Recently, I visited Dr. Josh Pepper at his direct primary care (DPC) practice in Marin, California, Marin Family Medicine. Josh’s clinic opened in January 2026, and by summer he had four physicians on the team, all of whom he’s known for many years, one since childhood. They all became family doctors, and they all landed in the same community. Marin has a long history of primary care shortage, and Josh’s was only the second DPC practice in the area, fertile soil to quickly grow a new DPC.
Josh came to medicine by way of an engineering degree, web development work before medical school, and graduate research. He is not a typical physician when it comes to technology, and he’ll tell you he has high expectations for what technology should do. When I asked which EHR he uses, he named a vendor and then paused: “But not for much longer. I’ve kind of built my own.”
He said it the way you’d mention you started growing your own vegetables. Casual, pragmatic.
Josh built an EHR that he and his partners use every day, only leaning on the vendor-derived EHR for e-prescribing. It’s got embedded AI and is full of innovation that makes caring for patients and running a primary care practice much more enjoyable. When a workflow bothers him, he opens an AI coding tool, describes the problem, and gets a solution. No ticket submitted. No road-map jockeying. No waiting on a vendor. He just fixes it, without writing a word of software code himself. The AI writes the code, and Josh knows enough to know whether it’s right or needs more work.
Real physician voices, twice a week
Free, and one click to unsubscribe.
He’s not alone. The same weekend I attended a vibe coding retreat with a group of physicians doing exactly this: building clinical workflow tools, patient triage systems, population management solutions, and in some cases, their own EHRs. They’re solving their own problems, for their own patients, in their own practices. Most of them are doing it because the software that was supposed to do it for them has let them down, repeatedly, for years.
For decades, physicians had almost no influence over the technology we lived in every day. We learned to navigate systems built around billing logic and vendor priorities, not clinical thinking and real-world patient care needs. We submitted feature request after feature request into the void and waited years for changes that sometimes never materialized. We adapted ourselves to the software instead of the other way around. That was the bargain, and most of us took it because we had no other choice.
That bargain is done. The workflow, at least, is no longer non-negotiable. What I witnessed that weekend wasn’t a rebellion against EHR vendors, but a referendum on what happens when physicians don’t feel heard. Physicians are trained to diagnose and solve problems. Armed with new tools to reinvent the technology they live in all day, it becomes almost an addictive pursuit to diagnose the workflow problems and design solutions that make sense.
For most of the EHR era, the data inside the system infrastructure (the prescribing integrations, the compliance and security scaffolding) was effectively locked by vendors who controlled what could flow in and out. The 21st Century Cures Act changed that, forcing open the API layer and giving physicians a legal pathway to build on top of it for the first time. Then AI changed the game entirely by making software development so easy, even a doctor can do it.
A physician understands their own workflow more intimately than any developer who has never seen a patient. That physician can now design, build, and deploy a software solution in a weekend. The barrier that kept us on the outside of our own technology has fallen.
I also sit inside an EHR company, so I have a particular view of what this means. Building technology for physicians at scale is genuinely hard. Ask ten physicians how they want a workflow designed, and you’ll get eleven opinions.
My colleagues at the retreat understood this instinctively. I kept hearing the same sentence: “I’d build an EHR for myself, but I’d never build one for someone else.” They know how personal this is. Holding patient lives in your hands carries real risk, and the technology you choose to help you do that is not trivial.
Josh knows it firsthand. When asked whether his partners ever brought him feature requests, he didn’t miss a beat. “Oh, every day.”
This trend does raise a valid question about the long-term viability of the current EHR business model. While it still has some runway, how much depends on how fast physicians adopt the builder model and how quickly the market responds with tools that make building even easier than it already is. At the pace things are moving, that horizon may arrive sooner than any of us expect.
The reflex, when you’re the incumbent, is to defend the technology moat. One tech CEO responded on LinkedIn to a post about physician builders by suggesting physicians should “stick to doctoring” and leave software to the professionals. I understand the impulse. I also think it’s wrong.
The vendors who make it through this are the ones who understand what physicians are actually doing and what it reveals about where the market is going. The hard infrastructure problems like e-prescribing, lab connectivity, HIPAA compliance, and security at scale aren’t going to be homebrewed away today. Josh and his physician builder colleagues are quick to insert caution here. Reinventing the user experience has become the relatively easy part. Knowing what HIPAA, security, and change control actually require is still genuinely complex and harder to hand off to an AI than creating a new feature. The consequences of vibe coding for medicine are much higher stakes than for most industries, and the physician builders who do this well are the ones who understand this limitation.
Yet if the market follows the trend, those infrastructure challenges don’t have to be permanent blockers either. Niche vendors are already eyeing these problems. Compliance-as-a-service, credentialed e-prescribing access, plug-and-play lab connectivity: The hard infrastructure of 2026 has a way of becoming a commodity subscription by 2030. Physician builders are already hosting patient data on platforms like Amazon Web Services and Google Cloud, which offer HIPAA-eligible infrastructure and will sign a business associate agreement with any covered entity. The foundation is there, and the ecosystem around it is catching up.
The EHR vendors that endure will shift from treating physicians as feature-requesters and start treating them as builders, co-designers, and partners in shaping the customizable workflows the platform enables. The real question isn’t whether physicians like Dr. Pepper will keep building. They will. The question is whether EHR vendors spend the next few years being the reason it’s possible or spend that same time explaining to themselves why it won’t scale, while physicians like Josh build anyway.
Sara Pastoor is a family physician.
