The Reference Protocol

The measurement discipline is the part worth stealing: blood panels every 3–6 months, annual whole-body MRI, annual skin and eye checks, dental every six months, continuous glucose monitoring, daily weight, body fat, muscle mass, hydration and arterial stiffness, multispectral skin imaging, and a protocol that changes when the numbers say it should rather than when a podcast does.

Roughly 90% of the plausible effect comes from the boring 10%: sleep regularity, daily training, whole-food plant-forward eating at a modest calorie level, zero alcohol, and relentless measurement. The exotic remainder absorbs nearly all the cost, all the risk, and all the attention.

NAD+ precursors remain the most oversold category in the entire longevity market relative to outcome data. Bioavailability studies get cited as though they were efficacy studies. They are not.

A mechanism you can draw on a whiteboard is not evidence. Both of these had gorgeous molecular stories and failed at the endpoint that mattered. This is the single most useful lesson on this page.

He stopped around late September 2025 after roughly five years. His stated reasons: lipid abnormalities, worsening glucose and insulin resistance, elevated resting heart rate, and recurrent soft-tissue infections that dose adjustment did not resolve — plus literature concern about beta-cell toxicity and NK-cell inhibition (a cancer-surveillance risk), reinforced by a preprint reporting accelerated aging across 16 epigenetic clocks. His framing: the maths changed for a healthy 46-year-old versus an aging mouse.

The interesting part is that this inverts the usual story. Matt Kaeberlein — among the most credible primary rapamycin researchers alive — is if anything *more* bullish than the self-experimenter now is, maintaining it remains the most robust and translationally viable geroprotective molecule of the last fifteen years, citing ITP data — the 2009 Harrison arm starting at 600 days produced roughly 9% (males) and 14% (females) median extension, with the larger ~30% figures coming from later, earlier-started, higher-dose arms. The best human RCT (PEARL, N=114, 48 weeks, 5–10 mg weekly) was null on its primary outcome of visceral fat, with real but modest secondary signal on lean mass in women and self-reported wellbeing.

A famous self-experimenter quitting a drug is an n=1 tolerance report, not a verdict on the molecule — and the leading domain expert reading the same literature reached the opposite conclusion. Hold both facts at once. This is the cleanest available lesson in why protocol changes by public figures should never be mistaken for scientific findings.

The two items with the best evidence-to-cost ratio are the two nobody posts about: sauna and dental hygiene.

Follistatin gene therapy (2023)

An AAV-delivered construct expressing follistatin, which binds and neutralises myostatin. Myostatin normally signals through activin type II receptors to activate Smad2/3, repressing muscle protein synthesis; removing that brake increases mass, as seen in myostatin-null cattle and rare human loss-of-function cases. Performed outside the US regulatory system. Gene therapy is not reversible, and durability, off-target expression and long-term immune consequences of AAV in a healthy adult are unknown.

Mesenchymal stem cells (~300 million, 2024)

Injected into knees, hips and shoulders. Current understanding is that MSCs do not meaningfully engraft or differentiate — they act paracrine, secreting extracellular vesicles and immunomodulatory factors that shift local macrophages toward an anti-inflammatory phenotype. Effects are therefore transient. Largely unregulated as an industry, with real variance in what is actually in the vial.

Plasma exchange

The multi-generational young-plasma transfer was publicly run and then abandoned. What has since accumulated real evidence is *dilution* rather than young-factor addition: a Buck Institute trial (Aging Cell, May 2025) found biweekly TPE + IVIG produced an average 2.61-year reduction in epigenetic age, TPE alone 1.32 years, front-loaded with diminishing returns. Diluting accumulated pro-aging plasma factors appears sufficient — a far more defensible claim than the parabiosis hype cycle, because it requires no magic molecule, only removal.

Cerebrolysin — a publicly reported null

A porcine-brain-derived peptide preparation given intramuscularly for three months. No measurable effect; discontinued. Worth flagging precisely because it is a clean negative result publicly reported — the scarcest artifact in this entire space, and the single strongest evidence that the protocol is run honestly.

Aesthetic device stack

1927 nm and 1550 nm fractional lasers, microfocused ultrasound and RF microneedling, each roughly every six months. All share one mechanism: controlled thermal injury at a dermal depth that spares the epidermis, triggering fibroblast activation, neocollagenesis and matrix remodelling. Real, well-characterised, genuinely effective — and the most expensive line item.

Auditing a protocol is mostly a sourcing problem. Almost nothing in this space is fabricated; it is real research stripped of its effect size, its confidence interval and its limitations by someone whose income depends on it sounding certain. Two questions have to be asked separately, and merging them is the single most common failure: how good is the evidence, and who is telling me and what do they want. A company-sourced claim about a well-designed RCT and a researcher-sourced claim about mouse data are different problems.

• Cites without engaging — a study named, its effect size never given. The most common failure, and the hardest to spot, because it looks exactly like rigour.
• Sells the certainty — confidence outruns the literature, and something is for sale at the end that the confidence was manufacturing demand for.
• Drifts outside their field — expertise does not transfer.
• Never publicly changes their mind — a spotless record in a live field means not paying attention.
• Claims operator status without disclosed, dated numbers.
• Only ever agrees with their own audience — audience capture is indistinguishable from expertise from the inside.

Applied to the protocol above, the filter is what separates the four researcher-tier critiques that land from the far larger volume of commentary that does not — and what puts the branded supplement blends at company tier however good the underlying biochemistry looks.

Source tier — Written up in full at /articles/finding-the-real-experts, and installed locally as a Claude skill so it runs automatically on any source-credibility question.

Morgan Levine — the headline number is a unit error

Levine, a leading epigenetic-clock researcher, has explicitly criticised the '31 years of aging reversed' framing. It derives from a DunedinPACE score of 0.66 multiplied against chronological age — but DunedinPACE measures a *rate* of aging at a point in time, not a quantity of years you can integrate backwards across a lifespan. The arithmetic simply does not work. She grants the lower epigenetic age probably does reflect genuinely lower disease and mortality risk; the specific number is unsupported.

Nir Barzilai — biological plausibility is not clinical evidence

The TAME trial PI's position: science is not done on n=1. A protocol with a hundred simultaneous variables and one subject cannot generate attributable knowledge, however well instrumented it is.

Andrew Steele — nothing has been proven yet

No intervention has been clinically proven to extend human life by targeting aging itself. Every geroprotective claim rests on animal lifespan data, human biomarker surrogates, or epidemiology — never a human longevity endpoint, because no such trial has finished.

Matt Kaeberlein — signal versus noise

Individual biomarker fluctuations in one person, measured frequently, generate apparent trends indistinguishable from noise without a control. Dense measurement makes this worse, not better, unless the analysis explicitly accounts for it.

Set against that: the criticisms are almost entirely about *inference*, not harm. Nobody credible argues that sleeping eight hours on a fixed schedule, training six hours a week, eating whole food, drinking no alcohol and measuring everything is bad for you. The argument is about what can be concluded — and whether the exotic 10% justifies its cost and risk.

Read it as a very expensive, very well-instrumented demonstration that the basics work, with an unattributable experiment stapled on top. Take the measurement discipline, the sleep regularity, the training floor, the food quality and the flossing. Leave the gene therapy.