Summary

The $101M XPRIZE Healthspan is not asking teams to prove that people live 20 years longer within 7 years, but to show that, within 1 year, therapies can shift muscle, cognitive and immune function in 50–80-year-olds by the equivalent of 10–20 younger years. The competition runs from late 2023 through 2030 and has attracted more than 600 teams from 58 countries; 40 finalists were selected in May 2025 and received $250,000 each, while 10 finalists will be selected in 2026 and receive $1M each before moving into human trials. Zhou Guangyu’s view is that “7 years is probably the biggest compromise currently allowed by the available time, observation windows and trial design.”
The competition’s biggest scientific and investment risk is not a lack of therapies, but the absence of a universally accepted yardstick for proving that “rejuvenation” has occurred. More than 100 teams may bring 40–50 different aging clocks, and the organizers are “letting everyone bring their own ruler”; the final assessment will need to synthesize clinical changes in cognition, muscle strength and immunity, while requiring the clocks used to undergo peer review and comply with publicly disclosed methods. Even winning the prize would not automatically establish a real lifespan benefit.
Aging is not a standalone target but a systems failure involving the genome, telomeres, epigenetics, proteostasis, mitochondria, stem cells, inflammation and cell-to-cell communication. A 2025 review expanded the hallmarks of aging from the original 9 to 14; an intervention that appears to target a single mechanism often changes 3–5 or more markers at the same time. Renhui stresses that hitting a target in animals does not imply a benefit in humans; once inside the clinic, a therapy may “hit some targets you never wanted to hit,” so trials must measure net benefit rather than isolated improvements.
AI aging clocks can already process millions, tens of millions or even higher-dimensional datasets, but their error in predicting chronological age is still around 3.5 years, leaving them even further from reliably measuring health age. Training labels are still mainly birth age, yet two people who are both 30 can have entirely different health states; elderly baselines face a seesaw between overcorrecting by selecting only the healthiest people and adding noise by including common diseases. Zhou Guangyu says the industry is still like “blind men feeling an elephant”: interpretable models are constrained by incomplete or even incorrect biological knowledge, while black-box models still need to offer credible explanations.
Human evidence today is still dominated by observational research, drug repurposing and specific disease indications, while interventions aimed purely at anti-aging endpoints remain extremely rare. The metformin longevity trial discussed on the show requires long-term tracking of mortality endpoints and has been difficult to advance; mesenchymal stem cells usually enter through indications such as acute graft-versus-host disease, pulmonary fibrosis or diabetes. A more realistic path for XPRIZE entrants is to designate immunity, muscle or biological age as secondary endpoints in compliant disease trials, rather than directly seeking approval for an “immune anti-aging drug.”
The anti-aging companies with the deepest pockets may actually be poor fits for the competition: according to Zhou Guangyu, Altos Labs and Retro Biosciences did not participate in XPRIZE. He suspects their regulatory filings and established R&D timelines could not be reconciled with the competition schedule. Altos is betting on cellular reprogramming that leaves the genes themselves unchanged but uses epigenetic controls to move highly differentiated cells from the “tips of the tree” back toward the “trunk or even the roots”; public updates remain scarce and standardization is still difficult, making it look more like a closely monitored, personalized breakthrough therapy than a drug that can be mass-replicated in the near term.
The most executable anti-aging goal for now is extending healthspan, not chasing immortality or blindly buying supplements. Accessible approaches discussed by the guests include emotional and mind-body management, exercise, different forms of fasting, and cold or heat exposure within one’s tolerance; they offered no one-size-fits-all answer on supplements such as NMN and PQQ. Zhou Guangyu’s bottom line is that “talking about efficacy without talking about dosage is bullshit,” and that safety and effectiveness still require an assessment of each individual’s condition.

Deep dive

1. The $101M prize compresses “10–20 years of rejuvenation” into a 1-year intervention window

Yushan described an exceptionally ambitious format: the $101M XPRIZE Healthspan aims to test, in relatively healthy people aged 50–80, whether muscle, cognitive and immune function can be shifted back by 10–20 biological years, with the therapy required to deliver the result within 1 year.
The competition runs from late 2023 through 2030, with more than 600 registered teams from 58 countries; the 40 teams selected in May 2025 each received $250,000, while 10 finalist teams will be selected in 2026 and receive $1M each. From 2026 to 2029, they will enter 1-year human trials, with the winner taking the remaining nearly $100M.
Some have called the competition the anti-aging “moonshot.” Yushan said it reportedly has backing from the Saudi royal family and multiple billionaires. The format tries to sidestep the problem of waiting a lifetime to observe mortality endpoints, at the cost of relying on surrogate endpoints such as functional changes and biological age.

2. Regenerative Bio trails in the first lap but endorses the forced common playing field

Regenerative Bio, where Zhou Guangyu works, submitted a complete proposal but did not make the top 40; in his words, the company is “temporarily a step behind in the first lap” of this long race, though it still has a chance to compete for the $1M prize later.
The application was “no less than a major research program,” requiring the team to lay out the research background, qualifications, prior results and full intervention design. Structurally, it resembled applications for NIH and other research programs, but XPRIZE pressed especially hard on how the team would ultimately prove that its intervention worked.
What Zhou Guangyu valued was not only the prize money, but XPRIZE’s effort to “force” otherwise incomparable approaches onto the same track in what remains a blue-ocean field. In an industry that has not even agreed on what the finish line means, organizing a direct comparison takes courage.

3. XPRIZE’s credibility comes from celebrity capital, while scientific review still lacks transparency

XPRIZE’s executive directors come from education, technology and investment and handle much of the operational work; its non-executive directors include public figures such as Larry Page and James Cameron, whom Zhou Guangyu described as members who typically “bring their own capital into the group,” contributing money, resources or influence.
Scientists participate more often as external advisers, while the specific roster has been disclosed only selectively. Zhou Guangyu believes prominent aging researchers and active industry figures may be advising the competition, but the program did not treat undisclosed identities as established fact.
The competition therefore has formidable convening power and a strong resource network, but the scientific credibility of specific therapies and aging clocks still has to rest on peer review, public methods and clinical results. A celebrity board cannot substitute for those.

4. The top 40 prioritized diversity of approaches over a single technology winner

Zhou Guangyu described the entrants using a “forced binary”: on one side are biotech approaches including gene therapy, stem cells and their derivatives, small molecules, peptides and reprogramming; on the other are diet, exercise, nutritional supplements, mind-body interventions and integrated health management.
Public team names are often based on targets or themes rather than disclosed company names, so it is not always possible to tell whether a GLP-1-related proposal comes from a major pharmaceutical company, an emerging biotech or a startup. Brand recognition was not a prerequisite for selection, and many of the top 40 names are unfamiliar to the broader public.
Reviewing the 40 finalists, Zhou Guangyu observed that similar solutions were generally represented by only 1–2 teams. This suggests the organizers wanted to cover 20–30 different directions rather than fill the field with large numbers of near-identical therapies.

5. Longeveron shows the clinical binary of anti-aging companies—and exposes the root problem that aging is not an indication

Yushan used Nasdaq-listed Longeveron as an example: she found that its market capitalization was around $20M, while its stock had once surged above $300 before falling to around $1 at one point. Zhou Guangyu said the company uses mesenchymal stem cells as its entry point and has multiple clinical programs.
The valuations of such companies depend heavily on clinical readouts: positive data in one phase can lift expectations of success, while a negative result can send the stock sharply lower. Failure does not necessarily prove the technology ineffective; it may reflect a poor choice of indication or flawed trial design.
The real structural problem is that drugs and cell therapies generally have to select a defined indication such as pulmonary fibrosis, while “aging” does not currently qualify as a strict indication. The industry is still developing a consensus on what constitutes effective anti-aging, with no mature system for pricing that endpoint.
Mesenchymal stem cells sit near the “root” of the stemness tree but are not the most primitive embryonic stem cells; they can be found in human bone marrow and adipose tissue. Stem cells with differentiation potential also exist in the placenta, umbilical cord blood and muscle. Their effects are not limited to differentiation: autologous cells may produce a homing effect, while allogeneic cells act more through factor release to remodel the microenvironment and stimulate the immune system to some degree.

6. The 14 hallmarks of aging form a circuit map, not 14 independent faucets

Zhou Guangyu noted that researchers have been interested in the mechanisms of aging for nearly a century, but the field’s integration with biomedicine and clinical applications has mainly occurred over the past 10–20 years. The hallmark system has expanded from the original 9 items to 12 and then to the 14 proposed in a 2025 review.
The primary hallmarks include genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis and disabled autophagy; antagonistic hallmarks involve dysfunction in nutrient sensing and mitochondria; integrative hallmarks include stem cell exhaustion, altered intercellular communication, chronic inflammation and dysbiosis.
Yushan pointed out that the three categories are not mutually exclusive: genetic and epigenetic changes propagate into mitochondrial function, inflammation and cell-to-cell communication. Zhou Guangyu agreed that the more accurate picture is a complex circuit with upstream switches, regulatory elements and downstream phenotypes, not a clean classification table.

7. Zhou Guangyu’s nearer-term ambition is to keep a young body alive to 120, not promise immortality

On the upper limit of lifespan, Zhou Guangyu distinguishes between two camps: one imagines humans living to 200 or 300, while the other wants people to reach 120 with the physical condition of someone aged 20 or 25. He personally favors the latter as a more attainable objective and a more worthwhile pursuit over the next 10–20 years.
Based on the figures cited on the show, the longest reliably documented lifespan is around 122.5 years, while the current human lifespan limit is generally estimated at 130–140 years. If core organs could be replaced in the future, or stem cells could replace each aging component as needed, the limit might theoretically be pushed higher, but the technical and ethical problems would escalate sharply.
Yushan asked whether non-shortening telomeres, permanently accurate DNA replication and continuously functioning mitochondria would amount to immortality. Zhou Guangyu compared the idea to a physics problem with “no friction and no resistance.” In reality, replication errors create disease risks but also preserve genetic diversity and adaptability to changes in temperature, humidity and other environmental conditions.
Renhui challenged lifespan metrics using clinical boundaries: dialysis, pacemakers, TAH, VAD, HeartMate and ventilators can all replace organ function, but unconscious dependence on machines is not an ideal form of longevity. Cancer cells that replicate continuously are close to “immortal,” yet they ultimately destroy the entire body system.

8. Once a single-target therapy enters humans, the endpoint immediately becomes multidimensional net benefit

Zhou Guangyu accepts the basic idea of “fixing the faucets,” but points out that targeting one hallmark of aging often changes 3–5 or more hallmarks at the same time; stem cell exhaustion, epigenetics and proteostasis cannot truly be isolated from one another.
Renhui’s clinical warning is that a targeted therapy that performs well in the lab or in mice may hit “targets you never wanted to hit” once administered to humans, damaging other organs. A clinical trial has to determine “whether the benefits outweigh the harms,” not merely verify that the original mechanism has been activated.
The basic chain of human function also dictates a multidimensional evaluation: cognition is the foundation of independent living, behavioral capacity is tied to muscle and energy metabolism, and the immune system is responsible for repair, while defective repair manifests as inflammation. A participant can make the most relevant item the primary endpoint, but safety and changes in other systems cannot be omitted.
Renhui explicitly reserved judgment on the literal claim of “reversing 20 years”: the anti-aging field has yet to produce a study proving that any treatment extends lifespan by 20 years, and the longest relevant follow-up mentioned on the show may have been only 5 years.

9. Routine blood tests distinguish normal from abnormal; aging clocks aim to create a continuous scale

Immune function can be assessed through C-reactive protein, complete blood counts and lymphocyte subsets, or by examining specific markers carried by immune cells. But routine measures mainly distinguish normal from abnormal, and the differences between healthy 20-year-olds and 65-year-olds may not be obvious within the normal range.
Zhou Guangyu sees epigenetics and proteostasis as two relatively consensus-based layers to observe: epigenetics acts like a set of switches controlling how genes work together, while proteins are the structures that constitute life and directly execute its functions. Both may help map molecular changes onto clinical phenotypes.
The ideal aging clock would first map the health scale of people aged 20 and 25 through 80 and 90. If a 60-year-old initially registered a biological age of 65 and fell to 60 after 1 year of treatment, researchers would have a way to answer in the short term whether the intervention was genuinely anti-aging.
This would still be a surrogate endpoint, not actual lifespan. The clock compresses a lengthy final outcome into a testable readout, while putting the question of whether the model truly represents health, disease and mortality risk front and center.

10. High-dimensional data makes biological age computable, but does not automatically make it real

Yushan compared the human body to a car with a large number of dashboards: collect data from people of different ages, then determine whether a given set of indicators looks more like 25 or 30. Zhou Guangyu agreed with the design but stressed that the key is not the number of dimensions; it is whether the indicators reflect real underlying states.
Epigenetic and protein assays can generate millions, tens of millions, or even hundreds of millions or billions of dimensions. The goal is eventually to compress them into an intelligible biological age while predicting disease risk, rather than merely restating the date of birth.
Larger samples generally make statistical effects and the characterization of narrower age bands more stable. But “measuring every human on Earth” is only an ideal state; with “money burning,” real-world studies can only move gradually from several thousand or tens of thousands of people, or from populations in specific regions, toward general rules.

11. Cohort selection and organ-level stratification determine whether an aging clock can enter the clinic

Zhou Guangyu cited the UK Biobank as a valuable example: it covers specific age groups and includes long-term follow-up. Such longitudinal data has a better chance than a one-off cross-sectional test of linking molecular signals to future disease.
Yushan raised a problem similar to the one seen in microbiome research: healthy people may “all be different in their own ways,” and their indicators may not follow a single age-related trajectory. Zhou Guangyu’s answer was that, as long as the method is stable, high-dimensional data may still reveal low-dimensional patterns, including biological age, disease states and future disease risk.
Organ age offers a more clinically oriented breakdown: a person’s heart and liver may age at different speeds, but there is no accepted consensus on whether the gap can truly reach 10 or 20 years. The advantage is that organ age can connect directly with existing organ-based medical systems such as cardiology.
Epigenetic or protein signals in blood may also appear before morphological changes. Zhou Guangyu cited cancer as an example: certain specific signals can emerge 1–5 years before imaging or clinical diagnosis, illustrating the potential of organ-risk prediction.

12. AI is still “blind men feeling an elephant”; the 3.5-year error is only the surface problem

Classic machine learning is mainly used for dimensionality reduction, denoising, classification and finding mappings. But when it comes to predicting actual age from test indicators alone, Zhou Guangyu said these models often perform poorly.
One route prioritizes interpretability by incorporating known pathways and biological rules into training. The problem is that human knowledge is limited in dimensionality and constantly changing: “some previous understandings were even wrong or the opposite of correct.” The other route delegates the work to unsupervised algorithms, which must still explain why their discoveries are credible.
At a collaborative meeting on aging biomarkers held at Harvard Medical School in Boston, industry, academia and regulators discuss progress together, and XPRIZE often participates. Zhou Guangyu put the current benchmark at around 3.5 years of error in predicting actual age, with leading teams and research groups such as Steve Horvath’s and Vadim Gladyshev’s operating roughly around that level.
The deeper contradiction is that training labels are mainly chronological age: two 30-year-olds may be one healthy and one sub-healthy, and should not necessarily be interpreted as having the same health age. When establishing a 70-year-old baseline, selecting only people who are unusually healthy and free of hypertension, hyperglycemia and hyperlipidemia can overcorrect the model. The tradeoff between clean samples and the real-world distribution is “always a bit like a seesaw.”

13. “Bring your own ruler” enables innovation to compete, but makes the champion harder to interpret

XPRIZE did not mandate a single aging clock; Zhou Guangyu summarized the approach as “letting everyone bring their own ruler.” Conservatively, more than 100 teams may have prepared 40–50 different clocks, some using public methods and others using proprietary systems.
Yushan’s challenge was blunt: “If you cut your track small enough, you are the champion of that track.” If every team uses the ruler most favorable to its own mechanism, cross-therapy comparisons could collapse into each team proving itself inside a closed system.
Zhou Guangyu believes the organizers can rely on only two layers of constraint: jointly observing clinical changes in cognition, muscle and immunity; and requiring self-selected clocks to undergo peer review, preferably with prior publication or publicly disclosed methods. The competition can compare approaches under imperfect conditions, but it cannot create an absolute ruler where none previously existed.

14. The compliant path still runs through specific diseases; the 1-year trial is a compromise within a 7-year schedule

Current aging research is still dominated by observational studies, while intervention research typically starts with drug repurposing. The show described the metformin lifespan-extension trial as the only current clinical study directly targeting aging, but it requires long-term follow-up of elderly people through their mortality endpoints and has been extremely difficult to advance.
Metformin is widely available as a generic, with most relevant patents expired; familiar drugs such as rapamycin are also being re-examined. Zhou Guangyu stressed that before testing efficacy, researchers must first make an “extremely, extremely thorough” case for safety.
The more realistic path for mesenchymal stem cells is to run trials for clear indications such as acute graft-versus-host disease, pulmonary fibrosis or diabetes, then designate improved immunity or overall anti-aging as secondary measures. Mature regulatory frameworks in the FDA, China and Europe generally still require a defined indication and target population.
Asked whether the competition’s format was reasonable, Zhou Guangyu did not answer yes so much as offer a compromise: “If it could have, say, 17 years or 27 years, that would certainly be better.” But the competition would lose public attention, so the 7-year overall cycle has to compress the actual intervention into 1 year.

15. Altos is betting on reversing cell fate, but is not suited to development on a competition clock

Zhou Guangyu observed that highly financed, closely watched companies such as Altos Labs and Retro Biosciences, to his knowledge, did not enter the competition. His hypothesis is that their regulatory discussions, clinical filings and existing R&D timelines could not be adapted to the XPRIZE schedule. Their funding and development cadence may also already be locked to their original financing plans.
Yushan noted that Altos Labs had received substantial investment from Bezos and had multiple Nobel laureates on its scientific team. Altos’s core program is cellular reprogramming, rooted in foundational research associated with Shinya Yamanaka. Zhou Guangyu explained it mechanically: genes are the machine, epigenetic regulation is the switch; reprogramming does not alter the genes themselves but changes the switches, moving a highly differentiated cell from “the end of the tree” back to the trunk or even the roots.
Public progress has been limited. After joining Altos, researchers have had less they can disclose publicly; papers, academic conferences and external communications have all faced restrictions. Claims that the company may pursue breakthrough routes or launch clinical trials quickly in different countries once its data matures were explicitly described by Zhou Guangyu as “rumors circulating outside.”
The core obstacle at this stage is standardization. Zhou Guangyu believes cellular reprogramming is more likely to become a personalized or breakthrough therapy delivered under extensive testing and continuous monitoring than a standardized innovative drug or biologic that can be rapidly replicated for the mass market.

16. The competition can screen for short-term functional signals, but cannot skip the evidence hierarchy

Renhui divided the strengths of participating teams into several categories: hard-tech teams may have solid preclinical data and phase 1 capabilities; clinical teams are strong in ethics, approvals and study design; integrated-management teams combine existing therapies, lifestyle changes and mind-body interventions.
If cognition, muscle strength and other relatively soft endpoints are observed over 1 year, integrated interventions may produce measurable changes suitable for comparison. A phase 1 study of a hard-tech therapy, by contrast, often answers safety questions first and may not simultaneously prove a “10-year reversal.”
With a control group, researchers can compare the treatment and control arms; without one, they can still observe changes in an individual before and after treatment. Epigenetic and organ clocks make short-term measurement possible, but Renhui cautioned that they are surrogate endpoints, not proof of actually living 10 years longer.
Zhou Guangyu summarized the industry’s different routes as “all roads lead to the same destination”: extending high-quality life. Renhui chose “a hundred flowers in bloom,” arguing that each direction should first demonstrate and validate itself in the competition before the field can assemble a more directional map.

17. For ordinary people, anti-aging still starts with low-risk habits and individual assessment

Renhui began with the “cheapest” options: maintaining a positive mindset, meditation and mind-body exercise may help reduce chronic inflammation. She did not present them as universal cures, but emphasized their accessibility and their suitability as a first step for the public to improve psychosocial and inflammation-related risks.
Zhou Guangyu’s principle is that “everyone is the chief person responsible for their own health”: first understand one’s health needs, habits and desired changes, since every small step may produce a positive effect. Organ testing, nutritional supplements and innovative therapies are all ultimately moving toward personalized matching.
Zhou Guangyu also mentioned that David Sinclair favors stepping moderately outside one’s comfort zone. Cold exposure, heat exposure, different forms of fasting and sauna may improve stress responses and chronic inflammation, but the condition is always that they remain “within what one’s own condition and tolerance allow”; more extreme does not mean more effective.
On supplements such as NMN and PQQ, Zhou Guangyu did not recommend a universal answer. Natural origin does not mean the absence of a safe dose: “talking about efficacy without talking about dosage is bullshit.” A population-wide approach may cover 80%–90% of people, but pursuing both safety and efficacy still requires individual adjustment.

Robinhood Founder & CEO, Vlad Tenev: Robinhood’s $85BN Resurgence & Tokenizing SpaceX & OpenAI 125: Using AI to Recreate a “Me”: A Conversation With Mindverse’s 陶芳波 on Identity Models