122 was the ceiling.
We're building the
floor for 1000.

For all of recorded history, one number sat at the edge of the human story like a wall nobody could see past.

122 years, 164 days.

Jeanne Calment, born in Arles in 1875, died in 1997, having outlived the airplane, two world wars, and every doctor who told her she couldn't. She remains, as of this writing, the longest confirmed human lifespan ever documented. No one since has beaten it. Not with money, not with willpower, not with the finest medicine of the twentieth and early twenty-first centuries.

For decades, gerontologists treated 122 as a kind of biological speed limit, encoded so deep into our cells that arguing with it seemed like arguing with gravity. Some serious scientists calculated that the probability of anyone reaching 125 was vanishingly small. The ceiling was not just a record. It was treated as a law.

The First Tenet

This single reframing, unremarkable as it sounds, is one of the most consequential ideas of our century. For most of human history, we accepted heart disease, then decided to fight it, and cardiovascular death rates fell by more than half in the developed world. We accepted childhood infection as inevitable, then decided to fight it, and childhood mortality collapsed. We accepted aging as neither. We simply looked away.

The ceiling at 122 was never really a wall. It was a mirror. It showed us the limits of what unaided human intelligence, working one molecule and one clinical trial at a time, could discover about the deepest machinery of the cell before the researchers themselves grew old and died.

What changed was not human willpower. What changed was the arrival of a new kind of collaborator, one that does not age, does not sleep, does not forget a single failed experiment from thirty years ago, and can hold in its reasoning more biological data than any human scientist could read in ten lifetimes.

That collaborator is artificial intelligence. This book is about what happens when you point it, deliberately and without flinching, at the oldest problem there is.

Every real shift in human capability has a moment where the abstract becomes personal. For AI and aging, there were two people who made that turn concrete, from two very different directions, arriving at the same conviction: that the tools built to recognize faces and translate language could be repointed at biology itself, and that doing so was not a side project but the whole point.

The Founding Thesis — Wei-Wu He

Wei-Wu He built and sold companies in genomics and diagnostics for two decades before making the bet that would define the next chapter. He understood something structural: the bottleneck in medicine was never raw biological knowledge. Human researchers had, by the early 2020s, already produced an almost incomprehensible mountain of papers, structures, and trial data. The bottleneck was synthesis: the ability to hold all of it in mind at once and reason across it fast enough to matter before a patient's disease outran the search for a treatment.

That is an information-processing problem, not a laboratory problem alone. And information-processing problems are exactly what artificial intelligence exists to compress.

The company built on that thesis, Insilico Medicine, was structured from day one to do something almost no drug company had tried: run the entire pipeline, target discovery, molecule design, and even patient trial simulation, through AI systems built for the purpose, rather than bolting AI onto a conventional pharma workflow as an afterthought.

The Generative Turn — Alex Zhavoronkov

Alex Zhavoronkov's contribution was more specific and, in retrospect, more radical. In the mid-2010s, most machine learning applied to biology was classification: given a molecule, predict whether it works. Zhavoronkov, trained across computer science, biophysics, and gerontology, asked the inverse question that generative adversarial networks had just made newly answerable: instead of judging molecules that already exist, could a system generate an entirely new molecule, from nothing, engineered from the ground up to hit a specific biological target?

That question, applied to drug design starting around 2016, produced some of the earliest published work on generative chemistry for therapeutics. It also produced a longevity-specific companion project: building AI models that could estimate a person's true biological age from blood tests and other biomarkers, deep aging clocks, distinct from the calendar age on a birth certificate. If aging is a disease process, you first need a way to measure it that isn't just counting birthdays.

Both threads, generate the molecule and measure the age, fed into the same institution and the same conviction: the ceiling at 122 was a limit of method, not a limit of biology.

Belief without evidence is superstition. So here is the evidence, stated as plainly as the source data allows, current as of mid-2026.

Rentosertib: The Proof of Concept

In June 2025, Nature Medicine published results for a molecule called Rentosertib (originally ISM001-055), a TNIK inhibitor for idiopathic pulmonary fibrosis, a disease that scars the lungs and for which treatment options remain limited. This was, by the account of the researchers and independent commentators, the first drug where both the biological target and the chemical structure of the molecule itself were discovered and designed by generative AI to reach a mid-stage human clinical trial.

In a 71-patient, 12-week, double-blind Phase 2a trial (GENESIS-IPF, NCT05938920), patients on the 60mg daily dose showed a mean improvement in lung function (FVC) of +98.4 mL, against a −20.3 mL decline on placebo. That difference, in a disease that normally only gets worse, is the kind of result that gets a case study written about it at Harvard Business School, which happened in early 2026.

Phase III, a larger global trial, is on track to begin in the second half of 2026. A second formulation, an inhaled version designed to lower systemic exposure, received clearance to begin its own human trials in April 2026.

The Pipeline as Scripture of Evidence

A single successful molecule could be luck. A pipeline is a pattern. As of mid-2026:

• 40+ programs in active development
• 13 IND clearances, the regulatory permission to begin testing a new drug in humans
• 31 PCC nominations (Preclinical Candidate Compounds)
• 8 programs in clinical trials, including 2 in Phase II and 4 in Phase I
• A new 8-program cardiometabolic and obesity portfolio, including oral small-molecule GLP-1 receptor agonists
• ISM8969, a brain-penetrant NLRP3 inhibitor for Parkinson's disease, began first-in-human dosing in June 2026

In March 2026, Eli Lilly signed a licensing deal for a portfolio of preclinical AI-discovered molecules worth up to US$2.75 billion, including a US$115 million upfront payment. Thirteen of the world's top twenty pharmaceutical companies now license the underlying AI software platform for their own drug discovery.

What the Machine Actually Does

None of this works by magic. It works because specific bottlenecks, each historically consuming years, are compressed by systems purpose-built for them:

Target identification. Multi-omics AI models scan enormous datasets of gene expression, protein structure, and disease pathways to propose which biological targets are actually worth attacking for a given disease.

Molecule generation. Generative chemistry engines design molecules that have never existed, optimized simultaneously for potency, safety, and manufacturability.

Trial simulation. Models trained on historical trial data estimate a candidate drug's probability of clinical success before a single patient is dosed.

Robotic execution. Physical laboratories run the wet-lab experiments that validate the model's predictions, closing the loop between digital prediction and physical reality.

A doctrine without a code is just an essay. Here is the code. It asks nothing that requires belief, only participation.

The Eight Tenets

The Practice

There is no temple. There is a blood panel, a biological age test, a sleep tracker, a resistance training program, a decision to read the primary literature instead of the headline. The ritual of this doctrine is measurement. The confession is an honest look at your own biomarkers. The communion is the sharing of data, so that the models that will eventually break the ceiling for everyone have something real to learn from.

Say the number out loud and it sounds absurd. A thousand years. That is deliberate. The doctrine is not asking you to believe it. It is asking you to notice that every prior ceiling sounded exactly this absurd right up until it fell.

In 1900, global average life expectancy was around 32 years. By 2020 it had roughly tripled, mostly through public health, sanitation, vaccines, and antibiotics, tools that did not exist a century earlier. The ceiling of "you get maybe forty good years" was not defeated by philosophy. It was defeated by germ theory, refrigeration, and a vaccine schedule.

The ceiling of 122 will not be defeated by philosophy either. It will be defeated the same way: specific tools, specific diseases, specific trials, compounding.

The Honest Roadmap

The Closing Vow

122 was not a law of nature. It was the edge of what unaided human effort had managed to discover in the time available. That edge has already started to move. Whether it moves all the way to 1000 is not a question this book can answer. It is a question the next several decades of trials, published data, and disciplined work will answer, one verified result at a time.