ZDNET’s key takeaways
- Early Apple coding was powerful, but painfully manual.
- Today’s AI agents build working apps from conversations.
- AI may add slop, but it learned that from humans.
I think it’s fair to say Apple owes its multi-trillion-dollar success to its apps ecosystem.
Think about it. If the Apple II from the late 1970s could only run in-house developed apps, if that first Mac could only run MacPaint, if later generation Macs were limited to Apple’s own products, if the iPhone could only run a few basic apps for contact management and calendar, where would Apple be?
Also: After the vibe-coding rush comes the debugging hangover
Independent apps like VisiCalc, PageMaker, PowerPoint, and Photoshop drove Apple sales in the early days. The iPhone and the App Store model blasted the iPhone into the stratosphere with apps like Instagram and Uber. And now, we’re in the AI era, with chatbots and coding agents, and the apps built by humans with those coding agents.
I’ve built Apple software in each of those eras, including 40 super-simplistic early iPhone apps, which I wrote about for ZDNET in 2012.
The early App Store era
When the iPhone first came out, it was locked to its own apps. You could build a web page for Safari, call it an “app,” and put an icon on the screen, but iPhone connectivity in 2007 was garbage. Then, in 2008, Apple introduced the first iOS SDK along with the App Store.
About a week after that announcement, I bought an iMac and an iPhone. Four weeks after that, I submitted DaysTo Christmas to the App Store.
Development was…unpleasant. Apple banned sharing any information about iPhone development, so while underground forums existed, there were no tutorials. You had to figure out everything about iPhone development and its new UI from Apple’s materials.
My 40 apps were all variations on the same theme: a simple counter to an upcoming event. They were surprisingly popular in those early days of the App Store. I even got user requests for new apps specifically aimed at new categories of upcoming events, including the military-themed DaysTo Discharge, business-themed DaysTo Tradeshow, and even DaysTo Divorce.
In those days, though, it felt like Apple hadn’t bought into the idea of independent developers. Unlike today, when I have helpful conversations with App Store reviewers, Apple developer relations were essentially nonexistent. Even my paid support requests felt like shouting into the wilderness.
Also: I used Claude Code to vibe code a Mac app in 8 hours, but it was more work than magic
After about four years in the Apple Developer Program, I gave up my account to spend my side project time getting a graduate degree and growing a YouTube channel.
But that early App Store experience wasn’t my first rodeo as an Apple developer. Almost two decades earlier, in the very early Mac days, I built a bunch of successful products.
I even built two AI products back in the late 1980s. The big one was Intelligent Developer. This tool let non-programmers build expert systems. An expert system is kind of the opposite of today’s AI. You had to do all the training and embed your expertise in the system. The training corpus was only what you could provide. Intelligent Developer made that process easy.
I wrote a book about that period called The Flexible Enterprise. You can download it for free here.
That was then, this is now
Fast-forward to 2023. AI happened. Not the 1980s form of AI we had with Intelligent Developer. I’m talking about generative AI, the AI we have now that is changing everything.
Also: I built an iOS app in just two days with just my voice – and it was electrifying
Earlier this year, I wrote the starting bones of an iOS app in two days, without even touching the keyboard or writing a line of code. I just talked into my AirPods and Claude Code wrote a working app. Sure, it’s not ready for distribution. But it works.
Recently, in my spare time, not even as a full side project, I have written three functional apps. Well, technically, they’re nine apps because two of them have unique capabilities and interfaces for the iPhone, Mac, iPad, and Apple Watch. That Watch app took me something like 12 hours to create, instead of months and months.
Also: I used Claude Code to vibe code an Apple Watch app in just 12 hours
The app I’ve been working on for the past two weeks is Mac-only (for now). The tool actually functions. I’m using it for some of my list management, though it will require at least another month of hour-or-so-a-day tweaking before it’s ready for prime time.
The thing is, these are all apps that would never have gotten written without vibe coding. None of them could justify the year or more they would have taken to build. They’re all fairly vertical, fairly specialized, and mostly built for me (and one for my wife).
Sure, I’ll put them up on the App Store and happily use the proceeds to pay for my AI habit. But I couldn’t build a truly viable business plan around them. The beauty is, I don’t have to. I’m not investing in a team and tens of thousands of developer hours. I check in with Claude every couple of hours, or when I’m waiting for my computer to complete another activity.
That’s not to say this work doesn’t require technical chops. I wrote earlier this week about how there’s a post-flow drop, a debugging hangover, where you need technical understanding to guide the AI to design and architect the overall application.
Also: 75% of developers I surveyed prefer Claude Code over Codex – here’s why
But it’s nothing like the two years and full team it took to create Intelligent Developer back in the day. And it’s nothing like the surly lack of helpful information and the days it took to write apps that displayed a cute picture and calculated days between dates.
This AI-enabled process is productive and fun. I get to focus my product management and design skills, rather than desperately scouring the web for hints about how to get a percentage-done indicator to show the right icon, or how to get a hover effect to work just right.
Sure, I expect a lot more AI slop on the App Store. But it’s not like there wasn’t slop on the App Store even in the earliest days. Someone cloned DaysTo Christmas, and our other apps, and even word-for-word stole our lovingly written marketing description, then posted their own version of the tools to the App Store. We reported this uncool behavior. Apple did nothing about it.
My point is, AI isn’t the only path to slop. Humans have been producing slop since long before gen AI was a gleam in Sam Altman’s eye. AI probably learned everything it knows about producing slop from us.
Also: ‘I can’t stop’: 80% of developers find AI coding more addictive than helpful
Agentic AI in 2026 has made app development a whole new game. For larger development teams, AI collapses the time-to-market lifecycle. And for independent coders like me, AI makes it possible to produce far more capable applications in what little spare time we have, for niches that would otherwise be ignored.
Plus, it’s really, really cool. And now, if you’ll excuse me, I think Claude needs me.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.
