E15: AI and the App (Day 364)

Good morning. It’s November 19th, a Tuesday. Gray foggy morning. I love the gray foggy mornings. I’ll always love a gray foggy morning. The year I lived in Austin, Texas — I think we had something like 90 consecutive days over 100 degrees. A weather stat I say out loud and immediately doubt. But it was a very hot summer. And I remember when it finally dropped into the 90s and you stood outside in the shade and it just felt positively cool and pleasant. People say, oh, it’s a dry heat — well, not really in Austin. Air coming up from the Gulf. Anyway. It’s two degrees Celsius outside right now.

The news of the day: I made an app.

I woke up Saturday morning thinking about vocabulary. I used to have an app I loved studying Hungarian with. The problem with language apps is: phrasebook apps with pre-programmed vocabulary are useful for travelers and fundamental learners, but you quickly outgrow them. And the flashcard apps either draw from pre-made lists or have clunky, complicated interfaces for customization. A bunch of extra cruft.

I thought: what if I could just bring in my own vocabulary lists and turn them into flashcards? Because flashcards are a genuinely great learning modality — otherwise we wouldn’t have the flashcard. The apps where you click to create a card, a screen pops up, you type what you want on each side, you save it — having to do that individually for every single word is a barrier to entry for me.

So I woke up Saturday with one curiosity: if I could design the perfect app, and I know nothing about coding, is it true that I could use ChatGPT to code it for me?

Sat down and thought: maybe I just ask for the app and it spits out the code. It wasn’t quite that easy. First thing it said: download Xcode. So I asked — I don’t know anything about coding, is this going to be totally overwhelming? And GPT said: yeah, the interface is overwhelming, but I’ll talk you through every part of it and you’ll never be disoriented.

I downloaded Xcode. And damn, if it wasn’t right. If I’d opened it without any help, I would’ve taken one look, closed it, and deleted it from my computer. But step by step I set up my project, set up the files, and got down to the coding.

I started with a general description of what I wanted. The AI gave me all the central code — and what resulted was a discouragingly clunky, ugly app. I didn’t know what prompts would get me from clunky to nice, but I just attacked it in steps: backend data handling, then UI design.

What I wanted most was to just type up a word list — Hungarian word on one line, English on the next, paragraph break, repeat — and have the app parse that into flashcards. Sure, here’s the code for that. Hit a practice button to launch flashcard mode? Sure. Click a word and the app pronounces it? Sure. And so on.

Then I dialed in the aesthetics. That took a lot.

The interesting human side: every time you ask GPT to do something, it doesn’t always get it right. You try a coding strategy, it screws everything up, you go back, and you iterate. I’ve done HTML and CSS before — I know that’s how it works. And I realized: if I was sitting with an actual coder, they’d be doing the same thing.

Saturday I started right after coffee and went until about 11 o’clock that night — aside from a couple of breaks. By the time I was done I had a workable app, buggy but workable. And I was completely exhausted because of this weird thing that happens when you code: I get into a tunnel-focused obsessive state. Working out the dynamics of the text box, something’s messed up, I’m hungry — six feet from food in the kitchen — and I don’t pause, go make a sandwich, come back. I keep at it until I solve it. Thirsty, have to go to the bathroom, doesn’t matter. You sit and sit and sit.

Not very healthy. I understand how the Mountain Dew and multiple-screens cliché comes to be.

By the end of the night I was mentally fried and frustrated. Every time GPT messed something up I was very upset. But I closed it out, shook it off, recognized my mental state as completely irrational. And I apologized to the AI — sorry, I got a little testy there at the end. We ended friends.

Sunday I made more refinements. Monday, more. The upshot is: it’s a remarkably functional app. Not ugly. Presentable. I’m not winning any design awards, but I designed the icon, I did all the UI. And this is bananas to me. On Friday — the day before I started — designing my own iPhone app was right up there with doing my own dentistry. Not really an option. And lo and behold.

I signed up for a developer account to get it on the App Store. I can easily adapt it for Spanish or German or French. Not a lot of people out there learning Hungarian, but I made it for me.

There’s something really interesting about AI in this context. I have a Pro account, and one of the things I love — which the alarmists would flag as terrifying — is that I can go back into a conversation at any time and it remembers everything. Open GPT 12 hours later and say, hey, can I put these buttons below my text box, without re-explaining anything. Google is dumb in the sense that if I spend an afternoon researching something, I can’t go back the next day and say, “remember I was researching 19th century Hungarian composers — were there any female composers?” You have to recontextualize every search. With GPT, I can go in with something as complex as app development and pick up exactly where I left off. It really does feel like an assistant. Someone who remembers what you’re doing and what you need.

And if you set aside your aversion to AI — if you work in any knowledge or producing field, the tools are extraordinary. I’ve used it in video editing immensely. Creating subtitles, translating them into other languages — things that wouldn’t have been possible at any reasonable speed before. Subtitling and translating into Ukrainian is still laborious even with AI. But the same task without AI would take three or four days instead of three or four hours, and would require a native speaker.

Now — I hear all the hysteria about how AI is going to devastate industries, steamroll creatives, bastardize the news, enable propagandists. If you sit back and pay attention to this discourse, you notice there’s an echo chamber of perpetual worriers, and AI is just the thing they’ve grabbed onto. The same people were screaming about the evils of the internet in ‘96, ‘97, ‘98. You can go back historically and look at the advent of domestic electricity — the perpetual worriers were talking about it bringing the devil into your house, about electrical fires burning homes to the ground. What they never account for is that the implementation and broad acceptance of new technologies happens incrementally, and we figure out along the way how to make them safe and use them responsibly.

Most of the hand-wringing about AI and personal privacy is coming from people who have been living their lives publicly on social media for a decade. Your digital DNA is already out there, already gobbled up by corporations. Oh well. Could bad actors get hold of it and could I actually be screwed? Maybe. But tomorrow night my toaster might start an electrical fire and I’ll die of smoke inhalation. The possibility has been there since the first night I laid my head down in a house with a toaster.

I don’t believe in worst-case-scenario fear mongering. I benefit far more by taking advantage of the tools.

Actually — I just found a video of Ben Affleck sitting on a stage. November 16th, 2024. [Searches phone.] Go Google it — Ben Affleck talking about AI’s impact on the film industry (it’s here). He jumps in with real clarity: which film industry professionals have something to be concerned about, which do not, what the implications will be for consumers in the next few years. Coherent, cogent. I didn’t intend for today to become a deep rant about AI. Am I ranting? Or am I cogitating? But there it is.

Coming off the heels of designing my own iPhone app, I’m riding pretty high on AI’s capabilities.

Beyond that — yesterday I spent a lot of time on the documentary. Research on the Bullet project, reaching out to old friends and relatives, having a few conversations. A lot more of that to come.

Today I have my Hungarian lesson. I might record it. Probably nobody would want to sit in and listen to the whole thing, but maybe I’ll do it.

I’ll leave it there. Got other things to do. Take care. Have a good one.

Share Chip Warren