Biomarker-driven longevity protocols

A biomarker-driven longevity protocol is a health system that treats the body as a measurable control system: collect biomarkers, choose interventions, re-measure, and iterate. project blueprint is the most visible current example because bryan johnson publishes personal metrics, claims, routines, and productized versions of the approach.

What gets measured

In the Blueprint corpus, measurements include blood and urine biomarkers, biological age / speed-of-aging tests, imaging, fitness tests, sleep data, resting heart rate, glucose control, sexual/fertility metrics, and organ-specific results. Blueprint’s biomarker product advertises 100+ biomarkers, 160+ measurements per year, baseline testing, six-month retesting, lab-import, and AI health guidance.

The May 2026 Kate Tolo and Enhanced Games posts extend the measurement frame beyond Johnson’s own protocol: Johnson describes a cycle-aware female baseline protocol for Tolo, suspected endometriosis evaluation using ultrasound/MRI/labs, and Enhanced Games athlete/protocol commentary built around measurements and medical supervision. These are primary-source claims and should be treated as hypothesis-generating/public-positioning material, not clinical proof.

YEARS groups Johnson’s measurements into blood biomarkers (lipids, inflammation, ApoB, HOMA-IR, hs-CRP), imaging (MRI/ultrasound/EKG), functional tests (VO2 max, grip strength, balance, cognition), and epigenetic clocks. The article’s central caution is that “data volume and clinical relevance are two different things.”

Why it is attractive

The strongest version of the idea is not “take Bryan Johnson’s supplements.” It is: stop guessing, measure meaningful health variables, implement evidence-backed basics, and use follow-up data to refine behavior. This can make health behavior legible, catch disease risk earlier, and prevent blind supplementation. Johnson’s own rhetoric repeatedly emphasizes “trust data, not opinions,” “biomarkers in context,” and lowering RHR before bed as a high-leverage leading indicator for sleep quality and downstream behavior.

Failure modes

1. N=1 generalization. A protocol optimized for one wealthy, motivated, medically supervised individual does not prove population-level efficacy.
2. Confounding. Diet, sleep, exercise, supplements, drugs, devices, and experimental therapies change together, making attribution nearly impossible.
3. Surrogate endpoints. Better biomarkers or epigenetic clock scores may not necessarily translate into longer healthspan or lower disease/mortality risk.
4. Over-testing. More imaging and lab work can create false positives, anxiety, unnecessary procedures, or unclear actionability.
5. Under-investigated signals. A dense dashboard can still miss disease if a repeatedly abnormal but “non-urgent” marker is normalized, explained away, or not connected to the right diagnostic pathway.
6. Commercial incentives. When the same actor sells supplements, testing, app access, and protocol products, claims need conflict-of-interest scrutiny.
7. Medical risk. Prescription drugs, hormone manipulation, gene therapy, plasma exchange, and aggressive restriction require clinical supervision.

MDLinx takes a clinician-facing view: some pieces (sleep, exercise, selected monitoring) are plausible or supported, but the extreme regimen — 100+ pills, extensive testing, transfusions/experimental interventions, and intensive routine control — is not generally feasible and may not be necessary. YEARS emphasizes N=1, confounding, surrogate endpoints, and the Hawthorne effect from being intensely monitored.

The June 2026 Tadalafil/Cialis posts are a useful example of how Johnson’s protocol mixes biomarkers, prescriptions, epidemiological associations, and public caveats. He frames daily 5 mg Tadalafil as a blood-flow/longevity intervention and names mortality/cardiovascular/dementia associations, while acknowledging observational data cannot prove causation and that evidence in women is thinner. For evaluation, this belongs in the “prescription/intervention claim” bucket: potentially measurable and potentially medically supervised, but hypothesis-generating rather than independent proof of longevity benefit.

The June 16, 2026 sauna/HSP27 thread is a clearer example of biomarker-driven protocol design. Johnson used continuous ingestible core-temperature tracking and repeated blood draws to ask whether dry-sauna dose is better represented by minutes in the sauna or by time spent above a core-temperature threshold. The dashboard takeaway is the experimental structure, not the exact heat prescription: HSP27 movement is a mechanistic biomarker claim and does not establish clinical outcome benefit for readers.

A June 18, 2026 post shows the same readout-first instinct applied to circadian recovery: Johnson reported that after his return flight from Australia, he watched his body clock “come back online, live, via blood glucose” while using his caffeine + melatonin jet lag protocol. The notable evaluation point is the choice of readout — blood glucose used as a real-time proxy for circadian resynchronization — while the intervention remains an N=1 self-report rather than proof that the dosing caused the recovery.

A June 19, 2026 skin-aging post extends the readout-first pattern to a cosmetic endpoint: Johnson claimed one week of Australian sun increased his skin-aging readout by ~5% despite umbrella use and peak-UV protection, and separately repeated a claim that he has reversed his skin age by ~9 years since starting the project. For evaluation, the notable point is the quantification of a skin/UV endpoint as a protocol metric; the exact percentages should be treated as personal self-measurements, not independent validation.

On June 17, 2026, Johnson made the methodology argument more explicit while discussing a Nature feature: with the Blueprint & Immortals Medical and Science Team, he argued that randomized controlled trials remain necessary for average-effect and safety questions but are not sufficient for optimizing individualized stacks of interventions. The essay frames N-of-1 measurement as a complement to RCTs, enabled by AI, wearables, multi-omics, real-time tracking, and nearest-neighbor mapping against deeply instrumented people. The caution is equally important: its examples — heat-shock-protein response, sauna-related plasticizer clearance, fertility markers, microplastics reduction, and a psilocybin metabolic signal — are presented as first-in-human observations needing validation, not as proven therapies.

On June 23, 2026, Johnson extended the same methodology argument to structural imaging while endorsing the third-party “Midjourney scanner.” He framed blood draws as chemical data, wearables as functional data, and imaging as structural data, then argued that baseline + repeated scans can shift a finding from “what is this?” to “is this changing?” That is a useful answer to the over-testing failure mode above, but only at the level of a hypothesis: the scanner’s clinical value, regulatory status, and false-positive handling remain unvalidated, and Johnson’s screening statistics were lay citations without source links in the tweet.

The same day also showed two protocol outputs rather than new measurements. First, Johnson replied to Nassim Nicholas Taleb’s attribution critique — “we’ll never know which drug, or combination of drugs, did him in” — which restates the confounding problem any dense intervention stack must face. Second, he suggested “travel internationally one time per quarter, max” after measuring China, India, and Australia as weeks-long biological insults. Treat the travel cap as a personal N=1 rule produced by Johnson’s measurement loop, not as general travel medicine guidance.

On June 24, 2026, Johnson extended the same longitudinal logic to consumer wearables. Replying to a post comparing Apple Watch, Whoop, Oura, and Fitbit Air and finding divergent readings, he argued that wearables remain useful for relative tracking — yesterday versus today, or how sleep changes after alcohol — if the device is consistent with itself. This is a reliability-versus-validity stance: within-device trends may support behavioral feedback, but divergent absolute numbers still leave accuracy and clinical-decision questions unresolved.

On June 25, 2026, Johnson added germline inherited-cancer screening to the measurement stack. He said he ran a combined DNA + RNA panel covering 71 inherited-cancer-risk genes, including BRCA1/2, ATM, MLH1/MSH2, TP53, APC, PTEN, RET, VHL, and MEN1, and reported a negative result. The useful dashboard signal is a new measurement category — inherited-risk stratification — rather than another blood/imaging/fitness readout. The caveat is equally important: the cancer-survival statistics and RNA-splicing claims are Johnson’s lay citations from the tweet, the test does not address the large majority of non-inherited cancers, and his negative result is an N=1 self-report, not independent validation or reader screening advice.

On June 26, 2026, Johnson connected biological-age clocks to cancer-surveillance rhetoric by citing an unnamed study in which people whose biological age exceeded chronological age had higher early-cancer risk, especially under age 55. He quoted larger-gap associations of up to 57% higher lung-cancer risk, 31% uterine, and 17% gastrointestinal, and noted that two of three age-estimation tests used basic blood markers. This is best read as an argument extending the existing biological-age measurement theme, not a new validated screening modality: the study was not named in the tweet, the statistics are Johnson’s lay citation, and the post does not establish what an individual reader should measure or do.

On June 30, 2026, Johnson’s autoimmune-gastritis disclosure added the inverse failure mode: not over-testing, but under-investigating a persistent signal. He said an 11-year pattern of low ferritin, with normal hemoglobin and hematocrit, was repeatedly explained away until a rebuilt care team connected it to autoimmune thyroid history, APCA bloodwork around five times the upper limit of normal, a bi-directional endoscopy, and five stomach biopsies. For this dashboard, the useful signal is methodological: a biomarker-driven protocol is only as good as the clinical reasoning that follows abnormal-but-subtle markers. Johnson’s proposed “Immortals Care” roadmap for AIG moves from current monitoring/support to explicitly investigational immune-pathway, regulatory-T-cell, CAAR-T, and AI-designed-antibody concepts; it should be read as his research agenda, not as an approved cure or reader treatment pathway.

On July 3, 2026, Johnson added a more granular AIG measurement layer: single-cell immune-receptor sequencing of 1,000,000 immune cells. His stated goal is to move beyond ordinary immune-cell counts and read each cell’s antigen-receptor “key” to identify the autoreactive clones attacking his stomach lining. The same thread tied that sequencing to a large blood draw (198.5 mL, 33 tubes, 50 tests, roughly 100 biomarkers) and to follow-up markers such as ferritin, anti-parietal-cell antibody, intrinsic-factor antibodies, gastrin, chromogranin A, HLA typing, cytokines, and neurodegeneration markers. The dashboard signal is a new cellular/receptor-level readout in his measurement stack; the caveat is that it remains diagnostic follow-through and hypothesis generation, not validation that a targeted AIG therapy exists or that readers should pursue similar testing.

On July 7, 2026, Johnson turned the same AIG case into a sharper lesson about under-investigated signals and false negatives. He said his ferritin averaged 38 ng/mL for 11 years even when he ate meat, but normal hemoglobin and hematocrit let the problem be dismissed until AIG explained why oral iron would not “stick.” He reports choosing a 1,000 mg Monoferric IV infusion and seeing ferritin of 205 ng/mL at two weeks and 195 ng/mL at four weeks, with an 80 ng/mL target. A companion post states that “absence of diagnosis” is not the same as health and that he fears false negatives more than false positives. For this dashboard, the important update is his explicit stance on the over-testing tradeoff; the iron treatment details remain Johnson’s self-report and not reader medical advice.

On July 8, 2026, Johnson moved the AIG thread from diagnosis and measurement into a staged curative engineering roadmap. His plan is to pair the 1,000,000-cell immune map with a second stomach biopsy for live T-cells, then build an early-warning loop from biweekly blood draws plus wearable data, cryopreserve immune cells for an ex-vivo “Bryan in a dish” model, and test precision therapies first in a computer model and then on his own frozen cells. The operational signal is new: measurement becomes a therapy-selection and safety-screening pipeline. The caution is equally important: this is Johnson’s proposed personal experiment, not an approved autoimmune-gastritis cure, validated ex-vivo screening method, or reader treatment pathway.

On July 9, 2026, Johnson used Kate Tolo’s endometriosis workup as a high-signal example of multi-modality diagnostic triangulation. He said standard pelvic MRI and transvaginal ultrasound had returned negative, then a second pass confirmed endometriosis without surgery through three complementary modalities: an endo-specific ultrasound performed by trained operators, AI-assisted pelvic MRI through MatricesAI / Geneviève Institute, and a Heranova microRNA/protein blood test. The dashboard-relevant pattern is not that readers should copy the exact workup; it is the protocol logic that hard-to-diagnose disease may require complementary readouts because any single test can miss signal. Johnson’s accuracy statistics and “world-first” framing are his public claims around one person’s case, not independent validation or medical advice.

On July 14, 2026, Johnson announced a major scale escalation of Kate Tolo’s female-health protocol: a projected 90-day program spanning three menstrual cycles, 100 tasks a day, 6–10 hours a day, 50+ devices, 1,900 biomarkers, and 14.8 million data points. He framed cyclical female biology as historically under-measured and said the goal is to make Tolo “legible” first to herself and then to others. For evaluation, the distinction between protocol specification and evidence is crucial: the figures are Johnson’s plan for a future collection window, not a completed dataset, study result, or independently verified measurement. Dense data can improve longitudinal visibility, but volume alone does not establish clinical meaning, actionability, representativeness, or privacy safeguards, and one intensively measured participant does not define a general female-health protocol.

On July 20, 2026, Johnson extended the measurement-first frame to eye health, asking whether it has been absent from longevity discussions and clarifying that he meant robust measurement, diagnosis of dysfunction, and corrective protocols—not only diet or supplements. The durable signal is the proposal to treat another organ system as a measurable feedback loop. The evidence boundary is equally important: the thread names no eye condition, test, target, intervention, or outcome, so it is an agenda-setting question rather than a validated protocol or medical recommendation.

On July 21, Johnson added the first concrete ocular-fluid readout to that agenda: he reported collecting tears with paper strips for a $1,800 specialty-lab panel spanning 15 inflammatory, tissue-degradation, and regenerative-signaling biomarkers. He said the panel would quantify the ocular-surface molecular environment and support progress tracking, with an eye protocol forthcoming. The useful methodology signal is the move from a broad organ-system question to a repeatable molecular sample. The evidence boundary remains substantial: no result, diagnosis, intervention, reference range, clinical utility, or outcome was published, so this is an N=1 measurement plan rather than a validated screening panel or medical recommendation.

On July 23, Johnson expanded the June sauna/HSP27 experiment into a public checklist. The list mixes measurable dose ideas—frequency, cabin temperature, session time, and a personal 102.2°F / 39°C core-temperature target—with broader recommendations or claims about hydration, fertility, post-session cold exposure, recovery, microplastics, air quality, materials, and dry versus wet/infrared evidence. This is a useful worked example of a biomarker finding becoming protocol policy, but also of why the transition needs scrutiny: an N=1 HSP27 response does not independently validate the full checklist, a clinical longevity benefit, or a safe universal heat dose. His separate 42-to-44 bpm meal-timing comparison is likewise a personal observation rather than a causal estimate.

On July 24, Johnson said Immortals was adding iPSCs, organoids, deep cellular characterization, and personalized therapy development to its existing health-product and concierge stack. Methodologically, this proposes a wider loop in which dense measurement feeds individualized disease models and candidate therapies. The expansion is still a company strategy claim: no operational pathway, diagnostic performance, safety process, treatment, cure, or outcome was published.

On July 25, Johnson operationalized Kate Tolo’s female-health baseline as a 100-day, pre-intervention collection program with 14 million claimed menstrual-cycle data points, 100+ daily tasks, 50+ devices, and a 12-person medical team. The schedule spans repeated samples, hormonal and metabolic measures, wearables, functional and sensory testing, imaging, microbiome work, exercise, sleep, mood, and routine care. This makes measurement burden, modality coverage, and sequencing visible, but volume does not establish validity or actionability. One participant also cannot define a representative female-health protocol, and publication will need clear metric definitions, cycle-phase context, missing-data methods, privacy governance, and interpretation limits.

On July 31, Johnson and Tolo added menstrual-blood sampling to that measurement program. They said roughly 10 mL was frozen at -80 °C and proposed menstrual blood as a repeatable, non-invasive view of the uterine environment, naming possible analyses for diseased tissue, microplastics, endocrine disruptors, PFAS, and endometrial, immune, and stem cells. The returned X reply is truncated after “We collected to…,” and no collection protocol, laboratory method, result, reference range, diagnostic performance, or clinical outcome was published. The dashboard therefore records a novel N=1 biosampling proposal—not a validated diagnostic test, established alternative to surgical biopsy, general female-health protocol, or medical advice.

On August 3, Johnson added a pharmacology claim rather than a new measurement modality. He quote-posted a summary saying Viagra plus statins may blunt cancer spread, cited 18% lower colorectal-cancer mortality and 15% lower metastasis, and proposed that a sildenafil mechanism could apply to tadalafil. His own caveat is decisive for classification: the underlying study was preclinical and the human evidence observational. The dashboard therefore treats this as a low-confidence intervention hypothesis linked to Johnson’s prescription-product narrative—not evidence of causality, proof that tadalafil prevents metastasis, a reason to combine drugs, or medical advice.

On August 4, Johnson made a new operational claim about transferring the measurement system: his own longevity infrastructure took five years to build, while a team assembled Kate Tolo’s protocol in 90 days and, in his words, made it “better.” This suggests a shift from one founder’s bespoke N=1 loop toward repeatable, team-mediated protocol infrastructure. It does not yet show that transferability works: the post supplies no definition of quality, comparative measurements, completed outcomes, external validation, or evidence that the protocol generalizes beyond Tolo.

On August 12, Johnson reported another large collection event: 47 tubes and 250 mL of blood for single-cell sequencing of circulating immune cells alongside inflammation, oxidative-stress, vascular, metabolic/lipid, glucose-regulation, and brain-related measurements. Compared with his July AIG draw, this is an expansion or repetition of the cellular and multi-domain measurement layer rather than a new clinical result. The source publishes no assay list, findings, diagnosis, intervention, or interpretation, and collection volume is not evidence of validity or actionability.

On August 13, he paired that laboratory-scale posture with a public Baseten event listing biological-age evaluations for brain, skin, strength, balance, reaction speed, and mobility. This broadens the visible functional-testing surface but does not establish that the tests measure one coherent “biological age,” predict outcomes, or support treatment decisions. A separate post that day summarized a mouse brain-clearance gene-therapy result without naming the paper; it belongs in preclinical research watch, not in the validated measurement or intervention tier.

On August 16, Johnson described food discipline as another biomarker-governed feedback loop: he said he is accountable to biomarkers, changes the protocol according to data, and views years of dietary control as mastery while acknowledging that the pursuit can become unhealthy. The useful methodology signal is the stated decision rule—measure, adjust, repeat—not a clinical conclusion about his eating. The post is a personal account, does not establish nutritional or psychological safety, and does not make the regimen appropriate for readers.

On August 18, Johnson supplied the clearest example yet of the measure–intervene–retest loop reaching an organ-specific diagnosis: after a stye visit, meibography and Schirmer testing showed advanced meibomian-gland dysfunction, and he published a four-therapy eye stack with a plan to track repeated imaging until it plateaus. The loop is also turned on itself—he named likely drivers inside his own protocol (a topical anti-androgen, low insulin/IGF-1 signaling, thyroid history) as candidate contributors to the gland loss. The methodology signal is the closed feedback design; the evidence boundary is that all therapies were simultaneous, the 30% improvement figure is self-reported imaging, probing is explicitly unsettled science in his own words, and no independent assessment was published. This is an N=1 account, not a validated eye-care protocol or medical advice.

On August 19, Johnson quote-commented third-party oncology news: an mRNA therapy that instructs the body to produce a tumor’s genetic fingerprint so the immune system trains T-cells against it, paired with Keytruda releasing the brakes cancer places on T-cells, which he says has reached a Phase 3 trial for the first time. He named no trial, paper, sponsor, or data. Like the mouse brain-clearance post, this belongs in research watch—third-party commentary inside his longevity narrative, not a Blueprint intervention, an available treatment, or medical advice.

On August 24, the female-protocol narrative moved from collection scale to lab interpretation. Johnson amplified Kate Tolo’s thread arguing that women in their 30s labeled with high cholesterol may be misdiagnosed because cholesterol is not one number: it reads higher in the follicular half of the menstrual cycle, falls after ovulation, and hits its lowest just before the period, with a claimed ~19% cycle-dependent swing. Tolo cited the FDA’s 1977 exclusion of women of childbearing age from early drug trials, called a 2010 study the “first rigorous, large-scale proof,” and announced her own $2.6M experiment collecting 14 million cycle data points. The claims track real literature: the NIH BioCycle study (Mumford et al., 2010, J Clin Endocrinol Metab; 259 women, up to 16 fertility-monitor-timed draws over two cycles) measured ~19% mean within-woman total-cholesterol variation, and more women crossed the ≥200 mg/dL boundary when tested in the follicular phase (14.3%) than the late-luteal phase (7.9%); a 2011 clinical-lipidology review by the same group recommends cycle-aware test timing with repeat draws near boundaries. The dashboard keeps the boundary explicit: the thread names no study, its “~6% mislabeled” figure does not precisely match published results (5% of BioCycle women were above 200 mg/dL at all visits; 19.7% at least once), and phase-specific reference ranges remain a proposal rather than an adopted standard. This is attributed claim-plus-context, not a directive to reinterpret lab results without a clinician or medical advice.

On August 25, Johnson moved the exercise claim lane into a concrete protocol with a quantified mortality figure. He posted 3 workouts a week for 8 weeks, each built from four 4-minute intervals at 90–95% effort separated by 3 easy minutes, and claimed a “lower risk of dying from any cause by 11% in 8 weeks,” closing that “your improved cardiorespiratory fitness is the risk reduction.” The protocol is the Norwegian 4×4 from Helgerud et al., 2007 (Med Sci Sports Exerc; 40 moderately trained men, workload-matched, 3 sessions/week for 8 weeks: +7.2% VO2max for the 4×4 group versus no change for moderate continuous training), and the mortality arithmetic tracks the Kodama et al., 2009 JAMA meta-analysis of 33 studies and 102,980 participants (13% lower all-cause mortality per additional MET of fitness, RR 0.87, 95% CI 0.84–0.90). The dashboard keeps the boundary explicit: no trial has measured mortality outcomes from eight weeks of interval training—the 11% compresses an RCT fitness outcome and a long-run observational association into an unattributed figure, and the post names no study. This is attributed claim-plus-context, not a demonstrated 8-week mortality outcome, an exercise prescription, or medical advice.

On August 26, the biomarker lane turned to LDL cholesterol as “one of the most actionable things from a blood draw.” Johnson cited “a meta analysis of 49 randomized trials and 312,175 people” finding about 23% fewer major vascular events per 39 mg/dL (1 mmol/L) LDL reduction, and said nutrition can do some of the heavy lifting—“dropping LDL levels by 13-14% over six months by combining viscous fiber, nuts, plant protein, and plant sterols”—with “medications” as a possible role. The meta-analysis matches Sabatine et al., 2016 (JAMA; 49 trials, 312,175 participants, 39,645 major vascular events: RR 0.77 per 1 mmol/L reduction, 95% CI 0.75–0.79, pooled across statin and LDL-receptor-upregulating nonstatin therapies), and the diet figure matches the Jenkins et al., 2011 portfolio-diet RCT (JAMA; 345 hyperlipidemic participants over 6 months: −13.1% routine to −13.8% intensive LDL reduction versus −3.0% for the low-saturated-fat control). The dashboard keeps the boundary explicit: the post names neither study, and the 23% is a trial-population relative effect accumulated over years of treatment, not an individual guarantee—LDL interpretation belongs with a clinician. This is attributed claim-plus-context, not a lab-interpretation directive, a diet prescription, or medical advice.

On August 27, the measurement loop added a sexual-health biomarker claim. Johnson published a self-reported NTE “personal best”—4 hr 2 min of nighttime erections at 93% strength, ranked against a personal database of 55,000 measurements—and called NTE “a tier 1 longevity biomarker as critical to systemic health as VO2 max, resting heart rate, HRV, and blood pressure.” He claimed low nighttime-erection scores mean “2x risk of heart attack and stroke in the next 4 years,” and listed a protocol of high sleep quality, cardiovascular fitness, a calm nervous system, and daily tadalafil 5 mg (“I take this longevity medication daily”). The direction is documented—erectile dysfunction precedes and predicts cardiovascular events—but the post’s specifics outrun the literature: the Vlachopoulos et al., 2011 JACC meta-analysis (12 prospective cohorts, 36,744 men) pooled RR 1.48 for CVD and 1.35 for stroke, and the Krimpen population cohort found HR 2.6 for MI, stroke, or sudden death only in severely reduced rigidity over roughly six years of follow-up. The post names no study; its flat “2x in 4 years” is stronger than the meta-analytic averages; consumer NTE scores are not a validated clinical risk tool; and daily tadalafil is a prescription medication requiring clinician oversight. This is an attributed N=1 biomarker claim with literature context—not a validated risk prediction, a monitoring recommendation, or medical advice.

On August 29, Johnson extended the NTE claim from personal data to population reference tables: age-binned nightly-duration figures of 190 minutes at age 20, 103 at 50, 81 at 60, and 50 at 75+, again repeating that “ED predicts cardiovascular events yrs before symptoms” and the “2x risk of heart attack and stroke over 4 years” figure. The decline itself is among the better-documented observations in sleep and sexual medicine—Karacan et al., 1975 (Am J Psychiatry; 125 healthy males aged 3–79, EEG-verified) established age-normative NPT data; Schiavi et al., 1988 (J Gerontol; 40 healthy men aged 23–73) showed NPT frequency and duration fall progressively with age independent of sleep changes; and Horita & Kumamoto, 1989 (189 healthy males aged 3–84) measured 189.6 minutes of total nightly tumescence at age 20 declining to roughly 62 minutes at age 70—putting Johnson’s age-20 anchor almost exactly on published values. But the post names no study; its intermediate bins match no single published cohort (older-age values differ between series—Horita found ~62 minutes at age 70 where Karacan’s series reports ~102); and the repeated flat “2x in 4 years” remains stronger than meta-analytic averages. Consumer NTE scores are not a validated clinical risk tool. The dashboard records an attributed normative-data claim with literature context—not a validated risk prediction, a monitoring recommendation, or medical advice.

On August 30, the measurement loop moved to the brain. Johnson posted regional percentiles from his own Kernel measurements—amygdala 95th, putamen 99th, caudate 97th, globus pallidus 93rd, and frontal gray matter at the 78th percentile, up 1.35% over two years “when it should be going down with age”—and mapped each region onto his personality, from processing hate as information rather than threat to near-effortless habit formation. He cited twin-study heritability of roughly 0.75–0.89 for the subcortical regions and gestured at plasticity literature without naming it: adults who learned to juggle grew measurable gray matter (Draganski et al., 2004, Nature; a transient, selective expansion in visual-motion cortex hMT/V5 over roughly three months that receded once practice stopped), and six months of aerobic exercise grew frontal gray and white matter in sedentary adults aged 60–79 (Colcombe et al., 2006, J Gerontol A; 6-month randomized trial, 59 participants). The post names no study, the normative cohort behind the percentiles is undisclosed, and Johnson himself closes with “speculative for my n of 1 context.” The dashboard records an attributed self-report with literature context—not a validated brain-health metric, evidence that his protocol grew his frontal cortex, or medical advice.

On September 2, the measurement story became a headline outcome claim: “in many ways, I am 18.” Johnson listed fourteen systems he says are indistinguishable from an 18-year-old—sleep quality, erection function, fertility, resting heart rate, vascular health, cardiovascular health, blood pressure, metabolic health, blood glucose control, bone mineral density, muscle, fat, “undetectable” inflammation, and shitposting—conceding only hearing and somatic mutations as age-typical, framing the result as “the first rep” achievable by others in five years. This is the biomarker-driven protocol’s thesis compressed into one claim: measure enough systems, act on the readings, and the body stays young. The evaluation boundary is the same one this page applies throughout: no dataset, assay list, reference cohort, or per-system comparison standard accompanies the post, so each “indistinguishable” is Johnson’s self-characterization of undisclosed-norm test results, not a validated equivalence. Notably, the two conceded systems—hearing and somatic mutations—are ones his measurement stack cannot yet correct, which is itself informative about where dense measurement stops translating into outcomes. The dashboard records an attributed N=1 outcome claim—not evidence of youthful equivalence, a replication promise, or medical advice.

On September 4, the biomarker lane produced its first product-level measurement claim. Johnson announced an AI model built on sleep data that “accurately predicts your age,” developed with Eight Sleep and available on their platforms as a membership “Biological Age” feature co-developed with his Immortals. The claim is unusually checkable because a preprint exists, and the numbers match the post: BCG-FM (Kjaer et al., arXiv:2606.07692) pretrained with participant-level contrastive learning on 2.04 million hours of nightly bed-sensor ballistocardiography from 136,575 participants, reporting 3.26-year MAE biological-age estimation (the lowest reported for any ambient, contactless modality), 92.5% Rank-1 identity retrieval from one night’s signal, and disease AUROCs of 0.852 for diabetes, 0.822 for heart failure on three independent external clinical cohorts, 0.810 for hypertension, 0.792 for sleep apnea, 0.751 for snoring, and 0.734 for general heart conditions, with representation quality scaling log-linearly in contrastive batch size (R²=0.982). This is the biological-age measurement theme of this page reaching a consumer product: a passive nightly signal converted into a weekly-updated age estimate and disease-risk readout. The evaluation boundaries are the page’s own: the model is vendor-trained and vendor-evaluated, the internal disease labels are self-reported, the external cohorts are small, the preprint is not peer-reviewed, and the post’s “diabetes better than Apple’s model” comparison has no corresponding table in the preprint (its Apple-Watch PpgAge benchmark is age accuracy, not disease detection). Johnson’s own post concedes “internal labels are self-reported and external cohorts are small” and calls it “a research milestone, not a diagnostic device.” The dashboard records an attributed product-launch and research claim with verified preprint context—not a validated diagnostic, clinical risk tool, purchasing guidance, or medical advice.

Practical evaluation checklist

When evaluating a biomarker-driven longevity protocol, ask:

• Is the metric clinically meaningful, standardized, and actionable?
• Is the intervention backed by human outcome data or only mechanistic/animal/early-stage evidence?
• Are variables isolated enough to attribute effect?
• Is there medical supervision for drugs, hormones, imaging, and invasive testing?
• Does the protocol prioritize basics before expensive/experimental layers?
• Are conflicts of interest disclosed?
• Are claims framed as personal data, hypothesis, clinical recommendation, or product marketing?

Related pages

• project blueprint — concrete case study.
• bryan johnson — primary actor and source of the dataset.
• dont die — community/philosophical wrapper around the protocol.