Same Pattern, Different Stakes

It's late on a weeknight in February. My daughter Jessie is in the ER for a nosebleed that won't stop. She'd had a tonsillectomy two weeks earlier, and we were worried the scab had broken through her sinuses.

It hadn't. Just cracked sinuses. The doctor wanted to give her something to help and wrote up Afrin. I agreed. I texted my wife: "They're going to give us some afrin and give us some to take home."

Not a question. A status update.

She texted back: "It's ok with long QT?"

Jessie has Rett syndrome. She's four, nonverbal, and has a heart condition called Long QT, which can come from the Rett itself and/or be inherited. That makes her medication avoidance list stricter than most. I didn't know Afrin was on it. I didn't know where I'd check. My wife Googled it from the couch and caught it before the doctor came back.

That's the safety net. A text message and a Google search. And a little bit of luck.

She can't tell us what hurts. When she wants food, she climbs into your lap until you share. When something's wrong, she just looks at you. A doctor hears "she's looking at me" and thinks that's normal. We know it's the opposite.

We watch for a grimace. Less energy. A twitch that isn't usually there. Tiny changes that map to a wide range of possibilities. And then we have to translate those observations into something a doctor can act on in a fifteen-minute appointment.

There are two separate skills here: noticing and translating. We're getting better at the noticing through presence. But the translating is a different problem, and it's one that gets harder as the medical complexity grows. More doctors, more appointments, more medications, more research that changes every day.

A few months before the ER visit, I'd built a product sourcing tool for my merchandise company. A client sends a brief. The agent scans vendor catalogs, too many for a person to search efficiently, and returns the best options with reasoning.

After the Afrin scare, the pattern clicked. I had an input: a medication name from a doctor. There was a source of data I didn't know existed, a database called CredibleMeds that's the gold standard for QT-prolonging drugs. And I needed the right answer immediately, in an ER, in the middle of the night, with a four-year-old who didn't want to be there.

I didn't build a new tool. I applied a pattern I'd already learned. I used an LLM to survey what databases existed, what APIs were available, and how to connect them to the agent framework I'd already built. Within days it was on my phone.

Two years ago I wouldn't have known where to start. Not because the data didn't exist. CredibleMeds has been around for years. The Rett research is out there. The ingredients were always available. What changed is that I can now pull them together into something that works.

Not coding. Not prompt engineering. The skill is pattern portability. A vendor catalog problem and a drug interaction problem have the same shape. Recognizing that, then articulating it clearly enough for the tools to build the bridge, is the new literacy.

I assume the answer is out there somewhere. That assumption changes which questions I'm willing to ask. And now, even when the answer isn't out there, I can pull together enough information to build it.

The tool checks a drug name against Jessie's full profile: her conditions, her current medications, the QT risk database, Rett-specific interactions. It returns a verdict with sources and alternatives. It's not a replacement for her doctors. It's the layer between a parent's gut feeling and a medical decision that didn't have a safety net before.

We've talked a lot about the cognitive load that comes with a medically complex child. The constant question of whether you're doing enough. Whether you're feeding her the right things, giving her the right medications, catching the things the system misses. The tool doesn't make those questions go away. But it gives us back some of the space to be present. For her, for her brother, for each other.

Technology that reduces cognitive load frees people for the thing they're actually good at. For my team at work, that's serving clients. For my wife and me, it's being parents.

I write at jordanbartlett.co about what one person can build now that couldn't be built two years ago. Same kind of thinking as this piece, applied to companies, teams, and the occasional ER visit.