Transcription
AI just designed a cancer vaccine, and almost nobody is talking about it.
A man in Sydney, with no medical degree and no biology background, used artificial intelligence to help design a personalized mRNA cancer vaccine for his dying dog. The tumors ended up shrinking 50 to 75%. The heads of OpenAI, DeepMind, and XAI all shared this story.
Not only is this an unbelievably huge deal, but 80% of the headlines that cover this story got the most important part of the story completely wrong. I've been tracking AI and its real-world applications for 14 years now. I spent every single day watching how these technologies converge and what they mean for regular people. So, when I saw this story break, I didn't just see a feel-good dog story. I saw something that is going to change how medicine works for every single one of us.
So, let me walk you through what actually happened, what the media got wrong, as per usual, and why this matters more than people realize.
Okay, so the story starts with a guy named Paul Cunningham. He's an AI consultant based in Sydney, which is in Australia. He's been in the machine learning space for about 17 years. So, he knows a thing or two about AI. Maybe he knows data science. He knows how to talk to these models, most importantly, but he has zero formal training in biology, medicine, or anything related to healthcare and vaccines. He is a tech guy, probably a tech bro.
In 2019, Paul adopted a rescue dog named Rosie. She's a Staffordshire Bull Terrier, Shar-Pei mix, cutiepie, about eight years old. And in 2024, she was diagnosed with aggressive mast cell cancer. Now, mast cell tumors are actually the most common type of skin cancer in dogs. The vet basically told Paul that Rosie had maybe 1 to 6 months left to live. She had tennis ball-sized tumors, including a massive one on her hind leg. This dog was in bad shape.
Now, most people in that situation, and I'm not saying this to be harsh, but it's just a reality, most people will accept the diagnosis. They'd be devastated, and you try to make the dog as comfortable and as happy as possible as you prepare for the worst. That's what most people will do. But Paul didn't do that.
Instead, he went to ChatGPT. Now, before you roll your eyes, 'cause I know some of you are already rolling your eyes here, let me walk you through what he actually did step by step, because the details here are extremely, extremely important.
Step one, he paid about $3,000 Australian to have Rosie's genome sequenced. He went to the Center for Genomics at UNSW Sydney. That's a University of New South Wales, one of Australia's top research universities. They sequenced both Rosie's healthy DNA and her tumor DNA. The reason why they did both is because by comparing the two, you can identify the specific mutations that are driving the cancer. These mutations create what scientists call neoantigens. Basically, little protein markers that are unique to the cancer cells. Sort of like a fingerprint that says, "Hey, I'm the bad guy here. I'm the one that's causing the issue."
Then step two, he used ChatGPT to help him navigate the biomedical literature. It helped the guy who doesn't know biology at all to figure out what questions to ask. It helped him create a data analysis plan. It helped him process the genetic data and identify which mutations matter. It helped him brainstorm treatment approaches. It even generated what he described as a half-page sequence, basically an initial blueprint for an mRNA vaccine.
So, think about what happened there. A guy with no background walked up to an AI and said, "Hey, my dog has cancer. I've got all this data. What do I do?" And within hours, he had a working blueprint for a personalized cancer vaccine. That's the part that should blow your mind. AI gave a non-expert the ability to navigate one of the most complex fields in all of science in a matter of hours that used to require a decade or more of education and training.
Now, step three, he used Google DeepMind's AlphaFold to model the 3D structure of the mutated protein that was driving Rosie's cancer. It's called the c-Kit protein. If you don't know what AlphaFold is, it's one of the most important AI systems ever built. It can predict the 3D shape of proteins with incredible accuracy. Understanding the shape of a protein is absolutely critical in drug design because the shape determines how the protein functions, how it interacts with other molecules, and most importantly, how you can design something to target it specifically. Paul used AlphaFold to figure out exactly which parts of the mutated c-Kit protein to target with the vaccine.
Then step four, according to Paul, XAI's Grok designed the final vaccine construct. Elon Musk actually posted about this on X, saying too many headlines are only talking about ChatGPT when Paul himself stated that the final mRNA vaccine construct was actually designed by Grok.
So far, clearly, this is an unbelievable story. There's so much doom and gloom about AI, but the fact is, we've already reached a point where consumer-grade products like AI chatbots can actually help design cancer vaccines. So, why does this matter? Because it tells you something about where the AI race actually is right now. You've got OpenAI, Google DeepMind, and XAI all contributing different pieces to what might be a genuine medical breakthrough. Greg Brockman from OpenAI shared the story. Dennis Hassabis from DeepMind also shared the story. And of course, Elon Musk shared it as well. Three competing companies all pointing at the same dog and saying, "Look at what AI can do." That almost never happens.
Now, real quick. About 80% of you watching this video right now aren't subscribed. It goes a long way to showing my videos to most people, which helps me keep doing this. Thank you so much.
Okay, now that Grok has created the vaccine, let's go to step five. Next, Paul worked with actual world-class researchers. Professor Paul Thordarson, the director of the UNSW RNA Institute, which is Australia's first dedicated RNA research center. This guy has over 190 publications in journals like Nature, over 12,000 citations. He's a legitimate expert in RNA and lipid nanoparticles, which is the delivery mechanism for mRNA vaccines. Basically, the tiny little bubble that carries the mRNA into your cells. He also worked with Professor Rachel Allavena from the University of Queensland's School of Veterinary Science. She's a professor of vet pathology who leads canine cancer immunotherapy trials and actually co-founded a company called Animal Solutions. And Dr. Martin Smith from the Ramaciotti Center, who handled the genomics. These are serious people at serious institutions doing serious work.
But the ethics approval alone took 3 months and required a 100-page document. Paul said something that I thought was really telling. He said, quote, "I had to do everything by the book because you can't just willy-nilly create a vaccine in Australia. The red tape was actually harder than the vaccine creation." Now, think about that for a second. The AI-assisted design of a personalized cancer vaccine was easier than filling out the paperwork to get permission to use it. That tells you something about where the technology is versus where the regulations are. The technology seems like it's getting ready, but the paperwork process seems way behind.
Lastly, here are the results. Rosie got her first dose on December 25th. She got a booster in January and February. The tennis ball-sized tumor on her hind leg shrank by half. Overall tumor reduction was about 75%. She went from barely being able to move, and I'm quoting here, to jumping fences to chase rabbits. That's pretty crazy.
But we should also highlight some caveats here before we keep going. Number one, Rosie was receiving conventional immunotherapy at the same time as the mRNA vaccine. That means there is no way scientifically to know for certain how much of the tumor shrinkage was from the mRNA vaccine versus the conventional treatment. This is the single biggest scientific weakness of the entire story. N=1, which means there's only one dog, one case with no control group. That's technically an anecdote, but it's pretty obvious that the timing of the vaccine plus the shrinkage of the tumor seems much more than coincidence, right? From a non-scientific perspective.
Number two, the real cost wasn't really $3,000. A lot of headlines ran with "Man Cures Dog's Cancer for $3,000." That $3,000 Australian dollars was just the genome sequencing. The actual cost, when you factor in the mRNA production and UNSW's RNA Institute, the researcher time from multiple professors who donated their expertise, the ethics approval process, the vet administration, the real cost was probably somewhere between $20 to $100,000. That's according to Egan Pelton, who's a Stanford PhD in chemical biology and a biotech co-founder. He puts it at $20 to $50,000 for the specific case. Cancer.org estimates around $100,000 per person if you try to scale this. But that's with today's technology. The much bigger takeaway here is that we're freaking potentially curing cancer. Holy cow.
And number three, one tumor did not respond to the vaccine at all. Paul is working on a second vaccine targeting different antigens for that remaining tumor. He's been very upfront about this. He said, quote, "I'm under no illusion that this is a cure, but I do believe this treatment has bought Rosie significantly more time and quality of life."
Now, let's cover some of the skeptic angles. Egan Pelton from Stanford made several strong points. He said, quote, "Everything here could have been done without ChatGPT." His argument is that basic literature reviews, sequencing recommendations, and protein modeling are routine work that experts do every day without AI. He also pointed out that mast cell tumors can spontaneously regress, meaning sometimes they shrink on their own without any treatment. But, and this is an important but, if you look into the vet literature, it's described as very rare. It's documented in rare cases involving puppies, but European consensus guidelines for adult dogs don't even mention it. Untreated high-grade mast cell tumors in adult dogs have a median survival of less than 4 months. So, while Pelton is technically correct that spontaneous regression is possible, the vet evidence suggests it's extremely unlikely in Rosie's case.
Pelton also said, quote, "I hope it's real. I love dogs, but this looks like an AI hype PR stunt." I understand where he's coming from. The AI hype cycle has also been wild, and healthy skepticism is appropriate, but two world-class professors staked their reputations on this project. Professor Thordarson isn't just some random guy. He's a former president of the Royal Australian Chemical Institute with 190+ publications. It doesn't seem like you would risk your career for a PR stunt.
Patrick Heiser, who's a cell and gene therapy researcher, raised another concern that proteins in the body can look similar to each other. So, a therapy targeting tumor proteins could accidentally hit healthy organs like the heart. That's a real safety concern and something that was heavily researched during the COVID pandemic, and it's part of why this stuff often takes years of clinical trials before it reaches humans.
But I don't want to get lost in the weeds of who's right and who's wrong about this because the real story, the thing that I think you should be paying attention to, is the convergence that made this possible and the fact we're about to enter an unbelievable age because of AI.
Okay, so here's the thesis. Three technologies converged to make this happen.
Number one was AI. You had ChatGPT for research navigation. You had AlphaFold for protein structure modeling, and Grok for vaccine design. You have three different AI systems from three different companies, each contributing a different capability.
Number two, mRNA technology. This is the same platform that was heavily used during COVID. Moderna and BioNTech spent years and billions of dollars building the infrastructure to rapidly design and manufacture mRNA therapies. That infrastructure now exists and is sitting in labs around the world.
Then we've got number three, cheap genomics. Rosie's genome was sequenced for $3,000 Australian dollars. The Human Genome Project, the first time we ever sequenced a complete human genome, took 13 years and cost $3 billion. And that was in 2003. 23 years later, you can sequence a dog's tumor genome for the cost of a laptop. That's what exponential cost decline looks like in practice. And those costs keep plummeting.
Now, put those three together. AI that can navigate complex biology, mRNA technology that can be rapidly customized, cheap genomics that can identify the exact mutations driving a specific cancer. When you combine those three things, you get something that didn't exist even 5 years ago: the ability for a non-expert to access the frontier of personalized medicine. The barriers between knowledge and action are collapsing.
Now, I want to talk about what's actually happening in human personalized cancer vaccines. Because this context is critical for understanding why Rosie's story matters beyond one dog. There's a clinical trial called Keynote 942. It's a partnership between Moderna and Merck. They're testing a personalized mRNA vaccine. It's called V942. I don't even know if I said that right. That's the official name. But you can just think of it as a custom-built mRNA vaccine for each individual patient. They're testing it in combination with Keytruda, which is Merck's blockbuster immunotherapy drug for melanoma patients who've had their tumors surgically removed but are at high risk of recurrence. The 5-year results from the phase 2b trial showed a 49% reduction in the risk of recurrence or death compared to Keytruda alone. 49% at 5 years. That's a massive result. The phase 3 trial is already underway and was fully enrolled in early 2026.
BioNTech and Genentech are testing a similar approach for pancreatic cancer, one of the deadliest cancers there is. In a phase one trial, six out of eight patients who responded to the vaccine were cancer-free at three years. Pancreatic cancer. If you know anything about pancreatic cancer survival rates, you know how remarkable that is. There are trials underway for kidney cancer, for glioblastoma. That's the brain cancer, one of the absolute worst ones. And there's even a trial at the University of Florida testing mRNA vaccines for glioblastoma specifically in dogs.
So, personalized mRNA cancer vaccines aren't really science fiction. They're in phase three human trials right now. What Rosie's story adds to the picture is the AI acceleration piece. The idea that AI can compress the timeline and lower the expertise barrier for designing these therapies.
If you want to hear more of my thoughts on how these converging technologies are reshaping entire industries, not just in healthcare, but in energy, transportation, manufacturing, and many other fields. Check out my new book, "Abundance or Collapse," which is designed to help everyone fully understand the massive disruption that's underway and how to position yourself on the winning side. Link for the book in the comments and the description below. Thank you.
Okay, so let me bring this back to why I think this story matters for you specifically, not just for all the beautiful doggos out in the world, but for all of us. I think we're in an inflection point in medicine that's very similar to where we were with self-driving cars about 5 or 6 years ago. Back then, people would argue about whether autonomous vehicles were even possible. Is the technology ready? Can it handle edge cases like weird things? What about regulation? What about liability? Does that sound familiar?
Those same arguments are happening right now about personalized medicine. Is a vaccine actually working? Can it scale? What about the regulatory burden? Is it safe? What about the cost? But the underlying technology keeps advancing regardless of whether the debate is settled. Just like Tesla's FSD went from a demo to coast-to-coast autonomous drives, while people were still arguing about whether it was vaporware, the personalized medicine stack is advancing while people argue about whether Rosie's tumors would have shrunk on their own.
The speed compression is the signal. The Human Genome Project took 13 years and cost $3 billion. Rosie's tumor genome was sequenced for $3,000 in a few weeks. The COVID mRNA vaccines went from sequence to injection in about 11 months, a process that used to take a decade. Paul went from genome data to first vaccine dose in about 2 months, plus 3 months of paperwork. In an optimized world, he can probably get the vaccine shipped to his doorstep the next day. Every layer of the stack is getting faster. Sequencing is faster. AI analysis is faster. mRNA design is faster. Manufacturing is getting faster.
The only thing that isn't getting faster is the regulatory process. And that's actually a really important observation because it means the bottleneck is shifting from technology to bureaucracy. The thing that slowed Paul down the most wasn't the AI or the biology. It wasn't even the manufacturing. It was the 100-page ethics document that took 3 months.
Now, I'm not saying we should eliminate safety regulations. We shouldn't. Obviously, drug safety is serious and important, but there's a real conversation to be had about whether regulatory frameworks designed for the era of blockbuster drugs where one drug treats millions of people and needs to go through a decade of trials. If that makes sense for an era of personalized therapies where each treatment is unique to one patient. That's a genuinely hard problem that governments are going to have to figure out. And I think AI is going to force that conversation sooner than anyone expects.
But what's actually going on is way more interesting than "ChatGPT cured cancer" or "Grok cured cancer." The reality is three different AI systems from three competing companies, each contributed a different critical capability. I think this is actually how AI is going to work in practice going forward. Not one super-intelligent system that does everything, but specialized AI tools that excel at different tasks being combined by humans who know how to orchestrate them, at least for the time being.
The person in the middle, the orchestrator, that matters enormously. Paul Cunningham couldn't have done this without the AI, but the AI couldn't have done this without Paul. At least until AIs are able to autonomously make decisions. But that opens up a Pandora's Box that I don't think humans are ready to tackle just yet. Paul was the one who had the motivation, the determination, the agency, the 17 years of machine learning experience to know how to use these tools effectively. He spent months navigating ethics approvals, coordinating with multiple universities, and pushing through bureaucratic resistance. The AI didn't do any of that.
Now, one side says AI cured cancer, and the other side says AI didn't do anything special. The reality is that this story is much closer to an unbelievable breakthrough than most people realize. Even if AI didn't actually cure cancer all on its own, but the reality is AI enabled a non-biologist to navigate one of the most complex fields in science, identify the right approach, design a therapeutic candidate, and connect with the researchers who could make it real in a matter of months, instead of years, decades, or potentially never. That capability, bridging the gap between expertise and action, is what changes everything. Not just in medicine, but in energy, in material science, in climate research, in every field where the barrier to entry has been "you need a PhD and a decade of training before you can even ask the right questions." We're watching that barrier dissolve in real time.
Rosie the dog is just the most visible example, and the direction is clear. I think five years from now, we're going to look back at this moment and this story the way we look back at the first successful SpaceX landing. It was messy and imperfect, kind of, but it was obviously and unmistakably the beginning of something enormous. And maybe I'm wrong about that. Maybe this turns out to be a one-off feel-good story that doesn't scale, ends up killing freaking every dog on the planet. Okay, maybe the costs never come down enough to reach regular people, but I highly doubt that. I've been watching exponential technologies long enough to know what the early innings look like. They look exactly like this. You've got a scrappy proof of concept that the experts dismiss. Real results that can't be fully explained away. Underlying technology curves that are bending sharply downward on cost and upward on capability. Multiple independent actors converging on the same breakthrough from different directions. That's the pattern every single time.
We seem to be living through one of the most exceptional periods in the history of science and technology. The tools that used to be locked behind decades of training and millions of dollars of funding are becoming accessible to anyone with the determination and agency to use them. What Paul did for his dog, someone else will do for their mother, their father, their child. The only question is how fast the institutions catch up to the technology.
So, if you've been watching what's happening with AI and feeling overwhelmed by all the noise and all the hype, the doom articles, the "AI will take your job." Just zoom out. Look at what's actually being built. A guy in Sydney used three different AI systems to help design a cancer vaccine for his dog, and the tumors shrank. That happened in the real world. Whatever else you think about AI, that is unbelievably important.
And speaking of long-term thinking and conviction, if you've been holding through all the volatility in the AI and tech space and wondering whether any of this is actually going somewhere, Bradford Ferguson, a rebellion, has a really interesting video about why most people could never hold through moments like these. It's about the psychology of staying the course when the world is still catching up to what's happening. I'll link that video in the description in the comment section below. Their team does a lot of deep research on the psychology of long-term investing, and they've got a very solid track record of thinking through these scenarios, especially before the crowd catches up.
Thank you so much for watching. I hope this video was informative and helpful, and we'll see you on the next one. Take it easy, everybody. Bye-bye.