More than a decade after Human Longevity Inc. (HLI) raised hundreds of millions of dollars around a vision of data-driven preventive medicine, the company is doing something very different: taking clinical-grade whole-genome sequencing, once part of an $8,000 executive health program, directly to consumers for $599, while spinning out a separate AI company built on the biological data HLI has spent years collecting.
For CEO Dr. Wei-Wu He, the $599 genome is less a product than an entry point.
'The genome is the beginning of a lifelong relationship with your health data,' he says. 'Your DNA essentially doesn't change, so sequencing it once gives us a permanent foundation.
The much larger opportunity is creating an intelligent health platform that continuously integrates changing data with your genome.'
That platform is taking shape through Human Life Foundation Models, Inc. (HLFM), a separate entity HLI spun out in May 2026. HLFM has since entered a multi-year, multimillion-dollar co-development deal with Insilico Medicine, while HLI's advisory board includes Nobel laureates Geoffrey Hinton and Michael Levitt.
The aim is not another chatbot but what Dr. He calls a computational model of an individual human being, 'a continuously improving digital representation of someone's biology,' integrating genomic data, imaging, blood biomarkers and wearable data to produce personalised risk predictions and intervention plans.
'I don't want HLI to become a company that adds AI onto healthcare,' Dr. He says. 'I want us to build healthcare in which AI and longitudinal biological data are fundamental infrastructure.'
Dr. He is an unusual person to be running it. He invested $40 million in HLI's Series B in 2015, then took operational control in 2019. He restructured it, narrowed its focus, and has since put around $70 million of his own money in without drawing a salary.
His interest in medicine goes back to childhood: as a boy in China, he watched his grandmother die of late-stage cervical cancer, a disease that is now largely preventable.
HLI was founded in 2013 by genomics pioneer Craig Venter, with Peter Diamandis and Robert Hariri, on a simple premise: combine whole-genome sequencing, advanced imaging and AI to predict and prevent the diseases that kill most people.
It raised roughly $500 million, drew investors including Celgene and Illumina, and became an early standard-bearer for data-driven health. It also tried to do too much at once, spending roughly $100 million a year on projects that ranged from cancer vaccines to predicting people's faces from their DNA.
For Dr. He, the difference between 2013 and now is not the idea but the tools around it. 'The vision hasn't fundamentally changed,' he says. 'The technology has finally caught up with the vision.'
When HLI was founded, he notes, three things were missing: sequencing was expensive, the datasets were immature, and AI was not powerful enough to integrate different forms of biological information.
All three have since changed, which is what makes the second act, the $599 genome and the AI models built on top of it, possible now rather than a decade ago.
Building the Dataset That Matters
What survived the turnaround was the asset that makes HLI unusual: a deeply phenotyped cohort of more than 10,000 clients, with up to 13 years of follow-up data. The health assessment offered to clients includes clinical-grade whole-genome sequencing, whole-body MRI on a Siemens Vida 3T scanner (with dedicated pancreatic imaging), and blood work covering proteomics, metabolomics and standard biomarkers.
Dr. He is candid about what that dataset is and is not. 'Our advantage isn't that we have more genomes than UK Biobank. We don't,' he says, drawing a distinction between wide data and deep data.
Public resources such as the UK Biobank and All of Us have breadth, hundreds of thousands of participants, while HLI's edge is depth: for many clients it holds whole-genome sequencing, whole-body and organ-specific imaging, cardiovascular measurements, blood biomarkers and years of clinical follow-up on the same person.
Because much of that originated in clinical care rather than population research, it is tied to actionable findings. 'For AI, the label can sometimes be more valuable than another million relatively shallow samples,' he says.
Dr. He points to the clinical record. In 13 years, he says, HLI has had no case of prostate cancer first detected at stage four among its members, and the company backs the claim financially: if an eligible member who completes the required screenings develops stage-four prostate or pancreatic cancer that HLI should have caught earlier, it will put up to $1 million toward their care.
Better Prediction, Not More Testing
The case for AI, according to Dr. He, is about complexity.
Conventional statistical models work well with a defined question and a limited set of variables, but human biology does not cooperate: a single person carries millions of genomic variants, thousands of laboratory measurements accumulated over time, billions of imaging pixels and a lifetime of clinical history.
'Modern AI can learn relationships across those modalities without requiring us to specify every interaction in advance,' he says. Foundation models can also reuse what they learn across many medical problems rather than starting from scratch for each disease.
The point, he says, is 'modeling the complexity of the individual rather than reducing the individual to a handful of variables.'
More powerful screening raises an obvious risk: finding abnormalities that would never have caused harm, and driving anxiety, testing and treatment along the way. Dr. He is direct about it. 'More detection does not automatically mean better medicine,' he says. 'Almost everyone will have abnormalities if you look hard enough.'
The goal, he argues, is to flag the abnormalities that meaningfully change a person's odds of illness or death, and to act only where there is evidence of benefit. He sees AI as part of the answer rather than the cause: by weighing context such as age, genetics, imaging characteristics, biomarkers and rate of change, it can cut unnecessary intervention rather than multiply it.
'The future shouldn't be more testing, more anxiety, more procedures,' he says. 'It should be better data, better risk stratification, fewer unnecessary interventions and earlier intervention when it truly matters.'
Shortening the Sickness Span
Where the founding vision was about extending lifespan, Dr. He's current focus is narrower: shortening the years people spend ill. The average American spends around 13 years in declining health before dying. He wants to bring that down to two.
The approach centers on the four major causes of illness and death: cardiovascular disease, cancer, dementia and diabetes. Instead of one screening protocol for everyone, HLI aims to use AI to work out which conditions a given person is most at risk of, based on their genome, and build a monitoring plan around those.
'There is no universal optimal testing solution,' Dr. He says. Over time, the aim is to help clients keep answering three questions: what are my biggest risks, what should I do about them, and is what I'm doing actually working?
On price, he reaches for an analogy from computing. 'The greatest value wasn't created by making computers more expensive,' he says. 'It came from making computing accessible to billions of people. We want to do something similar with genomics.'
That echoes the argument Craig Venter made from the start: precision medicine for a broad population, not a few thousand wealthy clients. For Dr. He it is only the opening move.
'I want to change the fundamental architecture of healthcare,' he says, describing a future in which every person has a continuously updated biological model and AI keeps asking what is most likely to make them sick in the years ahead, and what can be done now to prevent it.
'The goal isn't simply longer lifespan. It's to dramatically shorten the period of life people spend sick.'
The $599 genome, in that telling, is where it starts.
