🐴 Passion Projects and AI
How Mike uses AI to build passion projects
☀️ Happy Friday!
Today's guest post is brought to you by Mike, a high school teacher I connected on Substack. He’s also part of our Design with AI community.
Mike isn’t a designer or an engineer, yet he has leveraged AI to bring many passion projects, like Movie Quote of the Day, to life.
There was a lot he didn’t know, but he dove in, learning by doing.
Isn’t that inspiring?
The reason might lie in his own words: he’s “insanely curious”.
With that, I'll hand it over to Mike!
Hey friends!
I’ve been asked to share some thoughts on how I approach passion projects with help from AI. So here goes:
First off, who am I?
My name is Mike Macfadden. I work in the north suburbs of Chicago as a high school teacher of business, technology, and design. I’m 39 years old, a husband, and father of two young boys.
The intersection of my instructional responsibilities closely match my personal interests. I love fiddling with computers, making things (both real and digital), and learning about startups. Consumer AI is precisely in my wheelhouse, and I’ve recently started integrating AI into my passion projects, dramatically changing how I approach them.
What’s a passion project?
To me, a passion project is something NOT directly related to my professional responsibilities. It’s something that I work on because I think it’s awesome, not because I expect monetary reward (though I am not opposed to riches). It’s something I work on out of curiosity, to share with family and friends, and as an excuse to learn something new. In short: a hobby.
How about a few examples?
Below is an abbreviated list of some of my more notable passion projects I’ve worked on in recent years:
American Handkerchief: an e-commerce brand in the “Everyday Carry” space.
Crypto Swatch: a global collaborative art project featuring 100+ artists from around the world.
Movie Quote of the Day: Wordle, but for obscure movie quotes.
Meaningful Attendance: a question of the day app for teachers to use with students during morning attendance.
Windsor: a work-in-progress digital greeting card app
So what’s my process?
It’s only in the last three projects or so that I have leveraged AI, so my “process,” if you could call it that, is still evolving. Previously, I would spend hours “Googling the Algorithm” spending inordinate amounts of time on forums and blogs trying to get over whatever hurdle was in my way. Today, things are much different, and I’ve become 10 times more productive thanks to AI.
Let’s dive deep on Movie Quote of the Day:
The idea for a gamified movie quote app came at about the same time Wordle was blowing up. I married that concept with a common pastime among men my age, quoting obscure lines from movies to no one in particular. But where to start?
I wanted the app to feature movie art and be able to offer hints like: genre, release date, lead character, etc. but I didn’t want to gather all of that data myself. I figured a database like this already existed somewhere that I could access with an Application Programming Interface (API). The problem was, I didn’t know which database to use or HOW an API actually worked*.
*I should clarify. I am not a developer. I took one HTML class 100 years ago, but I am not a software engineer or computer scientist by any stretch of the imagination. I am technologically savvy, and insanely curious though.
So I turned to ChatGPT for help and gave it some version of the following prompt
Are there any APIs that exist that I could use to pull movie data for a movie quote of the day app—hopefully one that is free?
Sure enough, there were five! After further prompting and additional non-AI assisted research, I narrowed it down to one: The Movie Database. It offered all of the features I thought I might need. It was free (at my usage level), and it seemed like ChatGPT would be able to help me get it set up.
Takeaway: AI is a great tool for discovering what is possible even if all you have is a vague hunch or intuition to go off of.
Minimum Viable Product
At this point, I didn’t care how the project looked. I wanted to make sure the thing actually worked. Even if it was the coolest looking app in the world, if it didn’t do what it was supposed to do, the project would have been dead in the water.
Borrowing from the likes of Eric Reis and Steven Blank, I set out to get the simplest version of the app functional as quickly as possible. My goal here was to prove that it was in fact doable, so I could show it to someone, gauge their response, and determine if it was worth pouring more time into.
For this particular project, that meant wiring up the API—something I had never done before. For something this technical, and specific, I found ChatGPT was good at giving me an overview of the steps involved, but since the model is trained on an historical scrubbing of the web, not every step was precise, likely due to site updates.
It got me most of the way there, and with a few references to the API’s documentation, a was able to dynamically display the title of a movie and its movie poster art by referencing its numeric ID. This was enough to convince me I could build out the rest of the app.
Which is exactly what I did.
The first version was bare bones. Completely absent of style or character. A mostly white app that displayed plain text quotes with an ugly input field, and a completely forgettable submit button, but by god the damn thing worked. If you didn’t know the answer right away, you could ask for a hint, and it’d give you one. And if you guessed correctly, it would show you a short YouTube clip of the quote itself.
Takeaway: identify the biggest challenge in your passion project and tackle it first. You don’t want to get too far down the road only to discover that the thing you were dreading was in fact totally worth dreading about.
Share the thing
I thought the app was pretty cool, but I also built a handkerchief company in the 21st century, so I am not always the best judge of these things. So, I sent it out to my oldest friends—still completely unpolished. I hope everyone has friends like mine who have absolutely no problem telling you when you’re being an idiot. My thinking was if they didn’t shred me to bits with the app in its ugly state, it probably meant it was a better than average idea, and worth pursuing even further.
Guess what? They liked it.
Takeaway: Share your work along the way to validate its worth.
Make it look good
With the idea itself validated, and functional, I turned to making it look nice. At this point there was plenty of additional work that needed to be done unrelated to design: adding content, figuring out shareability, procuring a domain name, etc., but all of that was stuff I was confident I’d be able to figure out later. It also wasn’t super exciting to me. Making it look rad was.
I consider myself much more of a designer than I do a programmer. I know my way around Photoshop and Illustrator, and I think I have a decent eye for design, but I was curious to see how else AI could assist me in this project, so I turned to a Dall-E 3 powered text-to-image generator for some layout inspiration.
I had a blast prompting different concepts for what it could look like, but ultimately settled on a dark theme with neon accents. The images below are recreations of similar prompts I used at the time.
They were far from perfect, but directionally useful. You can see the influence of the generative AI in the final product below:
Takeaway: sometimes AI is more of a thinking partner, than a tool that can produce something that is ready-to-ship.
Do the human part
For my app to be any good, it needed to feature movie quotes that I actually enjoyed, which are often subtle throwaway lines in movies that are seldom written about and therefor not really part of the neural network. No matter what I prompted, I couldn’t get AI to give me what I was looking for. Additionally, it couldn’t find the links to the quotes I wanted to use on YouTube with precise start and stop URL parameters. Worse than that, it would hallucinate and give me incorrect URLs that went nowhere. So, all of the quotes and clips had to be dug up the old fashioned way—manually.
Takeaway: AI isn’t a magic wizard that knows everything. There’ll always be human work to do. Combining what humans do well with what machines do well may be where the magic lies.
Add polish
I discovered that when building an app, there are thousands of things you have to do beyond the original concept and design. User Experience/User Interface (UX/UI) is super important. It needs to be intuitive. It can’t require instructions. Users should understand what to do just by looking at it. Color choice, font, font size, font weight, layout, sequence, hierarchy, and on and on are all factors to consider when designing an app.
This is inherently human, but implementing these decisions often requires code. Implementing shareability on mobile was tricky, for example. I had seen on Wordle that one could leverage the operating system’s share settings, but I had no idea how to get that to work. Prompting questions like…
How do a share using the operating systems built in features on mobile?
…helped me understand from a high level what to do, and further prompts helped me to actually implement it.
Takeaway: It might take several hundred prompts to figure out all of the things that you want to do.
Conclusions
What I learned through this process is that building with AI is like having a really, really, REALLY smart friend with infinite patience. It will answer any question you have, no matter how rudimentary. It will help you with whatever task you need assistance with, but it’s up to you to ask the right right questions.
I am certain that a “real developer” would vomit at the site of my code base, but I didn’t build this project for code review. I built it to exist, and with the help of AI, now it does. I learned a lot along the way, and it made me excited about what I might build in the future (a fitness app that generates high intensity interval training kettle bell exercises).
Takeaway Summary
If you made it this far, congratulations. Here are all of my takeaways neatly packed in a single bulleted list:
AI is a great tool for discovering what is possible even if all you have is a vague hunch or intuition to go off of.
identify the biggest challenge in your passion project and tackle it first. You don’t want to get too far down the road only to discover that thing you were dreading was in fact totally worth dreading about.
Share your work along the way.
Sometimes AI is more of a thinking partner, than a tool that can produce something that is ready-to-ship.
AI isn’t a magic wizard that knows everything. There’ll always be human work to do. Combining what humans do well with what machines do well may be where the magic lies.
It might take several hundred prompts to figure out all of the things that you want to do.
Want some more?
Thanks so much for reading. I hope it inspired you to build a passion project of your own.
I hope you will consider subscribing to my Substack at mmacfadden.substack.com where I write about creativity and innovation in teaching and learning.
Thank you, Mike!
Hope you all enjoy the guest post today. If you're passionate about boosting your creativity, check out his writings—I’m sure you’ll find valuable insights.
Have a wonderful weekend.
Xinran
-
P.S. I realized that I’m only a month and a half away from the half-marathon I registered for at the beginning of the year. I’m not feeling ready at all, so I need to start training seriously from now on…
Great post.. I really enoyed the sentiment around just making something without worrying about it being done the perfect way. My codebase 'is meant to exist, not for peer review'.. love it. Interesting see how all the tools came together. A guest writer was/is a really neat idea Xin. Keep it coming. I have all the designers at work reading your stuff when it comes in.