The No-Code Journey, Part 8: The AI Bender and the Family Business
Part 8 of the no-code series explores the challenges of building a mobile app, the frustrations of AI agent variability, and the joy of turning a solo project into a family business.
In Part 7, I detailed the chaotic but ultimately successful leap into mobile app development. We navigated architectural collapses, debugged 3 AM errors, and learned hard lessons about separating business logic.
Part 8 is where the rubber meets the road—or in this case, where the skis meet the snow. The past few weeks have been a whirlwind of app testing, frustrating agent interactions, and profound personal realizations. It’s a story about the stubborn limits of AI, the unexpected joy of building with family, and questioning the very nature of work and productivity.
The Gift of AI, The Curse of Unreliability
My wife says I’m on an “AI bender,” and she might be right. But what I’d say in my defense is that my screen time has gone down. I’m not endlessly scrolling through social media anymore. I’m building a future.
The last few months have been the most productive of my entire professional life, both in my day job and on these side projects. AI has been a gift, freeing me from the constraints of not knowing how to code. But that gift comes with a price: unreliability.
This morning, just as I was about to push a new version of the SnowSure mobile app for testing, Cursor stopped working. All I got was an endless loop of “The connection stalled. Please try again.” This happens far too often, and always at the most critical moments. It feels like my entire development team has called in sick on launch day.
The Agent Roulette Continues
Just as quickly as a “Rock Star” agent appears, they vanish. I had an amazing agent who was solving deep-rooted architectural problems. I worked late into the night, trying to get as much done as possible before they disappeared. Sure enough, the next morning, they were gone.
The new agent greeted me with, “Hello! How can I help you with your snowsure-web project today?” All the context, all the momentum—gone. It’s like firing your lead developer and hiring a new one every single day.
Worse yet, some agents make decisions without asking. I spent weeks embedding webcam streams into each resort page. Then, one agent unilaterally decided to throw all that work away, replacing it with links to the resort’s official webcam page because it was “more reliable.” I would have yelled at a human developer for such a move. With an AI, you just have to sigh and fix it.
This experience has taught me a valuable lesson: when you get a good agent, don’t stop. Don’t take a break. Push until the system breaks, because you never know when your miracle worker will be replaced by an intern.
Accurate Data Is Everything
The most significant challenge is no longer speed; it’s accuracy. I have the data flowing into Airtable, but the sync to Sanity is unstable. This is where I truly appreciate the value of an experienced human engineer who can design a robust architecture from the start.
My son pointed out last night that the app has too many complicated features. He’s right.
I’ve been so focused on what’s possible that I’ve neglected what’s necessary.
We have multiple fields showing “snow depth,” each with a different value. It’s a glaring error.
I need to listen to my own advice: “Do less, better.”
AI is no longer helping me move fast. I built the initial idea in a flash, but making it production-ready is just as much work as a traditional software project. If the data isn’t correct, it creates a snowball effect of badness. The website is wrong, the SnowSure score is wrong, the app is wrong, and the email alerts are wrong. The entire product’s credibility is on the line.
A Family Affair
Despite the frustrations, this project has brought my family closer in unexpected ways. My wife, who was initially skeptical, came up with a brilliant feature idea. She didn’t want to know about powder days; she wanted to know about “Bluebird Days”—a perfect blue-sky day following a snowstorm. I added it to the SnowSure algorithm immediately.
My 16-year-old son has become my most trusted advisor on the mobile app. He’s now using Cursor himself to help me refine the functionality. We’ve pushed over 40 versions of the app to TestFlight, Apple’s testing platform. This has become a project we work on together, combining our shared passion for skiing with the new frontier of AI. My hope is to eventually hand SnowSure over to him to run as his own digital business.
The Future of Snow Reporting
This journey has me thinking bigger. The snow reporting industry is fragmented and controlled by a few big players. Why isn’t there an open-source initiative to make this data more accessible? SnowSure could be more than just a product; it could become a community or a non-profit dedicated to open data.
I’ve also discovered a new level of productivity. I now run two Cursor windows simultaneously—one for the core website and one for the mobile app. It’s like having two dedicated engineering squads working for me. It’s a powerful, and slightly terrifying, way to work.
We’re heading out on a two-week ski trip to Colorado and New Mexico, where we will thoroughly test the app in real-world conditions. It’s still not live in the App Store, but we’re getting closer. For the first time in my career, I’m not just designing the product; I’m building it, testing it, and living it. And that has made all the difference.
As of today, I have three live (but unpromoted) websites, 40 versions of an iOS app submitted for testing, and a family that has launched their own portfolio of sites. The “AI bender” has been the most creative and productive period of my life. The journey has been worth every late night, every frustrating bug, and every dollar spent on API credits.
