The tool you build is temporary. The clinical judgment behind it is not.
The cost of turning clinical insight into a working tool just collapsed. Doctors should notice.
You no longer need an engineer to turn what you know into something patients can use.
“Tools will come and tools will go. Only the vibe coder remains.” A legendary music producer said it about software. It is also the best argument I know for why doctors should learn to build their own clinical tools. Here is what it means for medicine.
Andrej Karpathy, a founder of OpenAI, gave it a name in February 2025. Rick Rubin, who has produced more hit records than almost anyone alive, gave it a philosophy: “Tools will come and tools will go. Only the vibe coder remains.” Collins Dictionary named it the Word of the Year for 2025. Vibe coding has arrived.
Most doctors are watching from the sidelines. They should not be.
What vibe coding actually is
Vibe coding means building software by describing what you want in plain language and letting an AI write the code. You do not memorize syntax. You do not hire an engineer. You say what the tool should do, the machine builds it, you test it, and you correct it. Karpathy’s original line was to forget that the code even exists. You see something, say something, run it, fix it.
Rubin’s point cuts deeper. He produces records without playing the instruments. His value is taste and judgment, knowing what good sounds like and steering the artist toward it. The tool in front of him keeps changing. His ear does not. That is the part that lasts.
For a doctor, the ear is clinical judgment
Fifty years at the bedside, thousands of patients, the feel for what a number means and when it matters. That is the scarce thing. Until now, turning that judgment into a working tool required someone else. You explained the idea to a programmer who did not know obstetrics, waited months, paid a great deal, and often got back something that missed the clinical point. The translation step was the bottleneck. Vibe coding removes it.
My own numbers
I have built more than 100 clinical tools this way. Risk calculators, screeners, patient education pages, decision aids, all on tools.obmd.com and liveevidence.com, all created by directing AI rather than writing code by hand. To my knowledge I am the only ObGyn in the United States with a tool library of this size built this way. I am in my seventies. I do not say this to impress you. I say it because if I can do it, the resident who grew up with a phone in her hand certainly can.
The tools are the easy part. What doctors get wrong about vibe coding is the part that actually matters.
The judgment is the product
A software engineer can build you a calculator. He cannot tell you whether the threshold is right, whether the population it was validated in looks like your patient, or whether the output will frighten a woman at two in the morning over a number that does not warrant it. You can. The clinical judgment is not a feature you add at the end. It is the whole point.
This is why doctors, not engineers, should hold the controls. The tool is only as good as the question behind it, and the question is clinical. When I build a tool, I decide what counts as a meaningful number. I write the plain-language explanation. I choose what the tool refuses to say because the evidence does not support saying it. Those are not coding decisions. They are medical ones.
Where vibe coding gets dangerous
Now the honest part. Karpathy’s advice to accept everything and forget the code exists is fine for a personal weekend project. It is reckless for a tool a patient will trust. A tool that computes a risk score is making a medical claim. It deserves the same scrutiny as a drug or a device.
So the rules are simple. Every number in the tool traces to a published source. I check the math against the original model, by hand, before anything goes live. I do not ship a black box: if a tool gives a result, it shows its reasoning and cites its evidence. And no patient data is ever stored or transmitted. My tools run entirely in the browser and keep nothing. Privacy is not a setting you switch on. It is the design.
Here is the paradox worth sitting with. The thing that makes a doctor safe at vibe coding is the same thing that makes vibe coding dangerous in untrained hands: judgment about what the numbers mean. A layperson who vibe codes a pregnancy risk tool does not know the threshold is wrong. You do. That is not a small advantage. It is the entire case for why this skill belongs to clinicians.
What this means for patients
When your doctor uses a tool to estimate your risk, you have a right to ask three things. Where do the numbers come from. Has the tool been checked against the published evidence. And what happens to the information you typed in. A good tool answers all three plainly. A tool built by a clinician who understands the evidence can answer them. A black box bought from a company often cannot.
The skill outlasts the tool
The tool I use today will be gone in two years. Karpathy himself, one year after coining the term, already calls vibe coding passe and renames the professional version agentic engineering. The label keeps changing. The brand of AI keeps changing. None of that matters. What Rubin understood, and what doctors should understand, is that the durable skill is not the tool. It is knowing what good looks like and being able to direct the machine toward it. For a doctor, that is simply clinical judgment pointed at a new instrument.
Conclusion
I am not telling you to become a programmer. I am telling you the opposite.
The era of the doctor as the person who memorizes ten thousand facts is ending, because the machine remembers better than any of us.
What it cannot do is care for the specific patient in front of you, or decide what a number should mean for her.
That is yours.
Vibe coding lets you build the tools that judgment deserves, without waiting for permission, a budget, or a programmer who has never set foot in a delivery room. Rubin reimagined the Tao Te Ching for the age of code. I will borrow from a source he would appreciate, Marcus Aurelius: the obstacle is the way. For doctors, the tool was never the obstacle. The translation was. It is gone. Build the thing only you can build.
ObGyn Intelligence is independent because readers keep it that way. If this changed how you think about the next decade of your work, subscribe. The evidence, and now the tools, should travel.
References
1. Rubin R, Andreessen M, Horowitz B, Midha A, Torenberg E. Rick Rubin: Vibe Coding is the Punk Rock of Software. a16z Podcast. 2025. Available from: https://a16z.com/podcast/rick-rubin-vibe-coding-is-the-punk-rock-of-software/
2. Karpathy A. Post on X introducing the term “vibe coding.” February 2, 2025.
3. Collins Dictionary. Word of the Year 2025: vibe coding. 2025.
4. Taft DK. Vibe coding is passe. Karpathy has a new name for the future of software. The New Stack. February 10, 2026. Available from: https://thenewstack.io/vibe-coding-is-passe/
ObGyn Intelligence - Evidence Matters | obmd.com