ChatGPT helps GitLab co-founder Sid Sijbrandij turn a flood of cancer data into sharper questions, faster research, and a treatment loop designed to move faster than his disease -- a process that ultimately saved his life. Sid received his cancer diagnosis in November 2022, the same month ChatGPT launched. For nearly two years, he pursued what standard medicine could offer: surgery, radiation, chemotherapy, and an experimental therapy he had backed as an investor. The cancer returned. Eventually, the standard options ran out.
He responded by going “founder mode” on his own disease. Cancer, he reasoned, was not a static target. It was an evolutionary system that could adapt under pressure, which meant his team needed to observe each change, understand it, and act before the next change arrived.
That approach drew on habits Sid had developed while co-founding GitLab and helping grow it from an open-source project into a global, publicly listed software company: create a single source of truth, shorten the feedback loop, bring in distributed experts, and keep the work moving. He began applying those habits to his care.
Sid gathered every useful signal he could: frequent imaging, monthly blood tests for residual disease, DNA and RNA sequencing of the tumor, immune monitoring, tests of drugs on cultured tumor cells, and single-cell sequencing of the tumor microenvironment. As he described the inputs ChatGPT needed: “DNA, RNA, protein expression, etc.”
The data arrived from different laboratories in different formats, with the same gene sometimes represented in several incompatible ways. Sid and a small team built a lightweight ontology so a gene observed in DNA, RNA, or protein would resolve to the same concept. They combined that biological record with a machine-readable timeline of clinical decisions and results.
ChatGPT helped write the glue code, normalize the formats, and hold the unified context as new results arrived. When the team uploaded gene-expression data and asked for the most important patterns with supporting citations, the first pass recovered signals their experts had already found and surfaced new hypotheses to investigate. “It’s frankly unbelievable,” said Jacob Stern , Sid's business partner and the geneticist who supported him through much of his recovery. Each new result created another research problem. ChatGPT could condense a month of changes, flag immune signatures worth watching, assemble reading lists around an anomaly, and help the team ask targeted questions such as: if a particular marker is highly expressed, what published methods have been used against it? Stern used the system to cross unfamiliar domains faster and arrive at specialist conversations ready to press on the details.
Each month reset the questions they had to answer: A new scan, blood panel, or tissue result could change which mechanism mattered most. Because the system kept earlier tests, treatments, and papers in view, a new anomaly did not arrive alone: the team could compare it with the tumor’s prior behavior, the immune response, and the interventions already tried. The result was a running account of what had changed and where to look next.
Sid organized the work as an observe-orient-decide-act loop: draw blood and gather new data, use ChatGPT to synthesize the changes, put the output through expert review, choose the next action, and repeat. That cadence let the team survey approved, off-label, and experimental paths while tracking how Sid’s cancer responded.
The research fed a growing “therapeutic ladder” of plausible treatments that included targeted radioactive therapy, a personalized mRNA vaccine, and engineered cell therapies. Sid and his collaborators worked with clinicians, laboratories, and outside specialists to evaluate the ideas and, where appropriate, manufacture small-batch treatments around the biology of his specific tumor. “I went from having no options to having this therapeutic ladder today,” he said.
After targeted radioactive treatment and surgery, Sid reported that there was no evidence of disease in his body. He continues to monitor it aggressively because he sees the absence of a visible tumor as a moment in an ongoing contest, not permission to stop gathering data.
The system Sid built may be beyond what most patients can assemble, but he believes its parts point toward a future of patient-driven medicine. People with rare or non-standard diseases can increasingly obtain their own biological data, use AI to study it in context, and share what they learn with others whose conditions may be too uncommon to attract traditional drug development.
“We’re allowing people to become biotech companies,” Sid said. “That’s what AI’s enabling.” Each treatment that reaches production can make the next one easier and less expensive for another patient. He believes AI is bringing a level of personalized care that once seemed decades away to a small number of people now—and pulling the wider arrival date closer.
