How Preg came to exist.

Megan tells me she's pregnant

She wanted to tell me in person, so she did, on a Thursday evening, with a small wrapped box on the dining table. I was working a string of night shifts that week and had come home in scrubs that smelled like the kind of long day I won't bore you with. She handed me the box. I knew what it was as soon as I picked it up, because she had been very nervous in the way she'd been moving around the kitchen, but I did not say so. I opened it. We hugged for a long time. Then I cried, then she cried, and then I made eggs.

The first thing I did after we sat down to eat was open the App Store on my phone and download every pregnancy app with more than a hundred reviews. We tried six that night. Three of them wanted Megan to upgrade for $49.99 to see the second trimester. Two of them had a baby-as-vegetable section that compared her baby to a lentil. One of them had a literal pop-up asking her to celebrate her bump. We deleted all of them.

I told her, half as a joke, that I could probably build something better. I had been playing with AI coding tools at work for a few months — mostly to write small SQL queries and unfamiliar config files. I had never built anything that ran on a phone. I said I'd try. We agreed I'd give it a month.

The first month

The first version was a single screen. It told Megan how many weeks pregnant she was. It was an Expo React Native app, which I picked because every AI coding tool I'd used seemed to know it well, and because it ran on both her phone and mine.

The next version had a journal. The version after that had a kick counter, which I'd built because Megan said she wanted one, and because the ones in the apps we'd tried had been either too clinical or too cutesy. I had no idea, when I built it, that the kick counter would be one of the screens we'd use the most.

Within about three weeks the app had nine screens. I showed Megan every Sunday. We sat on the couch and she opened things and frowned at me, and I would scribble notes in a small notebook, and the following Sunday I would have made changes. This is, more or less, still how we work.

I am building this app the same way I think about being a nurse. It is not about being smart. It is about paying attention to one person, every day, on purpose.

The Village

Our families were the next thing. Megan and I are both close with our parents and with each other's siblings, and the idea that we would post bump photos to Instagram for them the way we were supposed to was not appealing. I wanted to give them a porch instead. A place that was just for them and for us. So I built it.

I decided early that it should look like a cork board. Not as a gimmick, but because a cork board is exactly the right metaphor — a quiet, private, asynchronous place where you can pin a picture and somebody can come by later and leave a heart on it. I wrote a CSS masonry layout with little pushpins set at slight random angles, and I gave each card a tiny rotation so it never looked perfectly straight. I called it the Village.

Our families took to it instantly. The mom-in-law leaves a 🥹 on everything. My brother writes one-line comments that are funnier than they have any right to be. My dad, who is seventy-two and has never used a touchscreen with any confidence, learned how to leave a heart reaction in about thirty seconds.

The librarian

Around month four I started reading Megan's questions over her shoulder when she'd Google something at night. The Google results were not great, and the apps' built-in ask an expert features were either gated behind another paywall or staffed by people who, as far as I could tell, had no idea what they were doing.

I bought six books — a mix of clinical references and trusted lay-audience titles — and spent two weekends building a retrieval system on top of them. Vertex AI's Gemini 2.5 Flash as the model. Pinecone as the vector store. A Firebase Cloud Function as the backend. I called it Ask The Books.

It was the most technically ambitious piece of the app, and it was also the part I was the most cautious about, because the line between useful and dangerous is narrow when you are giving anyone medical-adjacent information. The whole thing is grounded in the chunks of the actual books, every answer carries a citation, and every answer that even brushes against something that should be evaluated tells Megan to call her provider. It is the most cautious AI I have ever built, and I think it is also the most useful one.

Labor and postpartum

Around month six I started writing Labor mode and Postpartum mode. I had been putting them off because they are the parts of the app I am most afraid of getting wrong. Labor mode, because Megan and I will not be reading carefully when we use it. Postpartum mode, because the things that get missed during postpartum — particularly mood — are the things that I, as a nurse, see the consequences of the most.

Labor mode is simple. Bigger buttons. Fewer screens. The contraction timer, the hospital bag, the birth preferences, partner coaching cards I wrote for myself after asking the L&D nurses I trust the most for their best advice, the family update sheet, and the big amber button that says Send Labor Alert.

Postpartum mode is the one I am the most proud of. It is the smallest of the three modes, and that is intentional. Recovery tracking for Megan. A feeding log, a sleep log, a diaper log. Mood tracking for both of us, because postpartum depression in the partner gets missed even more often than it gets missed in the mother. Daily prompts for me — written in my voice, with help from a couple of postpartum doulas I read carefully — that say things like take the night feed tonight and tell her one specific thing she did well today.

The trends

The Trends Dashboard was Megan's idea, although she did not call it that. She kept asking me what the journal entries were for. Once we had a few months of them, I built a small chart for mood. Then she asked about symptoms. Then about workouts. Then kicks. Then a calendar heatmap. Then a trimester breakdown. Then she pointed out that her mood was always lower the day before an appointment, and I built the line that would have shown her that.

Nine sections now. Every one of them is the answer to a real question she actually asked. I think this is the screen I am the most proud of as a piece of design — not because it is pretty, but because it is the screen most directly built out of a conversation.

About the AI in this

I will say this clearly, because I think it matters: I used AI coding tools heavily to build Preg. I would not have shipped it without them. I do not have a CS degree. I am a nurse. The way I have built this app is the way a lot of people will build apps in the next decade — with a question, a real problem, and a set of tools that lets the question carry more weight than the formal training.

But I want to be specific about what those tools did and did not do. They wrote a lot of code. They taught me a lot. They did not pick what to build. They did not decide how the Daily Journal prompts should sound. They did not decide that the kick counter needed an undo. They did not decide that Postpartum mode should be smaller, not bigger, than Pregnancy mode. Those decisions were ours, mostly Megan's. The taste in this app is human. The grunt work is automated. That is the right ratio.

Where we are now

Megan is twenty-two weeks along. The app is on her phone and mine. It has thirty-some screens. It has the Village. It has Ask The Books. It has the games we play across the day. It has a Trends Dashboard that has caught one or two real patterns I would have missed.

For a long time I told people I didn't know what we'd do with it afterward — maybe nothing, maybe it would stay a private artifact of a year we won't get back. That answer changed slowly and then all at once: enough people saw it over a shoulder and asked when they could have it that we stopped saying maybe. So Preg is heading to the App Store. It is not there yet — there's a waitlist and, since September, an install guide for the web version, but no store button, because I'd rather be honest than impressive. The subscription plumbing is wired, the listing is the next milestone, and the app it will ship as is the same one that's on Megan's phone tonight.

The thing that mattered hasn't changed, though. I got to spend a year of evenings building something for Megan, and the thing worked. Everything after that is extra.

— Nathan, somewhere in the second trimester.