An app that generates a custom to-do list is entirely different from an app for playing a first-person shooter. An app that lets you order a personal taxi or book a hotel room is vastly different from an app that lets you design 3D objects.
Many apps are much more than just a pretty interface. Take, for example, Instacart. Sure, the app has a database of products to choose from, an e-commerce component for managing purchases and billing, and a messaging interface between customers and shoppers.
Also: The best AI for coding in 2025 (and what not to use – including DeepSeek R1)
But it also has a vast infrastructure of deals with food outlets that allow it to keep inventory updated and deals that allow its shoppers to roam those outlets’ aisles. It has mapping and route optimization capabilities for managing the shoppers and optimizing their experience.
Writing code is not only complex, it’s multi-disciplinary. At least it is for fairly big projects.
The Apple vision
Which gives you some context for today’s topic: Vibe coding apps with Siri.
The fine folks at 9to5Mac recently ran an article, “Apple wanted people to vibe code Vision Pro apps with Siri.” The idea came from a report describing how “Apple hopes that even people who don’t know computer code could tell the headset, via the Siri voice assistant, to build an AR app that could then be made available via Apple’s App Store for others to download.”
Apparently, Apple execs discussed such a feature, but it has not yet been implemented.
On one hand, the idea seems ludicrous. How many times has Siri screwed up merely transcribing a text message to a friend? To assume Siri can do something as complex and powerful as creating an app would assume Siri is not the fairly simplistic AI we all know and love.
Also: How I used this AI tool to build an app with just one prompt – and you can too
But let’s assume Apple Intelligence can eventually rise above the crushing disappointment it’s been, and that Siri eventually has AI skills on a par with ChatGPT or Google Gemini. Then we’ve got a starting point.
I’ve repeatedly shown that AIs can code. In my recent tests, ChatGPT and Gemini Pro 2.5 hit bases-loaded home runs.
It’s not unrealistic, then, to think that Apple will (some day soon?) have a Siri that works at least at the level of its competitors.
What would it take for Apple to be able to vibe code — aka, AI code — apps with Siri? We must discuss three major factors: technology, Apple’s relationship to coding, and managed expectations.
The technology is here
There’s some precedent for the idea that you could describe an app in a sentence, and an AI could write it. Last week, I showcased how GitHub Spark could build a code analysis tool out of a single sentence. Yes, the interface was ugly, and yes, I tried somewhat fruitlessly to refine it, but the fact was, an AI did build a working app from a single sentence description.
Shortly after ChatGPT hit the big time, I asked it to create a full WordPress plugin, including the user interface, and it did. The plugin was fairly simple, and it took more than one sentence, but ChatGPT amazed me then with its ability to get the job done.
So, while it might take some time to get it right, the technology is there to do the job.
Apple’s history with empowering citizen developers
Apple has a long history of empowering developers, but also misunderstanding what development entails. I know. I was there. The Apple II was the original consumer computer hit, not just because of its friendly shell, but because Apple included a programming language, BASIC, accessible to new users.
When Apple introduced the Mac, it also introduced a vast library of books that included interface guidelines and coding guidelines so third-party developers could make Mac apps that looked like Mac apps.
Apple introduced each of these products with the awareness that developers, developers, developers would drive hardware acceptance. After all, it’s what you can do with the machines that makes them worthwhile, right?
Also: Brace yourself: The era of ‘citizen developers’ creating apps is here, thanks to AI
Apple’s first major low-code product was the revolutionary HyperCard, a tool that let you draw a user interface and connect the modules with minimal code. (I know, because I started the first company that built tools for HyperCard developers and ran a HyperCard project for Apple.)
But there was a big disconnect within Apple. I remember sitting in the office of HyperCard’s product manager at Apple and hearing him tell me that nobody wants to make custom applications. Apple users don’t want custom, he told me.
Yet, every day, I talked to school teachers, sports coaches, doctors, tradespeople, small business owners, and even the occasional big-budget movie director and sitcom star, all of whom most definitely wanted to build their own apps to help them get their jobs done.
Also: The most popular programming languages (and what that even means)
Other low-code tools Apple has introduced include Automator, Shortcuts, Playgrounds (as part of Swift), and the Xcode Interface Builder. Apple even toyed with AR creation tools like Reality Composer, introduced in 2019, which allowed developers to do drag-and-drop 3D asset placement, animations, and basic interactions without writing code.
I think of these tools as resources that empower citizen developers. These people are not necessarily developers first, but are willing to learn whatever skills they need to get the job done. While not everyone wants to build an app, and there are a bunch of people who want to develop apps just because they think they can strike it rich, there are also a surprisingly large and varied group of people who want to build apps because they just want to make their computers do specialized tasks on their behalf.
Managing expectations
This brings us to the crux of all AI coding for the masses. The naïve newbie wants to be able to utter a single line command and suddenly be at the helm of the next billion-dollar Uber.
That’s never going to be possible, but it’s entirely likely that AI coding tools can help Uber developers maintain and improve their code.
It is possible for AI tools to vibe code an app. We saw it with the GitHub Spark and WordPress plugin examples. It’s also possible to create drag-and-drop interfaces for interactive experiences. Reality Composer helped folks do that as far back as six years ago.
The real question is: What kind of app can an AI build for you? How much work are you willing to put in? How well does the AI handle iteration and stepwise improvement?
Also: This AR headset is changing how surgeons see inside their patients
So far, AIs are terrible at incrementally improving their work. They function far better when they’re asked to recreate something entirely, but with a few new elements. This makes it particularly difficult to get an AI to make incremental changes without something fundamental also randomly changing between iterations.
Some projects are not practical regarding the sort of your-wish-is-my-command coding that vibe coding seems to imply. For example, while it may well be possible for non-coders or low-coders to build AR and VR environments, it takes a team of very experienced engineers to build an AR experience that allows spinal surgeons to confidently cut open and repair patients’ spines.
As we consider vibe coding to create apps, it’s important to realize that such tools work well for certain applications (forms-based apps in particular) and not so well for other kinds of applications, particularly the big and complex titles that drive billion-dollar businesses.
Painting the vista
In marketing, the term “paint the vista” refers to the practice of presenting a marketing message so vivid that it creates a mental image, thereby capturing both the essence of what you’re trying to sell and the imagination of your prospects. Often, the practice exaggerates the actual experience of using the product, but it resonates with prospects and helps boost sales.
So, is it silly to expect people to be able to vibe code Vision Pro apps with Siri? That’s some serious vista painting there, I’ll tell you.
First, Vision Pro sales are struggling. For those who need the device, it would be better if they could build their own applications, because the device hasn’t sold enough to justify the development effort of a more commercial coding shop.
Also: Apple Vision Pro’s killer feature is finally here
Second, Siri still needs a lot of work before most of us trust it to send a text properly.
But is low-code, AI-supported, vibe coding-ish application development for AR and VR experiences part of a possible future? Heck yeah! None of that is impossible. The technology is already here (just not in Apple Intelligence). The rest is a matter of stepwise improvement, figuring out what works and needs help, and then waiting for that to be implemented.
Bottom line
Just keep your expectations in check. Get to know the tools, what they work well for, and where they run into a wall. While I doubt that Apple would approve a ton of amateur-coded VR and AR apps for the App Store, undoubtedly, there will be some great work by folks who don’t code for a living, and we may get to see some amazing works.
Also: My favorite XR glasses for productivity and traveling got 3 major upgrades
Bottom line: AI vibe coding a Vision Pro AI app with Siri is not an unrealistic expectation. But it does need some work before we get there, and you need to keep expectations under control.
What do you think? Could you see yourself building an app just by describing it to Siri? Have you tried any low-code or no-code tools like HyperCard, Shortcuts, or Reality Composer? Do you think Apple is on the right track with this vision, or are we just painting the vista a bit too vividly? Let us know in the comments below.
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.
Get the biggest stories in tech every Friday with ZDNET’s Week in Review newsletter.
