We Just Reversed Aging In Humans（原声文稿）

You've been told that getting old is inevitable, that you get roughly 80 years of life and then you die.

In all of human history, nobody has ever escaped death, but that may change.

In 2026, doctors injected a patient with a treatment built to make old sales young again.

And for the first time ever, immortality might be possible, because what we're building with AI to stop aging looks just like a science fiction movie.

October 2026, New Orleans, a company called Life Biosciences show the first results of something nobody had ever done to a human being before.

Three patients, all of them going blind from glaucoma.

You see, glaucoma kills the nerve running from your eye to your brain.

That nerve doesn't grow back for any species, including humans. Once it dies, you're blind for life.

Every single drug on the market can only slow it down, not stop it.

Four months earlier, these people had three genes injected into one eye. Genes are just instructions. Each one tells a cell what to build and how to behave. These three genes told the cells in that eye to go back to acting young again.

On day 56, no safety problems in anyone, and in two of the three, the field of vision had actually improved.

Normally, patients with glaucoma only get worse until they eventually go blind. In this case, they got better.

Three people is almost nothing. This is the smallest kind of trial there is.

The kind that only asks whether a treatment hurts anyone. It's a signal more than proof.

Nobody has ever reversed anything about human aging, not even once.

This is the first time a treatment built to do it was put inside of a human patient.

Can your discovery one day be used to extend the human lifespan?

We are doing our best to expand our health span.

If we can live to 120, that would be great.

We have more than 20 ,000 genes in our body.

Three of them can take any cell and wind it back to the state it was before you were born. Let's take a skin cell, for example.

Even if it has decades of damage on it, it can become a blank cell again.

These three genes were found by a scientist in Kyoto named Shinya Yamanaka, and he actually won the Nobel Prize for it.

It took him 14 years of testing combinations by hand one at a time, just to find three.

Nothing unusual here, that is the normal speed of biology, until AI came along.

You see, your body is built out of tiny machines called proteins.

They carry oxygen, fight infection, repair damage. Every single drug we have works by grabbing onto one of them.

Proteins are far too small to see. So for 50 years, scientists had no idea what they looked like.

Just to get one single picture of one protein, you had to freeze it solid and shoot x -rays through it. That could take years of work.

Some proteins were never figured out at all.

This is the real reason aging has never been solved. The ideas took decades each to test.

Now, what's what happens when you remove the time?

The drug that we are developing for a chronic lung disease called idiopathic pulmonary fibrosis may also have a role in aging.

In 2021, a company called Encyclical Medicine asked an AI a question that normally takes a research team years to answer.

Out of everything inside of the human body, what should we attack in order to stop pulmonary fibrosis?

Pulmonary fibrosis is a lung disease where scartish use slowly replaces the lung until a person cannot breathe anymore. Most patients die within three or four years.

The AI went looking for the thing causing the damage.

It came back with one specific protein out of thousands that no scientist had ever linked to this disease.

Then a second AI built something to shut that protein down.

That stage normally takes anywhere between 4 to 5 years. They did it in 18 months.

And then we have the three genes from the eye that I was talking about earlier.

It took Yamunaka 14 years to find them. Nobody had made them work any better since.

So in 2025, OpenAI trained an AI on proteins instead of words and gave it a job. Don't find new genes.

Just take two of the ones we already have and write better versions. It produces versions in weeks.

Retro Biosciences, the company that tested them, reported that they worked over 50 times better at flipping a cell into a younger state. Forget every single one of these results for a second. And let's look at what the speed actually does to your own lifespan.

For the last century, every day you've been alive, medical science has added roughly six hours to your own life expectancy.

Things such as clean water, antibiotics, vaccines, heart surgery, 6 hours a day every day without most people noticing.

That was the science moving at the speed of 14 years per discovery.

And as you just saw, AI is now doing that same work in weeks.

Every step is getting faster at once, finding the target, designing the molecule and rewriting the genes. 6 hours a day was the old rate.

Nobody has any idea what the new one is.

There's a rule that shapes this entire field, and almost nobody outside of it knows this rule even exists.

You cannot get a drug approved for aging, nowhere on planet Earth.

Regulators approve treatments for diseases, and aging is not on the list. It has never been on the list. Aging is not a disease.

Getting gold is considered normal, and you cannot treat normal. So every company in this story had to find a side door.

Life biosciences did not file paperwork saying it wanted to make human cells younger.

It filed for glaucoma, a specific eye disease with a specific test in specific patients who are specifically going blind.

The treatment inside of that syringe was built to reverse aging. The permission slip says eye disease.

In silico did the same thing with lung scarring and retro biosciences, the company with the re -written genes, they got its first drug into people last December through Alzheimer's.

Different companies, different diseases, all of them circling the same thing because none of them are actually allowed to say what they are trying to do.

Now here's what that rule costs.

Altos Labs launched with $3 billion.

A lot of Nobel Prize winners on staff. It's one of the best funded biology labs that has ever existed.

Almost five years later, it has not put anything into a human being.

New limit, which is backed by the founder of Coinbase, is valued at above of $3 billion.

Its first human trial is planned for next year.

Isomorphic labs. They spun out of the team that solved protein shapes.

They raised money at $2 .5 billion. It promised to drug in human trials by 2025. It still doesn't have one.

So we have billions of dollars, we have the smartest people alive, so the bottleneck is not money or talent.

It's that proving something works in a human takes years, and proving that something slows aging down would take decades. Nobody has that long.

So, that's why all of these companies are picking up diseases that kill people fast enough to measure. This is why the eye was first.

Glaucoma gives you a patient who is visibly going blind, a test that measures exactly how much they can see and a clear result in 8 weeks.

It's the fastest and most accurate answer available anywhere in the body. Three people got that injection and two of them improved. That is not a cure for aging.

It's something much more useful right now, because the eye was just a first door to open.

We have three more doors that came after it.

The second door is a drug.

Remember the lung disease from earlier?

One AI found the protein causing the damage, and another one built something to shut it down. That drug is called RENTOSERTIB.

In 2025, the results from 71 patients were published in Nature Medicine.

On the highest dose, lung function was improved, while in the placebo group, it got worse.

In September of this year, it entered its final round of human testing. 320 patients, 47 hospitals.

There's no drug where an AI chose the target and designed a molecule that has ever gotten this far.

If this passes, an idea that an AI had becomes a medicine that you can be prescribed.

And there's something else even crazier in these trials.

They took blood from those patients and ran it through aging clocks.

These are pretty much tests that estimate how old your body is, separate from your birthday.

After 12 weeks on the drug, the patient's blood actually looked younger than when they started, at one single dose, about 3 years younger.

The study was led by Alex Zavoronkov, who founded the company that makes the drug, so he has every reason to want a big number. Instead, he wrote the opposite.

He said plainly that he cannot tell whether the drug slowed aging, or whether fixing a disease lung just makes the whole body look younger. Nobody has untangled that yet.

But it's the first time in history a drug designed by AI has moved a number that tracks human aging.

The third door is a gene editor that doesn't even exist in nature.

Gene editors are tools that we use to cut and change DNA. The one we all know is called CRISPR.

Jennifer Dudna and Emmanuel Charpentier found it and they won a Nobel Prize for it.

A company called ProFluent trained an AI on every gene editor we know about and then asked it to write a new one. And it did. And what it wrote was nothing like the ones it learned from. No living thing on earth has ever had it.

So they built it in a lab, it got human DNA just as well as the real thing and also made fewer mistakes while doing it. Nothing wrote that except AI.

And the fourth door to longevity is an organ from another animal.

In January 2025, a man named Tim Andrews received a kidney that had been grown inside of a pig.

10 changes had been made to that pig's genes so that his body would not destroy it.

He had been on dialysis. Tim lived 271 days with that pig kidney during the work on its own.

Then it started to fail. His body was basically slowly rejecting it.

And in October the surgeons took it out. He went back on dialysis.

Three months later a human kidney became available and he got it.

A pig organ kept a human alive for 9 months just long enough to reach the transplant that saved him. These gene edits brought him 9 months.

Once those 10 genes were chosen by scientists, one at a time the slow way. Now let's think about what this means.

If AI can write a gene editor that never existed in nature, it can also work out which changes an animal organ needs in order to stop a human body from attacking it. And I'm not talking about 10 edits.

Those 10 edits were made by hand and it took forever.

I'm talking about hundreds of edits tested inside of a computer by AI long before anyone even touches an animal. Nobody has done that yet.

But the organ that can last multiple years inside of a human instead of a few months, that is just a design problem now.

And design problems are exactly what AI is for.

So we have four doors. An eye, a drug, an editor, and an organ.

Every one of them was impossible when you were born. And all four of them opened in the last three years.

We're going to be entering an era in which there are drugs that change the rate at which we age.

Now, I need to tell you about the other side of this, because this field has buried a lot of good ideas. Let's start with Unity Biotechnology.

Some of your cells stop dividing but refuse to die.

They just sit in your tissue, leaking chemicals that inflate everything around them. Scientists call them senescent cells. Everyone else calls them zombie cells.

If you clear them out of a lab rat, the lab rat gets healthier. Better kidneys, stronger heart, less frailty.

That result is real, and there are multiple labs that have repeated it.

Unity Biotechnology raised hundreds of millions of dollars to do it in people.

They had noble -level science, serious investors, and a drug that worked perfectly in lab rats. In 2025, the company was dissolved.

Its drug performed no better than a placebo. Then there's BioAge.

Their lead drug was pulled out of testing in 2024 because patients' liver enzymes started climbing. streaming.

The stock lost 75 % of its value in a day.

And then we have Isomorphic Labs, which is the company that spawned out of the team that cracked protein shapes.

They promised to drug in human trials by 2025. As of today, it still doesn't have one. So here's the picture.

We have billions of dollars spent, we have brilliant people, and yet the graveyard is still full of these companies.

But let's look at why these companies failed, because it's not what most people think.

Unity's drug didn't fail because the zombie cell idea was wrong. The idea is still correct.

It failed because clearing those cells in a lab rat is a completely different problem from cleaning them in a human.

And nobody could tell the difference until 200 people had already taken it.

That is what every failure in this list has in common. It's not bad science, it's only bad guessing.

For a full century, that was the only way to do it.

You had a promising idea, you spent 8 years and 400 million dollars finding out if it survived contact with a human body.

Most of the time, it didn't.

9 out of 10 drugs that enter human trials never make it out.

And I want to stop here for a second, because those 9 are not wasted.

Every single failed trial, every drug that did nothing, every treatment that worked in an animal and fell apart in a person, somebody wrote all of it down, decades of it.

And that's how we have more than a century of careful records of things that did not work. And that pile of failure is now training data.

An AI learning what makes a drug work needs to see what makes a drug fail.

And the only reason it can see that is because thousands of scientists kept writing down their losses.

People who spend their entire careers on compounds that went nowhere, they are the reason a machine can now rule out a bad idea in an afternoon. And that pile keeps growing.

Every single trial that runs this year.

failure, no matter if it's a success or failure, becomes something the next model will learn from.

Now, think about what AI actually changes.

It doesn't make biology simpler. It just makes guessing cheaper.

Every one of these dead companies had to build their best guesses, put it inside of a person and wait.

The drug in its final trial right now had its target chosen by AI and the AI had to read more biology than any human could in a hundred lifetimes.

The failures were the price of working blind, but we are not working blind anymore.

Here's something that should bother you more than it does.

Aging is not a law. It's not like gravity, where everything obeys it the same way.

Different animals age at wildly different speeds, and some of them barely age at all.

A mouse gets two years. You get 80. We have the same organs, same cells, same basic machinery.

Yes, we are very similar, that's why lab rats are used in experiments.

But still, there's 40 times the difference in lifespan.

Now let's go further up. There's a whale in the Arctic called the bowhead.

Hunters have pulled stone harpoon points out of living bowheads, weapons that stopped being used in the 1800s.

Those animals were swimming around with a 19th century spear in them.

They lived past 200 years. A bowhead has something like a thousand times more cells than you do. Every cell is a chance for cancer.

By every rule that we understand, a creature that size should be full of tumors by its 50th birthday. But it isn't. Bowheads never get cancer.

Then we have the Greenland shark swimming in near freezing waters.

The best estimates put the oldest one somewhere between 2 and 400 years old.

There are sharks alive right now that were born before the United States existed.

And then we have the naked mole rat. An ugly hairless rodent the size of a hamster.

They should live three years like every other rodent.

It lives over 30. Its risks of dying do not go up as it gets older.

And this last sentence that I said pretty much breaks the definition of aging. So the answer is already written down.

It's in their DNA. It has been this whole time. And here's why nobody could read it.

A genome is 3 billion letters long.

Comparing one species to another just to find out which handful of differences matter is not something a human being can do by hand.

And for most of the last century, scientists just picked a gene they suspected, they studied it for a decade and found out it wasn't the one.

That's exactly the job AI is built for.

In March of this year, a lab called the ARC Institute released a model trained on 9 trillion letters of DNA. They were pulled from close to 130 ,000 different species.

Every form of life they could get their hands on read as one single language. Nobody told the AI what it meant.

It learned the grammar of living things the same way a chatbot learns English just by reading enough of it.

Which means that for the first time in history, we can ask a serious question and expect an answer. What does a Bau had well have that you don't?

Evolution ran that experiment for us.

Four billion years of trial and error. And the results are sitting in those genomes already working, already tested in a living mammal. We just couldn't read the file until AI came along.

Now, everything so far in this video has been about slowing the clock.

There is also a second road, and it's the one you're most likely to actually meet. Almost nobody dies of old age, that's a fact.

People die of one specific thing that just got a head start.

Maybe a tumor that nobody caught, a heart that was failing quietly for years.

So by the time there's a symptom, that thing has usually been there a long time. So what if you could see it coming in advance?

In Sweden, researchers ran a trial with more than 100 ,000 women going through breast cancer screening.

Half of them had their scans read the normal way, and the other half had an AI reading alongside the radiologist. Guess what? The AI group caught 30 % more cancers.

But the number that actually matters is not this 30 % figure. In screening, there's something called an interval cancer.

That's basically a tumor that was missed at your first appointment, which then grows for months and shows up as a lump you can feel.

Those are the ones that can kill you because they get a head start that's measured in months Fewer of those slip through when an AI was reading now I want to be precise here that trial proved better detection It has not yet proven that fewer women die because that takes years to measure but in cancer Finding it earlier is most of the fight Then we have the heart test it's called an ECG doctors put electrodes on your chest You wait 10 seconds and they get a print out of squiggly lines It's been in hospital since the 1920s. Cardiologists have read millions of them.

At the Mayo Clinic, researchers trained an AI on hundreds of thousands of these printouts.

Then they fed it scans that cardiologists have already looked at and called normal.

The AI picked out which one of those people had a weakened heart that nobody had noticed. Think about what that means.

We have a 100 -year -old test, millions of doctors, and all of this time there was very important information in those lines that no human eye could see.

The answer to how a bowhead lives two centuries is sitting in a genome nobody could read.

The wording that your heart is failing is sitting in a printout that nobody could read.

The reason 9 out of 10 drugs fail is also sitting in 100 years of records nobody could read. So none of this is new information.

All of it was already there.

We just built the first thing capable of reading it and that is AI and it's in hospitals right now as you're watching this video.

Now, if all of this sounds too big to be real, it's worth remembering that we've already done this once.

In the year 1900, the average person lived to about 47 years old.

That number gets misunderstood constantly. It doesn't mean people drop dead at 47.

It means children died. Roughly one in five children never reached their fifth birthday.

A scraped knee could kill a child because it could get infected and there was nothing anyone could do about infection. Inside of a single century, that was fixed.

We now have clean water, we have sewers which almost nobody thinks of as medicine. They actually saved more lives than anything else on this list. We then have vaccines.

Then in the 1940s we have the antibiotics which turned a scraped knee back into a normal knee.

Life expectancy in wealthy countries went from the 40s to 80s.

If you ask a doctor in the 1900s whether the average person would live to 80 years old, he would have told you it was fantasy. And he wouldn't have been stupid.

He'd have been reasoning from everything humanity has ever observed, and everything humanity has ever observed said that people die of infection, because a technology arrived that broke the pattern that he was reasoning from. That century of progress was done the slow way, by hand.

A scientist would have an idea, he would spend 15 years testing it, and someone else would pick it up a decade later. Penicillin was discovered in 1928.

It didn't reach patients until the 40s.

That's 16 years of sitting in a notebook while people died of infections it could have cured.

That was the speed of discovering things that saved us. And it still doubled the human lifespan. So let's apply this to the only comparison we have.

Yamanaka spent 14 years finding three genes. An AI rewrote two of them in weeks.

A drug that takes a research team five years took 18 months.

The first doubling happened with human beings reading papers and guessing.

Not one single person involved could read more than a fraction of what was known. AI has read all of it.

I'm not gonna tell you that this means we're gonna double again or how long it takes.

Nobody knows that and anybody who gives you a number is trying to sell you something.

But the thing that made the first century work was never genius. It was accumulation.

Enough small discoveries stacked on top of each other until the world your grandparents were born into stopped existing.

We just handed that process to something that never sleeps.

tomorrow morning you will wake up and nothing will have changed you'll be a day older so will everyone you love nobody's going to hand you an injection that stops it not this year and probably not this decade I'm not gonna sell you mortality because nobody has earned that word yet but something did change a man in a chair had three jeans put into his eye and his vision got better instead of worse a machine wrote a tool for cutting DNA that nothing alive has ever carried.

A pig's kidney kept a man off dialysis for 9 months.

A drug that no human thought of, it's in his final run of testing right now, tonight, while you're watching this video.

None of that adds a single day to your life. All of it might fail.

For every year you've been alive, the people who came before you were failing at this.

They were running experiments that went nowhere. They were just dying of the things they were trying to cure. Every one of these failures got recorded. All of it.

A century of losses sitting in libraries and databases waiting. We finally build something that can read it. Somewhere in that pile is the reason a whale lives for 200 years.

Somewhere in there is the thing that was wrong with every drug that failed.

None of it was ever hidden. It was just more than any human being could hold in their head at once.

So no, you're probably not going to live forever.

But there's a decent chance you alive during the exact years when that stopped being a stupid thing to ask.