Four days in December

Before December 4, 2025, I had never written a line of code. No bootcamp. No computer-science degree. No technical co-founder. I was a practicing veterinarian seeing twenty-plus appointments a day, spending one to two hours every night charting after the clinic closed, managing hospitalized patients on a dry-erase board, and tracking callbacks on sticky notes.

At 6:35 PM on December 4, this happened:

Initial commit from Create Next App

That is the entire beginning.

The push

I didn’t start with a plan to become a programmer. I started with discharge instructions. I wanted to automate them, because even when a client comes in only to pay for an exam, I spend a great deal of time educating them, and none of that was surviving the trip home. First I asked my brother to build it for me. He told me, in no uncertain terms: do it yourself. Then he set me up with Claude Code and told me to go crazy.

So I did.

Week one

• Day one: an AI-powered SOAP notes generator, sign-in, discharge instructions, multi-recording, email integration, API cost tracking.
• Day two: the mobile app, with native recording on the phone, multiple patients and uploads.
• Day three: an AI assistant with an offline fallback, two-tap confirmation for anything destructive, keyword search.
• Day four: a rebrand, and discharge instructions translated into ten languages.

One hundred forty-three commits in four days.

December

Two thousand one hundred thirteen commits in the month. Shipped on Christmas. Shipped on New Year’s Eve. Shipped every single day. By the end of it: subscription billing, iOS and Android builds, background recording, a public landing page, and the small legal machinery a real product needs.

What the numbers do and don’t prove

I want to be precise about authorship, because precision is the point. The problems were mine. I lived them in exam rooms. The direction, the problem selection, the acceptance pressure: mine. The implementation labor was substantially supplied by AI agents, with me driving, reviewing, and dogfooding every feature against the next day’s appointments.

That combination is not the AI invented it, and it isn’t I hand-wrote every line. It’s a third thing, and it’s the discipline this whole site is about.

And the numbers don’t prove the software was good. They prove the loop was real: every friction point in my clinic day became a feature, usually within hours, because the person hitting the friction and the person shipping the fix were the same person.

What it became

That December sprint grew into a clinical platform, built over the course of several months while seeing 2,000 appointments. I see five to ten more patients a day because of it, and every conversation makes it back into the record now. And in March, when I wanted to code from my phone, the tooling I built for that became a project of its own, which is another essay.

Roughly seven months from that first commit to a published, open-source continuity engine. There was no plan. There was a loop, and it ran daily.