The Sabbatical
Last year, after 23 years as a systems engineer in defense, I left my career. My role had progressed to Chief Engineer over a large portfolio of weapons systems. I hadn't written code in years, but I still orchestrated the entire process - what we call lifecycle engineering in defense, which I've since learned is analogous to what the tech world calls full-stack engineering. Requirements, architecture, development, testing, deployment, sustainment. I'd been doing it at scale for decades, just not with my own hands on the keyboard.
I left for many reasons, but most importantly, it just felt done. I didn't feel like I was learning quickly enough or growing. And then the deferred resignation program came along, which would basically give you a paid sabbatical through the end of the fiscal year to figure out what's next. It almost felt like a sign.
I took a seven-month sabbatical. It wasn't a full-blown midlife crisis. I didn't buy a sports car. I did buy a cool-looking jacket, and I almost bought a colonial apple orchard and cidery. I also planned out, in detail (even making webpages and interactive business plans), an organic micro-farming business, a house renovation and flipping business, a surf shop, and built tools for financial analysis.
I stayed plenty busy. But I kept coming back to one project.
Learning to Build With AI
During the sabbatical, I spent serious time learning the new AI development tools. I started with Cursor, eventually moved to VS Code with Claude Code, and watched the tools advance tremendously over the course of 12 months.
The concept of becoming a solo developer, tasking AI agents to write code, help with configuration management, requirements, roadmap development, transition to production, deployment and testing - that was second nature. I'd been orchestrating exactly this kind of work for my entire career, just with human teams and defense contractors instead of AI models.
I couldn't believe how quickly I could create and iterate. I established a development flow and process that works for me, and now I can integrate new features or fix bugs in minutes.
The early days had some rough patches though. With the first generation of coding tools, I spent a crazy amount of time (up until the wee hours of the morning) fixing silly bugs and pasting error messages back into Cursor. I'd switch from Claude to ChatGPT to Gemini as I got frustrated with one, or a new model came out. I'm not proud to say I said some very harsh things to a few of my AI assistants in those moments, and realized that yelling at a machine works about as well as yelling at a human as a motivation tactic (not well).
But when Claude Code was upgraded around the new year (and Codex too), things clicked. I'm not the first to write about this. But I spent my winter break just making tremendous progress. Very few deployment errors, no silly linter issues, I could almost skip straight to user testing. This was fun.
Why FuelTron
The project I kept coming back to was heavily influenced by two things: my oldest son, and a health scare that changed how I think about my body.
About two years ago, I woke up in a panic thinking I was having a heart attack. My heart was racing or fluttering, I couldn't catch my breath, and I felt faint. I woke my wife up and told her we needed to go to the hospital. We left the kids sleeping and drove to the ER. They hooked me up, ran tests, and were confident it wasn't a heart attack (it was palpitations, likely from electrolyte imbalance and dehydration), but they set me up with a cardiologist and more tests. My bloodwork came back not terrible, but not good. I probably wasn't going to die soon, or even in the next 10 years from a heart attack, but I certainly was not on the path to live to 100, or even 80. And I got shit to do.
That experience changed how I approached nutrition. I started using ChatGPT to help me understand my lipid levels and full metabolic panel, and using that analysis to figure out which supplements and foods had the highest bang for the buck. I also started learning about trade-offs. I want to lower saturated fat intake to bring down my LDL, but what am I giving up when I eliminate my favorite food (steak) from my diet, which also happens to be one of the most nutrient-dense foods I eat? That kind of nuance doesn't exist in a calorie counter.
I knew the nutrition app market was already crowded, and getting more crowded. But I'd tried many of the most popular tools, and some of the less popular ones, and none had the features or stickiness I was looking for. I wanted something more than counting calories. I wanted an app that looked at whole health. Something that would help me extend my healthy lifespan, and maybe even reach longevity escape velocity, even though I've spent the last 25 years not exactly treating my body like a temple.
The Influence of My Kids
Around the same time, I started thinking about what education and the job market were going to look like for my three teenagers. On a family trip, shortly after I left my job, I asked each of them to think about something they were passionate about, something I could help them start and build a business from.
I'm worried that the old world order - study hard, get into a good school, graduate, get a good job, work in an office - isn't feasible anymore. And I know my oldest son (and most of his generation) has no desire to follow that path, which was always unnatural and a quirk of this specific era anyway.
Both of my sons are passionate about nutrition, fitness, and a light form of biohacking. My oldest is in the gym at least five days a week, and he started to out-think and out-plan my wife and me when it comes to nutrition. He influenced our cooking styles, our ingredients, even the pots, pans, and utensils we cook with.
While I'm not a fan of TikTok, he managed to self-educate himself at age 16 beyond what 99% of adults know about nutrition, fitness, and health. I started listening. And I started building what I was hearing into FuelTron.
There wasn't one single moment, but many. He'd describe the benefits of fermented foods on gut health, or break down the nutrient differences between red meat and chicken, and I'd think "this should be in an app." Maybe at 16 he can keep all of this information and these trade-offs in his head, but I was getting lost. I needed something to help me make sense of it all. I tried listening to Peter Attia and Andrew Huberman and similar health influencers, and they had some good insights, but it also felt like they were selling something, and often talking down to their audience. I'm an engineer. I just wanted the data organized in a way I could make decisions. And that became the philosophy of FuelTron.
Building for Real People, Not Perfect Data
I read the Reddit posts and forums. People complaining about the accuracy of AI photo analysis, or even human guessing at portion sizes, cooking oils, and so on. The prevailing advice: weigh each individual food, eat simple meals, log every gram.
I think that's the wrong path. In my world, we call that local optimization, and it's a bad thing. You can get the most precise calorie count in the world on a single meal, but if the friction makes you quit after two weeks, you've optimized for nothing. Optimizing one variable at a time at the expense of the whole system is how you build something that passes the next test but isn't actually useful in the real world.
I took that same philosophy to FuelTron. This app needs to focus on entire-system optimization, or it's worthless. I wanted something good enough, something that people would actually stick with because it's so easy. I also wanted to upload my blood results and create an AI "Jarvis" (Iron Man, for the non-nerds) for health - one that would know my nutrition and fitness routine, my dietary preferences, allergies, inflammation triggers, and everything else that makes my body different from yours.
I also wanted something fun for the family to do together. So I built in a friends feed where we could keep each other accountable and entertained. Finally, a social app that's actually healthy for you.
FuelTron isn't 100% of the way there yet, but we're on the right path.
From MVP to Production
By the time my sabbatical was over, I had an MVP. I had it deployed on Render so the family could start using it. Then the new job started, and I had less time and energy for the app. The bugs in the MVP created enough friction that I started ignoring it. I also wasn't that interested in finding out that I was slipping back into my old ways with the 9-to-5 grind.
The improvements in Claude Code were part of what brought me back, but more importantly, that nagging question about what's coming as the world shifts beneath us, and what my kids are going to be left with, kept haunting me. My health wasn't getting better. I'd lost a few pounds, but my LDL and other biomarkers weren't going in the right direction. I was purposefully not logging meals because I didn't want FuelTron to report back to me what I was doing to my body. So I was in danger of leaving my family earlier than I wanted, while also not giving them a path in this new economic world.
So I'm back. I've used my nights, weekends, and vacations to fix the bugs, build in a ton of features, push the app into the Google Play Store (first for internal testing, now into production), and build this website you're reading right now.
Why I Think This Is Different
I did a ton of market research, and I'm convinced FuelTron is the best app for people trying to start a real health journey. Here's why:
- No annoying onboarding gauntlet. You don't sit through 10 minutes of questions before hitting a paywall. You're in the app, logging food, within a minute.
- Full-body health and wellness. This isn't a calorie counter. It's nutrition, bloodwork, wearable data, exercise tracking, AI coaching, and a social feed, all connected.
- A solo founder who actually uses it. I'm building this for myself and my family first. Every feature exists because I needed it. Feedback goes directly to the person writing the code.
- AI-native, and getting better fast. This is the worst AI will ever be. Which means this is the worst FuelTron will ever be. The app improves as the models improve, and I ship updates constantly. We're working on getting FuelTron into the Apple App Store, integrating more wearable and health-connected devices, improving our integration with the USDA database for nutrition consistency, and more.