This Pet Owner Used AI to Save His Dog

A guy in Australia with no biology background used AI and about $3,000 to help design an mRNA cancer vaccine for his dying dog. And it worked! The tumor is shrinking, the dog is doing well, and a lot of people in veterinary medicine are surprised. Welcome back to The Prompt. I'm Jim Carter III, and today I want to tell you about this story and why it's bigger than one dog's cancer journey. Back in 2019, a tech entrepreneur named Paul adopted a rescue dog named Rosie from a shelter in Australia. 4 years later, in 2024, vets gave him terrible news: Rosie had aggressive mast cell cancer. Mast cell tumors are the most common skin cancer in dogs. They make up about 16 to 21% of all skin tumors. When they're grade 3, the most severe grade, the 1-year death rate is around 91%. So when the vets told Paul Rosie had months to live, the data lined up with what they were saying. Like most pet guardians, Paul started with the standard route, surgery, chemo, thousands of dollars. Despite their efforts, the cancer still came back. Most stories stop there. People hear there's nothing left to do, and they brace for the end. But not Paul. He kept going and turned to AI for help. Paul isn't a biologist, a researcher, or a doctor. He's a tech guy. But he knew that some modern protein research tools use similar AI methods to those used in language models. So he set a goal: give Rosie more time and improve her quality of life. First, he paid around $3,000 to sequence the DNA of Rosie's tumor. That produced a huge amount of data. Canine whole genome sequencing at 30 to 100 times coverage can produce raw files around 100 to 300 gigabytes, and uncompressed alignment files can reach a terabyte. Then he used ChatGPT and Google's AlphaFold-like assistant tools. AI helped him find mutated proteins tied to Rosie's cancer and helped outline a custom mRNA vaccine matched to those mutations. That kind of work normally takes months of manual bioinformatics, and it takes skilled teams. Here, the first pass happened fast, with a laptop and a lot of persistence. To be clear, the AI was not perfect. Research suggests ChatGPT can help sort genetic variants with around 80% accuracy when it pulls together published info, and it can write code for tools like VEP and ANOVAR. It also has known problems. It can make up protein interactions at a rate of about 20-30%, and it can miss issues like tumor heterogeneity. AlphaFold is great at predicting protein shapes, but it is not the best choice for predicting peptide-MHC binding. So AI helped him get started, but it did not finish the job on its own. And let's be clear, he didn't try to make the vaccine himself. He took the AI output to real researchers and clinicians. With expert help, they produced a custom mRNA cancer vaccine in about 2 months and got permission to inject it into Rosie. 2 months from sequencing to injection is fast. In this case, Australia's compassionate use path through the APVMA made it possible. Veterinary biologics can move with fewer strict rules than human drugs as long as sterility and quality checks are handled. It's unusual, but it can be done. Within weeks of the first injection, the tumor started shrinking. Rosie's quality of life improved. Her immune system was being taught to spot and attack her specific cancer cells. She is not cured in the usual sense. There is no promise of a full win. Still, she gained time and comfort when standard care had run out of options. What Paul did is a smaller version of what many oncologists have worked on for years— personalized cancer vaccines. This is already happening in human medicine. The hard parts have usually been 3 things: speed, cost, and access. Paul's story shows that AI can shrink the early design stage from a slow, expert-only bottleneck into something that a driven outsider can start exploring quickly. The AI part of the design can be close to free compared to the million-plus dollars often tied to early drug discovery work. The most important part of the story is not the tech. It's Paul's mindset. He didn't give up just because he wasn't a biologist. He took action, used the knowledge he had, learned what he could, and then worked with experts to handle the parts that required labs, manufacturing, and safety controls. He focused on asking clear questions, checking answers, and finding people who could help do the work the right way. AI is improving fast, and we have access to more knowledge than ever before. What matters more is choosing a mission and moving on it step by step. $3,000, 2 months, and no biology background— that's where this story started. Whether you're building a business, solving a problem, or trying to figure out where AI fits in your life, the lesson is simple: you don't need to know everything at the start. You just need a clear reason to act and use the tools you have. If you want a space to do just that, come join my Slack community, Ctrl+Alt+Build. I share tools and workflows there and break down how people are putting AI to work in real ways. There's a growing group of curious, motivated builders, and I'm in there often sharing what I'm learning, and we work through projects together. I'll drop the link in the show notes. We would love to have you join us. Stories like this are such a cool example of what's possible with AI, and the future of AI gets shaped by the choices we make, one decision at a time. Thanks for tuning in, and I'll catch you in the next one.