How I Built a Parenting Tool Suite Using AI

I'm writing this at 4:47am. The baby is finally asleep on my chest, the toddler's white noise machine is humming through the wall, and I have about 22 minutes before someone wakes up screaming. This is when I built most of Zero Day Dad.

Let me be crystal clear upfront: I am not a developer. Before this project, my coding experience consisted of a Python tutorial I half-finished in 2018 and some HTML I learned by inspecting MySpace pages in 2006. I'm a dad with a laptop, a newborn, a toddler, a 5-year-old, and a wife who deserves way more sleep than she's getting. Zero Day Dad started because I was drowning, and I needed tools that didn't exist — or at least didn't exist in a way that made sense for an exhausted guy holding a screaming baby at 3am.

The Moment It Started: Why I Needed These Tools

Three weeks after our third baby came home, I hit a wall. Not a metaphorical wall — an actual moment where I was standing in the kitchen at 2am, holding a bottle in one hand, trying to remember when the last feed was, whether it was left or right breast last time (my wife was pumping), and whether the baby had pooped in the last 24 hours. The pediatrician had asked all these questions at our checkup and I'd just stood there like an idiot, scrolling through a Notes app that said things like "fed at 11ish???"

I downloaded five different baby tracking apps that night. Every single one did the same thing: they either wanted $9.99/month for basic features, buried the useful stuff behind three menus, or — and this is the one that really got me — looked like they were designed by someone who'd never held a baby at 3am. Tiny buttons. Pastel color schemes that blinded me in a dark nursery. Graphs I didn't need. Ads for postpartum yoga pants.

So I did what any sleep-deprived person with just enough technical knowledge to be dangerous would do: I opened a code editor and started typing.

Building the Baby Log: Start Simple, Solve One Problem

The Baby Log was the first tool because it was the most urgent. I needed one tap to log a feed, one tap to log a diaper, and a clear, glanceable timeline of the last 8 hours. That's it. Not developmental milestone trackers. Not "share with your pediatrician" features. Just: tap, log, done.

Every baby app I tried failed the "one-hand test." If I couldn't operate it with one hand while holding a baby, it was useless.

I used AI — specifically Claude and ChatGPT — to write the initial HTML, CSS, and JavaScript. Here's what that actually looked like: I'd describe what I wanted in plain English. "I need a button that says 'Feed' and when I tap it, it records the current time and shows it in a list below. Also make the button big enough to hit with my thumb at 3am. Also make the whole thing work offline because the nursery is a Faraday cage."

The AI would spit out code. Sometimes it worked on the first try. Mostly it didn't. I'd test it, see what broke, copy the error message, paste it back into the chat, and say "this doesn't work when I have no internet." The AI would fix it. I'd test again. We'd go back and forth five, ten, fifteen times. Each tool probably took 30-50 rounds of iteration.

The Baby Log now lives as a Progressive Web App. It stores data in localStorage so it works offline. It has dark mode by default because who wants a white screen at 3am? And it logs feeds, diapers, and sleep with exactly one tap each. My wife uses it. I use it. We stopped arguing about who fed the baby last because the log is right there.

The Contraction Timer: Built in a Hospital Room

The contraction timer has a slightly more dramatic origin story. During our third baby's labor, my wife's contractions started speeding up faster than we expected. I had a contraction timer app on my phone that I'd downloaded two years earlier for our second baby. I opened it. It was behind a paywall now. "Start your free trial." I actually laughed out loud in the hospital room. My wife did not laugh.

She was 6 centimeters dilated and I was standing there reading the terms of service for a freemium contraction timer.

I ended up using the stopwatch on my phone and scribbling times on a napkin. It worked. We made it. The baby arrived healthy. But as soon as we got home — literally the next day, while the baby was asleep on my chest — I built the contraction timer.

This one was trickier than the Baby Log. Timing contractions isn't just "record the time." You need to track the duration of each contraction and the interval between them. The UI needs to show you the pattern — are they getting closer together? Longer? That's the signal for when to head to the hospital. The AI helped me figure out the logic for calculating intervals, highlighting patterns, and displaying the last 6 contractions in a timeline view that a panicking dad could actually read.

The contraction timer has exactly one job: tell me when to get in the car. No accounts, no subscriptions, no "premium features." Just a big green button that says "Contraction Start."

The Sleep Tracker and Meal Planner: Tools I Wish I'd Had Sooner

After the Baby Log and contraction timer, I kept going. The sleep tracker came next because my wife and I had been doing the "how long did the baby sleep?" dance for months. I'd say 45 minutes. She'd say it felt like 20. Neither of us was right because neither of us was tracking anything — we were just exhausted and guessing.

The sleep tracker lets you tap once when the baby falls asleep and once when they wake up. It calculates total sleep time, tracks patterns over days, and shows you wake windows automatically. I didn't invent wake windows — pediatric sleep researchers did that — but I made it so you don't have to do mental math to figure out when the next nap should happen.

The meal planner was born from a different problem: my wife and I were eating absolute garbage. DoorDash three times a week. Frozen pizza. Whatever we could grab with one hand. I'd look in the fridge at 5pm, see ingredients but no plan, and reach for my phone to order takeout. The meal planner generates a weekly family meal plan based on what you actually have in your kitchen. You tell it your ingredients, your time constraints, and any dietary needs, and it gives you real recipes — not aspirational Pinterest nonsense, but actual food you can make while a toddler is pulling on your pant leg.

Building this tool with AI was fascinating because it required a different kind of prompt engineering. Instead of "make a button," I was asking the AI to act as a meal planning algorithm: "Given chicken thighs, broccoli, rice, and 15 minutes of prep time, generate 5 dinner ideas for a family of four where a toddler will also eat it." The AI nailed it most of the time. Sometimes it suggested risotto, which is insane when you have a newborn, but I learned to add "nothing that requires constant stirring" to every prompt.

What I Learned About Building Software With Zero Sleep

Here's the thing nobody tells you about building software with AI: it's not magic. It's a conversation. A long, sometimes frustrating conversation where you're both speaking slightly different languages and the baby might wake up and end the conversation at any moment.

Lesson 1: Start with the most annoying problem

Don't build a tool because it sounds cool. Build it because something is actively making your life worse right now. For me, that was the feeding log. I was losing track of when the baby ate and it was causing friction with my wife. Solve the thing that's burning.

Lesson 2: One-handed operation is non-negotiable

Every interface decision came down to: can I do this with one hand while holding a baby? If the answer was no, I redesigned it. Big buttons. High contrast. No menus within menus. No tiny tap targets. The AI kept trying to give me "elegant" interfaces with slide-out drawers and gesture navigation. I kept saying "no, bigger, uglier, works with my thumb."

Lesson 3: Offline-first or don't bother

Nurseries are dead zones. Hospitals have terrible cell service. Your phone drops WiFi when you're bouncing a baby in the hallway at 4am. Every Zero Day Dad tool works offline. The data syncs when it can, but it never fails when it can't. This was a technical challenge that took several rounds of AI iteration — localStorage has limits, IndexedDB is complex, and service workers are their own special kind of headache — but it was worth it.

Lesson 4: AI won't understand your users

The AI would sometimes suggest features that made perfect technical sense and zero parenting sense. "Add a share button so parents can post feeding milestones to social media." No. No parent at 3am wants to share anything. They want to go back to sleep. I had to be the filter — the domain expert who knows what exhausted parents actually need versus what looks good in a product demo.

Lesson 5: Ship before it's perfect

One of the AI's worst habits was trying to polish everything. "Let's add animations for the transition between screens." No. "What about a settings page with 15 customization options?" No. The contraction timer had one button when I first used it. It still basically has one button. Tools for parents need to work at 98% functionality, not 100% with 40% cruft.

The Tech Stack Nobody Cares About (But I'll Mention Anyway)

Everything runs in the browser. HTML, CSS, vanilla JavaScript. No frameworks, no build steps, no npm install. The tools are Progressive Web Apps that install to your phone's home screen. Data stays on your device. There's no backend, no database, no accounts, no passwords. This was a deliberate choice — I don't want your baby's feeding schedule on some server somewhere, and I don't want to maintain infrastructure.

I built most of this using Claude in Cursor and ChatGPT in a browser tab. The workflow was: describe what I want, get code, paste it into a file, test it, screenshot the problem, ask the AI to fix it. Repeat until it works. Sometimes I'd paste entire JavaScript files into the chat and say "this is supposed to do X but it's doing Y instead." The AI would find the bug, explain it, and give me the fix.

The HTML templates came from asking the AI to generate a clean, dark-mode, mobile-first layout. I tweaked the colors myself because the AI kept giving me corporate blue and I wanted something that looked like a dad made it — warm orange accents, enough visual personality to feel human, not sterile.

The Moment It All Clicked

About three weeks into using the Baby Log, something happened. My wife was out running errands with the toddler, I was home with the baby and the 5-year-old, and the baby had just woken up from a nap. I opened the log, saw the last feed was 2 hours and 15 minutes ago, saw the last diaper change was 3 hours ago, and thought: "Okay, feed now, diaper after." No guessing. No scrolling through text messages to see if I'd told my wife about the last feed. No waking a sleeping baby to check if the diaper was wet. Just a quick glance and a clear plan.

That's the whole point. These tools aren't supposed to be impressive. They're supposed to disappear. They do one job, they do it fast, and they stay out of your way so you can get back to the actual parenting.

The best parenting tool is the one you don't even notice you're using — until you don't have it anymore.

What's Next

I'm still building. The sleep tracker needs better wake window predictions. The meal planner could use more dietary filtering. The contraction timer actually works great and I hope I never need it again (three kids feels like enough). But the real next step is something I've been thinking about for months: making these tools work better together. Right now each one is standalone. I want them to share data — when you log a feed in the Baby Log, the sleep tracker should know the baby is awake. When the meal planner generates a shopping list, it should sync with your phone's reminders.

I'll build all of that. Probably at 3am. Baby on my chest, one hand on the keyboard, the other holding a bottle. Same as always.