The barrier to building a clinical calculator has collapsed.
The standard it has to meet has not — and the obligation just moved onto the clinician who described it.
The calculator that did not exist
A few years ago I wanted a specific calculator. Nothing exotic — a clean interface, a defensible formula, a result a woman could understand. It did not exist in a form I would put in front of a patient. The versions I could find were buried in journal appendices, locked inside apps that wanted a subscription, or built on an interface that assumed the user had a doctorate.
The traditional options were to find a programmer, find money to pay the programmer, and find six months. I had none of the three. So the tool did not get built, which is the usual ending to this story and the reason so much good obstetric evidence never reaches the bedside in a usable form.
The tools at tools.obmd.com exist because that ending changed.
I described what I wanted in plain English and iterated until it worked. I do not write code. I never learned and I am not going to.
What this actually is
The term people use is vibe coding, which is a silly name for something serious. You describe the behavior you want. The model writes the code. You look at what it produced, tell it what is wrong, and it revises. You are not programming. You are specifying, testing, and rejecting, which happens to be exactly what a clinician does all day with residents.
The reason it works for us specifically is that the hard part of a clinical tool was never the code. The hard part is knowing which formula, which population it was derived in, which inputs a clinician actually has at the bedside, and how to present a result so that it informs a decision rather than terminating one. That knowledge takes thirty years to acquire and no model has it. The code takes an afternoon and every model has it.
“How I build these, and the validation rule I apply before any of them goes near a patient. Paid subscribers read on.”
