N of One Is Not a Limitation. It Is the Model.

Watching the System Teach What Patients Already Practice

Yesterday I watched an FDA webcast on biostatistics. It was thoughtful and well presented. I learned from the presenter and respected the intent behind the lecture. But as I listened, I did what I always do. I went straight into research mode.

I opened my chat and started working through the concepts in real time. I applied them to galactosemia. I expanded them to cover ultra rare disease and n of one designs. I took what was feasible for galactosemia and pushed it outward to see how it could apply across ultra rare conditions and individualized care. I deeply believe in n of one care. I pushed past the examples being shown and asked what these ideas look like when they are applied to real patients with real heterogeneity. By the end of it, I had arrived at approaches that felt more flexible, more humane, and more accurate than what was being presented as the state of the art.

That moment stopped me. Not because the FDA was wrong, but because I realized how much I had underestimated AI.

Learning at the Speed of Reality

I come from a family where math and statistics were never intimidating. My dad is an accounting derivatives expert, and growing up these ideas were treated as puzzles, not barriers. Slope, change over time, rate of improvement or decline. These concepts feel intuitive when you live inside a chronic disease.

AI allows me to take that intuition and formalize it. I can move from a webcast to applied statistical thinking in hours. I can test ideas, challenge assumptions, and redesign frameworks without waiting for permission or institutional validation. That speed matters when you are dealing with a disease that does not wait.

Why Heterogeneity Is Not a Stop Point

Heterogeneity is often treated as a limitation in rare disease trials. It is described as a barrier, a complication, or a reason outcomes cannot be measured cleanly. But all disease is heterogeneous. Rare disease simply makes this visible because the population is smaller.

In my daughter’s disease, some patients have seizures and some do not. In a drug trial, my daughter became seizure free. That mattered. For another patient, seizures may not be part of their disease at all, but cognition, fatigue, speech, or motor function may improve. That matters too. The mistake is assuming that because outcomes differ, they cannot be measured within a shared framework. But they can…

N of One Does Not Mean N of Nothing

Each patient can have an individualized outcome profile grounded in their lived experience. AI makes this feasible. Slopes of change can be measured for each person relative to their own baseline. Improvement can be tracked across different domains without forcing every patient into the same endpoint.

This is especially important in chronic disease, where people often do not realize how unwell they were until a treatment removes the constant background burden. How do you measure feeling better when you did not know how bad you felt? You cannot define that perfectly at the beginning of a trial. But you can detect it over time. Slopes can show it. Patterns can reveal it.

Derivatives and change over time are not abstract math concepts here. They are lived biology.

AI the Bridge Rare Disease Has Been Waiting For

AI is not replacing science. It is allowing science to finally meet lived reality. It helps patients translate experience into structure. It allows caregivers and advocates to design better questions, better measurements, and better frameworks than those built around convenience or fear of complexity.

Rare disease communities have been asking for this kind of care for forty years. Individualized. Flexible. Respectful of heterogeneity. We have always been ahead of the curve because survival required it. What has lagged is not our understanding. It is society’s willingness to listen.

The Question I Could Not Ignore

After that webcast and that research session, I had a thought I could not shake. What happens if institutions cannot keep up with this pace of learning and design? What happens if patients and small teams can generate better solutions faster than centralized systems can evaluate them?

This technology is moving faster than any single person or institution can fully process, operating across layers of abstraction so quickly that by the time we try to describe it, it has already changed.

AI made something visible to me yesterday. Not that the FDA is irrelevant, but that it risks becoming late. And in rare disease, late is not neutral. We will move on to better models that actually cover our needs. The question becomes whether the FDA adapts or becomes obsolete.

This is not about replacing regulators. It is about recognizing where innovation is actually happening. N of one patients, undiagnosed patients, ultra rare communities. They are already at the front line. AI simply gives them the tools to show what they have always known. And anyone who knows rare disease knows these communities are innovative, deeply knowledgeable, and intimately aware of the failures in our medical system. They will be the ones solving it and presenting new models.

The future of medicine is not waiting to be invented. It is already here, built quietly by people who could not afford to wait.