Every week, AI gives us reason to panic. And look, there’s no denying that caution should be real. But I also fear the good doesn’t get enough light. Meet @paul_conyngham & his dog Rosie. He used AI to shrink his dog’s cancer tumors by up to 75%. Here’s how he did it and what it means for the rest of us: Paul is a tech entrepreneur out of Sydney. He spent 17 years in machine learning and data science and has zero background in biology. Rosie, his staffy Shar Pei cross, he adopted from a shelter back in 2019. She’d been abandoned in bushland. So he gave her a new home. In 2024, large tumors started showing up on one of her back legs. The diagnosis was mast cell cancer, the most common skin cancer in dogs and typically incurable through conventional treatment. So Paul threw everything at it because dogs are man’s best friend. Multiple surgeries. Chemo. The chemo even slowed the spread but the tumors wouldn’t shrink. Vets estimated she had one to six months left to live. Traditional medicine had nothing left to offer him. So he did something most people wouldn’t even think to try. Some would even call it unsafe and unreasonable. He used ChatGPT to research possible treatment paths. It directed him down a path of immunotherapy and genomic sequencing. From there he started building out a pipeline. Sequence Rosie’s regular DNA and tumor DNA, compare the healthy cells vs cancer, identify the mutations driving the cancer, and design a custom vaccine that targets those specific mutations. People told him he was crazy. But he didn’t care. He used Google DeepMind’s AlphaFold to model the protein structures, and even wrote his own algorithms for selecting which mutations to target. After that, he brought all of his research to scientific researchers at UNSW’s RNA Institute, who looked at his work and agreed to collaborate. The director of the institute, Pall Thordarson, took Paul’s sequence design and turned it into an actual mRNA vaccine in under two months. Same delivery method as the COVID vaccines. Lipid nanoparticles carrying instructions that teach the immune system to recognize and attack the cancer. This was no longer some garage science experiment or AI slop research project. It took Paul months of required formal ethics approval and writing a massive document submission on top of everything else just to get his dog’s health back. A researcher at the University of Queensland who already had the right approvals in place helped him get it across the finish line too. Rosie got her first injection back in December 2025. Within a month, the results started showing. Her main tumor shrunk by up to 75%. She went from barely moving in December to jumping a fence and chasing a rabbit by January. Her coat even got glossier. And her energy came back. The vets said she just looked like a completely different dog. Not every tumor responded though. So they’re already sequencing one that was resistant to figure out why. This is still very early and experimental. But the immune system activated against the cancer in a real and measurable way. Promising. Where traditional medicine gave up, AI powered medicine finally showed up. And here’s why this story could change how we all think about our own health. Personalized mRNA cancer vaccines are already in human clinical trials. Moderna and BioNTech are both running them for melanoma, pancreatic cancer, and other tumors. The pipeline Paul ran on Rosie is the exact same concept. Sequence the DNA, find the mutations, predict which ones the immune system can target, and design a custom vaccine. What used to take pharmaceutical companies and billions of dollars is starting to compress. A data scientist with no biology background designed the core of this vaccine using tools that are publicly available right now. The head of the RNA Institute said it himself. AI is democratizing the whole process. We all remember WebMD. You’d type in a headache and convince yourself you had a brain tumor or some rare disease by the time you closed the tab. Self diagnosis became a joke because the tools were bad and the information was generic. That era is ending. AI is already outperforming specialists in certain areas. Studies where AI beat dermatologists at detecting skin cancer. Imaging analysis where AI caught tumors that radiologists missed. And now a guy in Sydney used publicly available AI to help design a cancer vaccine that actually worked for his dog. We’re getting closer to a world where regular people with access to the right tools can catch things their doctors don’t. Not because doctors are bad at their jobs. But because no human can process the volume of medical data and research that these systems can. The person who WebMD’d themselves into a panic back in 2012 and the person running their symptoms through an AI model in 2026 aren’t as far apart as you’d think. That’s not a threat to medicine. That’s medicine getting better for everyone. And this isn’t me encouraging you to dump all your private or sensitive data into AI models, but rather to practically think about the infinite use cases and workflows now at your disposal with AI. Our health could rely on it one day. Stories like Paul and Rosie deserve the same energy and coverage we give the doom and gloom too. Because the future of your health might not always start in a hospital. It might start with a prompt. Yes, a prompt could save a life. Or yours.

$200/mo claude max bros when anthropic doubles their usage

RT @kloss_xyz: bros seeing Claude’s 1M context window