Thrive Capital’s Dr. Nworah Ayogu on Why Neuralink Is More Interesting Than SpaceX

The last 10 years, we've spent about 40 trillion on healthcare. We have not moved our life expectancy an inch. The life expectancy is dropping. Americans are not living as long as they used to. And if that is not an indictment of the failure of the healthcare system, I don't know what is. 40 trillion for no change.

More so than maybe any other time in history, the frontier of science and the frontier of what we've accomplished is moving faster than ever. We're seeing this explosion of AI and drug discovery, and I think that is changing the way in which they're going to deliver and make drugs, and it's also going to change the entire value chain. Two-thirds of physicians in America are using Open Evidence.

OpenAI is now one of the largest healthcare providers in America. They have 300 million people asking ChatGPT questions every week about healthcare specifically. Healthcare is the key hero use case for most people. Neuralink is a more interesting company than SpaceX. As a nation, I think we need a 10x nurse a lot more than we need a 10x engineer.

Thanks for having me.

Our first investment was a healthcare investment in Oscar, so it's been part of our DNA from the start. I just got back from Priscilla and Mark Zuckerberg's event for Biohub, where they came together with, you know, the Department of Energy and a handful of other big leaders. Isomorphic Labs was actually there on the stage too. And so it seems like we're in this complete inflection point of biology and drug discovery and AI and healthcare, and there's just a lot going on there. So I'd love to start with that and just kind of get your pulse on why now.

I would say more so than maybe any other time in history, the frontier of science and the frontier of what we've accomplished is moving faster than ever. We're seeing this explosion of AI and drug discovery, and I think that is changing the way in which we are going to deliver and make drugs. And it's also going to change the entire value chain. The same way that with AI, we saw that the labs, we've seen that trade flow through the entire supply chain — needs around compute, needs around memory, needs around kind of energy — we've seen that flow through. We're going to see the same thing with, you know, biotech, where it's not just drug discovery, it's going to be manufacturing, it's going to be sort of development and how you do your clinical trials, it's going to be commercialization. We're going to see that flow all the way through.

We're seeing these new trends in end-of-one medicine, so the ability to create actual customized treatments for an individual that can treat conditions that we could never sort of do before. We've seen that already occur and already have an effect, and we're only going to see sort of more and more of that. I think the last I'd flag would be this interesting intersection between hardware and biotech as well, that's allowing us to kind of unlock treatments we couldn't unlock before. I think of two of our companies, Neuralink and sort of Foretell. When you look at the problems that they're solving and the capabilities they let us have, they're almost biblical in proportion, right? The lame can walk, the mute can speak, right? It's quite amazing when you actually see what folks have been able to unlock.

So yeah, so if we think of, you know, AI for drug discovery and where people often think of companies like Isomorphic, companies like Chai, you know, the magic of it is saying, look, a lot of times you have a target. So you have some, you know, molecule that's out there that when we hit this target, it changes something within the body. But oftentimes we both need to understand what are the right targets to hit, right? Folks will call that, you know, biological risk. So if we do something to this target, if we change it, what happens to the rest of the body? If we upregulate it, if we downregulate it, you know, does it make you, you know, sleepier? Does it make you more awake? Does it make you lose weight? So we need to understand what happens with that target. And then we need to understand all the off-target effects. So what are the side effects that it happens? So we want to do something, we want to make you lose weight, but we also don't want to make you nauseous. We don't want to make your heart rate go up. We don't want to give you liver toxicity. Those are kind of the off-target effects. So there's that kind of biological risk is the thing we're trying to solve. We want to do the thing we want to do, and we want to make sure we don't do the things that we don't want to do.

Now the flip side of that is that we then have to make a molecule to bind that target. And when you talk of all these models that are protein folding models, you know, molecular design, kind of AI for drug discovery, people often think of that problem of how do we make a, how do we actually make this molecule? And some molecules are easier to make than others. You hear this term undruggable. But undruggable means that, you know, the pocket's small, or sometimes there are molecules where, you know, I have to bind for part of it because all these things are three-dimensional shapes. So if I bind part of it, right, the molecule itself may move so something else can bind. So they'll talk about kind of hidden pockets and things that make this stuff hard to do because you have to be able to predict in real time what it is now and how the shape changes over time.

So those are kind of the things that you're often trying to predict. So with these drug discovery models, the idea is, okay, can we first create a model that will let us understand what is the best thing to make to bind this? And that's kind of step one. The next step is that, okay, I've made something that can bind here. Can I also get something that will predict what happens after I bind this thing? And then after that, you also want to be able to predict, okay, but can I also predict what are the unwanted effects that I can get? And that's kind of the progression of the problems that we're trying to solve with these models. And I think the first step of what molecule binds well is the thing we've actually gotten, you know, better and better and better at doing. But what folks have realized is, okay, now that we're able to do that better, there's still sort of room to grow and room to improve. But the actual thing that we want to be able to solve next is being able to predict what those effects are.

And you know, with coding, with math, you know, those things are very verifiable. So I know whether my code compiles or not, right? For these things, I have to also verify, did my molecule work? So you know, does it bind to the other? I can actually test how well does this thing bind. But does this have the effect that I want? That's a more complicated question to ask. So I then have to be able to, you know, put it into a cell and see does this change the cell in the way that I want to. Or to figure out what side effects does this have in a person, I need to first put it into a mouse to figure out does it do what I want it to do in a mouse. And there's a saying we always have in research where it's, you know, we've cured cancer in mice 50 times. So everyone's like, look, you know, if you want to cure your cancer, the best thing to do is to be a mouse, right? Because we've cured that 50 million times and we have not cured it in all cases in humans. So oftentimes we'll put it in an animal and say, okay, how does it work in the animal? Because that's a good, you know, sometimes that's a proxy for the human. But the real best way to know how is it going to affect you as a person is to put it into another person. So those are kind of the levels of data that we need to figure out how do we test it in animals and humans more rapidly. Other people are looking at how do we actually get data that will give us good proxies for that, and that's what we actually want the models to be able to predict. But before the models can predict what effect it'll have, we have to have enough data to train them to make those predictions. And that's a bit of what you're hearing folks say when they say we need to generate the data to let us do that. There's both data to let us verify the experiments we're doing, but there's now another vertical of what's the data that will get the models good enough to actually make those predictions.

It's one of the things I love about ISO. You know, I think there are folks on Twitter who are like, well, you know, ISO, they're really good at sort of, you know, protein folding, but protein folding is not the only problem. I'm like, look, you know, ISO — one, ISO and team invented protein folding with AlphaFold, right? That is why Demis got the Nobel Prize, was because he created this thing that created this whole field that everyone is — it's hard to do, which everyone is now kind of enjoying and kind of running at, which is amazing. We want more people solving these hard problems. But you know, AlphaFold came out in 2018, so they've been doing this for a while. They have also realized a while ago and been talking at least internally for a while about as we solve this problem, what are the other steps and what are the other processes to work on?

So they've been thinking about all of those kind of other steps that you need to accelerate. So they've been thinking about, you know, what does it look like if we're going to get a 90% sort of success rate? If we're going to make this go from a 10-year process to a four-year process, it both means that we need to get better at our molecular design. We have to get better at our in vitro testing, so testing in the cells. We have to get better at our in vivo testing, our testing in the animals. And we have to get better at our sort of predictions of what works, so that, you know, we may not be able to one-shot everything all the way through all the phases of clinical trials. Rather than having a 5% success rate, I can have a 60% success rate. Like that's amazing, because that radically changes the things I'm willing to do all the way throughout the process. But it also means we've got to think a lot about what happens in the clinical trial process as well, around sort of phase one, phase two, and phase three. And that is not just companies. That is, you know, us as a society thinking about how do we kind of optimize around this process. And you know, credit to ARPA-H, they've been doing a lot of work recently around thinking about what could it look like to be able to drive efficiency around that portion as well, because what we don't want to do is get very good at sort of pre-clinical and still have the clinical be the bottleneck. And I think we have governmental agencies who are thinking about that. We also have folks within these companies that are also thinking about what needs to be true so the rest of the system is ready to kind of absorb these innovations going through. And even after we do that, again, the bottleneck continues to shift. And we talk a lot about oftentimes you have to plan for the end at the beginning sometimes. So we will also need to make sure that we're getting better at how do you manufacture these solutions as they're coming through. If we're going to be in a world where we can have new drugs coming out, you know, every sort of three years rather than every 10 years, that also means our manufacturing processes have to get better and faster and have the ability to absorb that demand.

We've always had this philosophy. I've always had this philosophy that we want companies that are bilingual. So we want companies that are, you know, truly the best at both of what they do. They speak two languages fluently. And I think ISO is such a perfect example of that. The AI researchers at ISO are world-class, right? Which makes sense given that Demis also founded and ran DeepMind. So these are some of the best AI researchers in the world. It is, you know, DeepMind and ISO and Anthropic and OpenAI, they all hold the same bar when it comes to their AI researchers. As chemists and biologists who sit within ISO, I've met these folks. These folks are world-class. They are as good as the folks you would find at, you know, a Lilly, a Pfizer, a Broad Institute, a Wyss Institute. These are world-class kind of chemists. They are truly world-class in both. They're bilingual, amazing at the biology, the chemistry, and the science, and amazing at the technology and the AI.

I think some of the magic of what they do is that in ISO they talk a lot about the ISO way and the fact that, you know, they're actually bringing folks together to be able to do both who are world-class. So they have these amazing chemists who are integrating these AI tools, but using them to solve problems and bringing their kind of knowledge of chemistry in the same way. So this intersection of both and having that culture of excellence in both worlds, but also being able to get that cross-talk so you can overlap, I think that's a lot of what makes that company so special. And I think for Josh, you know, he spent so much time kind of working with Demis, you know, letting them know that we appreciate kind of what they're doing. To us, healthcare is a space that we know and we're deeply familiar with, and also that we're as passionate about this space as they are. You know, I think they're firm believers that a lot of the benefit that we are going to get from AI is going to come in the space of healthcare. That's also something that we believe, and sort of I believe deeply to my core.

One of my biggest critiques of healthcare is that I think we often do a very good job of extending life, but we don't do a very good job of improving people's quality of life, right? Of giving them that sort of life back. And you know, for Neuralink, they actually now focus on a lot of populations, but one of the first populations they focused on were folks who have, you know, ALS or paralysis. So these are people who, whether through a disease, whether through a, you know, car accident, a lot of sort of veterans unfortunately who have been, you know, serving our country but maybe were hit with sort of an IED, no longer have usage of their arms and legs. And oftentimes these are also kind of young folks. So folks who are, you know, 30, 40, who now no longer have the ability to, you know, work because you often need your, you know, hands and legs to work. You know, they can't be outside kind of playing with their children.

And you know, what Neuralink does is they say, okay, well, what could we do to kind of give these people, you know, access to interact with the world again? They've created an implantable device that then lets these individuals interact and control a computer with their mind's motorized arm, so that folks can now control a sort of robotic arm to sort of pick things and kind of grab things. They have another trial where they implanted the device in the vocal center, the vocal sort of cortex of the brain, so that someone who wasn't able to speak could speak. So you know, that is the core of what they do, is kind of let's use this implant to give people the ability to kind of manipulate the outside world again. So you know, we may not be able to replace the kind of arms and legs that you've lost, but we can give you that functionality back. So that is what Neuralink does.

Now when you meet the folks at Neuralink, and to your point on why they're bilingual, Neuralink has these amazing hardware builders, which makes sense, right? Their hardware folks were at the same level of, you know, SpaceX and sort of Tesla. These are true kind of hardware builders who are building things with heavy complexity. An example of that would be one of the things they have is within the implant, which is, you know, very, very small, they have these kind of microscopic threads, and their thread count gives you the ability to kind of read from the neurons. So you have these kind of very small, like microscopic threads, and you know, I think at this point they're now up over a thousand of them. They've increased their count as they went. That's amazing hardware in and of itself.

Because the threads are so small, right, they actually created a separate robot to do the implantation. So, you know, because it's such a difficult sort of procedure to do with something that microscopic, and they wanted it done with extreme precision, extreme reliability, they created their own robot to do that. Now it's not surprising that, you know, when you have access to folks who are, you know, SpaceX, Tesla, Optimus level, you can also create a sort of surgical robot to do this elite kind of neurosurgery. It's just wild to think about the fact that there are other companies for whom that would be their whole business, is we have created the neurosurgical robot for these folks. This is one vertical of what they do and what they're doing.

But when you go in, you also meet the neurosurgeons who are sort of training with them. And I know some of these neurosurgeons, you know, they went to Harvard Med with me. These were the best like students and, you know, neurosurgeons in their classes, and these folks now are here doing that. These are sort of practicing neurosurgeons who are there sort of co-developing these tools with them. They have pathologists. The first time I went to their pathology labs, I've never met a company before that insourced their own pathology lab. Interesting. And most places, you know, you send it out, you get your slides, you send it out, and you go forth from there.

What they realized though is that pathology is so key to what they're doing because they want to understand whether that's, you know, in humans and their animals. You want to understand carefully whether or not there are kind of any effects this is having on the brain. They want to be able to get that data and understand it because safety is so critical to what they do. So therefore they're like, we are going to insource the pathology lab here because it's sort of what we do and it's critical to what we do. So they're, you know, they brought in all these kind of world-class pathologists. But this, you know, bringing home that thesis that I think you see in so many of these Elon companies of we're going to kind of vertically integrate through all parts of our supply chain that are important. If it's important, it's mission critical. We need to have it in house. And seeing them do that across things as diverse as not just the sort of hardware, but something like pathology, which most people don't think of, I think that was that impressive level of detail.

Yeah. I mean, I think one, you know, DJ, who is the, you know, president and kind of leads a lot of that company, he's just, he's a very compelling leader. I think when you meet him, you can tell that his passion is there in a lot of ways. I think passion is contagious. I also think excellence and talent density is contagious, right? People who are truly N of one want to work with other N of one people.

I also think they're working on such an important problem, and they were, you know, they were category-defining. I think when Neuralink was first doing this, you know, no one else was doing it. So therefore, you know, they didn't just step out into the lead, they created the category. So everyone who cared about this space, right, they brought them in, and now as they've gone forward, their lead has only compounded as they've gone forward. I think the other thing, and I think maybe this is one of my hot takes, I actually think Neuralink is a more interesting company than SpaceX. The brain is vaster than space.

We don't make it to space. So we don't invent silicon chips without our brain. That is the fundamental substrate that has unlocked all of this. We talk a lot about, you know, compute and intelligence. You know, the brain is still a more efficient engine of compute than, you know, every chip that Nvidia has ever made.

Yeah. So let's talk about hearing aids and hearing loss. So, you know, hearing loss affects 1.5 billion people worldwide. I think, you know, we all go to concerts. We all have headphones in our ears at all times. I would only expect that number to increase as we go forward. Unfortunately, you and I are probably both going to have hearing loss when we get older. And it's also an issue that it's the greatest modifiable risk factor we have for dementia.

So, you know, folks who have hearing loss also have a one-third increase in their likelihood to have dementia. And, you know, think about that for a lot of reasons, but your ability to kind of hear things is kind of anchoring for the brain. That's a lot of the ways in which, you know, vision and hearing are the brain's main sensory inputs. So having that and the loss of that, as well as the social isolation that comes from not being able to hear, a lot of people theorize that those are some of the main contributors we have to dementia. So if you want to keep your brain sharp, one of the most important things you can do is not lose your hearing.

Now most hearing aids, what they do is they just amplify all sound, right? Which that's actually not what hearing is. Hearing is both the sound, but it's also your brain dampens things in and out. That's part of why when you're in a coffee shop and having a conversation, you can have the conversation because your brain knows, focus on the person I'm talking to and tune everything else out. Most hearing aids just increase everything. Which is why it's so hard for folks with hearing aids to go out to dinner, you know, be on a crowded street, anywhere where there's background noise, it's actually very hard for them to be, which is why those folks often will just stay home where it's quiet.

You know, Matt De Jong, amazing founder, him and his team, they were these amazing hardware builders at a company called Butterfly, so the handheld ultrasound. He'd had grandparents who had unfortunately suffered from hearing loss for a long time, and he'd seen the devastating effect it had for him. So he'd spent 15 years thinking of this project and a lot of time working on it as kind of a side project. He was fairly obsessed with a space that most people aren't obsessed with. But around five years ago, he brought his team together to actually work on this full-time. A lot of it was kind of the changes that he saw in sort of two verticals. One was the changes he saw in silicon and in chips.

So they've got this story where, you know, they went out to TSMC and they, you know, taped out their own chips, which is so hard to do. I think you've talked to a bunch of these kind of folks who work at the new kind of chip fab companies, and it's so hard to tape out your first chip. It's unheard of to tape it out successfully the first time. But in a lot of ways, you know, because of the way that a hearing aid works, they knew they need their own custom chip because hearing aids are very small, right? It's the aesthetic. No one wants this giant thing on the side of their ear stuck a couple times.

So, you know, you want something small, but it has to last all day because you're using it all day. So you need the battery life to go well. You know, you actually have to have very fast inference because you're translating kind of incoming speech, right? So that's difficult. Well, there's just a lot of complexity that you have to have, which is why they had to actually design and tape out their own sort of custom chips.

The other thing was that they were thinking a lot about, okay, how do we actually do the sound processing such that we know what we want to sort of amplify versus not? And they talk a lot about how he got this idea for the semantic noise processing actually during the pandemic from Zoom. One of the things that folks were doing with all of the kind of noise cancelling we would develop is that they would say, look, we're not just going to — we're going to dampen up and down the sound by knowing what kinds of sound this is. So different kinds of sound have different tenor. So being able to tell what's the sound of a conversation that's in front of you versus, you know, a car going by. So what the actual noise itself is, processing and understanding that and using that to kind of dampen up and down is one of the strategies that they used.

So, you know, for them, you know, they took this hard technical project and problem and decided to sort of build a company around it, and that was the company of Forell. And I think for them, they've had, you know, this kind of amazing one uptick in consumer validation and kind of technical validation. So when they had their beta, they had some, you know, they had a range of folks. I think everyone in their beta loved it. I think it's — I'm always careful with my healthcare companies not to divulge information they haven't already divulged, but especially with their patients, but these ones are public though. You know, Steve Martin and Henry Kravis, these are both folks who live very active lives and are hard judges of what quality is. Anybody who's worked for them or with them would, I think, say that, you know, these are hard judges.

And I think for them, you know, they have loved the product. They have been both folks who have been users, but they've also passed along their friends and said how amazing this is and how much this has changed their lives. They had one of their beta customers who, after the beta ended, they said, okay, well, you know, the beta's done, he's letting us have this, you know, can we have this back? And the person said, absolutely not, over my dead body will I give this beta product back to you.

I think the other is there's this group that kind of is the third party that rates all hearing aids. So, you know, they have this like rating scale that they use. 4.5 was their highest rating that a hearing aid had ever gotten on their five-point scale. Forell recently did it and got a six.

They broke the scale.

They literally broke the scale, and I think that just shows both from a technical perspective as well as a consumer perspective just how well it is. And it's just, I think that's a perfect example of creating this hardware product to meet a need in which, you know, you understand this unique problem you're solving, and then you go forward and say, okay, what are the advances in hardware and AI that will let us solve this in a very different way? And I think we're seeing that, you know, the market is loving it. It's impressive to see people applying these tools to problems that are really debilitating for people's quality of life.

Yeah. Yeah. Me and Josh joke that we are very popular with the 70-plus crowds now because we have so many people who will call us and say like, hey, you know, my mother, my father, you know, my uncle really needs this. Like, can you help us get Forell on the wait list?

So, you know, so right now they have a wait list.

I know. I know. But they're, you know, they're working and ramping up. So the wait list should be going away very soon. But they have an — right now they have an office in New York. They also have, you know, their website in which, you know, because they fit folks, they want to make sure that folks get to sort of try it beforehand before they buy it. But they have an office in New York. They'll be expanding kind of rapidly since then. So they're, you know, they've created this great product now. They're doing everything they can to make sure they have enough of it for everyone to utilize.

I was born on a snowy day in Columbus, Ohio. No, so I think for me, I talk a lot about in a lot of ways I've spent a bunch of time making good things big. So I spent a lot of time as startups trying to scale very effective models, and I've also spent a lot of time trying to make kind of big things good. So bringing that kind of startup mentality of, you know, tackling hard problems rapidly to kind of large-scale problems. So I trained as a physician. I'm trained in, you know, internal medicine. I originally did neurobiology and health policy in undergrad. So I liked both the social science and the science. You know, I studied medicine and business in grad school. So went to business school and med school, and you did residency. So I'm a trained, you know, internist and physician.

And then after that, I joined the founding team of a company called Cityblock Health. It's actually a Thrive portfolio company. And what Cityblock does is they do integrated care for high-cost, high-social-need populations. So full-stack, you know, social care, behavioral health care, kind of medical care for the patients who need it the most, you know. So joined that company, you know, in its very kind of early days and saw it kind of scale out to now a national care infrastructure.

And then spent a bunch of time at Amazon, launching what I guess now is Amazon Health. So we turned PillPack into Amazon Pharmacy. I served as our first chief medical officer there. And then we launched a business called Amazon Clinic. So our first vertical with Amazon Pharmacy was saying, what's the space in which Amazon will have an advantage in healthcare? We're like, Amazon has this amazing logistics network. You know, they're very good at getting things from point A to point B, right? And like that would be our advantage in pharmacy is we're going to have a delivery pharmacy. It would make sense that we should be able to get these meds from point A to point B, being from the pharmacy to people's homes. And there was a lot of work on sort of pricing and, you know, selection that we did, but really that was the core of it.

And we said, okay, once that started working, we saw that's going well. The next question was, do we think we can get people — people trust you with their medications. Would they trust you with their care delivery? So that was, you know, when I launched Amazon Clinic and ran that business, that was a virtual care business saying, let's handle the kind of low-acuity, kind of more, you know, transactional portions of care because we felt like that fit more with the sort of shoppable model you had on .com. And once we got adoption there, we said, okay, well, the next step should be if they'll trust us with their kind of low-acuity transactional, well, they also trust us with their kind of low-acuity relational. And that's what drove us to then, you know, pull and buy One Medical and integrate that in. And all those pieces together became Amazon Health. So it was kind of bringing a large tech company into a new space being healthcare.

And then about, you know, what, two and a half years ago, I joined Thrive. So I've known, you know, Josh and Kareem since college and was familiar with their work in healthcare. I was familiar with them as people. Thrive had always been a group of people that I respected, a group of investors that I respected, but they've been making some pretty big bets around AI. I was a big believer in, you know, AI as an opportunity. So came over to focus on all things healthcare.

Yeah. So I think a lot of ways, one, if you're truly AGI-pilled as they say, right, you should be actually long healthcare. So I think we're only going to see more and more of these trillion-dollar companies in healthcare for a few reasons. One is that as these capabilities progress, right, what are the things that we as humans are going to want to spend our personal money on, our societal resources on as we go forward? You know, if you take AGI to the extreme, we all are going to want more time, and we're going to want to be able to enjoy the time that we have. So that means the things that we know will be increased buckets will be healthcare because that is the way in which we increase the time and the quality of the time that we have on this earth, and then, you know, some version of kind of entertainment, right? Like we can blump those two things. So I think that macro trend, if you believe in kind of AI and AGI, will continue to go forward.

I think there's also a point you could say that, you know, since time immemorial we've also been doing that. I had a mentor once who told me — this is a bit of an aside — but a mentor once who told me there are three kind of professions that have existed for all time, and it's, you know, he said it's pastors, doctors, and teachers, because as long as there are humans, you will need people to care for the body, you'll need people to kind of care for the mind, and you'll need people to care for the soul. So I think, you know, I would say that going forward we'll spend more on healthcare. I'd also argue that going backwards we also did the same.

I think that's one reason, is that the macro kind of tailwinds are there. I think the second is that we talked about this frontier of science. We're seeing it expand in such unique ways. So, Keytruda is one of Merck's drugs. You know, Keytruda's 2025 revenue of just that one drug was larger than the revenue of all the major labs combined, right? And that's one drug. So, biotech is big business already. So, Keytruda is what we call a checkpoint inhibitor. So, think of it as an immune booster. It can kind of boost up your immune system. But the interesting thing is that companies are continuing to find new use cases for a lot of these medications. So you probably heard about Merck and Moderna's new cancer vaccine, which I think is an example of the kind of amazing frontier that we're getting.

So, you know, the way that they do it is, with melanoma, right, you take the tumor out, you have your tumor surgery. What they do is they say, okay, we're going to take this tumor because each tumor is unique, right? Like, you know, your mother's cancer is likely different in a lot of ways than the cancer of someone else. So we talk about cancer as this broad thing, but not just cancer, melanoma, a specific type of cancer, skin cancer, each one is different because they mutate and change in different ways. So the idea is you take the cancer out and you sequence it and you say, 'Okay, what are the parts of this tumor and this tumor's expression, its tumor's own genome that makes it unique, and we're going to create a vaccine, an mRNA vaccine to target that exact cancer.' So what they say is, okay, we take the cancer out, we're then going to make this vaccine that we're going to inject into you, and then we're also going to give you Keytruda to kind of give you an immune boost at the same time to make it easier for your immune system to really recognize that cancer. So then any microscopic pieces of that cancer that are still kind of floating around in your body, your immune system is now boosted to just wipe all of those out. But it also, because your immune system has memory, so now if that cancer happens to come back in three years, four years, your immune system will remember it and your immune system will be able to fight that cancer while the cancer is still small and sort of get rid of it. And that was the thesis of cancer vaccines, and that's exactly how it played out. I think in their phase three clinical trial, they found that people who had that were 50% less likely to have a recurrence of that cancer in five years. So I think that's an example of both the frontier and the types of new treatments we're able to create, but also the fact that a key portion of these new treatments is now Keytruda, this immune checkpoint inhibitor. So I think we're going to see more growth because we're seeing, because the frontier is able to tackle kind of more and more problems. I think the third piece I'd throw in really goes back to the Lilly example. And for that, maybe it's worth kind of zooming out a bit on why I think at least Lilly won, right? There was, you know, Lilly, they weren't first. Novo was first. And yes, they had a drug that had better weight loss, but at the end of the day, really, if you remember, right, when GLP-1s first came out, the drug was so popular that there were shortages.

Right. Like, you know, no one could get them. So, you know, Lilly leaned into its manufacturing capabilities and kind of innovated across the value chain. We talked about that, we talked about how you've got to innovate across the value chain and, you know, project forward to where the bottlenecks will be. So they were able to have more supply and get supply on board faster, a whole year before, I think, you know, they took a year before Novo could unblock their supply. Lilly got there a year beforehand, so they had more supply out there. But I think the other thing is in commercial, we all know kind of Lilly Direct. They're able to say, 'Sure, we're a pharma company, but let's think in an innovative way around how we partner and how we lean into kind of care delivery.' And they unblocked that commercial aspect. I think that's actually why they've been able to be so successful and be the trillion-dollar drug, is because they're like, 'We're not just going to focus on the science. We're going to be great at the science, but we're also going to think about manufacturing. We're going to think about sort of commercialization.' I think that's actually the opportunity, is that we're seeing more and more of these companies, the same way that in, you know, software and hardware and AI, we're saying it's easier to build with AI, so therefore we're seeing companies that out of the gate have more surface area, right? The companies are more ambitious. I think we're seeing the same thing now in biotech, where we now have companies who say, 'Huh, it's actually a lot easier to kind of build things in a lot of these spaces, therefore we're going to have more surface area, we're going to do the science and the manufacturing, we're going to do the science and more of the care delivery,' which then gives them, one, it makes their businesses more defensible, but also gives them a much bigger TAM. And I think that's a big reason as to why we're going to see more of these trillion-dollar companies, is because folks' ambitions have gotten larger. You know, we often say that in a lot of ways what we're seeing now is that our problems haven't gotten smaller, right? The tools we have have gotten bigger.

So yeah, so our time with OpenAI was, gosh, I guess it was two, two and change years ago. I was talking with Brad and Sam about healthcare kind of at large and this push that, look, they should spend more time with more of the leaders in healthcare because this is going to be a key space for them. So therefore they should spend time with the incumbents both to understand how do they work with them and to understand the space better. So I organized a series of meetings with different leaders across healthcare. So it was, you know, some of the top biotech CEOs, some of the top payer CEOs, and some of the top health system CEOs as separate kind of events with the OpenAI leadership to actually just sort of talk through and say, here's the state of the art in AI. Here are these folks' problems and let's talk through. I remember with the health system leaders at the time, the o1 wasn't out yet, so reasoning was a brand new thing. It hadn't happened yet, but because we were close to them, I'd been able to play with the reasoning models myself. And I'd started putting the New England Journal of Medicine case studies, so that's, think of the House MD-level difficulty hard cases. I would put the first half of them in because the way they tell the story is, you know, here's their history, here's this, here's that, and then something happened, and then they'll tell you about what the expert said and what it was. So I'd cut it off right there at the end that it happened. So I just put that portion in every week and I was just blown away by the fact that, you know, every single week they would get it right. And these are cases that, you know, neither me nor my colleagues, where they were some of the best doctors out there, maybe we'd get it right, you know, two times out of 10, and they were, you know, they were OpenAI and the o1 was batting a 10 out of 10, which was amazing to me. I remember going to that meeting and saying like, look, I think these models are amazing, these models can, you know, diagnose, and everyone was treating me like I was a heretic. It was like, this is blasphemy. Like, what do you mean? Like, this is impossible. And I remember we then showed them and did a couple of them live, and just seeing the eyes of some of these like healthcare CEOs, some of whom are also doctors, it was just amazing starting to see it click for them that, oh wow, this is the capabilities of these tools. This is what's happening. This is what's, you know, coming down the pipeline. Same with one of the biotech CEOs. We were like, okay, what do you think of these tools? Let's talk through them. And they're like, 'Well, I think they can be really good for customer service. You know, I don't really think though these tools will ever be useful in drug discovery.' And now I'm seeing two years later, you know, one of those same CEOs has announced, you know, billions of dollars of spend in this. So it's been also very interesting seeing the progression of folks' belief in the tools in such a short amount of time.