In August, a cancer vaccine designed for one patient at a time succeeded in a phase 3 trial.
Merck and Moderna's personalized mRNA therapy is built from the mutations in each patient's tumor. A series of AI algorithms reads sequencing data from tumor and blood, then picks up to 34 targets most likely to provoke an immune response.
Added to Keytruda after surgery for melanoma, the therapy significantly improved recurrence-free survival over Keytruda alone in a trial of 1,137 patients, at a pre-specified interim analysis; overall survival is still being measured.
It was the first positive phase 3 readout for a personalized neoantigen therapy, and for any mRNA-based cancer therapy. Every patient who received the vaccine received a different medicine.
How AI and biology made one-patient medicine possible
The vaccine belongs to a broader shift in how biology is done. In 2024, the Nobel Prize in Chemistry went to David Baker for computational protein design, and to Demis Hassabis and John Jumper for protein structure prediction.
Their model, AlphaFold2, has predicted the structure of virtually all of the 200 million proteins researchers have identified. Biology that once had to be discovered through experiment can increasingly be predicted and designed.
The personalized vaccine brings that shift to the scale of one patient, and a phase 3 trial measured the result.
The labs building AI are pointing at biology
The people building the most capable models say so themselves. In October 2024, Anthropic CEO Dario Amodei put biology and health first among the areas where powerful AI could change human life.
He predicted that AI-enabled biology and medicine could compress "the progress that human biologists would have achieved over the next 50-100 years into 5-10 years."
In September 2026, in an essay arguing that AI companies must pace the frontier, he still wrote: "I believe that AI could cure most major diseases in the next 5–10 years."
His 2024 essay was just as clear about what holds that progress back: "data, the speed of the physical world, and intrinsic complexity." Data, it said, "is often lacking—not so much in quantity, but quality."
Intelligence is becoming abundant, but the evidence it needs is not.
Precision medicine was defined to include prevention
The Precision Medicine Initiative describes precision medicine as "an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person."
Oncology moved first on the treatment side, in part because a tumor offers something concrete to sequence and an outcome that can be measured. The prevention half has been gathering under another name.
What longevity medicine describes
A 2026 study in JMIR Medical Education describes healthy longevity medicine as "an emerging clinical framework aimed at optimizing health span through proactive, evidence-based diagnostics and interventions grounded in geroscience."
Read closely, that is precision medicine pointed at a different target: the unit of care becomes a person's healthy years rather than a single disease, and the work starts before a diagnosis. That is the shift the word "longevity" names.
Why the evidence takes years
Prevention is harder to prove than treatment. A tumor returns or it does not, but healthy years are measured in events that take a long time to happen, often across several organs at once.
Even in people who are already sick, outcome trials take years. Semaglutide lowered the combined risk of heart attack, stroke, and cardiovascular death by 20% in 17,604 people with cardiovascular disease and excess weight.
That trial followed patients for a mean of 39.8 months. In people with type 2 diabetes and kidney disease, the same drug lowered the risk of major kidney disease events by 24% over a median of 3.4 years.
RNA medicine is on the same clock. Vutrisiran, approved in 2022 for a rare inherited nerve disease, lowered the risk of death by 35% through 42 months in ATTR cardiomyopathy, a form of heart failure seen mostly in older adults.
Why a better biomarker is not proof
The setbacks in the field make the same point from the other side.
In September 2026, Novartis reported that pelacarsen lowered Lp(a), an inherited risk factor for heart disease, but did not meet its main goal of reducing cardiovascular events in 8,323 patients with high Lp(a) and established cardiovascular disease.
Real-world studies had linked GLP-1 drugs to a lower risk of dementia. Yet in two phase 3 trials of 3,808 people with early Alzheimer's disease, semaglutide improved disease biomarkers but did not slow decline.
A biomarker is where a hypothesis begins; for one person on a personalized protocol, the proof is what happens over the years that follow.
The record the field is missing
A longevity protocol is a human intervention study run one patient at a time. It has a defined intervention, a schedule, baseline genetics, serial lab panels, wearable data, and a patient reporting how they feel between visits.
Pharma spends thousands of dollars per patient to collect data of that shape inside a trial, while the largest biobanks are mostly observational: they show who people are, not how they responded to an intervention.
Longevity clinics generate that data every day, for interventions no sponsor runs at scale, yet most of it still ends up in spreadsheets, PDFs, and chat threads.
That is the record the field is missing, and the one AI models are missing too: a model can learn who people are from a biobank, but not how they responded unless someone kept the record.
What we built, and why
We built iksa to keep that record. It runs the protocol at the point of care, turning notes, labs, genetics, wearables, and conversations into one patient record, labeled by intervention, so what was intended can be compared with what happened.
The physician, the care team, and the patient work from the same live plan.
Until recently, reading every signal about a patient was too expensive. Running a language model across one patient's record for a year cost several hundred dollars in 2020; in 2026 it costs between thirty cents and twenty dollars.
Through our work with leading groups in regenerative medicine, genetics, and life sciences, one lesson repeats: new therapies arrive from trials, but the evidence of what they do for one person is built in the clinic, visit after visit.
When the record does not hold, that evidence is lost, and the field never gets it back.
An invitation
Longevity medicine is being defined now by clinicians, researchers, diagnostics and therapy companies, the labs building models on human biology, and the people who fund them.
If you are working on how personalized prevention gets practiced or proven, we would be glad to compare notes.
