You Can Just Build Things

Don't wait to build something until the idea is "good enough." Just build. You'll be surprised what comes next.

The moment gpt-3.5 launched, it was clear that it was a better coder than I was. I was a full stack guy, but heavily biased to server and front-end. I could build functional prototypes, but in no way considered myself a bonafide developer who could ship something scalable to production.

Nevertheless, gpt-3.5 inspired me enough to try to get some ideas out of my head and into the world, for fun.

When I sold my company, I was very interested in climate change investing, so I did as much research as I could to determine what the "low hanging fruit" was in the world of climate change tech. I built a big spreadsheet, but thought it would be good to share my research with the world, as I found a lot of counterintuitive things in my research (did you know that afforestation is the biggest impact we can make on climate change?).

I used AI to turn that spreadsheet into a beautiful website, climatechangetech.org. I used Midjourney to create all of the images. It's a simple static javascript site, but there's no way I could have finished it without AI. This was my very first personal project that I launched in over a decade, and it only took me a week to build.

I was inspired. What else could I make?

I was taking the ferry to the city fairly often, but the website that showed the ferry schedule was a clunker. It took 3-4 taps just to get to the schedule, which was a big, hard-to-read table. So I used AI to build marinfairy.com. It did one thing really well: wherever I was in the bay, it would tell me when the closest ferry was next arriving, and how long it'd take. In that process, I learned how to scrape a website consistently, how to parse dates and times, and how to run a modern Node app on a Linux stack.

Inspired by that early win, I quickly got to work doing the same thing for the Polo Fields in Golden Gate Park. Their website had an even clunkier interface than the ferry website! The schedule page was so bad, it couldn't be scraped consistently. I had to scrape the text on the page, then issue an API call to OpenAI with the text as the prompt, asking "look at this blob of text. Are the Polo Fields open to bikes?" It worked great. polofairy.com done. Even more date parsing and scraping experience.

Next, I noticed I was getting a ton of emails from my kids' school in my inbox, and it was becoming quite the task to read through them all, parse out the pertinent information, add the appropriate calendar items, etc. So I built parentwrap.com, a tool that sifts through your emails from teachers, coaches, after-school staff, and other parents, then gives you a single daily summary of just the important bits, with a calendar feed.

I learned so much making ParentWrap. Full user auth; an email router; email parser; digest builder; sending digests with Postmark; sending/receiving data from the OpenAI API; and of course, marketing SEO. This was the building experience that made me believe that anyone could truly create a business with AI. You just need a good idea.

Then I made Repth, an AI cycling coach. I'm an elite-level competitive cyclist, and spent over a decade being professionally coached. I took those learnings and built an AI coach that takes an athlete's physical profile, goals, and performance history, and builds a custom, one-of-one training plan. It adapts on the fly based on sickness, or other life events that get in the way, which isn't even something every human coach does. I've been following Repth's plan exclusively for two years now, and I've never been stronger. It's just a hobby project, but 500 other athletes are enrolled now, too. I didn't do any marketing; but a few Redditors mentioned it in comments. That's all it took. Not only is Repth an app entirely built by AI, it's has the potential to succeed as a business. If the competitive cycling market were 10x larger, it would be.

Confident in my new abilities as a product builder, I expanded my horizons to making a native iOS app. My wife was lamenting the fact that there were no good veggie serving trackers in the App Store. They were either awful to use, or way too complicated. So I made Brocly, an app that lets you track your daily veggie servings in a fun and engaging way. It's a simple app, but I learned the basics of SwiftUI, publishing to the App Store, and more.

One day I was in the gym, and noticed how gross everyone's phones probably were. It's a room full of people sweating, lifting, and texting. I had my phone with me to follow along with an interval workout, but I thought: what if there were an app that let me follow a workout without needing to constantly touch my phone? So I made Etern1ty. It creates custom workouts with timed sections, and then an audio coach that guides you through the workout. Imagine: your own personal Peloton instructor.

Then... I had my first billion-dollar idea. I'm working on it now. I'm not ready to talk about it just yet, but I'm excited to see it come to life.

Don't vibe code; vibe ship

People seem to think that waiting for a billion-dollar idea is the justification that's required to start building it. Or worse, they'll happily abandon a project that's 85% complete if a billion-dollar distraction falls on their lap; Shiny Penny Syndrome.

The important part is not the building, but the shipping. If you're consistently getting in the "reps" of shipping products to the world, you'll naturally get better at it. You'll start noticing opportunities where you never did before, and your ability to evaluate them will accelerate, meaning the odds that you're working on things that have a real shot at creating value for a lot of people will only increase.