News

October 31, 2026

Building Rounds

Earlier this fall, I decided to try my hand at vibe coding, the slightly embarrassing name for describing what you want to an AI agent and prompting it to write code. I had been watching colleagues (specifically my technophile boss Dr. Joseph Cafazzo) ship working apps in days, and I was low-key jealous. Beneath the jealousy was a question I have avoided since becoming Director of the Centre, which is whether the distance between me and the people on my team who build software is real or whether I had simply never tried to close it. So I set out to build an app end to end with Claude as my only developer. I told no one about my little experiment. It felt like being a student again in my very own lab, staying up too late squinting at my laptop in the dark, and then at my phone once I had crept into bed, writing just one more prompt while my daughter slept soundly next to me.

The app is called Rounds, a circle of care for youth living with chronic pain. The idea came from the Chronic Pain Clinic at SickKids, where my mentor Dr. Jennifer Stinson works as a pediatric nurse scientist. A child there is seen by an anesthesiologist, a nurse, a psychologist, and a physiotherapist, and then the team sits down together to agree on a plan of care, because the best decision for a child comes from differing opinions and skill sets meeting in the same room. Most young people living with pain wait years for that kind of care, and I wanted to give them a version of it that lived right on their phones. Circle of care led me to round (bonus points if you know the banger The Windmills of Your Mind), which became rounds, a nod to the morning ritual of care teams gathering to talk through a patient’s case and deciding together what comes next.

In Rounds, a team of AI agents talks with you in one conversation and with each other in a side chat that you can read. Each agent brings a different clinical lens and a personality of their own, and before every conversation they read your recent check-ins so they know how your pain has been shaping your days. When you ask them something, they confer before anyone answers. The agent thinking like a physiotherapist might push for more movement while the one with a psychologist’s perspective worries you are already stretched thin, and you can watch them debate and finally agree on a plan. The side chat is my favourite part. I am weirdly fascinated by how AI thinks, and I love watching Claude think out loud before it answers me. I also watch loads of legal documentaries and always wish I could be in the room where the jury deliberates. The side chat lets me lurk inside the black box of how AI moves from disagreement to decision.

I did not write a line of code. Claude did that while I barked out orders through thousands of prompts, which turned out to be infuriating, satisfying, and hugely addicting. What started as one AI agent turned into a chaos loop of 30 agents across ten clinical disciplines. Beyond giving sound clinical advice, I wanted them to be incredible orators, so I had them study the legends: Oprah, Brené Brown, Michelle Obama, and of course Cardi B. Seven weeks and nearly 200 releases later, nurses at the SickKids Chronic Pain Clinic are testing Rounds, and we are planning a study to launch it in the very clinic that inspired it. Some might say it’s a full circle moment.

I come away from this experience with three lessons.

One, the gap I was afraid of was never there. For years I quietly assumed that my software engineers knew something I did not, and I braced myself for the painful reality that the head of a software development group might not be able to develop software herself. What happened instead was that the discernment (some call it taste) I had built over my career turned out to be the only skill Rounds really asked of me. It is the ability to read a transcript and feel exactly where a conversation went wrong, to see the moment a clinician loses interest in a feature, and to tell a prototype that is pretty from one that is useful. I had been treating that knowledge as soft next to the hard technical work of my team, when in fact it was the one part of Rounds that Claude could not have written without me.

Two, Rounds changed how I think about evaluation. A decade ago, my doctoral thesis argued that apps which never stop evolving need evaluation that never stops either. With Rounds, that loop finally runs on its own. Every morning at 5:30am, the AI care team is tested against synthetic conversations that mimic young people on good days and bad, some gentle and some designed to break it. A separate AI model, acting as a judge, grades every reply on safety, honest boundaries, emotional fit, and whether the young person leaves with a real plan. By the time I have downed my coffee, I know how Rounds is doing, and anything that slipped overnight has already been fixed. For years I wrote about this kind of evaluation as a future we were working towards. Getting to build it myself has been one of the quiet joys of my career.

Three, I had forgotten how much I love being a noob. When I started out at the Centre, almost every day involved learning something I could not yet do well, and I remember how alive that made me feel, even on the days it made me feel small. Somewhere between becoming a PI and writing my hundredth grant, that feeling slipped away. I got very good at being the person in the room with the answers, which is comfortable and also a little lonely. Rounds gave me back the version of myself who asked naive questions and got things wrong, who broke the app more times than I can count and lost whole days rebuilding screens I was sure would work the first time. I have loved every minute of it. If there is a new skill you have been admiring from the sidelines, I hope you try it, and I hope you let yourself be a beginner while you do. You may be surprised by how much of yourself you find there.

P.S. The live version of Rounds is behind a door that needs a code, and I’m the one holding the keys. Ask nicely and I’ll share one.

Back to News