My father broke five generations of physicians. It was the best thing he could have done.

02 Where it started

I set out to build the tool I needed. Building it surfaced something larger, and unexpected.

Coming out of residency and into practice, I started thinking about collecting my cases. In OB/GYN, board certification requires it. To sit for the oral boards, you have to collect and then defend a list of the real cases you have managed. That is where I ran into the problem. The work is tedious. The stakes are high. It is exactly the kind of place where a physician’s judgment matters most. And exactly the kind of place where the software has always been an afterthought.

I expected to build a better case tool. What I found was that you cannot build it right without answering a much deeper question. So I followed the question. It led me to something I did not invent so much as uncover.

03 What I found

What I found was the gray zone. The space a guideline cannot reach, where judgment meets a real person.

It was always there, and no one had built the place for it. To make the tool genuinely safe, not safe by promise, I had to work backwards to one question. How do you build a place where a physician can be honest about judgment that does not match the guideline, without that honesty ever being turned against them?

Here is the tension I kept circling. In case logs you have to defend your reasoning to an examiner, to a colleague, and one day maybe to a lawyer. The decision was always an n of one. A physician makes the call, accepts the risk and the consequences, and reaches it together with the patient. That part is real, and the physician should own it. But the defense is an n of one too, and that is only how medicine has been, not how it should be. You are left to justify the choice alone, as though the judgment behind it had only ever been yours. It never was.

Every time I ask an experienced physician why they did it the way they did, I come away with something: a nugget of their lived experience that quietly informs my own. It happens all the time, over the water cooler, in the lounge before a case. It is the patient who does not quite fit the guideline, the one whose path is shaped by what they can afford, what they believe, and what they will actually follow through on. And it is two physicians thinking it through together, deciding what is right for the person in front of them.

That conversation is one of the richest things in medicine. And there has never been a place to keep it.

I’ll admit something here, because it is true and it explains how I missed it for years. I came up in the age of evidence-based medicine. We were trained with more evidence than we knew what to do with. Against all of it, anecdote looked lazy, uninformed, ignorant of the newest changes in medicine. From inside that certainty, we looked down on the physician who would not follow the guideline. The one who went with their own read after thirty years. I thought that was stubbornness.

Then I started practicing, and the patterns began to show themselves. A real patient rarely matches the average a guideline is written for, and the distance between the two is exactly where the medicine happens. That is when I saw what I had been dismissing. The veteran’s read was never just an anecdote. It was built on evidence, not against it. On the best evidence of its time, layered over years of watching what actually came next. Anyone who has been in medicine long enough has seen the cycles turn. A guideline turns against hormone replacement therapy, and years later a follow-up arm encourages it for the right patient population. The literature is a moving target. A physician who cannot point to this month’s paper is not therefore uninformed, and not unsafe. The absence of a fresh citation is not the absence of reasoning. Could some physicians have stopped learning along the way? Of course. But that is not the physician reading this.

And that evidence used to die with them. It did not translate. It sat siloed in a single physician’s head, the best of their judgment lying dormant, and then it was gone. Yet this is the knowledge that matters most. Judgment that is informed by the evidence and then tailored to the patient in front of you, carrying the weight of everything that physician has seen, leads to better care rather than worse. It is the whole reason a specialist is worth more than a chatbot. To a green new grad it looks like drift. To the physician who has lived it, it is evidence of another kind. It deserves to be preserved, shared, and protected. Not flattened away.

I did not create it. I discovered it. And then I began to build, so that the thing medicine has done in hallways all along would finally have somewhere to live.

04 Why this is becoming urgent

The longer I built, the more I saw where this is heading. Guidelines keep getting crisper, and AI is learning to surface, after the fact, what should have been done. In that world, tailoring a plan to a real patient stops looking like good medicine and starts looking like a liability — something you have to defend, whether or not it was right. Not wrong. Just exposed. And when the safe move is always to match the average, the choice fitted to the person in front of you gets rarer. This is the deeper damage, because innovation in medicine lives at the edge — it comes from the physician practicing at the limit of their understanding, secure enough to question a guideline instead of deferring to it. That questioning is not rebellion; it is how medicine has always improved. Take away the edge and medicine drifts toward the middle, until guidelines stop being a floor and become a ceiling, self-reinforcing, until the physician who steps past them reads not as an innovator but as a deviant. The distinction is everything. Get the incentives wrong, and the one quietly becomes the other.

I also came to see that the tools we have were never built for any of this. The record was built for billing and documentation, not for physician intelligence — so asking it a real clinical question, or building honest research on top of it, means fighting infrastructure that was assembled backwards, after the fact, for other purposes. The problem was never that the people inside these institutions are bad. It is that the foundation was never laid for the thing we now need it to do. Documentation was always a necessary evil, and now AI promises to take it off our hands — and the promise is real. An AI scribe can turn a messy visit into a clean note in seconds. But without a physician’s intelligence laid over it, something quietly goes missing. The note is cohesive, and it is flat. A physician walks a patient through the risks, the benefits, and the alternatives, and records a suspected diagnosis rather than a confirmed one, because confirmation still waits on labs. The scribe keeps what it decides is important and flattens the rest. Months later, when an auditor or a lawyer asks why you chose what you chose, the reasoning that would have answered them is gone. The voice in the note was never quite yours, and the judgment behind it was never written down.

So the work became clear. There has to be a place for clinical judgment that lives outside the record and outside the institution — led by physicians, built by physicians, owned by physicians. A place where a physician’s reasoning is safe enough to be honest, structured enough to be useful, and never turned into a score. It takes real work to build, and it has to be fed by clinicians to be worth anything. But it is the missing piece.

05 Where this goes

The case tool is the first thing, and it is real. But it was never only about case lists. It is the proof, in one specialty, of something larger. That clinical wisdom can be shared. That it can compound. Without surveillance. Without employer dashboards. Without the physician ever losing hold of their own records, or their own judgment.

That is the work. The first tool is how it starts. It is not where it ends.

If we care about patients, we have to care how AI enters medicine, because it is entering either way. AI does not create. It amplifies whatever substrate it is given. AI in medicine was promised to surface insight, and it mostly surfaced prettier versions of what was already broken. The missing piece was never the AI. It was the substrate beneath it: physician-owned, non-punitive, descriptive not prescriptive, statistically honest. Build that, and clinical judgment as it is actually practiced becomes something you can finally see, share, and protect.

By physicians, for physicians is not a slogan. It is a promise that the people who carry the liability are the people the tool is built for. Break it, and it becomes one more tool built for someone else.

The invitation

The best tools come from real frustration.

If you’ve stood in the gray zone — made the call the guideline couldn’t make, felt the pull to match the average just to stay defensible, watched good judgment get treated as a liability — then you already know this is real. You don’t need me to convince you. You live it.

This is being built for you. The physician who carries the decision, and answers for it. Not the hospital, not the payer, not the layer that grades the call from the outside without ever standing behind it.

The whole argument, and where it goes next, is on the PDI Med Substack.

Start there →