere in the chain. The willingness to pay sorts the cohort.

The cohort that signs up first is therefore the cohort that already values their data, which is the right early population for a remote clinical trial. The product progresses down the income curve as the protocol engine learns enough to ship a less labour-intensive version to a broader audience. The two-stage market is intentional; the first thousand pay for the data that lets the next ten thousand be served more cheaply. The Caucasian-data critique becomes the population-bootstrap strategy.

Beyond clinics. A premium first market is acceptable when the first market is the data engine that will eventually subsidise the broader one. The price is the cohort filter, not the long-term equilibrium.
13

The 10,000-step wearable does not mean you are healthy.

Vishal volunteers his own habit — ten thousand steps a day, wearable confirms, therefore healthy. Anirudh's correction is gentle and structural. The cheaper wearables count motion that is not walking — the steering arm of a driver, the gestures of a desk worker — and the step count overstates real activity. Even when the count is accurate, steps are a coarse proxy for movement, and movement is one of perhaps a dozen variables that decide whether the body's metabolic floor is in good shape. A high step count alongside a rising cortisol and a worsening HbA1c is a body that is paying for activity with stress; the step count is signal, but it is not sufficient.

The implication for the user is that any single-variable health practice is a leading indicator that has not been audited against the rest of the chemistry. The implication for an insurance company is that step counts cannot, on their own, justify premium discounts; the right discount is computed against the integrated picture Beyond is collecting, not against a wearable's stand-alone signal. The wearable is an input; it is not the output.

Beyond clinics. Any single metric that promises to summarise a multi-variable system is performing arithmetic the system does not support. The single number is a feature, not a verdict.
14

Build in India because regulation is light, urgency is high.

Anirudh grew up in the US, but he is building in India deliberately. The structural argument is two-sided. The US market is mature; access to a doctor takes two weeks, the panel that comes back is comprehensive, the regulation absorbs years of compliance time before a preventative-health product can even enter the market. India is the inverse. Urban Indian customers have the affluence and lifestyle profile of their Western counterparts, but the regulation is lighter and the healthcare layer above them is shallower. A preventative-health company can enter the market and begin shipping while a comparable US company is still in regulatory queue.

The personal half of the argument is that Anirudh wants to build here regardless. He grew up in the States, came back, and is unapologetic that the company is being built in India because he loves the country. The strategic case for it is real, but the founder's motivation predates the strategic case. The order is important. Beyond is built in India because Anirudh chose to build in India; the regulation arbitrage is a tailwind rather than a thesis.

Beyond clinics. A founder's geographic conviction precedes the strategic argument for the geography. The market analysis has to be done, but it follows the choice rather than producing it.
15

Principles, the power of ethics, a father who taught himself code.

The last ten minutes of the conversation are the founder's reading list and the family chemistry behind it. Anirudh is on his annual re-read of Ray Dalio's Principles and a slower trip through The Power of Ethics, the second of which he describes as a horrible book to read because there is so much on each page that you have to put it down and meditate on each chapter. The shape of his learning is principles laid down by someone else and then absorbed slowly enough to become his own — the same shape he is now trying to design into Beyond's protocol engine.

The family piece is small but load-bearing. His father is a chemistry professor in Bangalore who taught himself to write code when Anirudh was eight, partly to fix things at home, partly because there was no reason not to. The light fixtures and the speakers in the family house are his father's builds. Anirudh's mother is the optimist who held the line during the four months in Lucknow when the rheumatoid-arthritis diagnosis looked real. The founder did not arrive at preventative health from a clean slate. He arrived from a household that already treated agency over your own systems as the default.

Beyond clinics. A founder's editorial defaults are usually the household's editorial defaults, with one generation's distance. The investor due-diligence question that returns the most signal is not the resume; it is the household.

Lines worth keeping near your desk.

We’re basically trying to build a preventative Healthcare ecosystem that is more based on science. Anirudh · 01:48
Your cortisol and your HbA1c and your cholesterol form a nice vicious cycle, and then you end up on a very bad path. Anirudh · 05:14
Most preventative protocols are usually run on fifty to two hundred Caucasian people, and that’s the data that gets thrown at the whole world. Anirudh · 26:36
We’re about prevention. Just not diseased is not good enough. What is optimal. Anirudh · 14:23
You can’t build a Swiggy in health. You have to be a lot more thoughtful, and if you take care of the details, you end up just getting users that stick. Anirudh · 26:03

The jargon, unpacked.

Some of these are preventative-health terms specific to Beyond's product; some are lab-science basics. Skim, mark the unfamiliar, come back later.

Biomarker
measurement, noun
A measurable indicator of a biological state — a blood glucose level, a cortisol reading, an inflammation marker. Beyond's product is built around tracking biomarker trajectories rather than single-point readings, and around defining optimal rather than merely undiseased ranges per sex, age and activity profile.
HbA1c
blood marker
Glycated haemoglobin — a three-month average of blood glucose. The clinical diabetic threshold sits at 7.5 and above; the standard reference range pre-diabetic at 5.7 to 6.4. Anirudh's editorial line is that for a thirty-year-old, optimal is closer to 5.2; the gap between 5.2 and the clinical threshold is the gap preventative work lives in.
Anti-CCP
blood marker
Anti-cyclic citrullinated peptide antibody. Elevated levels are an indicator for rheumatoid arthritis. The marker is rarely retested once elevated because clinical convention assumes it does not return to baseline. Anirudh’s personal experience is that the underlying lab result can be a pre-analytic false positive, and that the convention against retesting is itself a quality-system failure.
Cortisol
hormone
The stress hormone; produced by the adrenal glands. Chronic elevation is one leg of the cortisol-HbA1c-cholesterol cycle Anirudh names as the under-appreciated metabolic spiral for high-output young professionals.
CGM
acronym
Continuous Glucose Monitor — a wearable patch that tracks blood glucose in real time. Beyond ships the FDA-approved Abbott Freestyle CGM (third generation), the same underlying device most Indian CGM products are repackaging. The differentiation is the software wrapper and the integration with the rest of the user's data.
NABL
certification, acronym
National Accreditation Board for Testing and Calibration Laboratories — the Indian standards body that accredits clinical laboratories. NABL certification regulates the analytical phase; the pre-analytic phase (sample collection, transport, time-to-analysis) sits largely outside its scope. Beyond uses NABL labs but controls the pre-analytic chain in-house.
Pre-analytic phase
lab science
Everything that happens to a blood sample between the needle and the analyser door — needle gauge, time of draw, sample transport, temperature, time-to-centrifuge. Anirudh names it as the lab system's most under-monitored failure mode and the likely source of the false positives that triggered his own four-month rheumatoid-arthritis episode.
VO2max
biomarker
Maximum rate of oxygen consumption during exercise — the textbook measure of cardiovascular fitness. Not mentioned by name in this episode but implied in Anirudh’s endurance assessment; the Beyond panel covers metabolism, cognition, strength and endurance as parallel scoring axes, of which VO2max is the standard endurance metric.
ApoB
blood marker
Apolipoprotein B — a measurement of the protein component on every atherogenic particle in the blood, increasingly used as a more precise cardiovascular-risk marker than standard LDL cholesterol. Implicit in the missing-parameters critique Anirudh names when US-based doctors review Indian lab panels.
Lab interoperability
data infrastructure
The capacity of one lab’s outputs to be ingested, normalised and compared against another lab’s outputs. Largely absent in Indian healthcare. Beyond digitises every prior report a user supplies on intake to manufacture a longitudinal baseline that the rest of the lab system does not, by default, provide.
Doctor-in-the-loop
ML architecture
A machine-learning system design where a clinician verifies each algorithmic recommendation before it ships to the user. Beyond’s protocol engine is built this way: candidate intervention from the model, clinical verification, verified set rejoins the training data. The architecture trades throughput for safety.
Protocol engine
Beyond term
The recommendation system at the core of Beyond’s product. Surfaces candidate diet, supplementation, fasting and movement protocols against a user’s integrated data — biomarkers, wearable signal, CGM curve, family history, prior reports. Each output passes through a human reviewer before reaching the user.
Randomised control trial
study design
The gold-standard clinical study design in which subjects are randomly assigned to treatment or control. Anirudh’s critique is not of the methodology but of its population: most preventative-health RCTs run on small Caucasian cohorts because of where the funding lives, and the results travel as if universal.
Remote clinical trial
Beyond’s framing
Anirudh’s description of what Beyond is operationally doing — assembling a longitudinal Indian cohort, applying candidate protocols at scale, measuring outcomes against the integrated baseline, updating recommendations. The unit is the population, not the individual case.
Blue zones
cultural-health concept
The five regions Dan Buettner identified as having outsized centenarian populations — Okinawa, Sardinia, Nicoya, Ikaria, Loma Linda. Cited in the broader preventative-health literature as the qualitative case for sustained lifestyle interventions over single-point clinical fixes. Adjacent to the “live longer and better” thesis Beyond pitches against.

Check what you actually retained.

Try to answer before you click. The point is to notice where the conversation is fuzzy in your memory, then return to the transcript.

Q1
What happened in Anirudh’s anti-CCP story over the four months he believed he had rheumatoid arthritis
A home blood test during COVID flagged anti-CCP above the cutoff, with rheumatoid arthritis printed at the bottom of the report. A telemedicine doctor confirmed and prescribed medicines. A second laboratory returned the same elevated value. A rheumatologist called it early onset. Four months later, after Anirudh had moved home to Lucknow because he could no longer stand without pain, a third lab returned a clean result. The marker had never been elevated; the original chain produced two correlated wrong answers.
Q2
Why does Anirudh argue that an HbA1c reading of 5.7 has almost no information on its own
Because the same point can come from a falling trajectory (5.9 to 5.7) or a rising one (5.4 to 5.7), and those are opposite signals about the same person. The actionable diagnostic is the trend, not the value. A platform that shows position without trajectory leaves the only useful signal on the floor.
Q3
What is the Caucasian-data critique, and how does Beyond plan to address it
Most preventative protocols are downstream of peer-reviewed studies run on fifty to two hundred Caucasian subjects, and the reference ranges used by Indian labs inherit the same calibration. Beyond proposes a remote clinical trial — collecting clean longitudinal data on Indian users and validating each Western protocol against that data before recommending it. The bet is that a few thousand thoughtfully studied users beat a hundred thousand noisy ones for protocol calibration.
Q4
What is the cortisol-HbA1c-cholesterol vicious cycle Anirudh names
Chronic stress raises cortisol, which contributes to glucose dysregulation (raising HbA1c) and unfavourable lipid profile (raising cholesterol). The metabolic shift increases physiological stress, which feeds cortisol back upward. Anirudh names this as the underlying mechanism of his anxiety since 2015, and the loop he broke by addressing the metabolic chemistry rather than the felt-state alone.
Q5
What does Anirudh mean by “optimal is not the same as not diseased”
The reference range on a standard lab report is a disease threshold — the cutoff where a doctor would prescribe medicine. Below that cutoff there is no clinical intervention, but for a thirty-year-old with decades of compounding ahead the gap between, say, an HbA1c of 5.7 and 5.2 is meaningful. The doctor cannot operate inside that gap because the tool is medicine; Beyond’s editorial choice is to score against a parallel “optimal” range, not the lab’s clinical cutoff.
Q6
Why does Anirudh say you cannot build a Swiggy in health
The consumer-internet playbook — light infrastructure, deep marketing, fast scaling — assumes a customer who absorbs small failure modes cheaply. In health, failure modes show up months later in lab results and they are paid for by the user’s body. The growth model has to match the cost-of-being-wrong, which is why Beyond is priced and paced like a services business with a software backbone rather than the other way round.
Q7
What is the difference between NABL certification and pre-analytic control, and why does it matter
NABL accredits the analytical phase — what happens inside the laboratory. The pre-analytic phase — needle gauge, time of draw, transport temperature, time-to-centrifuge — sits largely outside its scope. Most false positives Anirudh names are pre-analytic in origin. Beyond uses NABL labs but controls the pre-analytic chain in-house, which is part of why the product is expensive.
Q8
What is the doctor-in-the-loop architecture, and what does it rule out
Beyond’s protocol engine surfaces a candidate intervention from the integrated data; a clinician verifies the candidate; the verified output rejoins the training data. Over time the engine learns the verifier’s judgement. The architecture rules out any product that ships the model directly to the user. The expert-in-the-loop design is the durable shape for any category where the cost of an unaudited recommendation is paid by the body.
Q9
What does Anirudh actually mean when he says Beyond is “a data company”
Beyond is not selling lab tests, wearables, supplements or services as the primary product. Those are the inputs. The product is the integrated longitudinal record of the user across blood biomarkers, wearable data, CGM curves and family history, scored against the optimal range and tracked over years. The long-term asset is the cohort-level data; the short-term touchpoint is the report and the protocol.
Q10
What is the structural argument for building Beyond in India rather than the US
Urban India has the affluence and lifestyle profile of the Western markets but a lighter regulatory layer and a shallower preventative-care infrastructure. A preventative-health company can begin shipping in India while a comparable US company is still in regulatory queue. The motivation predates the strategic argument — Anirudh wanted to build in India regardless — but the structural case for the geography is real.
Q11
Why does Anirudh push back on the ten-thousand-step-a-day-equals-healthy framing
Cheaper wearables count motion that is not walking — gestures of a desk worker, the steering arm of a driver — and the step count overstates real activity. Even with an accurate count, steps are a single coarse proxy for movement, which is one of perhaps a dozen variables that decide metabolic health. A high step count alongside rising cortisol or HbA1c is a body paying for activity with stress; the wearable is an input, not a verdict.
Q12
What is the eventual partnership posture Beyond imagines with insurance, pharma, and the existing healthcare players
Anirudh imagines Beyond as the customer-obsessed layer that integrates with existing players rather than replacing them. Insurance could adjust premiums against the Beyond-tracked picture instead of generic step counts. Doctors could read the integrated dashboard rather than the static panel. Pharma fits behind the prevention layer rather than around it. The Holy Grail he names is doctors using Beyond’s data as the input to their own consultation.

Five prompts. Notes save in your browser.

Use these to convert the reading into your own decisions. Nothing is uploaded; storage is local to this device.

Anirudh broke a chronic state by measuring its underlying chemistry rather than reframing his thinking about it. Where in your own life are you applying a cognitive intervention to what is probably a physiological signal

The trend rule says a single reading carries no information; the trajectory does. What metric in your own work are you tracking as a snapshot when you should be tracking it as a slope

The Caucasian-data critique says a recommendation engine cannot be better than the population it was calibrated on. What default assumption in your category is downstream of someone else’s data, and what would your equivalent of a remote clinical trial look like

Anirudh refuses to build a Swiggy of health because the cost-of-being-wrong is paid by the user’s body. Which growth playbook are you currently running on a category whose failure modes don’t match the playbook’s assumptions

The doctor-in-the-loop architecture treats the algorithm as a draft and the expert as the gate. Where in your stack are you shipping a model output directly to the user without an expert review you should be inserting

The strongest version of the disagreement.

Four counter-arguments that an honest sceptic would press on this conversation. Each is written to be persuasive, not to win.

Preventative health is a premium-segment hobby, not a national-scale category.

The market counter, articulated charitably.

The honest version of the argument is that the customer Beyond is built for — affluent urban Indian, thirty to fifty, paying real money for a lab plus a CGM plus a clinician’s time — is a sliver of the market that any preventative-health company globally has ever served. The bottom of the income curve is not buying optimal HbA1c; it is buying treatment for already-established disease, and that is what the insurance and hospital infrastructure is built for. A thoughtful company at the top end may build a beautiful product and a durable brand, but the addressable market in India remains small enough that the data engine never gets to the population scale Anirudh’s remote-clinical-trial pitch requires. The honest read is that Beyond becomes a high-margin boutique business; the structural claim that it rewrites Indian preventative health is the part that needs evidence beyond the first cohort.

The Caucasian-data critique is rhetorically powerful but operationally hard to act on.

The methodology counter.

Anirudh’s argument that preventative protocols are run on small Caucasian cohorts is correct and important, but the corollary — that Indian-cohort data will produce materially different protocols — is not yet demonstrated. Most biomarkers we measure (cholesterol, glucose, blood pressure, inflammation markers) have well-characterised biology that does not vary dramatically across populations at the level of the recommendation. The Caucasian-data problem matters most for rare-disease risk, specific pharmacogenomic responses, and a handful of inflammation profiles that may indeed differ. The systemic recommendations — exercise more, sleep more, eat less processed food, reduce stress — are likely to look very similar after a thousand Indian users as after a thousand Americans. The cohort-collection thesis may end up validating Western protocols rather than overturning them, in which case the moat is the integrated picture, not the population recalibration.

The doctor-in-the-loop architecture is the right shape but caps the company at boutique scale.

The unit-economics counter.

A protocol engine that requires a human clinician to verify every output is a service business with a software backbone, not a software business. The unit economics are bounded by clinician hours, which do not scale exponentially. The competitors who ship model outputs directly to the user will get the unit economics wrong for safety reasons but right for growth, and the brand-trust gap Anirudh describes may not be as decisive as he believes — once the unverified competitor is large enough, the market will accept some false-positive rate as the price of accessibility. Beyond’s discipline is real and probably correct ethically, but the question is whether the discipline produces a sustainable institutional shape or a beautiful niche. The honest read is that the latter is more likely than the former, even if Anirudh wins on every editorial point.

The pre-analytic chain critique is real but the customer cannot evaluate it.

The trust-asymmetry counter.

Beyond’s premium price point is partly justified by end-to-end control of the pre-analytic phase — the needle, the transport, the time-to-centrifuge. The argument is technically correct, but a customer cannot, by inspection, distinguish a well-controlled pre-analytic chain from a poorly controlled one. The visible artefacts — the phlebotomist’s uniform, the cooled transport box, the polished app — can be replicated by any competitor at any quality level. The trust signal is therefore narrative and brand-driven, not falsifiable by the customer. In the long run, Beyond’s claim about pre-analytic discipline becomes a marketing position rather than a defended technical edge, and the company has to live with the fact that the cheaper competitor with a similar narrative will look identical to the customer who has no way to audit either. The remediation, if there is one, is publishing the pre-analytic protocols and the QA results — which most labs, Beyond included so far, do not do.

Three readers, three different jobs to do.

Each card is a checklist for one role this conversation is most useful to. Pick the one that fits your week.

S

If you are someone trying to get a real read on your own health

  • Audit your last three lab reports. If you cannot lay them next to each other and see a trajectory for HbA1c, lipid profile, vitamin D and inflammation, you have data and no picture. Build the picture this week, even if you have to manually copy numbers into a spreadsheet.
  • Resist the single-test-from-the-cheapest-vendor habit. A panel from a price-led aggregator can produce a false positive that costs months of stress; the cheap test is not always the cheap option.
  • Track the slope, not the score. Any biomarker that has moved more than ten per cent against its own baseline in six months is worth a conversation with a clinician who looks at the panel as a whole, not as line items.
  • Treat your wearable’s step count as one input among many. A high step count alongside rising cortisol or HbA1c is a body paying for activity with stress; the wearable is signal, not verdict.
  • If a marker comes back outside the range, get the test repeated at a different chain before you start the medication conversation. Replication inside the same vendor is correlation, not corroboration.
F

If you are a founder building in a high-stakes consumer category

  • Choose your growth playbook against the cost-of-being-wrong, not the addressable market. If the failure mode is paid by the user’s body or money, the e-commerce model is the wrong shape; build the services-with-software-backbone version.
  • Price the product as a quality signal at the top end of the market until the protocol engine has the data to ship a less labour-intensive version downstream. The first cohort funds the calibration; the second cohort buys the calibrated product.
  • Be honest about which parts you did not invent. Stating the FDA-approved hardware and the NABL-certified lab in the marketing buys more trust than the proprietary claim, and the customer who notices is the customer worth keeping.
  • Build the expert-in-the-loop architecture from day one. Shipping the model output directly to the user is fast and unsafe; the right curve is to use the expert’s verifications as training data and shrink the expert’s role over years.
  • Treat the first thousand users as collaborators in the data set, not as acquisition targets. Their reports become the cohort. Their willingness to pay sorts the data hygiene.
I

If you run an insurance, pharma, or hospital partnership desk

  • Stop pricing premiums against single-variable wearable signals. A step count in isolation overstates activity for the casual user and understates risk for the metabolically deteriorating one; the integrated biomarker picture is the only honest input.
  • Pilot a preventative-health discount lane separate from the standard reactive-care discount lane. The unit economics of catching pre-diabetes at 5.2 are very different from the unit economics of treating diabetes at 7.5, and the pricing has to reflect both.
  • For pharma partnerships, model the prevention layer as upstream of prescription, not adjacent to it. A user whose HbA1c is held below the clinical threshold for ten years is not a lost customer; the customer redistributed to a different revenue line.
  • Audit the labs in your network for pre-analytic discipline, not just NABL certification. The certification is the floor; the pre-analytic chain is where the false positives that drive avoidable claims are born.
  • Move the underwriting metric from one-year incidence to ten-year trajectory. The customer is more profitable to underwrite against the trend than against the snapshot, and the data infrastructure to do that exists today.

The whole conversation, searchable and timestamped.

Search the text, highlight what matters, and export your highlights and notes together. Click any timestamp to jump in the video.

Tip: press / to focus search, N / Shift+N to step matches, H to toggle highlight mode, then click any segment to mark it. Notes and highlights save automatically to this browser only.
Transcript is generated from YouTube’s auto-captions. Names, regional terms and proper nouns may be lightly imperfect — HbA1c appears as “hba1c”, “hpa1c” and “sba 1” in places; anti-CCP as “anti-sccp”; NABL as “nabl”. The audio is the canonical source.