For the Dog We Refused to Give Up On

On the same day Moderna's cancer vaccine news sent Wall Street into a frenzy, another smaller story reached a turning point: Paul S. Conyngham, an AI founder with no medical background, used ChatGPT, Gemini, and Grok to design a personalized mRNA cancer vaccine for his dying dog Rosie. The story went viral months ago — and now it has grown into a company.

Rosie was a Staffordshire Bull Terrier mix diagnosed with mast cell tumors, the most common skin cancer in dogs. Surgery and chemotherapy slowed the disease but could not stop it; immunotherapy failed, and the vet's prognosis was measured in months. Faced with what looked like a dead end, her owner chose not to give up — he decided to bet on a different approach.

Conyngham had 17 years of machine-learning experience but no formal biology training. He spent about $3,000 of his own money on paired sequencing of Rosie's tumor and healthy DNA at UNSW, then used AI to screen thousands of mutations for neoantigens and protein structure prediction to judge which ones the immune system might recognize: ChatGPT and AlphaFold for mutant protein identification, and Grok for the final construct design.

He did not make the vaccine himself — researchers at UNSW's RNA Institute reviewed the data and manufactured the mRNA construct in the lab. Gaining ethics approval in Australia took three months and a 100-page document. Rosie received her first dose in December 2025, a booster the following month, and several tumors shrank significantly with noticeably better quality of life. It is believed to be the first fully computationally designed mRNA cancer vaccine ever given to a dog — yet he consistently refused to call it a cure.

The Cost of Being a Pioneer

The ending was not a happy one. Rosie entered remission, but after a follow-up surgery the cancer returned: within four weeks tumors spread rapidly through her hindquarters, she was in pain, and she was no longer the dog he knew. He made the decision to euthanize her — the hardest decision of his life, he said. The first line he wrote when announcing the funding was: “I lost my dog because we couldn't build her next mRNA cancer vaccine fast enough.”

But the story did not end there. Conyngham founded Gamgee, a member of Y Combinator's Summer 2026 batch, and raised a $4 million seed round led by Founders Fund. The company's mission is to turn the process he had hand-crafted for Rosie into a repeatable, industrialized pipeline delivering personalized mRNA cancer vaccines for more dogs.

In his own words, Rosie was a pioneer in the truest sense — the first dog to receive a fully computationally designed mRNA cancer vaccine. Her medical record, and the months of painstaking ethics approval that preceded her first dose, became the blueprint for everything Gamgee is now trying to industrialize.

From Handcraft to Assembly Line: N-of-1 Medicine

Gamgee's path is clear and concrete: sequence each dog's tumor and matched healthy tissue, find the mutations unique to that cancer, identify tumor-specific neoantigens, design a personalized vaccine that teaches the immune system to target them, and manage the entire chain from veterinary intake, sequencing, design, production, treatment, and follow-up. Unlike peers like Torigen and ELIAS offering autologous vaccines, Gamgee's differentiator is N-of-1 — one bespoke vaccine per dog, not a universal target.

Founders Fund partner Amin Mirzadegan framed the investment rationale in terms of the person rather than the technology: “Paul crossed several disciplines for a dog, taught himself an entirely new field, and pushed forward where most people would have stopped.” Conyngham's long-term goal also extends beyond pet medicine: making personalized mRNA vaccines deliverable by any veterinary oncologist, and establishing N-of-1 medicine as a better treatment model across species — including humans.

Cancer Vaccines Are a Race Against Time

Put the Moderna news and Rosie's story side by side, and they ultimately circle the same problem: time. Merck and Moderna's Phase 3 results proved the feasibility of making a custom drug for each patient — but it took five years from protocol to Phase 3 results, and ten years from project start to today. Rosie went from diagnosis to first dose in about two years, three months of it spent on that 100-page ethics filing, while she only had months to wait.

What AI Can and Cannot Change

The core tension of personalized treatment is never just whether a design is possible — it is whether the total duration of the design-approval-production-delivery chain can be compressed to fit within what the patient (or the dog) can still wait for. AI compresses the first part of the chain — finding targets and designing sequences. The rest has not been compressed to the same degree. Conyngham's line about not building the next vaccine fast enough is exactly about this.

For those of us who build technology, the story is also instructive: AI is turning a single person's obsession into a repeatable process — just as we see in smart hardware and IoT, where AI is making high-barrier expertise accessible to ordinary builders. At the same time, clarity matters: a viral case is not a proven treatment, and professional work ultimately belongs to professionals and to time.

Rosie did not live to see the next chapter of this story. But her medical record is becoming an opportunity for many more lives — the $4 million Gamgee raised is essentially buying the ability to run that pipeline smoothly and produce evidence that regulators and veterinarians can accept. Sometimes the meaning of technology is simply making it in time. For the ones we refuse to give up on, humans will go very far.