689 companies. How exactly are they trying to beat aging?
The technical classification (19 fields plus modality) stays in the database. Here every company is filed under one of ten ways humanity is attacking the problem.
Senolytics, mitochondria, extracellular matrix, clearance of waste.
5. Reprogram cells
Gene therapy, epigenetic reprogramming.
6. Restore tissues
Cell therapy, regenerative medicine.
7. Replace what can no longer be repaired
Organs, bioprinting, xenotransplantation.
8. Preserve the body until it can be repaired
Cryonics, biostasis.
9. Build tools for everyone else
Research tools, contract research organizations (CROs), data, infrastructure.
10. Bring it to people
Clinics, preventive medicine, consumer products.
Source: a live cut of the Eternal Search database, section “The numbers” on eternalsearch.net. The same breakdown, with the company count per row, opens when you click the “companies” tile on the first slide. 258 companies are not yet classified and appear there as a separate row.
The numbers · sources
988K sources. Where do we know this from?
Every stored document is filed under one of eight kinds by the nature of the evidence: who is making the claim and what backs it up.
Source: a live cut of the Eternal Search database, section “The numbers” on eternalsearch.net. The same breakdown, with the document count per row, opens when you click the “sources” tile on the first slide. Not counted: search-engine result lists, summaries extracted from documents already stored, our analysts’ notes, translations and chunks.
THE SPECTRUM
What do we do, depending on the difficulty level we assign to the problem of aging?
Are the remedies for old age already at the pharmacy?
Consider the scenario: the drugs are known, the targets are known — all that remains is to carry a combination therapy for aging through clinical trials.
Pharmacy
No field has higher hopes for drug repurposing than aging
Pharmacy
Drug repurposing means taking an already-approved drug and aiming it at a new target — here, aging. Such a drug has years of safety data behind it, costs pennies, and is sitting on pharmacy shelves today. The standout examples are statins, PCSK9 inhibitors, metformin, and rapamycin.
It activates the enzyme AMPK, the cell’s low‑energy sensor, damping excess metabolism and inflammation. Diabetics on metformin have fewer age‑related diseases.
The first direct human test of an “anti‑aging pill” — the TAME trial.
They make the kidneys excrete excess glucose in urine, protecting the heart and kidneys along the way. Canagliflozin extended the lifespan of male mice in the ITP.
They slow bone breakdown; in several RCTs they unexpectedly reduced all‑cause mortality too, so they are being studied as a senolytic — a drug that clears senescent cells.
Patents on most of these drugs have expired, so major investment stays away from this research: the precise “gero-dose” and dosing regimen have to be worked out from scratch, and there is no budget for clinical trials. And on its own, none of them delivers a radical effect.
Survival of female mice across the 10 RMR‑1 groups (% alive by age, in days). The red “All” line (the full four) and the orange “Rapamycin” line track together, both above the “None” control — the combination did not beat rapamycin alone. Source: Gowing Life’s analysis of LEV Foundation data.
Aubrey de Grey combined 4 therapies in old mice — no synergy emerged: all the lifespan gain came from rapamycin alone
In Robust Mouse Rejuvenation 1, the LEV Foundation, led by Aubrey de Grey, gave 4 different interventions simultaneously to 1,000 already-old mice (C57BL/6J strain, treatment starting at ~18–19 months). The bet was on synergy — that independent mechanisms would add up to more than their sum. The result: only rapamycin delivered a meaningful lifespan gain, and the full combination did not beat it.
The combination’s 4 interventions and what each one delivered in RMR-1
Intervention
What it does
Result in RMR-1
Rapamycin · 42 ppm
Inhibits mTOR — the most reproducible longevity node in mice
The only one to deliver a consistent lifespan gain; the combination’s entire effect rests on it
Senolytic Gal‑Nav galacto‑navitoclax
Kills aged (“senescent”) cells, which poison the tissue around them
Renews blood stem cells and the aging immune system
The gain beyond rapamycin was statistically insignificant
What the combination actually added
The full four gave mice ~4 months of mean lifespan over controls. Maximum lifespan did not grow — the species ceiling stayed where it was.
In de Grey’s own words
“In females the picture is unpleasant: all the benefit we see comes from a single intervention — rapamycin.”
The takeaway
Stacking interventions blindly is not enough: synergy does not appear on its own. A combination has to be built from independent, reproducible nodes and tested pairwise — each pair on its own. That is exactly the discipline of the ITP →
The ITP is the best program for testing drug effects on mouse lifespan
Interventions from the NIA ITP (National Institute on Aging Interventions Testing Program), which measures effects by the survival of genetically heterogeneous mice. These data do not yet prove direct benefit in humans. Their value lies elsewhere: they are a reproducible foundation for selecting candidate drugs for combinations.
Median lifespan extension in the ITP — %, relative to controls; strong signals
The most reproducible pharmacological node, acting through mTOR; the effect holds in both sexes and even with a late start.
Acarbose
Acts through metabolism and works better in males; it is a distinct mechanism with no overlap with mTOR.
17α‑estradiol
Shows sex specificity: a strong signal in males, with no matching effect in females.
Canagliflozin
Yet another metabolic node: glucose, ketogenesis, inflammation, and cardiovascular risk.
Rapa + Aca
The ITP has already tested this pair: the combination's lifespan gain (+34%) beats rapamycin or acarbose alone — the mechanisms add up.
The key takeaway
Combinations should be built from independent, reproducible nodes — mTOR, carbohydrate metabolism, sex-specific regulation, the SGLT2/metabolic pathway.
ppm = parts per million; here it is the concentration of the compound in the feed. n.s. — the effect is not statistically significant: no reliable lifespan gain in that sex.
The ITP can be trusted because three labs test every drug independently
Since 2004, the NIA has funded a program that tests the same drug in parallel at three centers — in genetically heterogeneous UM-HET3 mice, in males and females, with replicates. A result counts only if it reproduces at all three sites. These are the scientists who run them:
Co-author of the ITP protocol and head of the Michigan site. He set the program's strict rules — triple replication, both sexes, late-life dosing — so a result cannot be written off as chance.
Canagliflozin·17α-estradiolhis group was the first to show these extend mouse lifespan
Heads the third site and the program's pharmacology: how to calculate a dose, blend a compound into the feed, and keep it stable over the mice's entire lives.
The field is shifting from preclinical studies to clinical RCTs and deliberate polypharmacy: GLP-1 + SGLT2 combinations, HRT + rapalogs, senolytics in short courses.
One short course extends mouse lifespan almost as much as lifelong dosing — and the effect persists after the drug is stopped. That means drugs can be given in rotation: a short course against one target, then another.
Rapamycin · a single 3-month course in midlife
·Mice were given rapamycin at 20 months of age for just 3 months — then it was stopped.
·Males (8 mg/kg injections): +60% in expected remaining lifespan after the course.
·In feed (126 ppm): +13% in total lifespan in both sexes.
·The authors note the effect is comparable to taking rapamycin for life.
·One caveat: high-dose injections caused blood cancers in females — dose and formulation matter.
Our takeaway: since one course works and its effect outlasts the drug, it pays to rotate drugs — hitting different targets with short courses, one after another. Over a lifetime that covers more mechanisms than a single lifelong pill, without accumulating side effects or tolerance.
The companies designing combination therapies for aging
Only a handful of companies are attempting to build a combination therapy for aging outright. Some of them go beyond repurposing existing drugs and engineer new combinations from individual molecules.
rhGH + DHEA + metformin (the TRIIM protocol): growth hormone kick-starts thymus regeneration, while DHEA and metformin blunt its blood-sugar side effect — in Phase 1 the combination rolled epigenetic age back by ~2.5 years
Mature players are still scarce. The most advanced clinical example is Rejuvenate (Phase 2). BioAge shows the sector's resilience: the azelaprag combination trial stopped in 2024, but the company regrouped — moved BGE‑102 into the clinic, signed a Novartis alliance worth up to $550M, and built its cash position to ~$385M.
The Cat Health Company is testing anti-aging combinations in cats and already has its first clinical data
An image from The Cat Health Company's site. Here the cat is both patient and model: a faster way to test geroprotectors meant for humans.
The company develops drugs against the very mechanisms of feline aging and selects candidates computationally (ML + omics data). Veterinary approval is a shorter path than human approval — a fast proving ground for geroprotectors.
first data, per the CEO: sarcopenic cats are gaining weight and several biomarkers are improving, with no serious side effects; a full publication is still pending
Timeline
Q3–Q4 2026
the goal: gather the data for a conditional-approval filing with the veterinary regulator
Who is behind it
Founded in 2024 (Alex Voda, CEO; Alex Bacita, COO). A $1.2M SAFE round (Oct 2025) is led by Portfolia (Active Aging & Longevity Fund II); investors include Ani.VC (a pet-longevity fund, Garri Zmudze), Alex Zhavoronkov, 100 Plus Capital, and Early Game Ventures.
Why this matters for humans
Cats age much like we do: the same sarcopenia, kidney disease, cognitive decline. Their short lives and faster route to approval make them a living test of geroprotectors — success in cats speeds up testing the same targets in humans.
Intervene Immune has already reached Phase 2: a three-drug combination regenerates the thymus
The thymus trains naive T cells, but it involutes with age — its working tissue atrophies and is replaced by fat, and immunity weakens. A combination led scientifically by Greg Fahy sets the reverse process in motion — thymus regeneration; in Phase 1 it turned biological age backward.
Phase 1 (TRIIM): 10 men aged 51–65, one year of treatment (9 completed)
−2.5 years
of biological age in one year (Horvath clock). The industry treats methylation clocks with reserve — but here MRI and T-cell data backed them up
Doses are personalized after the first 3–4 weeks based on individual response (IGF-1, insulin). Growth hormone drives thymus regeneration, while metformin and DHEA hold back its side effect — rising blood sugar.
85 participants aged 40–80, now including women, with a control group (metformin + DHEA). This is the step from pilot to proof of concept.
Interim data are encouraging: gains in aerobic capacity (VO₂max), leg strength, and immune markers — which the authors read as a return of function lost to years of aging. The full publication is still ahead.
Why this may outpace epigenetic reprogramming.Partial reprogramming (Sinclair and others) is still mostly in preclinical studies; a Phase 1 will likely arrive in 2026. Thymus regeneration is already in Phase 2 — a proof of concept in healthy people with functional biomarkers. If it shows a meaningful effect size, that is a major step forward.
Fig. 1 from the review by A. Panchin, M. Batin, et al. (Aging, 2024): combination-therapy strategies mapped to the hallmarks of aging.
Fig. 1 · Panchin et al., Aging 2024 — 4 classes of strategies: without editing the embryonic genome, with maximum hallmark coverage, those proposed by LEVF, and combinations already tested.
The logic of the map
1Aging is a complex of parallel, overlapping processes;
2A single intervention leaves some hallmarks of aging untouched;
3So compatible interventions are stacked to hit several hallmarks at once, counting on an additive or synergistic effect.
What has already worked in animals — combinations from the same paper
Combination
Which hallmarks of aging it targets
Lifespan effect (mice)
Pravastatin + zoledronate
Proteostasis and vascular calcification. In a mouse model of progeria (accelerated aging, HGPS), not in normal mice.
+77% median, +80% max.
p53/Arf + telomerase (TERT)
Genomic instability and telomere shortening. Transgenic mice: reinforced cancer protection plus activated telomerase.
+40% median
Rapamycin + acarbose
Nutrient sensing and autophagy — the cell's recycling of its own waste.
+28% ♀ / +37% ♂ median
Dasatinib + quercetin · senolytics
Clearing aged ("senescent") cells and reducing chronic inflammation.
Mitochondrial dysfunction. Transgenic overexpression of both enzymes.
+19% median
All combinations and figures are compiled in Tables 2 and 4 of the review — Panchin et al., Aging 2024. "Median" is the lifespan gain of half the group; "max." is the gain among the longest-lived.
What the authors propose to assemble: dasatinib + quercetin, FGF-21, acarbose, methionine restriction, young bone-marrow transplantation, and gene therapy with VEGF and telomerase (TERT), plus switching off NF-κB in the hypothalamus — one regimen hits nine hallmarks of aging at once without touching the embryonic genome.
Source: Panchin AY, Ogmen A, Blagodatski AS, Egorova A, Batin M, Glinin T. Targeting multiple hallmarks of mammalian aging with combinations of interventions. Aging (Albany NY). 2024;16(16):12073–12100. doi:10.18632/aging.206078.
Pharmacy
Combinations have already changed the outcome of disease
Breakthrough regimens were usually built as a deliberate assembly of different mechanisms: several points of attack, incomplete cross-resistance, controlled toxicity. The one-molecule-against-the-whole-disease approach worked far less often.
1960s
Leukemia — VAMP
Four drugs struck different mechanisms driving the growth of leukemic cells. This became one of the earliest arguments for multi-agent therapy. MD Anderson / Freireich
1970
Hodgkin lymphoma — MOPP
Vincristine, mechlorethamine (nitrogen mustard), procarbazine and prednisone together showed that advanced cancer could be made curable in a share of patients. DeVita et al.
1996 →
HIV — HAART
Triple antiretroviral therapy attacked the virus from several sides at once and turned HIV into a manageable chronic disease. ACS HAART landmark
2015 →
Melanoma — nivo + ipi
Combining PD‑1 and CTLA‑4 inhibitors releases two different brakes on the immune response and widens the share of patients with a durable response. NCI / FDA approval
The same route is open to geroscience: assemble a testable combination that acts on several mechanisms of aging at once, exactly as oncology does with cancer.
Almost no one tests combinations systematically: rigorous trials are expensive and slow, while unproven “protocols” are already on sale.
What the scientific route demands
·State the mechanism: which nodes of aging you are touching and why they are independent.
·Assemble the combination: doses, sequence, duration, stopping criteria.
·Test the interactions: effect, toxicity, synergy or antagonism.
·Publish the null result: otherwise the field keeps repeating the same mistakes.
expensiveslowrisky
Source: Panchin et al., Aging 2024
Why the market runs to Wellness
·The product is simple: a program, a subscription, a supplement stack, a personal protocol.
·The story is fast: people buy a promise they understand, and a long experiment cannot be sold that way.
·The evidence bar is lower: biomarkers and self-tracking sell more easily than a survival endpoint.
·Attention is easier to monetize: media and subscriptions scale faster than preclinical studies.
fastcheapmeaningless
Source: Blueprint products / Huberman Premium
Sources: Panchin et al. 2024 · Blueprint official · Huberman Premium
Pharmacy
Combination therapy against aging is already thriving — under the name “Wellness,” with no evidence
People already buy “longevity protocols”: supplement stacks, peptides, trackers. That is combination therapy against aging with no evidence base — the money and the attention flow into sales and skip the testing of hypotheses.
Gray — sold with no evidence base · green — clinical data exist (GLP-1) · black — funding for evidence-based science. All figures are annual, 2024.
≈ 500×
The gap between market and science
The entire wellness economy runs at $6.8 trillion a year. All evidence-based aging science together (NIA + Hevolution + venture capital) gets about $14B. The market takes the money and the attention, while the hypotheses stay untested.
The exception · GLP-1
Semaglutide (Ozempic) is a rare drug that is both sold at mass scale and tested rigorously in the clinic. In the SELECT trial semaglutide cut cardiovascular events by 20% in people without diabetes, and it is already being examined as a tool against aging. It could well turn out to be part of a genuine combination therapy — and an enormous body of data on it already exists.
We have the targets and we have the drugs. What we lack is a way to deliver them precisely to the right tissue, cell and compartment.
Delivery
Examples of modern delivery platforms
We may have the targets and the drugs, but not the method of delivery.
1 · ADC
Antibody–drug conjugates
An antibody–drug conjugate (ADC) is an antibody with a toxic payload stitched to it through a chemical linker. The antibody recognizes its antigen on the surface of a tumor cell, the cell pulls the conjugate inside, and there the payload is released and kills it. Chemotherapy arrives at one address instead of hitting the whole body.
A targeting molecule first images the target, then delivers a radioisotope to it. The therapy is therefore tied directly to imaging where the dose accumulates.
Example:Pluvicto (PSMA target in prostate cancer).
3 · LNP
LNPs for RNA and CRISPR
Lipid nanoparticles carry mRNA (messenger RNA), siRNA (short interfering RNA) and CRISPR components. The central idea is organ selectivity: getting the cargo into the organ you want, not only into the liver.
GalNAc is the sugar N-acetylgalactosamine. It is stitched onto an siRNA as an address label: it grabs the ASGPR receptor on liver cells, and they pull the siRNA inside. One of the most mature formats of chemical RNA delivery.
Example:Leqvio (inclisiran) — an siRNA against PCSK9 that lowers cholesterol.
5 · BBB
Delivery across the BBB
For the brain there are two routes: focused ultrasound, which temporarily opens the blood–brain barrier (BBB), or receptor “shuttles” that cross the endothelium.
Extracellular vesicles and membrane-coated nanoparticles combine the biocompatibility of a cell with the controllability of a synthetic carrier. The platform is less mature so far, but very promising.
Microneedles are an array of microscopic needles that painlessly pierce the top layer of skin and place a drug or vaccine into the dermis, with no injection and often no cold chain.
Peptides are normally injected, because the stomach and gut break them down. A permeation enhancer briefly opens the intestinal wall, and the peptide is absorbed straight from a tablet.
Examples:Mycapssa (octreotide in capsules, normally an injection); Rybelsus (semaglutide in tablets).
The key takeaway
Developing a drug against aging comes down in large part to building an effective platform that delivers the therapy into the right tissue.
Every example above is a clickable link to a primary source: the manufacturer's site, an FDA release or a peer-reviewed review. Examples are current as of mid-2026.
Position 02 · delivery
Which targets are we trying to reach?
“We know the targets and the therapies; the problem is delivering them precisely”
1 · Elastin and collagen crosslinks
The target sits in the extracellular matrix: dense tissue that renews slowly and resists selective delivery. Who:Revel ($1.7M NIH)
2 · Mitochondrial DNA
The cargo has to cross the cell and both mitochondrial membranes, then land in the right intracellular compartment. Who:Pretzel ($72.5M)
3 · Amyloids, the BBB and lysosomes
Aggregates and intracellular debris are often locked away in hard-to-reach tissue, in the brain or inside lysosomes. Who:Capsida (>$300M) · Aliada ($1.4B M&A)
4 · Senescent cells and SASP
The drug has to reach senescent cells specifically, without suppressing the whole immune system or damaging the surrounding tissue. Who:Rubedo ($40M)
5 · RNA, DNA repair and PAPP-A
mRNA, siRNA, genome-editing cargo and nuclear targets all need carriers, protection for the cargo and controlled release. Who:ReCode ($260M) · Evox (£69.2M)
A concrete example in aging: targeted delivery of rapamycin
The logic here is easy to see: the target is already known, the molecule already exists, and the bottleneck is getting it to the place where it is actually needed.
1 · What we know
mTOR is one of the best-validated targets in aging, and rapamycin is one of the best-known geroprotector candidates.
2 · What the problem is
Systemic rapamycin acts on many tissues and cell types at once, which quickly raises the question of side effects and of how precisely it hits.
3 · What is being tried
Nanoparticles, liposomes and senescence-targeted carriers try to steer the drug into the right cells and release it locally. Clinical:Emtora / Biodexa — eRapa, an oral form of rapamycin with pH‑sensitive polymers (>$20M in grants). Academic proof of concept:CD9-targeted rapamycin nanoparticles — addressed delivery into senescent cells.
4 · Why this matters
Sometimes it is enough to deliver a known drug more precisely to its target; no new molecule is needed.
A protein capsid carries genetic cargo into tissue. Capsid engineering sets the tropism, the passage across the blood–brain barrier (BBB) and the steering away from the liver (detargeting).
Dyno Therapeutics — AI capsid design, $100M Series A Capsida — CNS and the BBB, >$300M 4D Molecular Therapeutics — directed evolution of AAV, public company / ≈$821M+ Voyager — TRACER for the CNS and muscle, ≈$439M+
These platforms are already past preclinical: 4DMT and Capsida have clinical programs, and Dyno and Voyager partner heavily with pharma.
2. LNP lipid nanoparticles
LNPs carry mRNA, siRNA and genome-editing cargo. The bottleneck is getting past the liver and steering the cargo into the lungs, T cells or other tissues.
ReCode has clinical programs; Generation Bio and Aera remain platform plays at the preclinical stage, but with large partnerships and funding.
3. Chemical conjugates
An address label is stitched onto the cargo: a sugar, an antibody, a peptide or a ligand. That routes an RNA or oligonucleotide into a specific tissue without a bulky nanoparticle.
Alnylam / Arrowhead / Silence — GalNAc-siRNA: addressed to the liver through ASGPR Avidity — AOCs (antibody–oligonucleotide conjugates); muscle and heart, >$1.74B Rubedo Life Sciences — ALEMBIC: prodrugs switched on by enzymes of senescent cells (a sugar address label); senolytic, $40M
The most mature chemical addressing is liver GalNAc-siRNA; Avidity's AOCs are already in late-stage trials.
4. ADC antibody–drug conjugates
The antibody finds an antigen on the cell, the linker keeps the payload attached in the bloodstream and releases the toxic payload in the tumor or inside the cell.
This is no longer just a delivery concept but a mature drug class with real sales and many approvals.
5. RLT and theranostics
A ligand or antibody carries a radioisotope to cells that display the target; a diagnostic analogue shows in advance where the therapy will accumulate.
Clinically and commercially mature; the clearest working version of “see the target, then treat it.”
6. PLV protein–lipid / fusogenic particles
A particle carrying fusogenic proteins merges with the cell membrane and releases DNA/RNA cargo into the cytoplasm; the selection can then be made by a promoter inside the construct.
Oisín Bio — senolytic constructs, p16/p53 → inducible caspase-9 (iCasp9) Entos Pharmaceuticals — Fusogenix PLV for DNA/RNA delivery
An earlier-stage platform, important as a separate non-viral route for delivering large genetic constructs.
The first group answers the question “what carries the cargo”: a viral capsid, a lipid particle, a chemical conjugate, an antibody or a radioligand.
A transport module is attached to the drug; it binds a receptor on the vascular endothelium of the brain and ferries the cargo across the BBB without opening the barrier mechanically.
Denali — Transport Vehicle across the BBB Roche — Brainshuttle, an antibody fragment that crosses via BBB receptors Aliada / AbbVie — the MODEL platform; $1.4B acquisition JCR Pharmaceuticals — pabinafusp alfa: a TfR shuttle, approved for Hunter syndrome (MPS II)
Clinical programs in the central nervous system (CNS); for large proteins this is one of the key routes to making the brain druggable.
8. Physically opening the BBB
Focused ultrasound with microbubbles raises the permeability of brain vessels temporarily and locally, so that a drug already given can pass into a chosen region.
Carthera — SonoCloud, an implantable ultrasound device; €37.5M Series B Insightec — focused-ultrasound platform
Clinical trials are running, especially in neuro-oncology and neurodegeneration; the method is device-based, with no molecular shuttle involved.
9. EVs and exosomes
Natural or engineered cell-to-cell parcels are used to carry RNA, proteins and genome-editing complexes into hard-to-reach tissue.
Evox — ExoEdit / exosome platform; £69.2M Series C Vesigen — ARMMs (microvesicles via ARRDC1); $28.5M Series A Capricor — cell and exosome platform
Very promising but less mature: the main barriers are standardization, scale-up and reproducible composition.
10. Living cells and in vivo
The patient's own cells become the therapeutic factory: T cells, for instance, are turned into CAR-T inside the body. Simply delivering a drug is not enough here.
Capstan — targeted LNPs for in vivo CAR-T; $175M Series B / AbbVie deal worth up to $2.1B Umoja — in vivo CAR-T; $100M Series C Sana — fusogenic delivery into T cells
A transitional class between delivery and cell therapy; pharma interest is strong, but clinical validation is still taking shape.
11. Bacteria and microbial carriers
Engineered bacteria can colonize a tumor or a mucosal surface, produce a therapeutic protein locally, deliver an antigen or carry a genetic payload.
Prokarium — microbial immunotherapy / bactofection; $30M Synlogic / Senti Bio — programmable living systems and genetic circuits
A niche but distinct delivery class, of particular interest for tumors, the gut and local immune modulation.
12. Local depots and implants
The molecule itself stays the same; what changes is the route and the place of release: an implant, an injectable depot, inhalation into the lung, the skin, micropatches or a local catheter.
ReCode — inhaled mRNA delivery to the lung Vaxxas — microneedle patches DelSiTech — a silica matrix for sustained release Emtora Biosciences / Biodexa — eRapa: encapsulated oral rapamycin (pH-sensitive polymers)
Not always high-tech biotech, yet the route of administration is often what decides whether a therapy works in the tissue you care about.
The second group answers the question “how do you get past a barrier or put a living system to work”: a BBB shuttle, a physical window, an exosome, a T cell, a bacterium or a local depot. A special case is when the bottleneck lies in the target itself rather than in the carrier: Revel (cleaving collagen crosslinks in the matrix, $1.7M NIH) and Pretzel (the mitochondrial genome, $72.5M).
The capsid's structure decides which tissue it reaches
By changing the capsid — the protein shell of an AAV vector — Dyno learned to route the same gene into chosen regions of the brain (the cortex, deep structures, the spinal cord) rather than into the liver. Each image is a single capsid, and you can see how differently each one covers the brain.
Crosses the blood–brain barrier through the ALPL receptor. After intravenous dosing it covers the cortex and other CNS regions — up to 50% of cortical neurons in primates — while reaching the liver less than AAV9 does.
Enters the brain through the human transferrin receptor hTfR1. In hTfR1 mice it switches the gene on in 93% of neurons — 113 times the brain coverage of AAV9 — and is strongly steered away from the liver.
Selected for even coverage of the brain after intravenous dosing. Up to 30% of neurons in primates — 280 times the coverage of AAV9 — with little going to the liver.
Dyno's first capsid for delivery to the brain: AAV9 with 7 substitutions in the shell protein. Brain coverage 100 times wider than AAV9, liver uptake 10 times lower; 5–20% of neurons in primates, reaching deep structures and the spinal cord. Weaker and less selective than the newer variants.
How to read the images: blue marks the nuclei of every brain cell, green marks the cells that received the delivered gene and switched it on. The capsid is the protein shell of the vector; by changing its amino acids Dyno changes the route — which receptor the particle uses to cross the barrier, which cells let it in, and how much of it ends up in the liver.
The Oisín Bio particle. Destroying senescent cells selectively
How it works. The envelope delivers the cargo. The targeting happens after delivery, and it depends on whether the promoter inside the DNA construct switches on.
1 · The carrier
A PLV particle (a protein–lipid carrier) studded with FAST proteins (small transmembrane proteins that drive membrane fusion); the DNA/RNA cargo is packed inside.
2 · Entry
FAST proteins fuse the envelope with the membrane → the cargo enters the cytoplasm; the DNA then has to reach the nucleus.
3 · Targeting
The promoter fires only once inside the nucleus. If p16 or p53 are active there, transcription factors settle on the p16/p53-dependent promoter (p16/CDKN2A) and switch on transcription of the iCasp9 gene.
4 · Activation
The cell makes iCasp9. A dimerizer activates iCasp9, the caspase cascade fires and the cell dies by apoptosis. If the promoter never switched on, the cell survives.
PLV particle (Fusogenix)
FAST proteins — fuse the envelope with the membrane
lipid envelope — carries the cargo into the cell
DNA/RNA cargo — promoter + iCasp9 gene
The particle can enter almost any cell, but apoptosis only fires in senescent cells, where the right promoter is active. Outside is the envelope with its FAST proteins; inside is the DNA/RNA cargo (diagram at left)
The inverse problem — restoring the body's barriers
Delivery keeps learning to cross the body's barriers. But by one theory, aging itself is the breakdown of barriers and boundaries. That flips the task: boundaries need to be restored, not breached.
1 · Nucleus
Nuclear lamina
The nuclear envelope weakens: lamin A and progerin stop holding the genome's boundary. DNA fragments escape into the cytoplasm and trigger chronic inflammation through the cGAS-STING sensor.
2 · Cell
Cell membrane
The cell's outer boundary loses integrity and lipid order — the cell gets worse at holding in its contents and keeping its ionic balance.
3 · Tissue
The cell collective
Cells lose their shared identity and their tight junctions with one another; the tissue stops acting as one whole — altered intercellular communication is one of the hallmarks of aging.
4 · Organism
Gut barrier
“Leaky gut syndrome”: the intestinal wall becomes permeable, gut contents seep into the bloodstream and feed systemic inflammation (inflammaging).
The delivery problem
Cross the barrier. Get the cargo inside: a capsid, an LNP, a BBB shuttle, or ultrasound temporarily opens a boundary — the BBB, the membrane, the epithelium.
The inverse problem
Reinforce the barrier. Bring back separation: restore the nuclear envelope, the cell membrane, the tight junctions in tissue, and the gut wall.
Both problems sit on one axis — the boundary: some therapies need the barrier opened, others need it restored. Aging can be seen as the breakdown of boundaries at every level: nucleus → cell → tissue → organism.
We stand here if we believe the problem of aging is solved in theory — all that's left is to find the cure for old age.
Drug discovery
Where drugs come from: nature and accidental finds
Each row is a specific organism, the scientists, the place, and the year of discovery. The “searched for → found” column shows how often a discovery was made while hunting for something else entirely.
Source (species)
Searched for → found
What it grew into
Who · where · when
From unexpected organisms — the tools of biology and geroscience
Heloderma suspectum Gila monster
Studied venom and saliva peptides and their effect on insulin → found exendin-4 (a 39-amino-acid peptide, ~50% similar to GLP-1, resistant to enzymatic breakdown)
exenatide (Byetta), FDA 2005 — the first GLP-1 drug. Ozempic and Mounjaro grew out of the same “GLP-1 as a target” logic
John Eng, Bronx VA (New York), 1990–92
Tetrahymena thermophila ciliate
Studied how chromosome ends are built and extended → found the enzyme telomerase (an RNA–protein complex, a reverse transcriptase)
Telomere biology: the replicative limit, cellular senescence, cancer, telomere syndromes
C. Greider and E. Blackburn, Berkeley, 1985 · 2009 Nobel (with Szostak)
Zea mays maize
Studied why kernel color is unstable and chromosomes keep breaking → found the mobile elements Ac/Ds (“jumping genes”)
Mobile DNA; in humans, LINE-1 retrotransposons, genomic instability, inflammaging
Barbara McClintock, Cold Spring Harbor, 1948–50 · 1983 Nobel
Streptococcus thermophilus · S. pyogenes bacteria: the yogurt starter and group A strep
Worked out how S. thermophilus fends off bacteriophage viruses → found the CRISPR-Cas system; the workhorse protein Cas9 came from S. pyogenes (SpCas9)
Nature supplies the raw material — then the protein you need is “bred” in a test tube through rounds of mutation and selection
Directed evolution is a method for producing a protein with a desired property: the protein's gene is randomly mutated over and over, the variants with the best version of that property are selected, and the cycle repeats until the property reaches the goal. It is natural selection itself — just in a test tube, over weeks rather than millions of years. Frances Arnold won the 2018 Nobel Prize in Chemistry for it (together with George Smith and Gregory Winter).
1 · Diversity
One gene is turned into millions of variants. Error-prone PCR copies the gene while introducing 1–2 random mutations per copy; DNA shuffling cuts successful genes into pieces and reassembles them, mixing the useful mutations.
2 · Selection
The variant library is screened for the desired property and the best are kept. In phage display, each protein variant sits exposed on the surface of a bacteriophage — that is how, out of billions, the ones that bind a given target get fished out.
3 · Amplification
The genes of the best selected variants are multiplied — they become the starting material for the next round.
4 · Repeat
The mutation → selection → amplification cycle runs several times. With each round the desired property grows until it hits the target.
What was “bred”
How it was done
The real drug / result
Therapeutic antibodies phage display
A library of human antibodies is displayed on phages, and the ones that bind the target tightly (here, the inflammatory protein TNF-α) are selected. Gregory Winter's method.
Adalimumab (Humira) — the first fully human antibody, approved in 2002; ~$20B a year in sales at its peak.
An enzyme for drug synthesis biocatalysis
A transaminase enzyme that barely worked on the target molecule was evolved until it could assemble the active ingredient and withstand factory conditions (high temperature, solvent). Codexis + Merck.
AAV capsids for gene-therapy delivery vector evolution
Millions of variants of the AAV viral shell are run through selection directly in living tissue — the survivors are the ones that reach the right organ and evade human antibodies.
A machine-learning model uses the variants already tested to predict which mutations are worth trying next — so the selection cycle proceeds with aim rather than blindly.
The point for us: directed evolution can build a protein with exactly the property you need — binding a target, working in the body or in a reactor — even when no such protein exists in nature. And this method has already reached aging itself: the same variant-by-variant search bred an enzyme that strips age-related damage right off human proteins — next slide.
CML (Nε-carboxymethyl-lysine) is the residue left when sugars react with a protein — the “browning” of tissues; it builds up in long-lived proteins over decades. CMLase oxidizes CML and restores the native lysine.
500M
enzyme variants searched by directed evolution; the scaffold is a bacterial glycine oxidase
−70%
CML in arterial tissue from a 75-year-old donor after enzyme treatment (ex vivo)
−55%
CML in an elderly donor's skin — the level dropped below what is typical for 31-year-old skin
For decades, chemical aging of proteins was considered irreversible: the sugar “weld” permanently alters long-lived proteins. Directed evolution produced an enzyme that removes it. Honest about the stage: demonstrated in vitro and in human tissue samples; not yet tested in live animals.
Find which genome variants are linked to a disease, then which transcriptome shifts stand behind them, and pick molecules that turn that transcriptome around. Then test in cells and models.
1 · Polymorphisms
GWAS: which DNA variants are associated with the disease and its course.
2 · Gene and transcriptome
Which gene and which expression changes stand behind the association.
3 · Molecule matching
Search for compounds that reverse the disease expression signature (Connectivity Map / LINCS).
4 · Validation
Cells and animal models, then clinical trials.
≈ ×2.6
how much higher the odds of approval are for a target with genetic support (Minikel, Nature 2024; refining Nelson 2015)
Human genetics works as a priority filter: it screens out targets with no causal grounding up front and raises the share of programs that reach clinical trials. It is a cheap way to choose what to test.
The genetics of aging already knows far more than has been tested in mice
The practical potential of the knowledge accumulated in aging genetics far exceeds what has actually been tested in animals. The OpenGenes database links thousands of genes to aging — but only a handful have made it to an intervention that extends mammalian lifespan.
2402
genes linked to aging — knowledge assembled from more than 1,700 papers
247
of them have made it to at least one lifespan experiment (in any model animal)
78
genes have actually extended lifespan in mammals — mice and rats
8
are solidly confirmed: lifespan changes both when the gene is boosted and when it is switched off in mammals
The bottleneck is application. Thousands of genetic leads are already known, but only a handful have been tested as interventions in mice. Practice trails knowledge by orders of magnitude.
How OpenGenes is mined for candidate cures for old age
In Nature Aging (2026), Barabási's group overlaid the OpenGenes database on the network of all human proteins and built SHARP (Systematic Hallmark-based Aging Repurposing Pipeline): take the aging genes, find the spots on the network where they gather into hallmarks, and select existing drugs whose targets land on those spots and push aging into reverse.
1 · GENES
OpenGenes: 2,358 longevity genes, each with a confidence level for its link to aging (1 highest, 5 lowest).
2 · HALLMARKS
1,250 genes are annotated with 11 hallmarks of aging (the fundamental signatures of aging): 860 with one, 390 with several at once.
3 · NETWORK
The genes are mapped onto the interactome — the network of all known human protein interactions: 18,223 proteins and 524,156 links between them.
4 · MODULES
The genes of a given hallmark converge into one connected cluster (a module). For 9 of the 11 hallmarks, that connectedness is statistically significant.
5 · PROXIMITY
6,442 existing drugs (DrugBank) are ranked by how close their targets sit to each module. 370 remain significantly close.
6 · pAGE SIGN
For drugs with expression data (CMap), pAGE is computed: does the drug reverse the age-related gene shift or amplify it.
Steps 1–4. Aging genes from OpenGenes on the protein network: the 11 hallmarks converge into a single “longevity module.” The large nodes are hub genes TP53, FOXO1, SIRT1, ATM, AKT1, tied to several hallmarks at once. Fig. 3 · Gross et al., Nature Aging 2026.Step 6 · pAGE (Pro-Age). Compares the age-related shift in gene expression with the drug's effect. pAGE > 0: the drug reverses the shift (a geroprotector); pAGE < 0: it amplifies the shift (accelerates aging). The stronger a gene's link to aging in OpenGenes, the greater its weight. Fig. 4 · Nature Aging 2026.
The metric already predicts geroprotectors — not yet perfectly
8 / 8
of the ITP compounds that extend mouse lifespan (and have a measurable pAGE) score positive on at least one hallmark. Among those that failed to extend lifespan: only 3 of 7.
1.24 : 0.52
the ⟨pAGE+⟩ / ⟨pAGE−⟩ ratio for lifespan-extending vs. non-extending compounds — the successful ones push the hallmarks in the positive direction notably more often.
8 / 9
of the compounds from clinical studies of aging (with a measurable pAGE) positively affect at least one hallmark.
5 / 5
successful geroprotectors in an independent mouse test (published after the predictions were made) score positive; among the unsuccessful ones: 2 of 3.
A ready-made candidate list. Out of 60 compounds not yet tested in mice that have a measurable pAGE, the method picked 21 with a positive effect on at least one hallmark — priority geroprotector candidates. Judging by the validation above, some of them really will extend lifespan (but not all).
As a side effect, the database confirmed the hallmarks themselves. Genes of a given hallmark are more tightly connected in the network than chance would predict (significant for 9 of 11), and different hallmarks consistently overlap in genes — meaning they are distinct yet describe one process. On top of that, OpenGenes genes land non-randomly in 5 independent classifications: age-dependent pathways, aging genes from 7 major studies, 5 age-related diseases, 8 “age-related” cancers, DNA repair, and progerias.
The takeaway for us. The more complete OpenGenes is, and the more accurate its confidence levels for each gene's link to aging, the more accurate the geroprotector predictions. Growing the database directly raises the odds of finding a cure for old age.
OpenGenes and OpenDrug reduce the fight against aging to problems on a graph
Write biology down as a graph: nodes are genes and proteins, edges are their interactions. The big question—how do we treat aging—then breaks down into rigorous mathematical problems, each with a verifiable answer.
Aging as a graph. Nodes are genes and proteins; edges are their interactions. The orange cluster is the aging module; ringed nodes are driver nodes; the diamond is a drug found by the proximity of its targets to the module.
PROBLEM 1 · CLUSTERS
Where does aging sit?
Community detection on the graph: genes of one hallmark converge into a connected module, and aging reads as a shift of the whole module. Network medicine, Barabási, 2011.
PROBLEM 2 · THE DRUG
What will work?
Network proximity: rank drugs by the distance from their targets to the aging module—a shortest-path problem on the graph. Guney et al., Nat Commun 2016.
PROBLEM 3 · LEVERS
Which nodes to push?
Network controllability: find the minimal set of driver nodes that dictate the dynamics of the whole network—a maximum-matching problem. Liu, Slotine, Barabási, Nature 2011.
PROBLEM 4 · VALIDATION
Do the predictions hold?
The method becomes a testable problem with a known answer: 8 of 8 compounds that extend mouse lifespan in the ITP were predicted correctly. Nature Aging, 2026.
OpenGenes—labels on the nodes
2,402 genes of aging, each with a confidence level for its link to aging (1 = highest, 5 = lowest). That is exactly the labeling of the graph's nodes. Nucleic Acids Research, 2024.
OpenDrug—the “drug → target” edges
Our platform maps compounds to their protein targets in the same network: which drug hits which nodes. These are the edges over which proximity and controllability are computed.
Bottom line. The more complete and accurate these two graphs, the more accurate the solutions. Building out OpenGenes and OpenDrug directly improves the input data for the math that searches for geroprotectors.
PCSK9 inhibitors: sharply lower “bad” LDL cholesterol → fewer heart attacks and strokes, the leading cause of death in old age. Repatha and Praluent are antibodies against the PCSK9 protein; Leqvio is an siRNA that silences the PCSK9 gene, injected twice a year.
All three are approved. In VESALIUS-CV (Nov 2025), Repatha cut heart-attack risk by 36% even in people with no prior heart attack or stroke.
Metformin—a cheap diabetes drug under study as a geroprotector: it may slow several diseases of aging at once.
Approved for diabetes for decades. The TAME trial (3,000 people aged 65–79) is meant to test whether it slows aging itself—and is still waiting for funding.
Boosts regulatory T cells (Tregs), which damp the chronic inflammation of the nervous system—one of the engines of aging.
COYA-302 (a Treg biologic) for amyotrophic lateral sclerosis (ALS): Phase 2 ALSTARS is underway; FDA Fast Track status granted in 2026.
small-cap ~$100–125M
The cautionary tale: Unity Biotechnology—the most advanced public senolytics company, with positive Phase 2b data—was delisted from Nasdaq (August 2025), and in September its stockholders voted to liquidate the company. Even a late clinical stage and a stock listing do not guarantee a drug will reach patients.
Hybrid molecules (the TOBI platform) rejuvenate the immune system: they push it to clear senescent cells and damp the chronic inflammation of aging (inflammaging).
Epigenetic reprogramming, AI design of transcription factors
$130M+
scoring in progress
Eternal Search · Powered by Open Longevity · eternalsearch.net · the score is a sum of scientific and evidence dimensions · snapshot as of July 5, 2026
Drug discovery
Where the money is: longevity investment by year and by theme
Annual venture funding for longevity swings year to year, but the trend points up. Inside it, themes trade places: cellular reprogramming and rejuvenation broke out precisely in 2024.
Longevity biotech investment by year, per the Annual Longevity Investment Report 2024, $B (number of deals)
Cellular reprogramming jumped from 21st place (2023) to 4th (2024)—$1.6B in a single year. Rejuvenation and organ replacement follow right behind. This is what an incoming trend looks like: capital flowing into a field on the early upswing of interest.
They recognize an external target precisely and leave almost everything else alone; low immunogenicity. Rarer, pricier, and not suited to every target type, but with a higher chance of success.
Switches on, in an adult organism, a program that has already extended lifespan in animals. Expensive, done once and forever, with a high price for a dosing mistake.
Example: the telomerase gene (AAV-TERT) extended mouse lifespan by up to 24%, with no rise in tumor counts (Bernardes de Jesus, 2012).
Peptides (oligopeptides)
They mimic the body's own signals; the familiar case is the GLP-1 drug class. Example: semaglutide and tirzepatide (Ozempic, Mounjaro) grew from a natural peptide into a therapeutic platform at scale.
Which therapy class to pick depends on which mechanism of aging we believe is the main one.
Drug discovery
We are heading toward unthinkable drug prices
The world's most expensive drugs come in two kinds. A gene therapy is bought once—$2–4M per course. A recombinant enzyme or antibody is bought every year, for life—and the record holder here, Strensiq, runs about $1.8M a year. The harder the biological problem, the more expensive the drug.
REcell: building a pool of replicatively immortal stem cells
The idea: raise protein output by 5–10% from six of the cell's own genes in stem cells, return those cells to the body, and restart renewal of the blood-forming system — a root cause of age-related mortality.
Core idea
1Name
A genetic overhaul of blood-forming cells
REcell — ex vivo editing of human hematopoietic stem cells (hHSC) to rejuvenate the blood and the immune system.
2Concept
Six genes turned up slightly — the promoter stays untouched
We edit the non-coding regions of the mRNA — the 5′ end (the Kozak sequence) and the 3′ end (the polyadenylation signal) — which set how much protein each RNA molecule yields. The output of each of the 6 genes (TERT, KLF4, p53, p14/p16) is raised by a set percentage.
Technology and plan
3What's missing · distance to go
The bottleneck is multiplexing across 6 loci
Simultaneous prime editing in human cells has already been shown for 3 loci; going beyond that number is the next step.
4What it takes
The preclinical plan
We select the best combinations of edited genes and expression levels. We test the effect on slowing telomere shortening, on preserving stem-cell identity (the capacity to self-renew and to differentiate), and on lowering the rate of cancer formation, both spontaneous and induced. Then come myeloablated chimeric mice: in the test group the bone marrow is repopulated with modified human stem cells, in the control group with unmodified ones.
Outlook and market
5Odds of success
The precedent is encouraging
In Serrano's mice, telomerase combined with tumor suppressors delivered +40% median lifespan with no rise in cancer. Cell and gene therapies win approval from Phase 1 ~19% of the time — twice as often as the 7.9% average.
6Spillover value
A multi-gene editing platform
A platform of stable stem cells with designed properties. It lets us set the traits we want at the level of stem-cell genetics and metabolism — traits that can carry over into the differentiated cells they later become.
Colossal is the most successful combinatorial gene therapy so far
Colossal is bringing extinct species back: the mammoth, the dire wolf, the dodo. But its main asset is the method itself: dozens of genes edited at once inside a living animal. That is exactly what gene therapies for aging need, in projects like REcell.
Brand identityThe dire wolf — 20 edits across 14 genes
1Platform
Large-scale multiplexing plus computational genome design
Aging runs on many genes, so it takes genome edits at scale
Dozens of genes drive aging at once, so a single edit neither halts it nor turns it back. Church built exactly the technology for editing genomes at scale.
Colossal's value is in the method. It can help many companies deliver their own approaches to aging. REcell, for example, needs multiplexing across 6 loci — and Colossal already shows that 8–20 simultaneous edits work in a living animal.
At Eternal Search we gather analysis on life-extension ideas and on the methods needed to realize them, and we assemble those into routes to a goal. By our data Colossal, at first glance not a longevity company at all, opens the road for many others.
Every couple of years the record for the world's most expensive drug is rewritten. A one-time gene or cell therapy today costs between $2M and $4.25M for a single dose — and the ceiling keeps rising.
Casgevy — $2.2M sickle cell disease, the first CRISPR therapy
This is why we need a cheap route to medicines. Combinatorial gene therapies for aging (REcell among them) are technically harder still — and on this pricing logic they will cost even more. At millions per dose, a treatment reaches tens of patients instead of millions. So we take the opposite route: find cheap, already-known molecules and repurpose them.
AI has compressed early development — finding the target, predicting protein structure, designing and optimizing the molecule — from years to months. Below is a map of the field by class of AI application: who does what, exactly, and on whose money.
Class of AI use
How it works / what it gives
Companies — differentiator — money
Stage and maturity
1. End-to-end platforms from target to molecule
AI runs the entire early chain on one platform — from finding the target to the candidate molecule — closing the loop of predict → synthesize → test.
Insilico Medicine — the end-to-end Pharma.AI suite: PandaOmics finds the target, Chemistry42 draws the molecule — ~$293M IPO (HKEX: 3696); rentosertib for pulmonary fibrosis has cleared Phase IIa Recursion — phenotypic “cell maps” at scale, plus Valence Labs — absorbed Exscientia (~$688M); partnerships with Roche, Bayer, Sanofi Schrödinger — physics-based molecular simulation plus ML, selling both the software and its own pipeline — ~$256M revenue; SGR-1505 and SGR-2921 in Phase 1 insitro — ML plus stem-cell (iPSC) disease models for target discovery (founded by Daphne Koller) — ~$400M Series C; partnerships with BMS, Lilly, Gilead XtalPi — AI plus quantum physics plus robotic labs, sold as a service to pharma — listed in Hong Kong (2228.HK)
The most mature class: several already hold clinical assets. The core of the bet is to make the expensive search through early hypotheses cheap.
2. Structure-based molecule design physics + AI
The model predicts the target's 3D structure and fits a molecule to it, computing binding strength instead of screening compounds at the bench.
A tooling layer, with part of the models open. Molecules are now computed directly, without blind screening.
3. De novo design of proteins and antibodies molecules from scratch
A generative model designs a new protein or antibody against a chosen site on the target, instead of picking one out of a ready-made library.
Xaira Therapeutics — generative protein models (out of David Baker's school, RFdiffusion) — >$1B at launch (ARCH + Foresite) Nabla Bio — JAM-2: fully computed antibodies against hard targets (KRAS, peptide-MHC, GPCR) — partnership with Takeda, confirmed in the lab Absci — ABS-201, an AI-designed antibody in Phase 1/2a — a $100M round (Lilly among the backers); public (ABSI) Arc Institute — Germinal: open AI design of antibodies against a chosen epitope — a nonprofit institute Chai Discovery — Chai-2: de novo antibodies with a ~20% hit rate and no screen — $70M Series A at a ~$1.3B valuation EvolutionaryScale — ESM3, a frontier protein model (sequence + structure + function) — a $142M seed (Amazon, NVIDIA) Cradle — generative AI protein engineering as SaaS for biologists — $73M Series B; customers include Novo Nordisk and J&J
A fast-growing front, running from open software to the clinic (Absci). Proteins are already designed from scratch against a chosen target.
4. The virtual cell simulating interventions
AI predicts how a cell will respond to a switched-off gene or a drug — the experiment runs in silico before the bench, narrowing the list of targets.
Xaira X-Cell — a virtual cell with 4.9B parameters plus the open X-Atlas/Pisces perturb-seq dataset CZI / Chan Zuckerberg Biohub — the virtual cell program: models of how a cell responds to an intervention; the CZI Cell Models platform PerturbGen — predicts a tissue's future state after an intervention, trained on >100M cells — an academic preprint (bioRxiv, 2026)
The earliest stage — almost all of it academic. The virtual cell is precisely what could replace expensive cell experiments.
5. AI for cell reprogramming and rejuvenation rejuvenation
AI searches for a safe set of genes or factors that returns an old cell to a young state while it keeps its function and stops short of becoming a stem cell.
NewLimit — AI plus epigenetic reprogramming of liver, immune, and vascular cells — $130M Series B (founded by Brian Armstrong); preclinical Shift Bioscience — an AI virtual cell plus aging clocks hunting for a single rejuvenating gene — $16M; asset SB000 Retro Biosciences × OpenAI — the GPT-4b micro model redesigned the Yamanaka factors — a >50× rise in markers of cell rejuvenation Yuva Biosciences — the MitoNova platform picked molecules that raise ANT1 in muscle cells (mitochondria, the GLP-1 market) — partnered with NorthStrive Gero — physics-informed AI on longitudinal data targeting the “noise” of aging — a Chugai/Roche deal potentially worth >$1B; preclinical
Preclinical, yet the most direct bridge to a drug for old age. This is where longevity funds and crypto billionaires place their bets.
6. Scientific AI agents targets, assets, experiments
An AI agent runs the research itself — planning, finding targets and assets, writing code, setting up experiments — and takes part of the manual work off the scientist.
Anthropic — Claude for Life Sciences: an agent for genomics, single-cell, proteins, and chemistry; 60+ scientific skills and connectors Owkin — K Pro: a co-pilot agent for finding targets, assets, and clinical rationale — licensed to AstraZeneca and Sanofi Biomni — a general-purpose biomedical agent (Stanford) that plans and runs bio tools — open Medra AI — its “AI Experimentalist” turns a goal stated in plain language into experiments in an autonomous lab — a project with DARPA Bioptic — vertical agents for scouting drug assets (~80% accuracy) — used by biotech funds Lantern Pharma — the RADR platform for rare cancers plus withZeta.ai, a multi-agent “co-scientist” by subscription — clinical stage Noetik — OCTO-VirtualCell: a spatial-omics model that picks out immunotherapy responders from a tissue slice — licensed to GSK Rockefeller (Sean Brady's lab) — DNA from environmental samples → computed molecules → synthesis → testing for antibiotics (cilagicin)
The newest direction (2025–2026): from open academic agents to enterprise licenses. It changes how R&D itself is organized.
7. Mechanistic AI for safety and dose in silico toxicology
A physiological AI model computes safety, dose, and tolerability in advance, moving part of the testing off animals and people and into calculation.
VeriSIM Life — the BIOiSIM platform for safety, dose, and pharmacokinetics — $15M Series A; a research agreement with FDA/NCTR (2026)
A narrow but strategic direction — a bridge to regulatory science, where the clinic remains the bottleneck.
8. Multi-omic models of health biomarkers of aging
AI folds blood work, the genome, and clinical data into one model that scores biological age and risk — and points to where to intervene.
Human Longevity — Health Nucleus: whole genome plus whole-body MRI plus ML for early disease screening and longevity Helix — turns blood biomarkers into a map of biological age (epigenetics and multi-omics are on the roadmap) AI4L / Forever Healthy — open “audit prompting”: AI itself checks the reviews on senolytics, NAD, and mTOR, verifying every reference live; it feeds the Evipedia encyclopedia
The applied end, closest to the patient — it measures aging and disciplines the evidence, leaving the hunt for molecules to others.
9. Data, benchmarks, and infrastructure for bio agents
Agents fail without shared data, honest tests, and adapters into the models — this layer sets what to train them on and how to measure whether they are good enough.
OpenAI GeneBench-Pro — 129 tasks on messy real-world bio data: it tests whether an agent can genuinely analyze them (best model ~29%) LatchBio — TxBench — open evaluations of agents on preclinical tasks (TxBench-PP, 100 tests) Open Reaction Database — an open reaction database for training synthesis planning; a dataset of 50,688 C–N reactions (Princeton + Merck) NVIDIA BioNeMo — the Agent Toolkit turns protein, molecule, and genome models into tools agents can call Astera Institute — funds open science and requires machine-readable negative results
The foundation under everything above — shared data and honest benchmarks decide whether AI science becomes reproducible.
≈ 200
AI-derived programs already in clinical development (early 2026): ~94 in Phase 1, ~56 in Phase 2, ~15 in Phase 3.
30 months
from target to clinic for rentosertib (Insilico), against the usual 6–8 years.
One important caveat: no fully AI-made drug has been approved yet. AI speeds up the cheap early stages, where the work is searching through hypotheses; testing in humans remains the main bottleneck. That is why the bet is strongest at the entrance to development.
Also on the radar (weaker so far on verifiable biological results): Xellar (organ-on-a-chip + AI) · Sanyou Bio (AI antibody library) · Denovo Sciences (small-molecule generation, partnered with Mankind Pharma) · MindWalk (HYFT fingerprints) · Inocras (cancer genomics) + AimedBio (AI antibodies for ADCs) · CHARM Therapeutics (3D protein–ligand folding) · Terray Therapeutics (chemistry at scale, the tNova platform) · PeptAI (autonomous peptide design, a DeSci project).
Put most of his liquid net worth into Retro Biosciences, whose goal is to add 10 healthy years to human life. OpenAI, meanwhile, trained GPT-4b micro specifically for the biology of aging: it reengineered the Yamanaka factors that reprogram cells back to a youthful state.
AlphaFold predicted the shape of nearly every known protein — a ready-made map for drug development. His company Isomorphic Labs designs drugs with AI; Hassabis himself says AI could help "defeat all diseases" within roughly a decade.
$600MIsomorphic Labs round (2025); Lilly and Novartis deals worth up to ~$3B
In his essay "Machines of Loving Grace," he laid out the bet behind this entire section: powerful AI will compress 50–100 years of progress in biology into 5–10 — up to doubling the human lifespan and eliminating most cancers.
His bet is AI itselfno direct longevity investments; Anthropic raised a $65B round at a $965B valuation (2026)
Back in 2013 he launched Calico, the Alphabet company that studies the biology of aging (a Time cover asked "Can Google Solve Death?"). One of the first tech titans to place a major bet on longevity.
Explaining what already happened is easy; naming the right probability in advance is hard. A prediction market forces a calibrated forecast, backed by money or reputation, and then checks it against reality.
How it works
A contract pays $1 if the outcome is Yes. The contract's current price is the probability the crowd assigns. Money or points punish overconfidence and reward accuracy.
What biotech already has
Endpoint Arena — a betting market on clinical trial outcomes (paper trading for now). Bio Protocol — prediction markets decide which research gets funded next. Eternal Search — our 43 forecasts on who will extend life.
Why it works
Markets aggregate scattered information better than individual opinions. In the Reproducibility Project they correctly predicted ≈71% of replications, beating an expert survey (Dreber, 2015). Metaculus is calibrated to within 2–3% across thousands of questions.
Caveats
Thin liquidity distorts prices; there is a risk of insider trading and manipulation (bets placed on non-public information have happened). And an ethical concern: betting on medical outcomes can affect the trials themselves.
Why this matters to us: it is a way both to understand — seeing the honest probability instead of a convenient retelling after the fact — and to influence, steering attention and capital toward what can actually take off. A forecast backed by money or reputation disciplines a field more than any commentary.
Methods are what bring revolutionary change to science
The chromatograph, the mass spectrometer, the DNA synthesizer, the sequencer, the MRI. When does the next class of instruments and methods arrive?
Methods
Making research cheaper: attack the cost at every step of the experiment
Every experiment runs the same chain: set up → measure → get data → repeat. Make any step cheaper, and testing hypotheses at scale stops being a luxury. Everything built on top gets cheaper next: target discovery, antibodies, biomarkers, cell therapies. Below are six levers, one per step of the chain.
1 · Measure
Sequencing: we read the genome, and every cell individually. The price of reading a genome has fallen roughly 30-million-fold.
2 · See
Next-generation instruments: cryo-EM, thousands of proteins from a drop of blood, organ-on-a-chip. More data per experiment — and often no animal testing.
3 · Run experiments
Automation and cloud labs: robots do the pipetting for the scientist, and instruments are rented over the internet — no manual drudgery, no equipment to buy.
4 · Manufacture
Bioreactors grow cells that work as factories. Antibodies are dropping from $180–500 to <$40 per gram; cell therapies, severalfold.
5 · Compute
AI predicts protein structure and cell response in silico, before anything touches a test tube — screening out unneeded wet-lab experiments. Covered in depth in the separate "AI" section.
6 · Reuse
Open data: never rerun an experiment someone has already done. Raw results get preserved, found, and put back to work.
Next: for each lever, a table of what exactly it makes cheaper and who is doing it. The companies here are infrastructure: they make research faster and cheaper for everyone developing drugs against aging.
Methods
Sequencing: the price has collapsed millions-fold
Reading a genome used to cost $3B. Today it costs $100. Sequencing reads biology at the level of DNA; when it gets cheaper, every method built on it gets cheaper too.
$3B
one human genome, the Human Genome Project, ~2003
$1,000
the price of reading a human genome fell to $1,000 — Illumina HiSeq X Ten, 2014
$100
Ultima Genomics UG 100, 2024 ($1.5M machine, ~20,000 genomes/year)
≈ 30M×
how far the price of a genome has fallen: $3B (2003) → $100 (2024)
Reading cell by cell. Sequencing used to give an average across a tissue. Now it reads each cell separately, showing exactly which cells are aging. And newer methods also show where those cells sit within the tissue.
Read class
How it works
Companies — differentiator — money
Stage and scale
1. Short reads short-read
The genome is chopped into short pieces of 100–300 "letters" and read all at once, millions in parallel. Each piece read out is a "read"; short reads are the cheapest approach, and the entire race for the cheap genome happens here.
Illumina — the industry standard, ~80% of the market, a genome for ~$200; public (NASDAQ: ILMN), $4.4B revenue Ultima Genomics — drove the price down to $100 per genome, ~$1.5M machine; raised ~$600M Complete Genomics / MGI — DNA nanoball chemistry (DNBSEQ), ultra-high throughput; the main non-Illumina alternative (listed in Shanghai) Element Biosciences — the benchtop AVITI, challenging Illumina on price and quality; >$680M raised ($277M Series D, 2024)
A mature commodity market: the technology is standard, and the fight is over price ($200 → $100 per genome).
2. Long reads long-read
Reads long, continuous stretches — thousands to millions of "letters" in a row. This reveals the large genomic rearrangements and repeats that short reads miss.
Oxford Nanopore — real-time sequencing, ultra-long reads, the portable MinION device; public (LSE: ONT), £183M revenue PacBio — HiFi reads: length plus short-read-level accuracy; public (NASDAQ: PACB), $154M revenue
A mature but narrow niche: technically irreplaceable, yet the market is smaller and grows slowly.
3. Cell and tissue single-cell & spatial
Reads every cell individually, showing exactly how the cells of one tissue differ; spatial methods additionally preserve where each cell sits in the tissue.
10x Genomics — the leader: Chromium (cells) and Visium/Xenium (tissue maps); public (NASDAQ: TXG), $611M revenue Parse Biosciences — cell barcoding without microfluidics, up to millions of cells from an in-a-tube kit; $50M Series C Vizgen — MERFISH/MERSCOPE: spatial transcriptomics across whole tissue; ~$170M raised
A fast-growing frontier: both science and venture money are moving here, with fresh rounds in 2026.
Automation and bioreactors: robots run the experiments, cells make the drugs
Robots handle the pipetting, cloud labs rent out their equipment, and bioreactors grow cells that work as factories. Experiments get faster, more precise, and cheaper.
What a bioreactor is
A tank where living cells or microbes are grown under precisely set temperature, acidity, oxygen, and stirring. Inside, the cells work as miniature factories, making proteins, antibodies, vaccines, or the cells themselves for therapy. Conditions are dialed in on a small tank first, then transferred to a large one — from milliliters to thousands of liters.
Class
How it works
Companies — differentiator — money
Stage and scale
1. Lab robots liquid handling
Robots pipette, move tubes, and run reactions on their own — more precisely than a human and without tiring, taking the manual drudgery off the scientist.
Opentrons — affordable benchtop pipetting robots (Flex, OT-2); $200M Series C · Automata — links instruments into a single automated line (LINQ); $45M Series C (2026), with Danaher among its investors
Working products with customers, but still on venture money: Opentrons leads the mass market; Automata is at the late integration stage.
2. Cloud labs cloud labs
A lab hundreds of miles away: you mail in samples and define the experiment in code; robots on site run it and send back the data.
Emerald Cloud Lab — a remote lab fully driven by code, 200+ instrument models; private (Founders Fund among its investors) · Strateos — a robotic lab-as-a-service for drug discovery; acquired by Daiichi Sankyo
Technically proven but still a niche: ECL is operating, while Strateos never scaled on its own and was folded into pharma.
3. Bioreactors for rent bioreactors-as-a-service
Ready-made bioreactors rented over the internet: cells are grown under specified conditions, with runs launched and monitored from a browser — no equipment to buy.
Culture Biosciences — cloud bioreactors (250 mL and 5 L), launched and monitored from a browser; $80M Series B, >$100M total
An emerging category with a single clear leader; the service is real and well funded.
4. Organism foundries organism foundry
An automated factory designs, builds, and tests cells and microbes to order — cycling through thousands of genetic variants and keeping the ones that work.
Ginkgo Bioworks — a custom cell and microbe foundry (Foundry + Codebase); public (NYSE: DNA), ~$174M revenue (2024)
The most mature by stage (publicly listed), but under financial pressure: ~$174M revenue with nearly zero growth.
Antibodies are getting cheaper
Conventional antibody manufacturing runs $180–500 per gram. The continuous platform from Enzene (EnzeneX) pushes the cost below $40 per gram.
So is cell therapy
A standard CAR-T (a cancer therapy made from the patient's own cells) costs ~$95K to manufacture and sells in the US for $373–475K. Manufacturing right at the clinic cuts the cost to ~$35K.
This is the infrastructure of longevity and biomedicine: these companies don't treat aging themselves, but they make target discovery, experiments, and manufacturing markedly faster, more reproducible, and cheaper.
The cheapest experiment is one somebody has already run: its result is sitting on someone's drive. But "data available on request" gets lost over time, and negative results almost never get published. Data infrastructure solves exactly this problem.
− 17% / year
the odds that a dataset still exists fall by this much for every year after publication (Vines 2014)
~ 3×
positive results are published roughly three times as often as negative ones — the literature is skewed
$412M
raised by Benchling — the cloud where labs keep all their data; valuation reached ~$6B
What it is
First Approval is a platform where biological data is posted, annotated, and exchanged. Its bet is on openness, reuse, and the "failed" results that usually stay hidden.
data + annotationpermanent linknegative resultsreplicationsexchange · sale · donationcheap entry
Why it matters for target discovery
1Removes the skew toward only "pretty" results: negative and replication data come into view as well.
2Standardized annotation makes small datasets comparable, which means they can be aggregated and reused.
3Lowers the cost of entry: it gets easier to collect many cheap observations and discard weak hypotheses before expensive preclinical studies.
4Training without moving the data (Owkin, ~$334M, backed by Sanofi): AI is trained inside the hospitals themselves — the data stays put, and only the trained model travels between nodes.
5This is still early infrastructure: data and connections first, targets and molecules later. A finished pharmaceutical machine is a long way off.
The key idea
If the aging problem looks like "few targets in view, too little data, and poor comparability," then First Approval removes exactly that bottleneck. Its logic is to make biological data cheaper, broader, and fit for reanalysis first — inventing a drug comes later.
Note: this is distinct from the regulatory "first approval" of Nir Barzilai and TAME. This is data infrastructure that makes early target discovery cheaper; FDA approval plays no part here.
·The point: find the signal earlier and select patients more precisely by pairing the test with the drug.
For pharma giants, target discovery gets cheaper through a data flywheel — one good theory is not enough. For smaller players, open data, cheap experiments, and reusable annotation stand in for that flywheel.
Without a general theory, we work only on the consequences
Theory
If leprosy had been studied the way we study aging, no cure would ever have been found
Imagine a disease — leprosy, say — that afflicts nearly every human and animal, but whose source cannot be seen. The field would earnestly set about describing damaged tissues, compiling lists of hallmarks, counting “clocks,” swapping organs, and testing combinations. All of this can yield data and even partial benefit, but the central question stays open: what sets the process in motion?
Leprosythe microbe never found
Hallmarks of leprosya list of features
Leprosy clockscounting the “clocks”
Replacing tissue leprosy destroyedswapping organs
An intervention against the rotfighting a consequence
The alternative: find the pathogenMycobacterium leprae
There is plenty of activity, but the true cause has not been found. A theory of aging is the working alternative to today’s approaches.
Theory
No theory explains the lifespan differences between species
Compare closely related species: lifespans diverge tenfold, sometimes a hundredfold. In bees, ants, and termites they differ just as sharply even on the very same genome. The biochemistry is nearly the same, yet aging runs at different speeds. That difference is exactly what no theory can yet predict.
Even renouncing theory means, in practice, working from an implicit, unstated model. The moment a researcher picks a target, an experimental design, a success criterion, or a funding direction, they are already acting on some picture of how aging works.
visible decisions
what counts as a target · which data to collect · where to invest resources
the hidden model of aging
Theories abound, but most stay at the level of intuition — and intuitions cannot be honestly compared, strengthened, or refuted. That, not their number, is the problem.
A theory of aging is not meant to replace drug discovery, delivery, tissue replacement, reprogramming, or combinatorial approaches. It is there to tell us when those efforts act on the cause and when they merely improve individual symptoms.
Theory
Why is there no unified theory of aging?
In mature sciences, rival theories are pitted against each other and tested until one is left standing. Aging works differently: back in 1990, Zhores Medvedev counted more than 300 theories, and the count has only grown since. Many of them flatly rule each other out, yet the field goes on living with all of them at once.
Dispute 1 · what comes first
Damage or signaling?
Mikhail Blagosklonny: “Aging is caused not by the accumulation of molecular damage but by the inappropriate activation of signaling pathways such as mTOR.”
Tom Kirkwood and Simon Melov: “Compelling arguments reject the idea that death is generally programmed by genes for aging.”
Vladimir Skulachev: the contrary — aging is a program, and therefore it “can be slowed, prevented, and perhaps even reversed.”
What surveys of the field itself show
The oddity is not even the number of theories — the field cannot agree on its own foundations. A survey of 71 specialists — Cohen et al., 2020 — was titled an “agreement to disagree”: “there is no clear consensus even on the most basic questions.” A repeat survey — Gladyshev et al., 2024 — found that “no question reached a majority — not even the question of whether consensus is needed.” For a natural science this is an anomaly: the field has no shared paradigm — a rarity in the natural sciences.
Why no one takes on the big question: the structure of grants
Grants and PhD defenses reward narrow, safe, incremental results. Nicholson and Ioannidis (Nature, 2012) showed that funders systematically pay for conformist science at the expense of risky new ideas. The big question of aging does not fit that frame.
The price of delay — a concrete example
Take metformin. The drug has been known since the late 1950s, yet the basic question — does it slow aging — is still unanswered: since 2015 the TAME trial has been unable to raise the funding to reach completion.
What the survey’s authors propose
The survey’s authors do not demand a unified theory — by their own data, the field is split even on whether consensus is needed. Their recommendation is more concrete: every paper should state explicitly which process it calls “aging” and give that definition right in the text — so that researchers studying different processes stop labeling them with a single word. In parallel, they call for shifting the focus to aging itself, not only its consequences — and they note that merely spelling out the disagreement can point to new experiments.
Nothing is more practical than a theory that works
Here is a theory bidding to be the unified framework for aging. Gero (Singapore), together with Peter Fedichev, builds on a result from statistical physics: over time, a complex system is left with just a few slow variables that govern its dynamics — the handful of quantities that change most slowly. For aging there are three — z₀, Z, and D₀ — and, per the model, they determine what a drug can achieve. On this theory Gero built a drug-target discovery technology that has already led to research deals with Pfizer and Chugai (Roche Group) (up to ~$250M in milestone payments), while Gero itself uses it to develop its own drugs.
Original figure from the Gero paper · the energy landscape
The ball is the state of the organism. Well depth = resilience z₀. Accumulated damage Z lifts the ball and lowers the barrier on the right; noise D₀ rocks the ball until, sooner or later, it tips over the edge — into chronic disease and death.
z₀ · stress response (resilience)
The fast variable: how quickly the body returns to normal after a shock. With age, recovery slows — and that slowing is the loss of resilience.
Z · entropic damage
Irreversibly accumulated damage (configurational entropy). It grows linearly with age and gradually wipes out the resilience reserve z₀. This is the second law of thermodynamics at work inside the body.
D₀ · regulatory noise
The amplitude of random physiological fluctuations. It is the noise that pushes the organism over the death threshold before the damage alone would reach the limit.
Humans are a stable species: we hold steady until linearly accumulating damage Z carries us to a final limit of ~120 years (Pyrkov & Fedichev, 2021). Short-lived models (mice, flies) are unstable from birth — which is why they overstate how durable a drug’s effect will be in humans.
Three tiers of drugs, by the variable they act on
Tier 1 · tune z₀
Targets = the hallmarks of aging: senolytics, fasting mimetics, reprogramming. The effect is moderate — the model puts it at ~10 years — and mostly lasts only while treatment continues.
Per the model, this is the only way to move the maximum lifespan itself (by ≳40 years and more). Not yet technologically within reach: removing irreversible damage is very hard.
Whole areas remain outside the reach of the dominant theories. We work in them as a hedge, in case today's theories of aging fail.
Blind spots
Levin: aging is the loss of the goal a collective of cells holds
Michael Levin and colleagues (Advanced Science, 2025) shift the level at which aging is described. In their view, cells are competent agents: together they solve one problem — what shape to maintain. Evolution selects hard for cells to finish building the body and reach reproduction, but barely selects for holding the shape after that. So once development is complete, there is no anatomical goal left — and the collective drifts off the shape. A computer model shows the key result: aging emerges on its own the moment the body is finished — even with no damage or noise added. Molecular damage merely speeds up the drift.
Levin's computer model (Advanced Science 2025) · aging appears as soon as the shape is complete
What the model shows: each cell is a tiny neural network; together the cells build out a target shape (neural cellular automata trained by neuroevolution). Once the shape is complete and there is nothing left to maintain, it starts to drift — aging emerges on its own, with no damage and no noise. But the “blueprint” of a lost organ stays written in the tissue: a precise signal wakes that memory and rebuilds the organ.
Evolution selects for cells finishing the body and reaching reproduction. Holding the shape after that is barely selected for. With the goal fulfilled, the collective has nothing left to strive for, and the shape drifts. In the model, that is aging.
Differentiation errors, loss of cell “competency” (their ability to solve the shape problem), communication failures, and genome damage all speed up the drift. But in the model, aging begins even without them — on its own, once development is complete.
Even after an organ is lost, its “blueprint” stays written in the tissue. A precise regenerative signal wakes that memory and rebuilds the organ — most powerfully when it reaches both the damaged cells and the neighboring tissue.
How this changes the thinking
The treatment target is the goal itself, held by the collective of cells; it is governed by an upper layer — bioelectricity, where the anatomical goal is stored. Restore the right goal, and the tissue rebuilds the young version on its own. That this goal is real and rewritable is visible in the planarian: resetting the bioelectric network rewrites the shape, and two heads are inherited with no change to the genome. The same lever is already assembling new bodies out of cells — xenobots made of frog cells and anthrobots made of human cells move and build copies of themselves, genome unchanged. Regeneration and rejuvenation then become one and the same task: restore the goal.
Transposons: a mechanism of antagonistic pleiotropy?
What happens
LINE-1 (“jumping” genes) become derepressed with age, switching on cGAS-STING and inflammaging — the same antagonistic pleiotropy seen in the genetics of diabetes.
How it is tested
Reverse-transcriptase inhibitors (censavudine, lamivudine) have passed their first human test on the logic of “block it, lower the inflammation.”
Who is doing it
Transposon Therapeutics; the Gorbunova and Sedivy research groups. A cheap precursor: lamivudine (3TC).
Source: De Cecco M, Ito T, Petrashen AP, et al. L1 drives IFN in senescent cells and ageing. Nature. 2019;566(7742):73–78. doi:10.1038/s41586-018-0784-9.
Blind spots
The hallmarks are a catalog of features, not a causal arrow
The hallmarks are a systematized catalog of features. The upside: a shared language for the field and a handy checklist that organizes data and interventions.
The downside: it is a description of correlated features with no explanation — a catalog without a causal arrow. It does not say which feature is primary and which is merely a consequence (critique: Gems & de Magalhães, Ageing Res. Rev., 2021).
A list of features helps the field talk about the problem, but by itself it does not explain the problem's causal architecture. The hallmarks are useful for organizing observations and agreeing on which levels of biology are under discussion. But if the list is mistaken for an explanation, the field starts treating consequences as independent causes.
Theory is what draws the line between the primary mechanism, secondary changes, and technical intervention points. Without that line, even good experiments add up poorly into a strategy.
DNA damagesenescenceinflammationmitochondriaepigeneticsstem cellscausal model
Blind spots
Hallmarks: a map of features, not a theory of causes
The list of aging hallmarks gives the field a vocabulary, but by itself it does not say which feature is a primary mechanism, which is a secondary consequence, and which is merely a convenient technical intervention point.
STEP 1
Theory · explains
states the primary mechanism: selection, damage, information loss, systemic instability.
STEP 2
Predictions · test
specify which changes should come first and which should be shared across different models.
STEP 3
Hallmarks · classify
organize the observed features: genomic instability, senescence, inflammation, mitochondria, and so on.
STEP 4
Interventions · act
choose what to test: eliminate cells, retune signaling, restore delivery, alter metabolism.
Good use of the hallmarks
The list helps the field agree on levels of biology, choose measurements, and assemble a protocol where several features are tested within one logic.
Bad use of the hallmarks
If the list of features is mistaken for an explanation, the field starts treating consequences as independent causes and loses track of which combinations should add up.
The resulting structure: the hallmarks provide the vocabulary; theory sets the causal arrow; experiments test which arrows actually drive aging.
Sources: Hallmarks 2013 · Hallmarks 2023 · Hallmarks as framework 2024 · Causality of hallmarks 2026
Blind spots
The search for a key mechanism
With no unified theory yet, the working alternative is to hunt for a key mechanism of aging. Below are the leading candidates for the root cause; the “weight” of each mechanism remains an open question.
Root-cause candidate
What happens
Authors · source
Protein cross-links
Irreversible cross-links (glucosepane and other AGEs) and elastin fragmentation stiffen collagen and elastin. The damage accumulates on its own, with no signaling-pathway failure — hence “the missing hallmark.”
Capillary rarefaction, arterial stiffening, and endothelial dysfunction degrade oxygen delivery to every tissue at once. A mechanism shared by all organs.
Which of these is primary and which is secondary is still unknown.
Blind spots
A key mechanism illustrated: vascular aging
Take one candidate from the previous slide — vascular aging — and look at who is working on it in practice. From acellular vascular grafts to regenerating the body's own tissue and restoring the endothelium.
An acellular bioengineered vessel (ATEV / Symvess): human extracellular matrix grown from aortic smooth-muscle cells, then decellularized; after implantation it is repopulated by the patient's own cells.
TRUE AVC — a 100% biological, acellular vascular conduit on the TRUE Tissue platform, grown in vitro in a bioreactor and remodeled by host cells after implantation.
Engineered endothelial cells (E-CEL, E4ORF1+): restoring the vascular niches of organs via angiocrine (autocrine-juxtacrine) endothelial signals; lead drug AB-205.
Even the single mechanism of “vascular repair” splits into a spectrum of approaches: grow a replacement vessel, trigger regeneration of the body's own tissue, or restore the endothelium. Angiocrine's negative Phase 3 result is a reminder that the field's maturity is still taking shape.
H. pylori gastritis: a familiar disease turned out to have one specific cause
A strong theory often looks implausible at first precisely because it is so specific.
For decades, chronic gastritis was blamed on stress, diet, acidity, and lifestyle — the condition seemed too ordinary to have one specific cause. Barry Marshall and Robin Warren showed otherwise: the main cause of chronic gastritis (and, in some people, of stomach ulcers) is the bacterium Helicobacter pylori. To convince the skeptics, Marshall drank a culture of it in 1984 and gave himself gastritis; the discovery earned them the 2005 Nobel Prize.
For aging, this is an important analogy. When a new theory explains a general process through one specific mechanism, the first reaction is almost inevitably emotional: “it can't possibly be that simple.” But that very feeling is not an argument against the theory.
We may not even know where to look. “Science allows for large blind spots.” What if we are using the wrong functional unit and the wrong logic of regulation? What does science barely see?
Bioelectricity
Morphogenetic space
Network theory
Aging as a disease of regulation
07Replacing
our projectBioconstructor
What if aging can't be slowed?
If aging itself can't be altered, what remains is building with biology: replace worn-out tissues and organs, restore lost functions.
Replacing
The Replacing companies
“Replacing” means swapping out cells, tissues, organs, and critical transplant infrastructure — a map of who is building the alternative to “unfixable” aging today.
Company
What it does in “Replacing”
Total raised
Stage / maturity
eGenesis
Genetically modified pig organs for xenotransplantation. The focus: making kidneys, islet cells, and other organs compatible with humans.
$416M (B)+(C)+(D)
Preclinical / preparing for first-in-human use: in 2024 the company said it was advancing its lead kidney program toward a first human study.
BlueRock Therapeutics
iPSC-based cell therapies that replace lost cells. Most visibly: replacing dopaminergic neurons in Parkinson's disease and replacing retinal cells.
$225M (Series A)
Clinical stage: bemdaneprocel has advanced into registrational Phase III for Parkinson's disease; OpCT-001 is in clinical trials in ophthalmology.
bit.bio
Programmable human cells on demand — the foundation both for cell replacement and for scalable manufacturing of standardized cells for research and future therapies.
≈$225M
A commercial platform and products for research and drug discovery; the therapeutics arm is growing but not yet clinical.
iTolerance
Makes “Replacing” more realistic rather than replacing organs itself: local immune tolerance for cells, organoids, and tissues without lifelong immunosuppression.
≈$19.0M ($17.1M)+($1.0M)+($0.85M)
Preclinical / pre-IND: programs for stem-cell-derived islet cells and for combination with organoid/tissue therapy.
OrganOx
Key infrastructure for “Replacing”: normothermic machine perfusion keeps donor organs viable longer and assesses them better, raising the number of usable transplants.
$160M (2025)
Commercial stage: an FDA-approved liver perfusion platform, with more than 5,000 liver transplants to date.
TISSIUM
A biopolymer platform for tissue reconstruction: nerve repair, hernia repair, cardiovascular sealants. This is true “repair and restore” for tissues — deeper than treating symptoms.
€50M (Series C)
Entering the market: in 2025, FDA De Novo authorization for COAPTIUM® CONNECT in peripheral nerve repair.
“Replacing” is a whole stack of solutions: a cell source, protection against immune rejection, tissue engineering, organ preservation, and in some cases direct fabrication of the replacement biomaterial. No single technology is enough.
The replacement spectrum: from organs down to cell state
Here “Replacing” is a family of strategies: replace an organ, grow a tissue, engraft new cells, restore mitochondrial function, or return a cell to a younger state. No single method covers it all.
Replacement level
How it works
Companies / money
What actually gets replaced
Organ xenotransplantation
Pigs are gene-edited: strong antigens are removed, human compatibility genes are added, and the risk from PERV retroviruses is reduced.
eGenesis ($191M, Series D) · United Therapeutics / Revivicor (public company)
kidney, heart, potentially other organs
Tissue bioengineered organ
Functional tissues, vessels, and organoids are grown or printed; sometimes a lymph node serves as a “bioreactor” for a new piece of organ.
Humacyte (public company) · Aspect ($115M, Series B) · LyGenesis ($19M, Series A-2)
vessels, tissue segments, liver / kidney function
Cells replacement
Stem cells are turned into the needed cell type and engrafted in place of cells that are lost or dysfunctional.
BlueRock / Bayer (up to $1B) · Aspen ($147.5M, Series B) · Sana ($588M IPO)
neurons, β cells, cardiomyocytes, immune cells
Mitochondria and bioenergetics
The aim is to restore function through mitochondrial augmentation, transfer of healthy mitochondria, or therapies that change the mitochondrial program.
Minovia ($20M upfront) · Pretzel ($72.5M, Series A)
mitochondrial function, sometimes the organelles themselves
Cell programs and living cell factories
Cells are more than an injected drug: they are programmed to find targets on their own, manufacture therapeutic molecules, or rewire the immune response.
Capstan ($165M launch) · AbbVie is acquiring Capstan for up to $2.1B
immune cell function, cellular programs
Organ bypass replacing the function
Methods that perform an organ's function while bypassing the organ itself: cardiopulmonary bypass (the heart-lung machine) and ECMO for the heart and lungs, hemodialysis for the kidneys, blood oxygenation; parabiosis — exchanging the circulating environment between organisms — sits right alongside.
CARMAT and Abiomed (artificial hearts and pumps), plus ECMO and hemodialysis machines, replace the function of the heart, lungs, and kidneys externally, bypassing the organ itself. Altos Labs ($3B) and NewLimit ($130M, Series B) work at a different level: reprogramming cells back to a young state.
The function of an organ or system
The extreme end — head / body transplantation — isn't shown as an industry: it is an ethical and technical frontier, and no mature field with real clinical companies exists beyond it yet.
How much can we replace — and how much do we actually need to?
The maximum: almost everything is already being replaced
·Heart: conventional transplants have long been routine; in 2022, David Bennett lived about two months with a genetically modified pig heart. It was the first xeno-replacement of a vital organ in a human.
·Kidney: in 2025, Tim Andrews lived 271 days with a genetically modified pig kidney. That same year, the FDA cleared the first clinical trials of kidney xenotransplantation.
·Face: the first full face transplant was performed in Barcelona in 2010.
·Uterus: in 2014, the first baby was born in Sweden following a uterus transplant.
·Cornea, skin, bone marrow, heart valves: today this is routine clinical practice, long past the realm of futurology.
The minimum: sometimes one node is enough
·Thymus: in the TRIIM study (Fahy et al., Aging Cell, 2019), thymus regeneration in 9 men was accompanied by improved immune markers and an epigenetic-age shift of roughly 2.5 years against the expected march of time.
·The systemic blood environment: in heterochronic parabiosis (Conboy et al., Nature, 2005), a young circulating environment partially restored regeneration of aged muscle and liver in mice.
·Sometimes replacing one key control system is enough — the thymus, say, or the composition of circulating factors — with no need to replace a whole organ.
The record for organs replaced at once: Andy Voge (Cleveland Clinic, 2021) — a multivisceral block of 5–6 organs in a single 17-hour operation; triple heart-liver-kidney transplants: McPharlin and Smith (2018), and Sams — one of ~46 in the US since 1987. The brain is never among them. In practice, we can already replace nearly any organ or tissue, but there is no full brain replacement today. So the boundary of this approach runs not only through how many parts can be swapped, but through whether aging can be shifted from one systemic control point. Sources:UMaryland — heart · eGenesis / MGH — kidney · NKF — clinical trials · Barret et al., face · Brännström et al., uterus · Fahy et al., TRIIM · Conboy et al., parabiosis
Replacing
Transplanted young tissue already takes hold and restores function
In experiments, embryonic and stem-cell-derived neural tissue has already engrafted, wired into the cortex, and restored movement and behavior in animals. Replacing a piece of the brain is already a reproducible laboratory result.
Hébert, 2023: embryonic cortical tissue filled a removed patch of adult mouse cortex, matured into layered neurons, and grew out connections. The lab of Jean Hébert.
Pașca, Nature 2022: a human cortical organoid grew into the rat cortex, formed working connections, and began driving the animal's behavior.
Shinoyama et al., 2013: neurons derived from embryonic stem cells restored motor function after brain ischemia — measurably, on rotarod and beam tests.
Martínez-Curiel, 2025: human cortical neurons formed synapses in the rat cortex after a stroke and wired into the host network.
Primates: in a macaque model of Parkinson's disease, iPSC-derived neurons survived in the brain for more than 2 years and restored movement (Kyoto, 2017; no figure shown — the paper is under copyright). In humans, fetal dopamine neurons engraft and function (Lindvall, Science 1990).
NeoGraft: replacing a patch of brain cortex with young tissue
The idea: restore lost brain function by transplanting young (embryonic / fetal-like) tissue into a damaged patch of cortex; the graft takes hold, integrates into the host's neural network, and takes over the function of the removed patch. In the product version, the graft is grown from the patient's own genetic material — structurally identical to the patient's own tissue. The project is affiliated with Jean Hébert (Replacing Aging).
The core
1Name
Replacing lost cortical tissue with a young graft
NeoGraft transplants young cortical tissue, together with its pia mater membrane, into a damaged patch of cortex, so that the graft structurally integrates into the host brain and restores the lost function.
2Concept
The graft takes over the function of the removed patch of cortex
1Aspirate the patch→2Transplant the graft→3Retrain→4Verify
With targeted aspiration we remove the patch of motor cortex that controls three fingers of the right hand; we temporarily immobilize the limb, transplant an embryonic graft, and support its engraftment. Motor training then shifts the function onto the graft.
3Novelty
We prove cause and effect by reversibly switching off the graft's activity
For the first time, we prove that the graft itself carries the restored function: we reversibly suppress its neurons' activity chemogenetically (DREADD — an engineered receptor in the neurons that responds only to a designer drug); movement stops, and returns once suppression is lifted. This separates the graft's contribution from the brain's own plasticity.
Technology and plan
4What already exists
Every step has already been demonstrated in studies
Unsolved: scaling to the primate neocortex and graft size
The full chain has yet to be shown in the primate neocortex: aspirate a patch → transplant an embryonic graft → restore function → prove cause and effect by reversibly switching off the graft's activity. Vascularization and full graft maturation also remain unsolved. Distance to product: preclinical studies in primates.
6What it takes
The plan to a reproducible result
A primate model (targeted aspiration of a cortical patch + limb immobilization); an embryonic cortical graft with pia mater; a motor retraining protocol; a DREADD construct for reversibly switching off the graft's activity; longitudinal imaging over time and behavioral testing. The key parallel task: the technology to grow tissue from the recipient's genetic material, structurally identical to their own.
Economics · estimates and market data
7How much money
$20–40M
For neurosurgical preclinical studies in primates
Estimate · driversMore expensive than REcell's cell-therapy preclinical program ($10–30M): neurosurgery on every primate, animals in short supply, and years of behavioral follow-up and imaging. ARPA-H BRAINS provides $0.5–1M per team as seed funding.
8Odds of success
The biology is proven; the risk is the leap to primate cortex
Engraftment, integration, and functional recovery have been shown at the level of cells and tissues (see “What already exists”). The main risks: scale-up to the primate neocortex, the completeness of graft integration, and the lack of an established registration path for “brain tissue replacement.”
9Spillover value
Even a partial result leaves valuable assets
HighA graft-engraftment platform; DREADD-controlled therapeutic cells whose activity can be reversibly suppressed; a base of autologous cortical tissue. At the maximum: the technology to grow tissue identical to the recipient from their genetic material — from tissue all the way to cloned organs and a whole organism. Applicable to stroke, traumatic brain injury, and neurodegeneration.
10Price of therapy
$2–5M
One-time; the first generation sits at the top of the price range
Estimate · driversMore expensive than a cell infusion: instead of a cell suspension, a whole structured tissue is grown for the recipient from their genetic material (personalized GMP manufacturing), plus complex neurosurgery and rehabilitation. The benchmark: one-time autologous therapies at $2.1M–$4.25M; NeoGraft lands at the top of that range and above. As the process matures, the price falls.
11Market
Millions of people without a single regenerative therapy
ScenarioThe ceiling is set by throughput, not demand (>20M patients): tissue manufacturing for each recipient, plus neurosurgery. At a $3M price and a realistic ~3,000 surgeries a year, that is ~$9B/yr. US regulatory exclusivity for biologics is 12 years. It all depends on manufacturing scale.
Suppress the graft's activity and movement disappears — the graft carries the function
The flagship experiment of our primate preclinical program is designed to prove causality directly: reversibly suppress the graft's activity and the restored finger movement should vanish; lift the suppression and it should return. This separates the graft's contribution from the brain's own plasticity and delivers the reproducible result that Jean Hébert and grant programs are waiting for.
1Step 1
Aspirate a patch of cortex
We precisely aspirate the zone of motor cortex that drives three fingers of the right hand, producing a reproducible, measurable neurological deficit.
2Step 2
Immobilize the limb
We temporarily immobilize the limb so that compensatory plasticity in neighboring cortex doesn't capture the lost function before the graft takes hold.
3Step 3
Transplant the graft
Into the defect we transplant a young embryonic cortical graft together with the pia mater membrane that preserves its structure.
4Step 4
Engraftment
The graft vascularizes, matures, and grows connections into the host cortex; we allot a fixed period for this, monitored by imaging.
5Step 5
Motor retraining
We remove the immobilization and run motor training: the workload helps the graft take over the function and integrate into the cortex.
6Step 6
Verify by suppression
We reversibly suppress the graft's activity chemogenetically (DREADD): finger movement stops — so the graft carries the function; when suppression is lifted, movement returns.
Money is already earmarked for this — ARPA-H BRAINS. The program's TA2 track — engrafting fetal progenitor tissue into the adult brain — is exactly our experiment. $0.5–1M per team per technical track, 18 months. The next application window opens January 7, 2027 (sam.gov). The neighboring program FRONT (neocortex, no fetal tissue) has already gone to a performer — our door is BRAINS.
What we build in parallel. (1) Integrating the graft into host tissue. (2) Growing tissue from the patient's own genetic material — identical and immunocompatible. The maturity ladder, honestly: cells — already in the clinic (the patient's own iPSC cells (NEJM 2020)) → organoids — in the lab, stuck on vascularization and maturity → organs — experimental → a cloned organism — doesn't exist yet and is banned by law.
A field can have strong theories and good experiments, but without a funding architecture they never grow into large-scale research programs.
Resources
Who pays for life-extension technology
NIA Budget
1. Government funding
The NIA funds research on aging, Alzheimer's disease, and related areas. The scale is large, but it is locked into government rules of priorities, grant competitions, and institutes.
Calico · Calico–AbbVie $1.5B
2. Corporate capital
Google/Alphabet placed its bet through Calico; Calico and AbbVie then set up a separate R&D collaboration on age-related diseases, with joint investment of up to $1.5B on the table.
Hevolution · XPRIZE Healthspan
3. Philanthropy and prizes
Hevolution has announced an annual budget of up to $1B, and XPRIZE Healthspan offers a $101M prize for clinically measurable improvement of function in aging.
Kizoo · Juvenescence
4. Angels and super angels
Angels are private investors who put their own money into the earliest, riskiest longevity startups, often writing the first check. Super angels do this systematically and at scale: Michael Greve, through Kizoo, puts $1–10M into Seed and Series A rounds in rejuvenation biotech, and Jim Mellon has built an ecosystem of anti-aging companies around Juvenescence.
Khosla · Longevity Fund · Kleiner Perkins
5. Venture funds
Funds pool money from institutional investors and place professional bets on longevity biotech at the early rounds. Khosla Ventures backed both Loyal (longevity drugs for dogs) and NewLimit (cellular reprogramming). Loyal's first checks came from Laura Deming's The Longevity Fund — the first venture firm devoted entirely to longevity; NewLimit's $130M Series B was led by Kleiner Perkins together with Founders Fund.
Even the largest funds are small against the scale of the problem: Hevolution promises up to $1B a year but in its first 20 months actually disbursed ~$400M in grants, and XPRIZE Healthspan is a $101M prize stretched over 7 years. The rest is held back by a moral argument: Bill Gates called funding life extension "pretty egocentric" while malaria and tuberculosis remain unbeaten — so effective-altruism money flows into global health, not longevity. All the world's venture investment in longevity for 2024 came to $8.5B, about 0.17% of annual US healthcare spending.
The invisible barrier · reputation
It is not only the science that slows the money. The industry's image works against it: the field gets mistaken for fraud, and pop culture has handed the very dream of living longer to the villains.
The myth: "it's all scams and magic pills"
High-profile scams feed this image. The startup Ambrosia sold "young blood" transfusions at $8,000 each until the FDA warned in 2019 that the procedure has no proven benefit and that patients are being "preyed upon by unscrupulous actors." Stories like these leave many people doubting that longevity holds any real science at all, rather than elixir-of-youth salesmanship.
In pop culture, only villains chase immortality
Mass culture codes victory over death as vice and hubris: Voldemort murders to split his soul into Horcruxes for the sake of immortality, and Dorian Gray pays for eternal youth with the ruin of his soul. There is even a dedicated trope — "craving immortality = evil": a sympathetic hero almost never wants it. So the wish to live longer looks suspect from the start.
Resources
Aging needs bolder money
Resources
A vicious circle. A weak hypothesis
Aging is not recognized as a disease → so there are no specialist physicians, no approval pathway, no money, no biomarkers → so there is no evidence it can be treated → so, once again, it is not recognized as a disease. The circle closes on itself.
5Long, costly trials. No approved aging biomarkers exist to serve as endpoints — that is exactly what the TAME trial is meant to prove.
6No evidence of efficacy. There is nothing to prove that aging is treatable — so it stays "not a disease." The circle is closed.
Where the circle breaks: an approved aging biomarker as a surrogate endpoint makes the effect measurable — and therefore approvable. Then come the data → an indication → money → physicians.
Resources
No familiar explanation of the underfunding survives scrutiny
Ask why so little money goes into the fight against aging and you get a set of ready-made answers. Each points to a real difficulty, which is why each sounds convincing — yet against each stands one concrete fact that refutes it. If none of the familiar explanations passes the test, the true cause has not yet been named.
The usual explanation
The fact that refutes it
"The field simply has too little money"
Altos Labs raised ~$3B in a single round — the largest launch in biotech history — and has yet to publish a breakthrough. What holds the field in place are structural barriers that even capital of that size runs into.
"The horizon is too long — investors won't wait 15–20 years"
Fusion company Commonwealth Fusion raised $1.8B for a technology with the same payback horizon. A distant payoff, on its own, does not scare capital away.
The review "The Hallmarks of Aging" (López-Otín et al., Cell, 2013) became one of the most-cited papers in modern biology and set the research agenda for hundreds of thousands of studies. The biology of aging is a peer-reviewed discipline in the leading journals, not esoterica.
"There's nothing to treat aging with — no targets, no drugs"
Senolytics (dasatinib + quercetin) improved physical function in pulmonary fibrosis patients in their first clinical trial (2019), and the TAME trial is testing metformin against aging. The targets — and a way to measure the effect — already exist.
"No need: ordinary medicine is beating the diseases one at a time anyway"
Completely curing cancer or heart disease would add only ~2–3 years to a person's life, whereas slowing aging itself yields more healthy years and ~$7.1T in gains over 50 years (Goldman et al., Health Affairs, 2013). Treating diseases one by one is a strategy with a built-in ceiling.
"People don't want to live longer — there's no demand"
The world spends $6.8T a year on wellness (Global Wellness Institute, 2024) — four times the entire pharmaceutical industry. Demand for health and youth is enormous; it just leaks into supplements and trackers instead of therapies that target the causes of aging.
"There's no one to do the work — the field lacks scientists of the right caliber"
Researchers of the highest rank have come in: Nobel laureate Shinya Yamanaka is a scientific advisor to Altos Labs. Talent goes where the scale of the problem and the funding are — not the other way around.
"Aging is not a disease, so there is no approval pathway"
That is a link in the same circle, not its root: an approved surrogate biomarker makes the regulatory status fixable (see the previous slide). A consequence of the trap, not its cause.
The takeaway: each explanation points to a real difficulty, but none accounts for the scale of the underfunding — each is matched by a counterexample of equal force. If neither capital, nor timelines, nor the maturity of the science, nor demand, nor talent, nor regulation holds the field back on its own, the limiter sits one level deeper — in the very structure that ties these barriers into a self-sustaining circle. That is where we go next.
Sources: Altos Labs / C&EN · Commonwealth Fusion / CFS · Hallmarks of Aging / Cell · Senolytics IPF trial / EBioMedicine · Longevity Dividend / Health Affairs · Global Wellness Institute · Shinya Yamanaka / Altos · TAME / AFAR · FDA regulatory hurdles / NatLawReview
Resources
The dampening factor
Aging is a problem no institution is shaped for. To academia, venture capital, and the corporation it looks orthogonal to everything they do.
Big capital → a cautious bet
Academia, venture, corporations: billions and years — zero approved drugs
Works to make defeating aging a political priority.
The dampening factor. The bigger the capital, the denser the institutional filters it must pass through — and the more cautious the bet. Breakthroughs come from where courage outruns money. Tellingly, de Grey himself called Calico his "greatest disappointment" — for betting on studying aging instead of repairing it.
Sources: Calico–AbbVie / Pharmaphorum · AbbVie winds down the partnership / STAT · Altos Labs / C&EN · Altos and clinical trials / Longevity.Technology · NewLimit results / Longevity.Technology · NewLimit Series C / STAT · Robust Mouse Rejuvenation / LEVF · Open Longevity · Longevity Biotech Fellowship · Vitalism Foundation · de Grey on Calico / Fight Aging
Dark experts
The resistance comes from: conservative academic tradition · the political left's agenda · religious conviction.
Resources
What steals the will — and the money — from life extension
Money
$6.8T
Wellness swaps the problem for a market of habits
The Global Wellness Institute values the wellness economy at $6.8T for 2024 and projects nearly $9.8T by 2029. Under the "longevity" banner, attention drifts into supplements, trackers, and habits — instead of therapies that strike at the causes of aging.
Shame
2015
The moral argument makes aging "the wrong priority"
Bill Gates's answer in his Ask Me Anything. The argument runs: while malaria and tuberculosis exist, life extension for the rich looks like selfishness. It shifts the conversation from a medical question to whether the goal itself is permissible.
The norm
Culture's patterns have no stake in longer lives
Cultural institutions — the postal service, football, show business, supermarkets, pension funds — have no interest of their own in people living longer, and they impose their interests on people. Well-organized systems reduce a person to a functionary who neither recognizes nor defends his own interests.
Global Wellness Institute · Gates AMA transcript · de Grey, Rejuvenation Research
Resources
Why do people argue for death?
the "selfish rich" argument
01
Bill Gates
In his 2015 Ask Me Anything (AMA), he called funding life extension "pretty egocentric" while malaria and tuberculosis remain. This is a dispute over the right to pursue the goal at all.
the fear of hubris
02
Richard Dawkins
On Big Think he calls the fight against aging presumptuous and warns of overpopulation. A scientific question is swapped for anxiety about the consequences of winning.
the "dignified limit" narrative
03
Ezekiel Emanuel
In The Atlantic (2014) he wrote that he hopes to live to 75. Life extension is cast here as a refusal to accept a "natural" limit; treating the cause of aging never comes up.
the bioconservative line
04
Kass and Fukuyama
Leon Kass defends a life lived "in the rhythm of time"; Fukuyama called transhumanism a dangerous idea. This is the fear of losing the familiar image of the human being; no specific therapy is ever examined.
Gates AMA · Big Think / Dawkins · The Atlantic / Emanuel · The New Atlantis / Kass · Foreign Policy / Fukuyama
We will not move an inch until we defend the idea of staying alive.
If you are not acting from first principles, you are building a world that nudges you into mistakes.
Foundations
Can life extension advance without becoming a person's central idea?
Probably not: life extension competes with a person's other supreme values.
Staying alive is required for any goal
Whatever your ultimate goal, you have to stay alive to reach it. Omohundro (2008) showed that self-preservation is a convergent drive of any agent. Even an AI whose sole task is "compute π" resists being switched off — in Bostrom's version it is the paperclip-maximizer AI. On this view, life extension serves all of a person's values rather than competing with them.
Without an overarching idea, you never leave the local optimum
People optimize within their role — the successful lab director with a profitable business, the person selling useless supplements. Locally everything looks fine; globally it is a dead end. March (1991) showed that those who only refine their current competence lose, in the long run, to those who explore the new. An overarching idea supplies the external reference point that pulls you out of the local optimum.
Nothing is more practical than a theory that works
Kurt Lewin's (1943) aphorism: a correct theory saves years of trial and error because it tells you in advance what will work. As a public idea, life extension remains barely worked out.
A science of the fight against aging
This is a question for a new social science, not for the biology of aging: how people think, feel, and act about life extension. Without it, the field cannot scale. A Pew survey (2013): 56% of Americans would refuse treatments to live to 120, even as 68% are sure other people would say yes. By understanding and changing these attitudes, the field grows — and moves on to lobbying.
Lobbying
A study by Alexander, Mazza, and Scholz (2009): corporations earned $220 in tax benefits for every $1 spent on lobbying — a 22,000% return. And lobbying is already changing the rules: Montana (2023, SB 422) made it legal to offer experimental drugs to any patient after Phase I trials. Whoever tunes the laws sets the speed of the entire field.
Attracting geniuses
Breakthroughs are made by exceptional people, and they go where the scale of the problem fires the imagination. A National Academy of Engineering report (2005): framing a problem as a "grand challenge" captures young imaginations and draws the best minds, while a narrow framing loses the fight for them. Life extension must be described so that geniuses from physics, AI, engineering, and biology come to it.
Foundations
How to defend first principles
Philosophy walks right past the thesis "living is good, dying is bad." Life extension has not yet won the intellectual argument.
Style
The field lacks style. How we deliver the message is the message.
Knowledge
There is plenty of data, but it never becomes knowledge. The fight against aging may really be a set of instructions for how to act.
Everyday life
Stronger than philosophy: what you do every day matters more than what you want every day.
Ontology
Perhaps death needs a different language — changing the terms themselves, not the arguments.
Community
Who are "we"?
The canon — what holds the thesis up · 9 books on defeating death
In 12 years, an intellectual club turned a forum into an AI industry worth hundreds of billions
LessWrong (2009) is a forum about how to think without systematic errors. Out of it grew effective altruism, the conclusion that safe AGI matters more than anything, and people with money who founded OpenAI, DeepMind, and Anthropic. It is a working example of an intellectual club.
Yudkowsky writes the fan fiction Harry Potter and the Methods of Rationality—660,000 words. It became a funnel: thousands of readers found their way to rationality and AI safety.
Founded LessWrong, wrote the essay collection Sequences, and founded MIRI (2000)—the first organization devoted to the risks of AGI.
"Eliezer has done more to accelerate AGI than anyone else: he got many of us interested in it, helped DeepMind get funded when AGI was far outside the Overton window, and was critical in the decision to start OpenAI." —Sam Altman, CEO of OpenAI, 2023
Peter Thiel—money and a call for courage
Gave MIRI more than $1.6M and launched the Singularity Summit; an early investor in DeepMind.
"Brilliant thinking is rare, but courage is in even shorter supply than genius." —Peter Thiel, Zero to One
Oxford philosophers William MacAskill and Toby Ord (the term took hold in 2011) teach people to choose, on evidence, where they will do the most good—the mission of 80,000 Hours. Hence the conclusion: safe AI is priority #1.
A club of a few hundred people launched an industry worth hundreds of billions—OpenAI, DeepMind, Anthropic—but reaps none of the harvest: the money and power stayed with the companies, and the community stayed a community.
Foundations
Stress-testing the fundamental claims
We may be wrong at the most fundamental level, and the foundations we stand on deserve to be questioned.
·How do we turn a personal fear of death into effective public action?
·How much does a person who asks "what should I do?" actually need to know?
·What should the entry point be for non-scientists—engineers, designers, organizers, artists, communicators?
·What can one person accomplish, and which actions carry the most leverage?
·Can saving your parents be a stronger motivator than your own survival?
·Which aesthetics make the fight against death feel natural, and which push it to the margins?
·How do we make the topic as culturally contagious as football, entertainment, and celebrity content?
·Which role models actually help the field?
·What record-setting experiment could break the mass disbelief that aging can be defeated?
·If a copy equals the original, what are we obligated to preserve: memory, the body, the stream of subjective experience, the right to continuity?
10Autonomy
our groupSideloading
Act autonomously
If a person does not accept the basic premise—living is good, dying is bad—technical arguments will almost never persuade them. Stop spending serious effort on recruiting new people. Point our focus at solving the problem with our own hands.
This part shows how a single will scales—through protocols, trusted groups, digital agents, and long infrastructure bets.
Autonomy
Autonomy turns agreement into action and infrastructure
Work with the people who already share the premise. The classic mistake is to keep re-arguing the right to go on living with people for whom death is already built into their moral picture of the world. The autonomous move is different: find those who agree with the starting premise, and quickly convert agreement into a project, a routine, and infrastructure.
STEP 1
Filter
who is even worth talking to
STEP 2
Commitment
what the person promises to do
STEP 3
Protocol
how that becomes repeatable actions
STEP 4
Infrastructure
what outlives any one person
The practical test
An autonomous group has its own selection rules, its own language, its own tempo—and the right not to wait for permission from a culture that has already decided to do nothing.
What to cut
Endless debates, symbolic gestures, and public pledges with no enforcement mechanism. They create a feeling of motion but do not change the odds of survival.
Autonomy
How to scale yourself
Reduce dependence on other people's approval; speed up the decision → action → feedback loop.
STEP 1
A personal protocol
Health is run as a managed system: measurements, constraints, repeatable decisions, and regular course corrections—all within the principles of evidence-based medicine.
STEP 2
Multi-agent AI and robots
Digital agents take over search, documentation, coordination, and routine work. This is about growing the compute that works toward a single goal—not about immortality. Source: OpenAI Agents SDK.
STEP 3
A private club
A trusted circle belongs wherever publicity raises resistance. Build the enforcement mechanism first; communicate outward second.
STEP 4
Sideloading
A digital copy of you, created while you are alive, that acts in your interests.
Sources: OpenAI Agents SDK—agents, tools, and orchestration · Nectome—post-mortem human preservation · The Network State—the network as an institutional form
Autonomy
Small autonomous groups that solved enormous problems
Time after time, the big breakthrough came from a small, tight-knit core. Five examples:
Los Alamos
The Manhattan Project's narrow core of physicists built the atomic bomb in a handful of years.
Bletchley Park
A sealed-off team broke Enigma and changed the course of the war.
Fairchild's "Traitorous Eight"
Eight engineers who walked out launched all of Silicon Valley.
The Bourbaki group
A small circle rewrote the language of all of mathematics.
The Bolsheviks, 1917
A small, disciplined organization seized power in a vast empire.
A small core with shared principles, high trust, and a fast decision → action loop outruns big, loose structures.
11Hope for the future
our projectOrgan cryopreservation for transplantation
Count on the future alone
While today's medicine cannot yet cancel death, the task changes: do not vanish before the moment it becomes technically solvable.
Hope for the future
Counting on the future: four bets
If radical life extension is not ready yet, the task is to survive—and to preserve what the future will be able to restore.
Preventive medicine
Lower today's risks and preserve function while new technologies arrive.
Cryonics
Preserve the body or brain at ultra-low temperature until future restoration.
Humanism
Hold on to the value of every life as the ground for action.
Make as much of everything immortal as possible
Civilization has always preserved people—in monuments, books, museums, recordings. Now it can be done deliberately, during life.
"Freeze them—people smarter than you will figure it out." —Usovich
Hope for the future
Prevention is the longevity lever with evidence at the level of clinical trials—available today
Bet 1 is simple: don't die ahead of schedule. While radical therapies aren't ready, people change how they move, eat, and live so they survive to the moment aging becomes solvable. This branch of lifestyle medicine rests on randomized controlled trials (RCTs) and cohorts of hundreds of thousands of people—a standard of evidence most wellness habits lack.
Intervention
How it works
Evidence base
Effect on lifespan
1. Physical activity
Regular exercise builds cardiovascular endurance, insulin sensitivity, and endothelial function—lowering blood pressure, blood sugar, and chronic inflammation.
−31% mortality at the recommended ~150 min of moderate exercise per week; up to −39% at 3–5× that dose.
2. Mediterranean-style diet
Olive oil, nuts, vegetables, fish, and less refined food improve the lipid profile and vascular function and lower inflammation.
PREDIMED · Estruch et al., NEJM 2018—an RCT in 7,447 high-risk participants; random assignment supports a causal conclusion, stronger than observational correlations.
−30% major cardiovascular events: heart attack, stroke, cardiovascular death.
3. Quitting smoking
Removes the single largest source of carcinogens and of damage to blood vessels and DNA.
The harm-minimizing level is zero; the less you drink, the lower the risk.
+14 / +12
years of life · women / men at age 50
These habits stack. Five factors together—not smoking, a healthy weight, ≥30 min of daily activity, moderate alcohol, and a high-quality diet—add that many years of life (Li et al., Circulation 2018). Those years are the whole point of Bet 1: to live until the other approaches are ready.
Civilization is what remains of our attempts to build something eternal
Everything civilization builds is a way to outlast death—to preserve a name, a body, a face, a voice, a thought, a life's work. Even if all you do is put up a building and die, more of you remains in it than nothing.
Pyramids and tombs
The pharaoh's name and body were built to survive 4,500 years. Giza.
Monuments · kings, reformers
People who immortalized themselves through their deeds—carved into rock. Rushmore.
Culture is itself a mechanism of immortality: it carries a person across centuries. It used to happen spontaneously; now the same thing can be done deliberately →
Hope for the future
Long-horizon intentions
What is a long-horizon will, really?
It is the ideas, values, desires, and plans of yours that do not fit within your lifetime. The core purpose of a person is to be represented in the future, as far as that is possible.
A smart will
The person is frozen while their digital agent and trust keep acting in their interests: holding assets, running affairs, waiting for revival. The real-world prototype is Alcor and its Patient Care Trust, a fund that maintains patients for decades.
A book as a message
Professor Randy Pausch, dying of pancreatic cancer, delivered and published The Last Lecture—a message recorded for his children, so they could know their father in the future.
Art for a future AI
Artist Trevor Paglen makes images designed for machine vision: their addressee is a future AI that will learn to "read" them.
Work with the future itself—the tools for that now exist.
12Catastrophe
And what if none of it works?
What if even one of these is true: (1) the laws of physics forbid slowing aging; (2) the fight against death can never attract enough resources; (3) civilization will soon perish.
Catastrophe
"Death with dignity"—losing without surrendering yourself
EY
Eliezer Yudkowskythe essay "Death with Dignity" (MIRI / LessWrong, 2022). When the outcome looks doomed, the answer is to refuse false hope—and to keep your principles.
Catastrophe
01Pharmacy
02Delivery
03Drug discovery
04Methods
05Theory
06Blind spots
07Replacing
08Resources
09Foundations
10Autonomy
11Hope for the future
12Catastrophe
A no-regrets policy
Actions worth taking no matter which future arrives.
STEP 1
Build a model of the world
We gather data, pose testable questions, and compare hypotheses—Eternal Search runs as an answer-finding machine.
STEP 2
Defend first principles
We hold the simple ground: living is good, dying is bad. From it follows a duty never to normalize death as a convenient ending.
STEP 3
Extend everything we can
We extend people, knowledge, institutions, and projects wherever a real lever already exists. We promise no eternity where there is only a hypothesis.
STEP 4
Don't accelerate death
A personal baseline in the spirit of Don’t Die: first, stop doing what obviously hastens the destruction of your body, your environment, and the future.
