Survive the next ten years, lifespan could extend by 50 years
His logic: within 8 to 10 years, medical progress will accelerate to a critical point where each year lived brings new therapies extending lifespan by more than a year—this is longevity escape velocity.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Longevity escape velocity means each year returns more than a year
Unutmaz cites Aubrey de Gray's concept of ‘longevity escape velocity’: within about 8 to 10 years, medical progress will accelerate to where each year lived provides access to new therapies extending lifespan by more than a year. He gives GLP-1 drugs as an example—already adding 5 to 10 years for obese and chronically ill patients, with effects compounding. Eventually people in their 80s and 90s could return to the physiology of their 30s and 40s, potentially adding 50 to 100 years total.
— Derya UnutmazDigital twins will compress clinical trials to weeks
He envisions a ‘digital twin’: integrating a person's genome, metabolome, immune system, microbiome and all other data into an AI model, simulating drug response first in the virtual body, then running verification only on the small fraction of humans where AI predicts 100% efficacy—compressing years-long clinical trials to months or even weeks. He thinks this is achievable within 5 to 10 years, with the ultimate goal of ‘treatment-on-demand’: after AI analyzes your genome, you order from a factory and receive your personalized drug the following week.
— Derya UnutmazAGI's first level has already been achieved
Unutmaz breaks AGI into tiers, judging by whether knowledge learned in one domain can transfer to completely unrelated domains. He says if you gave GPT5 Pro a biology experiment and asked it to redesign it as a video game, it could do it—proof in his view that ‘first-level general artificial intelligence’ has been achieved. Further tiers require persistent memory, self-learning, and ‘physical intelligence’ emerging from the physical world before reaching true superintelligence.
— Derya UnutmazThe only existential threat is humanity itself
Against the pessimistic view that AI is an existential risk, he counters with a thought experiment: imagine a parallel world where anyone with IQ above 100 is dangerous and must be imprisoned or killed. That world would have no vaccines, none of what we have now. He argues that historically, human violence causes more suffering than anything else; AI is an enabling tool, not a replacement. The real threat is how humans misuse it.
— Derya UnutmazGPT 5.5 Pro has crossed the intuition threshold
He tested it with 30 years of accumulated lab intuition: give the model a complex experiment already done with known results, and ask it to predict the outcome independently. Earlier models hit about 90% accuracy, but GPT 5.5 Pro, after 112 minutes of reasoning on a massive immunology dataset and writing a 40-page report, achieved 98% accuracy—precision he'd bet his own money on. For the first time, the intuition layer he once found ‘mysterious’ has been crossed by a model.
— Derya UnutmazPersonalized cancer treatment requires AI to design individual therapies
He cites an Australian computer scientist who sequenced her dog's melanoma tumor and submitted it to AI, which designed an mRNA molecule to train the dog's immune system to recognize tumor mutations. Synthesized and deployed within three months, the tumor began to recede and the dog survived. He expects that different personalized therapies could emerge for every cancer subtype—thousands of variants—with AI selecting which one each person needs from the mutation map.
— Derya UnutmazDisease prediction now enables prevention in healthy people
He mentions a retrospective study using the UK Biobank (500,000 people's data) where an older model analyzing extensive biological data could predict roughly 1,000 diseases that would eventually occur in healthy people—and those people did develop those diseases later. He stresses current medicine is ‘sick care’—treating illness after it arrives—but predictions like these enable intervention through lifestyle or diet changes before disease forms.
— Derya UnutmazEpigenetic clocks significantly overestimate aging reversal
Skeptical of the growing use of Horvath and other epigenetic aging clocks to measure ‘reversal of aging’, he notes these blood tests mainly reflect shifts in the ratio of young naive T cells to senescent ones, not whether skin, liver or brain actually rejuvenated. He warns future companies may claim aging reversal based solely on improving this one immune marker in their drug.
— Derya UnutmazIn their own words · checked verbatim
So I think this is probably the most critical time in human history. So try not to die for the next 10 years.
Derya Unutmaz6:06
I make the counterpoint there's only one existential threat to humanity and that's humanity.
Derya Unutmaz25:18
It was almost 100%. I mean I would say 98% correct. What I would have predicted like I would not have bet against 5.5 Pro myself.
Derya Unutmaz33:23
the AI models are able to detect certain tumors like breast cancer years before a radiologist is able to to to see that.
Derya Unutmaz42:31
the tumor started to to regress and the dog is was was alive when it was supposed to to die.
Derya Unutmaz1:07:44
they took a lot of that data and they were able to predict about thousand diseases before they happened
Derya Unutmaz1:14:52
So that it has a very limited use in my opinion.
Derya Unutmaz2:03:40
Figures
| Longevity escape velocity expected timeline | 8 to 10 years | 4:04 |
| Complete aging reversal expected timeline | at most 15 to 20 years | 5:04 |
| US annual misdiagnosis cases | 12 million | 41:30 |
| GPT 5.5 Pro single inference duration | 112 minutes | 36:27 |
| Global population of supercentenarians | 300 | 1:41:18 |
Glossary
- longevity escape velocity
- The point where each additional year lived provides access to new therapies that extend lifespan by more than a year.
- digital twin
- A virtual human model integrating genome, metabolome, immune system, microbiome and all biological data for simulating drug response.
- Yamanaka factors
- Four protein factors that reprogram ordinary cells into pluripotent stem cells.
- partial reprogramming
- Transiently activating Yamanaka factors to make cells younger without fully reverting to stem cells.
How to listen
Doctors, researchers, and longevity enthusiasts interested in AI drug discovery, digital twin clinical trials, or which large language models to use for diagnostic assistance.
The final few minutes on how he documents research on X and podcast closing pleasantries—low information value, skippable.