A few weeks ago, I found myself reading about Sid Sijbrandij, the founder of GitLab.
In 2022, Sid was diagnosed with osteosarcoma, a rare and aggressive form of bone cancer. Since then, he has documented large parts of his treatment journey publicly. Reading through it feels less like reading a patient's medical history and more like watching a complex research project unfold in real time.
There are imaging studies, pathology reports, genomic analyses, blood tests, treatment protocols, specialist opinions, clinical trial evaluations, and outcome measurements. New information arrives constantly. Existing assumptions are challenged. Treatment decisions evolve as the picture becomes clearer.
Most people looking at Sid's journey focus on the AI angle. There have been interviews and essays about how he uses AI to analyze information, generate hypotheses, and better understand his disease. That is certainly interesting.
But it is not what caught my attention.
What struck me was the amount of information required to understand a single human being.
Sid became the system maintaining context
As I worked my way through his treatment notes, I realized that Sid was doing something most healthcare systems struggle to do. He was maintaining context. Information was arriving from different specialists, different tests, different institutions, and different points in time. Yet somehow it was all being brought together into a single, evolving understanding of his disease.
In many ways, Sid had become the system responsible for maintaining continuity across his care.
That is not just an AI story. It is not even just a cancer story. It is a story about the hidden work required to make personalized medicine possible.
Perhaps this resonated with me because I have seen firsthand how quickly a serious illness becomes a data problem.
The more I thought about it, the more I found myself questioning the way we talk about personalized medicine.
The usual story about personalized medicine is incomplete
For years, personalized medicine has been framed as a scientific challenge. We need better genomics, better diagnostics, better biomarkers, better AI, better therapies, and better scientific understanding.
All of those things matter. In many areas of medicine, they have already transformed what is possible. Cancer care in particular has changed because of targeted therapies, tumor sequencing, immunotherapy, and molecular diagnostics.
Yet despite extraordinary scientific progress, personalized medicine still feels surprisingly difficult to deliver in practice. Most patients still move through healthcare in broadly the same way they always have. They see a doctor, undergo tests, receive a diagnosis, and start treatment. The science is better, but the system still struggles to maintain a complete and evolving understanding of the person.
I increasingly think that is because we are looking at only half the problem.
Personalized medicine was never just a science problem. It was always a data problem.
A patient is not a genome
The way we often talk about personalized medicine suggests that once we know enough about a person's biology, the rest naturally follows. Sequence the genome. Identify the mutation. Match the therapy. Problem solved.
Reality is far messier.
A person is not a genome. A patient is an accumulation of thousands of signals collected over time: laboratory results, imaging studies, symptoms, medications, procedures, diagnoses, family history, treatment responses, lifestyle factors, and outcomes.
None of these things exist in isolation. A pathology report means something different when viewed alongside imaging. A genomic finding means something different when considered in the context of treatment history. A blood test means something different when compared against five years of prior results.
The challenge is not simply collecting these signals. Healthcare already generates an extraordinary amount of information. The challenge is connecting it.
The patient should not have to become the project manager
That is what makes Sid's story so interesting. What he appears to have built is not simply an AI workflow. He has effectively become the coordinator of an enormous and constantly evolving body of medical knowledge about himself.
Information is arriving from different specialists, different tests, different institutions, and different points in time. The value does not come from any single source. It comes from assembling those sources into a coherent picture and using that picture to make better decisions.
That is what personalized medicine actually looks like in the real world. Not a magical AI model. Not a single breakthrough test. Not one revolutionary scientific discovery. It looks like building the most complete possible understanding of a specific human being and continuously updating that understanding as new information arrives.
The uncomfortable part is that this approach currently works best for exceptional patients. Patients who are willing and able to become project managers for their own disease. Patients who can coordinate specialists, collect records, understand research papers, evaluate treatment options, and maintain context across years of care.
If personalized medicine requires founder-level effort from the patient, then we have not really solved personalized medicine.
We have simply shifted the burden onto the individual. Sid could do that. Most patients cannot. Nor should they have to.
Personalized medicine needs infrastructure
The healthcare industry often talks about personalized medicine as though it will arrive through a scientific breakthrough. One more discovery. One more model. One more diagnostic tool. I suspect the future will be more mundane than that.
It will involve infrastructure.
The most advanced AI model in the world cannot reason about information it cannot access. The most sophisticated diagnostic tool cannot connect records that remain fragmented. The most promising treatment cannot be personalized if the broader story of the patient is missing.
This is where Aether fits. We are building the longitudinal data layer that makes healthcare easier to understand, share, and act on. Reports, prescriptions, imaging, vitals, and clinical context should not live as disconnected fragments. They should become part of a living health graph that follows the patient over time.
That does not replace science. It makes science usable at the level of the individual.
The real bottleneck
The future of personalized medicine will not be built on genomics alone. It will not be built on AI alone. It will be built on our ability to assemble a complete understanding of an individual patient.
The science is advancing rapidly. The harder challenge may be making that capability available to everyone, not just the patients capable of building it for themselves.
Personalized medicine becomes real when the system can remember the patient as well as the patient is expected to remember themselves.