Transcription
All right, I have got a really interesting story for you today. So interesting, in fact, I'm just going to do a deep dive like instead of just kind of jumping from article to article that this guy's dog is dying of cancer. And he tried surgery, he tried chemotherapy, he spent thousands of dollars and out of desperation, he opened up Chat GPT. And I know this sounds like science fiction, but he used it to basically build a custom RNA cancer vaccine. He got approval to inject it into the dog. And the dog is actually making a ton of progress. Like I didn't even know if it was real, but once I dug into it enough, I just thought it was so fascinating. And I feel like if you understand what he went through, how he used LLMs and other AI tools like AlphaFold and and using like actual researchers and scientists to kind of help you along the way, you can do freaking fantastic things. And I think some of this capability is is just like people just don't know. So, I'm going to cover it in more of a deep dive fashion, but this is all 100% real. Like, this actually happened on the planet right now. And it makes me think, what else could be done like this? Makes me think that any of us might be like so much more capable if we're leaning on AI tools than we would even believe.
So, according to an Australian newspaper report, a Sydney tech entrepreneur named Paul Connham has used AI tools to translate DNA sequences into something that actually made sense to him. Then he leaned on even more tools and real researchers to build a custom mRNA cancer vaccine for his adorable rescue dog Rosie. So this is her. This is Paul's uh Twitter account, his ex account. "5 days of attempting to cure my dog's cancer using AI, we finally found a way to sequence Rosie's DNA a thread."
So, what's going on here is a tennis ball-sized tumor in Rosie's leg has shrunk dramatically, about 75% since he injected his own dog. He says that she got her energy back. She's starting to act like herself again. And of course, before we get carried away, I want to slow the story down. I'm going to go through it step by step because obviously the headline version is flashy and the real version has a lot of like kind of asterisks around things and has important ways to go about something like this. So, don't just take the headline from this one and go off on your own. Like, really think about what he did and and how it could be used as a guide for something if you're ever going to attempt something this ambitious. This is not a story about a chatbot curing cancer. It is a story about what happens when one determined person uses artificial intelligence like a map and then uses that map to team up with actual scientists, actual labs, actual medical oversight to try something that nobody has ever done before. And I'm Dylan Curious on YouTube. Hit subscribe and let's break it down together.
Okay, so Paul Coningham is a data engineer. He's a tech entrepreneur. He's spent 17 years working in machine learning and data analysis. So he definitely is somebody who already came with an understanding of code and algorithms, but not biology and not medicine. None, no history in any of those fields. But in 2019, he adopted this cute little guy or little girl. This is Rosie. She came from a shelter. He found her abandoned in a bush and like, of course, like any dog that's abandoned and falls in love with you. She became his best friend. He got to know her through lockdown, through tough personal times, all of it. Strong bond. Obviously, we all know what it's like. Well, not all of us, but you can imagine loving a dog can be one of the most gratifying things ever. Like animals are absolutely wonderful. But in 2024, Rosie got diagnosed with mast cell cancer. Now, this is an aggressive form of cancer in dogs and it spreads really quickly. So, Paul started right off the bat throwing thousands of dollars at chemotherapy and surgery. It did slow down the cancer, but the tumor never shrunk. And at the point where doctors were saying, or veterinarians were saying, like Rosie only has a few months left, this is where he decided to go in a direction that like pretty much nobody else has ever gone before. He's an idea guy. He's a data guy. So he goes to Chat GPT not to diagnose the issue but to brainstorm. He basically says prompting something like, "My dog has this type of cancer, what treatment approaches might work?" and then just sorting through the details, asking follow-up questions, referencing things over and over again, using multiple models. And he understood that one of the powers about the current generation of LLMs is not so much that they're doctors or they can examine your dog, but they're incredibly good at connecting dots across massive amounts of published research, way more than a human could handle. And he eventually ends up getting pointed towards this immunotherapy, specifically one that is mRNA based cancer vaccines. And remember, just like a regular guy. So he also notices that it points him to this specific lab, which is this institution, Ramicotti Center for Genomics. It's at the University of New South Wales, right? So he's about to reach out to this institution. And he, like, at this point, you have to understand, Chat GPT is not creating a cancer vaccine. It's just like that's like saying GPS is driving the car. It's just a guide to the route. Human beings are still needing to turn the wheel. Human beings need to like make these calls and connect these people and have experts look at it. And that guidance is helpful. It's a something that like, who would have known to go actually call this company or that they would specialize in something like this or that mRNA vaccines in the first place could solve it.
So, key pieces of information he has learned at this point. And maybe a normal person would have just thought, "Okay, it's still out of my realm because it's, you know, you need to be a doctor, you need to understand genetics or something." But Paul just picked up the phone. He calls this lab and he ends up talking to a world-class genomics research center saying like, "I want you to sequence my dog's DNA." So, he ends up meeting this guy, lead researcher Martin Smith, who's an associate professor of computational biology. And of course, he thinks this is kind of weird. Like, who's this guy? They don't normally do like direct to consumer sequencing. And even if they did, interpreting genomic data is incredibly complex. But Paul got his attention because he said, "You know, don't worry. I'll figure out the analysis myself with the help of AI." So, he paid about $3,000 and he got Rosie's DNA sequenced. And here's how it actually worked once he got a hold of it. And it's a beautiful concept. You think about taking healthy DNA from a dog's blood and then you take the DNA from the tumor and then you let artificial intelligence compare the differences. So, think about it for a second. Like comparing a clean instruction manual to a damaged version of that instruction manual that's full of like typos and and all like kind of misorganized and stuff. Cancer is often growing because the instructions inside of your cells got corrupted in some way. Like probably this, you know, stop multiplying signal got corrupted or something like that. And of course, the hard part is understanding the DNA, like which A's and G's actually matter. But that's where AI does some fascinating things that like a human can't or tools of the past just had a lot of troubles doing.
Once Paul actually had raw data, gigabytes of genetic information from the tumor cells and the regular cells, he used three different artificial intelligence tools stacked on top of each other. So first, he went to Chat GPT. This was like his research planner and it helped him figure out the overall strategy, identifying treatment directions and iterating on his like approach throughout the whole process. And then he tried to figure out which of those mutations would most likely be responsible for triggering an immune response. Okay. Now, in this targeted immunotherapy that's, um, called finding neoantigens, I guess, but basically they're molecular fingerprints that make a cancer cell look different from a normal cell. And then third, he referenced a tool that, you know, we've been hearing about for years now from DeepMind, AlphaFold. That's DeepMind's AI system that predicts protein structure. This is Demis Hassabis' baby from like a decade ago. And then he used it to model the 3D shape of the proteins that were the mutated proteins the cell was producing. So think about this now. We're not just comparing like a manual with clean DNA and a 3D protein shape that came from the mutated DNA. So you can now see something, you know, something that like might not be fitting the way it should or whatever. I mean, that's like super simplified, but you have these physical structures now. And one probably just looks different than the other. So genes are instructions. Proteins are the tiny machines that are built from those instructions. And the shape matters. A protein shape is like a key, right? If the key changes to the shape, it stops it from maybe like fitting into the right lock. If it's like a key going into a lock, it's like not opening the door. Or maybe it's opening the wrong door. Or maybe it's supposed to keep a door shut and it's like letting it open because it's not holding it tightly. That's the reason why sometimes just a single mutation can be so dangerous and cause cancer.
So Paul started just talking to Chat GPT about this like all day, every day. He had it help him plan his own algorithms, pick the best targets, used AlphaFold and uploaded data back to Chat GPT showing the shape of the problem, asking for thoughts. So the three layers of AI are each kind of doing something different. And then when he brought these results back to the university, the scientists, like, stunned is an understatement. Like, this guy with no biology background had essentially done the work of a computational biology team. Sam Altman's talking about there being a billionaire, like a one-person company billionaire. It's just like the power that you can leverage in the right situations for the right problems like this blows my mind. But here's what seemed to change in the minds of the researchers is that he kept showing his work. Every single time he went back, he had more analysis. He had more results and he showed real effort. It's like a student going to a professor like again and again, except each time the homework is actually getting done. So the geneticists were like, "This is crazy, but we should try to collaborate on this and get it done." And here's the point in the story where the human researchers with their knowledge kind of threw in a piece that wouldn't have just been totally possible by itself. But Paul and his team now, because he's collaborating, identified an existing immunotherapy drug that could work. And when they reached out to the pharmaceutical company, the company refused to supply it. So even though like this dog is dying and he's with actual scientists who are capable of getting a copy of this from the pharmaceutical company, they still wouldn't give it to them. So it almost felt like a dead end, like where the story could have ended. But instead, it just forced a pivot. Martin Smith mentioned that maybe they should try an mRNA vaccine, the same technology that powered the COVID vaccines. And then Paul jumped on it.
Okay, so just a quick refresher, mRNA, it stands for the M in front of the RNA, meaning it's the messenger RNA, the one that can come outside of the nucleus. So it's a messenger ribonucleic acid. You can think of it like a temporary recipe that you just like hold in your hands while you're actually making the ingredients. But it's important because it gives cells instructions on how to build specific proteins. So in in the case of COVID, those instructions were teaching your cells how to build a harmless version of the spike protein so that your immune system could learn to recognize the real COVID when it hit you and then just take care of it right away before it becomes a problem. So with this, with this uh cancer vaccine, the concept is similar, but there's another element. It's actually even harder because the virus is clearly foreign. With um COVID, it's easier to identify it, but cancer grows out of your own cells. So, it's like trying to spot a criminal who stole an employee badge or whatever and like walked back into the building. You're like, "Oh, you could be here, but I have to find you from within my own system." So, the immune system can like completely miss cancer cells. It makes them that much harder to target. But, there are, you know, errors and copies and changes in there that can be targeted. So, it's just it's possible, just more tricky. And that's why those tumor-specific mutations mattered so much that he was investigating before, because their unusual signals, their weird flags, that's like the person inside of the company acting funky, like not knowing something that everybody else in the company knows and you're like, "I can tell you're the fake one." So you have to train the immune system to look for that.
So Paul took his genomic analysis, he boiled it down to essentially a half-page formula and then he sent it to a new guy, new guy in the story, Paul Thordarson. I guess you'd say Thor Darson. Thor Darson. He is the director of the UNSW RNA Institute. And Thor Darson's team used it to create a bespoke mRNA nanoparticle vaccine. Not a mass market drug, but a custom-built treatment aimed at Rosie's particular type of cancer. And you want to talk about a a crazy fast timeline. Timeline from the moment that Thor Darson's team actually received Paul's formula to the moment of the finished vaccine was handed over to the vet: two months. And the director of the RNA Institute said it was the first time that a personalized cancer vaccine had ever been designed for a dog, right?
So, you think the story would be over now, but like obviously it's not because you can't just inject something like this into a dog. That's completely illegal. So, designing the vaccine didn't even seem like it was the hardest part. Then came the process of getting permission to actually use it. So, Paul goes back to AI. He spends three months, two hours every single night writing a 100-page ethics document trying to get it approved. Australia takes it like super seriously and rightfully so because you can't just inject homemade vaccines into animals. So, keep in mind, like I want to be very clear, this is like not garage science. Like if the headline ever makes it sound that way or you're hearing other people talk about this, like remember there was a lot of oversight here and he was using the AI tools as guides, doing important analysis, but always running these past experts and getting approvals. And then, you know, you just keep hustling at something like this and sometimes fate steps in. This is where a new woman enters the story, Mari Ma. She runs the K9 Cancer Alliance in the US. She just happens to stumble across a university article about Paul's quest and connected him to Rachel Alavina, a K-9 immunotherapy professor from the University of Queensland who, get this, already had the right ethics approval in place to do the administration. And that, you know, I got to say, that part kind of blew my mind because this is where the power of the crowds comes in. Mari Mari, yeah, Mari, she's like not supposed to be part of the story, right? She just happens to catch the news. So, this is kind of where wisdom of the crowds, or like getting something out into the open so like more people can connect, or even understanding just like how few degrees away we are from so many people who can solve problems is important. But anyway, she already had the right ethical approval in place. So, Rosie ended up getting her first injection in December 2025. She just got a booster in February and there's another one due soon.
And now you want to hear the results, 'cause that's the craziest part. This original Australian article said that the tennis ball-sized tumor on her leg shrank by roughly half. But guess what? It's outdated and it's gone away even more. 'Cause I found an article on this website called PupLink that was covering a Today Show interview where he reported, like from his own mouth, he said the tumor has shrunk by 75% to date. And that was said on March 14th, 2026, which is also today. So, as of today, that little guy has a 75% reduction in the tumor down there. Oh my god, I didn't even notice. He's got a little NASA scarf on. You little space space thing. Little space dog. So, he claims that Rosie's energy is coming back. Her coat is glossy again. At one point, she even jumped a fence at a dog park trying to chase a rabbit. Seems like a healthy thing for a wonderful dog to do. Something that she hadn't had the strength to do in months. But like also, like, get the fence bigger because you can't have dogs jumping out of those things. But anyways, that's the image that sort of sticks with me when I try to think of this whole story. And it's also about time and it's about buying more time and better time and like iterating towards things and working with people and using AI for ways that like humans can't do. I just really wanted to kind of deep dive into it and share this article because there's, it was like inspiring to me. I'm sure there's some things I could do, like the big picture things that I'm not being ambitious enough about. And probably for you, if you want to be an entrepreneur, if you want to shine at your company, if you just realize like how powerful it can be when you sew these things together. To me, this is what AI is actually for. This is the best of it. It's not replacing people. It's not like writing your homework. It's about giving a motivated person a ladder tall enough to look over the walls that used to be invisible. Paul didn't become a biologist overnight, but he worked hard at it. And AI gave him a bridge between what he already knew, data analysis, and advanced biology, which was like locked up before then.
Also, I kind of was thinking, gosh, mRNA cancer vaccines for humans, like there's a lot of them in clinical trials and like maybe we're closer to truly customized medicine or all sorts of cancer vaccines than I ever thought. I mean, if Rosie's case shows that it can be done on a personal level for a dog, like why wouldn't it humans be close behind? And we just, you know, I mean, obviously it's got to be safer before you're probably doing it in humans, but it's super helpful for cancer and probably all sorts of other stuff too. Anytime there's just misshaped proteins causing issues in your body, which probably is a ton, like a ton of issues. Like where genes create things that create problems, like that seems like it's not everything, but that's a huge, like there's a ton of health benefits. We get that figured out. And then of course, I, you know, I'd be remiss if I didn't also point out that it's like also kind of uncomfortable. Like if a guy with a laptop and a chatbot subscription can design a vaccine that shrinks tumors, like why aren't we scaling it faster? And also like what bad things could be created that way too? And do we have the right like safeguards in place that, you know, somebody wouldn't want to take money to create something that wasn't in the best interest of a person or a dog. So, tons of stuff probably needs to be thought through from like a geopolitical point of view, from like a governing point of view, from a health and and pharmaceutical point of view. I would love to see a bunch more people talking about this, like bringing it up, like thinking about how to prevent anything bad from happening when this becomes commonplace.
And yeah, if you like or don't like this kind of a format, let me know in the comments below. Obviously, like hitting subscribe to my YouTube channel would like make a big difference. The hype button, leaving some comments on your thoughts, like I'll, I'll read through it and kind of actively engage with everyone. So yeah, what do you think? And hopefully you learned something interesting from this or inspired you in some way. And as always, I will see you in the next video.