Alex Zhavoronkov, CEO of the AI drug-discovery company Insilico Medicine, recently ran an experiment that is worth paying attention to. He asked eight frontier AI models the same question, independently: if a healthy, well-optimized 50-year-old man wanted to live as long as possible, what drugs should he take — and how many years would they actually buy him?
The models largely agreed with each other. They also delivered an answer that almost nobody in the longevity industry says out loud. We think the experiment is revealing — though not entirely for the reasons it has been shared around the longevity world. Zhavoronkov is a sharp, deliberately provocative writer, and he is asking exactly the right question. The answer, we’d argue, is even bigger than the one he lands on.
What the AI Models Agreed On
Across all eight models, a remarkably consistent drug list emerged for the hypothetical healthy 50-year-old: rapamycin, statins, metformin, an SGLT2 inhibitor, a GLP-1 receptor agonist, and an ARB or ACE inhibitor, with most models adding acarbose, aspirin, and some form of NAD+ precursor. Of that entire list, rapamycin was the only compound most models credited with any potential to extend maximum lifespan at all. You can read Zhavoronkov’s full write-up here: https://www.forever.ai/p/how-much-can-you-extend-your-life
The sobering part was the size of the effect. When the models estimated what the whole optimized protocol would add to an already-healthy person, the numbers were modest — and the more capable the model, the more conservative it became. The most advanced reasoning model in his panel concluded the entire optimized drug stack adds less than two years of life expectancy. Not decades. Not a doubled lifespan. A couple of years, at the ceiling.
Why “The AI Agreed” Is Not the Same as “It’s True”
Here is the first thing worth slowing down on. When eight AI models converge on the same answer, that feels like strong evidence. It usually isn’t.
This point was made well in the discussion that followed on the longevity forum Rapamycin News, where one analysis put it plainly: the exercise is not a study, and leaning on the agreement between models as if agreement were validation is a mistake — models converge because they share training data, common assumptions, and the same internet-scale consensus. You can read that thread here: https://www.rapamycin.news/t/how-much-can-you-extend-your-life-with-drugs/24717
This matters for anyone using a chatbot to plan their health. Public AI models are trained on mainstream biomedical consensus and are bound, by design and by liability, to stay inside approved, conventional guidance. They will reliably hand you the standard pharmaceutical playbook. That is genuinely useful for what it is — but it is a mirror of the average medical opinion, not a window onto the frontier. Forum commenters were blunter still, with one arguing that simply including metformin on the list revealed the limits of the approach.
What Drugs Alone Cannot Do
Now the more interesting disagreement. Zhavoronkov is skeptical that drugs move the needle much — and he is also skeptical of “diet, exercise and sleep.” We part ways with him on the second half, but only partly, and the nuance is the whole point.
It is not that any single lifestyle factor beats any single drug. It is that interventions aimed at one system at a time — one pill for one pathway — keep running into the same ceiling, whether the intervention is pharmaceutical or not. The counterweight raised in the same forum discussion is striking: long-running data on the Loma Linda Adventist community is often summarized as roughly seven extra years of life for the men studied, and around nine and a half years for vegetarian Adventist men, compared with other Californians. That is a larger signal than most proposed longevity drugs have demonstrated in humans — and it comes not from one habit but from a whole pattern of living working together.
The honest reading of all of this is not “skip the drugs” or “just live clean.” It is that aging is systemic, and single-target interventions — of any kind — have a low ceiling on their own. As one forum participant summarized it, treatments that only address single body systems are not the answer.
What a Systems-Level Approach Looks Like
If single targets are the limitation, the logical alternative is not a longer list of pills. It is an approach that treats the body as an interacting system and is built around the specific person in front of you, rather than a generic 50-year-old.
In practice that means several things working together rather than in isolation: reducing the body’s toxic and inflammatory load so other interventions can actually work; supporting regeneration and repair, including through regenerative medicine; correcting the internal environment through targeted nutrition, the microbiome, and IV support; and using diagnostics to understand one individual’s biology instead of applying a population-average protocol. None of these is a miracle on its own. The argument is the same one the AI experiment accidentally makes for: the leverage is in the combination and the personalization, not in any single agent.
Our Approach at Santa Maria
This is the gap Santa Maria Health is built to address. We don’t hand patients a list of ten compounds and send them home. Our programs are personalized and systems-level by design — combining detoxification, regenerative medicine, diagnostics, and lifestyle work into a single coordinated plan over a focused period of time, calibrated to the individual rather than the average.
The takeaway from Zhavoronkov’s experiment, in our reading, is not pessimism. It is direction. The pharmaceutical approach, in isolation, has a known ceiling — and so does any single intervention used alone. The opportunity is in integration: understanding one person’s biology and restoring it at the systems level. That is the work we do. You can explore the full range of our programs on our health programs page, and read more about our regenerative medicine work specifically.