There has been an explosion of AI-powered prompt-to-code tools within the last year. The space began with full-on integrated development environments (IDEs) like Cursor and Windsurf. These enabled developers to use leverage AI assistants right inside their coding apps. Then came a tools like v0, Lovable, and Replit, where users could prompt screens into existence at first, and before long, entire applications.

A couple weeks ago, I decided to test out as many of these tools as I could. My aim was to find the app that would combine AI assistance, design capabilities, and the ability to use an organization’s coded design system.

While my previous essay was about the future of product design, this article will dive deep into a head-to-head between all eight apps that I tried. I recorded the screen as I did my testing, so I’ve put together a video as well, in case you didn’t want to read this.

It is a long video, but there’s a lot to go through. It’s also my first video on YouTube, so this is an experiment.

The Bottom Line: What the Testing Revealed

I won’t bury the lede here. AI tools can be frustrating because they are probabilistic. One hour they can solve an issue quickly and efficiently, while the next they can spin on a problem and make you want to pull your hair out. Part of this is the LLM—and they all use some combo of the major LLMs. The other part is the tool itself for not handling what happens when their LLMs fail. 

For example, this morning I re-evaluated Lovable and Bolt because they’ve released new features within the last week, and I thought it would only be fair to assess the latest version. But both performed worse than in my initial testing two weeks ago. In fact, I tried Bolt twice this morning with the same prompt because the first attempt netted a blank preview. Unfortunately, the second attempt also resulted in a blank screen and then I ran out of credits. 🤷‍♂️

For designers who want actual design tools to work on UI, Subframe is the clear winner. The other tools go directly from prompt to code, skipping giving designers any control via a visual editor. We’re not developers, so manipulating the design in code is not for us. We need to be able to directly manipulate the components by clicking and modifying shapes on the canvas or changing values in an inspector.

For me, the runner-up is v0, if you want to use it only for prototyping and for getting ideas. It’s quick—the UI is mostly unstyled, so it doesn’t get in the way of communicating the UX.

The Players: Code-Only vs. Design-Forward Tools

There are two main categories of contenders: code-only tools, and code plus design tools.

Code-Only

• Bolt
• Lovable
• Polymet
• Replit
• v0

Code + Design

• Onlook
• Subframe
• Tempo

My Testing Approach: Same Prompt, Different Results

As mentioned at the top, I tested these tools between April 16–27, 2025. As with most SaaS products, I’m sure things change daily, so this report captures a moment in time.

For my evaluation, since all these tools allow for generating a design from a prompt, that’s where I started. Here’s my prompt:

Create a complete shopping cart checkout experience for an online clothing retailer

I would expect the following pages to be generated:

• Shopping cart
• Checkout page (or pages) to capture payment and shipping information
• Confirmation

I scored each app based on the following rubric:

• Sample generation quality
• User experience (25)
• Visual design (15)
• Prototype (10)
• Ease of use (15)
• Control (15)
• Design system integration (10)
• Speed (10)
• Editor’s discretion (±10)

The Scoreboard: How Each Tool Stacked Up

Final summary scores for AI design tools for designers. Evaluations conducted between 4/16–4/27/25.

Here are the summary scores for all eight tools. For the detailed breakdown of scores, view the scorecards here in this Google Sheet.

The Blow-by-Blow: The Good, the Bad, and the Ugly

Bolt

First up, Bolt. Classic prompt-to-code pattern here—text box, type your prompt, watch it work. 

Bolt shows you the code generation in real-time, which is fascinating if you’re a developer but mostly noise if you’re not. The resulting design was decent but plain, with typical UX patterns. It missed delivering the confirmation page I would expect. And when I tried to re-evaluate it this morning with their new features? Complete failure—blank preview screens until I ran out of credits. No rhyme or reason. And there it is—a perfect example of the maddening inconsistency these tools deliver. Working beautifully in one session, completely broken in another. Same inputs, wildly different outputs.

Score: 43

Lovable

Moving on to Lovable, which I captured this morning right after they launched their 2.0 version. The experience was a mixed bag. While it generated clean (if plain) UI with some nice touches like toast notifications and a sidebar shopping cart, it got stuck at a critical juncture—the actual checkout. I had to coax it along, asking specifically for the shopping cart that was missing from the initial generation.

The tool encountered an error but at least provided a handy “Try to fix” button. Unlike Bolt, Lovable tries to hide the code, focusing instead on the browser preview—which as a designer, I appreciate. When it finally worked, I got a very vanilla but clean checkout flow and even the confirmation page I was looking for. Not groundbreaking, but functional. The approach of hiding code complexity might appeal to designers who don’t want to wade through development details.

Score: 49

Polymet

Next up is Polymet. This one has a very interesting interface and I kind of like it. You have your chat on the left and a canvas on the right. But instead of just showing the screen it’s working on, it’s actually creating individual components that later get combined into pages. It’s almost like building Figma components and then combining them at the end, except these are all coded components.

The design is pretty good—plain but very clean. I feel like it’s got a little more character than some of the others. What’s nice is you can go into focus mode and actually play with the prototype. I was able to navigate from the shopping cart through checkout (including Apple Pay) to confirmation. To export the code, you need to be on a paid plan, but the free trial gives you at least a taste of what it can do.

Score: 58

Replit

Replit was a test of patience—no exaggeration, it was the slowest tool of the bunch at 20 minutes to generate anything substantial. Why so slow? It kept encountering errors and falling into those weird loops that LLMs often do when they get stuck. At one point, I had to explicitly ask it to “make it work” just to progress beyond showing product pages, which wasn’t even what I’d asked for in the first place.

When it finally did generate a checkout experience, the design was nothing to write home about. Lines in the stepper weren’t aligning properly, there were random broken elements, and ultimately—it just didn’t work. I couldn’t even complete the checkout flow, which was the whole point of the exercise. I stopped recording at that point because, frankly, I just didn’t want to keep fighting with a tool that’s both slow and ineffective. 

Score: 31

v0

Taking v0 for a spin next, which comes from Vercel. I think it was one of the earlier prompt-to-code generators I heard about—originally just for components, not full pages (though I could be wrong). The interface is similar to Bolt with a chat panel on the left and code on the right. As it works, it shows you the generated code in real-time, which I appreciate. It’s pretty mature and works really well.

The result almost looks like a wireframe, but the visual design has a bit more personality than Bolt’s version, even though it’s using the unstyled shadcn components. It includes form validation (which I checked), and handles the payment flow smoothly before showing a decent confirmation page. Speed-wise, v0 is impressively quick compared to some others I tested—definitely a plus when you’re iterating on designs and trying to quickly get ideas.

Score: 61

Onlook

Onlook stands out as a self-contained desktop app rather than a web tool like the others. The experience starts the same way—prompt in, wait, then boom—but instead of showing you immediate results, it drops you into a canvas view with multiple windows displaying localhost:3000, which is your computer running a web server locally. The design it generated was fairly typical and straightforward, properly capturing the shopping cart, shipping, payment, and confirmation screens I would expect. You can zoom out to see a canvas-style overview and manipulate layers, with a styles tab that lets you inspect and edit elements.

The dealbreaker? Everything gets generated as a single page application, making it frustratingly difficult to locate and edit specific states like shipping or payment. I couldn’t find these states visually or directly in the pages panel—they might’ve been buried somewhere in the layers, but I couldn’t make heads or tails of it. When I tried using it again today to capture the styles functionality for the video, I hit the same wall that plagued several other tools I tested—blank previews and errors. Despite going back and forth with the AI, I couldn’t get it running again.

Score: 71

Subframe

My time with Subframe revealed a tool that takes a different approach to the same checkout prompt. Unlike most competitors, Subframe can’t create an entire flow at once (though I hear they’re working on multi-page capabilities). But honestly, I kind of like this limitation—it forces you as a designer to actually think through the process.

What sets Subframe apart is its MidJourney-like approach, offering four different design options that gradually come into focus. These aren’t just static mockups but fully coded, interactive pages you can preview in miniature. After selecting a shopping cart design, I simply asked it to create the next page, and it intelligently moved to shipping/billing info.

The real magic is having actual design tools—layers panel, property inspector, direct manipulation—alongside the ability to see the working React code. For designers who want control beyond just accepting whatever the AI spits out, Subframe delivers the best combination of AI generation and familiar design tooling.

Score: 79

Tempo

Lastly, Tempo. This one takes a different approach than most other tools. It starts by generating a PRD from your prompt, then creates a user flow diagram before coding the actual screens—mimicking the steps real product teams would take. Within minutes, it had generated all the different pages for my shopping cart checkout experience. That’s impressive speed, but from a design standpoint, it’s just fine. The visual design ends up being fairly plain, and the prototype had some UX issues—the payment card change was hard to notice, and the “Place order” action didn’t properly lead to a confirmation screen even though it existed in the flow.

The biggest disappointment was with Tempo’s supposed differentiator. Their DOM inspector theoretically allows you to manipulate components directly on canvas like you would in Figma—exactly what designers need. But I couldn’t get it to work no matter how hard I tried. I even came back days later to try again with a different project and reached out to their support team, but after a brief exchange—crickets. Without this feature functioning, Tempo becomes just another prompt-to-code tool rather than something truly designed for visual designers who want to manipulate components directly. Not great.

Score: 59

The Verdict: Control Beats Code Every Time

Subframe offers actual design tools—layers panel, property inspector, direct manipulation—along with AI chat.

I’ve spent the last couple weeks testing these prompt-to-code tools, and if there’s one thing that’s crystal clear, it’s this: for designers who want actual design control rather than just code manipulation, Subframe is the standout winner.

I will caveat that I didn’t do a deep dive into every single tool. I played with them at a cursory level, giving each a fair shot with the same prompt. What I found was a mix of promising starts and frustrating dead ends.

The reality of AI tools is their probabilistic nature. Sometimes they’ll solve problems easily, and then at other times they’ll spectacularly fail. I experienced this firsthand when retesting both Lovable and Bolt with their latest features—both performed worse than in my initial testing just two weeks ago. Blank screens. Error messages. No rhyme or reason.

For designers like me, the dealbreaker with most of these tools is being forced to manipulate designs through code rather than through familiar design interfaces. We need to be able to directly manipulate components by clicking and modifying shapes on the canvas or changing values in an inspector. That’s where Subframe delivers while others fall short—if their audience includes designers, which might not be the case.

For us designers, I believe Subframe could be the answer. But I’m also looking forward to if Figma will have an answer. Will the company get in the AI > design > code game? Or will it be left behind? 

The future belongs to applications that balance AI assistance with familiar design tooling—not just code generators with pretty previews.

I was sitting on a barstool next to my wife in a packed restaurant in Little Italy. We were the lone Kansas City Chiefs supporters in a nest full of hipster Philadelphia Eagles fans. After Jon Batiste finished his fantastic rendition of the national anthem, and the teams took the field for kickoff, I noticed something. The scorebug—the broadcast industry’s term for the lower-third or chyron graphic at the bottom of the screen—was different, and in a good way.

I posted about it seven minutes into the first quarter, saying I appreciated “the minimalistic lower-thirds for this Super Bowl broadcast.” It was indeed refreshing, a break from the over-the-top 3D-animated sparkling. I thought the graphics were clear and utilitarian while being exquisitely-designed. They weren’t distracting from the action. As with any good interface design, this new scorebug kept the focus on the players and the game, not itself. I also thought they were a long-delayed response to Apple’s Friday Night Baseball scorebug.

Anyhow, as a man of good taste, John Gruber also noticed the excellence of the new graphics. Some of his followers, however, did not.

It looks as if they just let an intern knock something up in PowerPoint and didn’t bother having someone check it first. Awful. 👎

The scorebug is absolutely horrible! I really hope they don’t adopt this for the 2025 season, or I will riot. Horrible design and very distracting especially the score, this looks like something out of Fortnite.

Gruber has a wonderful and in-depth write-up about FOX Sports’ new NFL scorebug. Not only does it include a good design critique, but also a history lesson about the scorebug, which surprisingly, didn’t debut until 1994.

Until 1994, the networks would show the score and time remaining when they cut to a commercial break, and then show it again when they came back from commercials.

I had totally forgotten about that.

Better look at the new scorebug displayed during a pre-game broadcast test.

In the early 2000s to the mid-oughts, every designer I knew wanted to be featured on the FWA, a showcase for cutting-edge web design. While many of the earlier sites were Flash-based, it’s also where I discovered the first uses of parallax, Paper.js, and Three.js. Back then, websites were meant to be explored and their interfaces discovered.

A grid of winners from The FWA in 2009. Source: Rob Ford.

One of my favorite sites of that era was Burger King’s Subservient Chicken, where users could type free text into a chat box to command a man dressed in a chicken suit. In a full circle moment that perfectly captures where we are today, we now type commands into chat boxes to tell AI what to do.

The Wild West mentality of web design meant designers and creative technologists were free to make things look cool. Agencies like R/GA, Big Spaceship, AKQA, Razorfish, and CP+B all won numerous awards for clients like Nike, BMW, and Burger King. But as with all frontiers, civilization eventually arrives with its rules and constraints.

The Robots Are Looking

Last week, Sam Altman, the CEO of OpenAI, and a couple of others from the company demonstrated Operator, their AI agent. You’ll see them go through a happy path and have Operator book a reservation on OpenTable. The way it works is that the AI agent is reading a screenshot of the page and deciding how to interact with the UI. (Reminds me of the promise of the Rabbit R1.)

Let me repeat: the AI is interpreting UI by looking at it. Inputs need to look like inputs. Buttons need to look like buttons. Links need to look like links and be obvious.

In recent years, there’s been a push in the web dev community for accessibility. Complying with WCAG standards for building websites has become a positive trend. Now, we know the unforeseen secondary effect is to unlock AI browsing of sites. If links are underlined and form fields are self-evident, an agent like Operator can interpret where to click and where to enter data.

(To be honest, I’m surprised they’re using screenshots instead of interpreting the HTML as automated testing software would.)

The Economics of Change

Since Perplexity and Arc Search came onto the scene last year, the web’s economic foundation has started to shift. For the past 30 years, we’ve built a networked human knowledge store that’s always been designed for humans to consume. Sure, marketers and website owners got smart and figured out how to game the system to rank higher on Google. But ultimately, ranking higher led to more clicks and traffic to your website.

But the digerati are worried. Casey Newton of Platformer, writing about web journalism (emphasis mine):

The death of digital media has many causes, including the ineptitude of its funders and managers. But today I want to talk about another potential rifle on the firing squad: generative artificial intelligence, which in its capacity to strip-mine the web and repurpose it as an input for search engines threatens to remove one of the few pillars of revenue remaining for publishers.

Elizabeth Lopatto, writing for The Verge points out:

That means that Perplexity is basically a rent-seeking middleman on high-quality sources. The value proposition on search, originally, was that by scraping the work done by journalists and others, Google’s results sent traffic to those sources. But by providing an answer, rather than pointing people to click through to a primary source, these so-called “answer engines” starve the primary source of ad revenue — keeping that revenue for themselves.

Their point is that the fundamental symbiotic economic relationship between search engines and original content websites is changing. Instead of sending traffic to websites, search engines, and AI answer engines are scraping the content directly and providing them within their platforms.

Christopher Butler captures this broader shift in his essay “Who is the internet for?”:

Old-school SEO had a fairly balanced value proposition: Google was really good at giving people sources for the information they need and benefitted by running advertising on websites. Websites benefitted by getting attention delivered to them by Google. In a “clickless search” scenario, though, the scale tips considerably.

This isn’t just about news organizations—it’s about the fundamental relationship between websites, search engines, and users.

The Designer’s Dilemma

As the web is increasingly consumed not by humans but by AI robots, should we as designers continue to care what websites look like? Or, put another way, should we begin optimizing websites for the bots?

The art of search engine optimization, or SEO, was already pushing us in that direction. It turned personality-driven copywriting into “content” with keyword density and headings for the Google machine rather than for poetic organization. But with GPTbot slurping up our websites, should we be more straightforward in our visual designs? Should we add more copy?

Not Dead Yet

It’s still early to know if AI optimization (AIO?) will become a real thing. Changes in consumer behavior happen over many single-digit years, not months. As of November 2024, ChatGPT is eighth on the list of the most visited websites globally, ranked by monthly traffic. Google is first with 291 times ChatGPT’s traffic.

Top global websites by monthly users as of November 2024. Source: SEMRush.

Interestingly, as Google rolled out its AI overview for many of its search results, the sites cited by Gemini do see a high clickthrough rate, essentially matching the number one organic spot. It turns out that nearly 40% of us want more details than what the answer engine tells us. That’s a good thing.

Clickthrough rates by entities on the Google search results page. Source: FirstPageSage, January 2025.

Finding the Sweet Spot

There’s a fear that AI answer engines and agentic AI will be the death of creative web design. But what if we’re looking at this all wrong? What if this evolution presents an interesting creative challenge instead?

Just as we once pushed the boundaries of Flash and JavaScript to create award-winning experiences for FWA, designers will need to find innovative ways to work within new constraints. The fact that AI agents like Operator need obvious buttons and clear navigation isn’t necessarily a death sentence for creativity—it’s just a new set of constraints to work with. After all, some of the most creative periods in web design came from working within technical limitations. (Remember when we did layouts using tables?!)

The accessibility movement has already pushed us to think about making websites more structured and navigable. The rise of AI agents is adding another dimension to this evolution, pushing us to find that sweet spot between machine efficiency and human delight.

From the Subservient Chicken to ChatGPT, from Flash microsites to AI-readable interfaces, web design continues to evolve. The challenge now isn’t just making sites that look cool or rank well—it’s creating experiences that serve both human visitors and their AI assistants effectively. Maybe that’s not such a bad thing after all.

Apple finally launched its Vision Pro “spatial computing” device in early February. We immediately saw TikTok memes of influencers being ridiculous. I wrote about my hope for the Apple Vision Pro back in June 2023, when it was first announced. When preorders opened for Vision Pro in January, I told myself I wouldn’t buy it. I couldn’t justify the $3,500 price tag. Out of morbid curiosity, I would lurk in the AVP subreddits to live vicariously through those who did take the plunge.

After about a month of reading all the positives from users about the device, I impulsively bought an Apple Vision Pro. I placed my order online at noon and picked it up just two hours later at an Apple Store near me.

Many great articles and YouTube videos have already been produced, so this post won’t be a top-to-bottom review of the Apple Vision Pro. Instead, I’ll try to frame it from my standpoint as someone who has designed user experiences for VR. 

Welcome to the Era of Spatial Computing

Augmented reality, mixed reality, or spatial computing—as Apple calls it—on a “consumer” device is pretty new. You could argue that Microsoft HoloLens did it first, but that didn’t generate the same cultural currency as AVP has, and the HoloLens line has been relegated to industrial applications. The Meta Quest 3, launched last October, also has a passthrough camera, but they don’t market the feature; it’s still sold as a purely virtual reality headset.

Vision Pro Home Screen in my messy home office.

Putting on Vision Pro for the first time is pretty magical. I saw the world around me—though a slightly muted and grainy version of my reality—and I saw UI floating and pinned to reality. Unlike any other headset I’ve tried, there is no screen door effect. I couldn’t see the pixels. It’s genuinely a retina display just millimeters away from my actual retinas. 

The UI is bright, vibrant, and crisp in the display. After launching a weather app from the home “screen” and positioning it on a wall, it stays exactly where it is in my living room. As I move closer to the app, everything about the app remains super sharp. It’s like diving into a UI. 

The visionOS User Interface

The visionOS UI feels very much like an extension of macOS. There’s a lot of translucency, blurred backgrounds for a frosted glass effect, and rounded corners. The controls for moving, closing, and resizing a window feel very natural. There were times when I wished I could rotate a window on its Y-axis to face me better, but that wasn’t possible. 

Admittedly, I didn’t turn on the accessibility feature. But as is, a significant issue that the UI presents is contrast. As someone with no accessibility issues, it was hard to tell half the time when something was highlighted. I would often have to look at another UI component and then back again to make sure a button was actually highlighted.

When you launch a Vision Pro app, it is placed right in front of you. For example, I would look at the Photos app, then click the Digital Crown (the dial for immersion) to bring up the Home Screen, which is then overlaid on top of the app. The background app does get fainter, and I can tell that the new screen is on top of Photos. Launching the Apple TV app from there would bring up the TV window on top of Photos, and I would run into issues where the handles for the windows are really close together, making it difficult to select the right one with my eyes so I can move it.

Window management, in general, is a mess. First of all, there is none. There’s no minimizing of windows; I would have to move them out of the way. There’s no collecting of windows. For instance, I couldn’t set up a workspace with the apps in the right place, collapse them all, and bring them with me to another room in my house. I would have to close them all, reopen them, and reposition them in the new room.

Working in Apple Vision Pro

I was excited to try the Mac Virtual Display feature, where you can see your Mac’s screen inside Vision Pro. Turning this on is intuitive. A “Connect” button appeared just above my MacBook Pro when I looked at it.

The Mac’s screen blacks out, and a large screen inside Vision Pro appears. I could resize it, move it around, and position it exactly where I wanted it. Everything about this virtual screen was crisp, but I ran into issues.

First, I’m a pretty good typist but cannot touch-type. With the Mac Virtual Display, I need to look down at my keyboard every few seconds. The passthrough camera on the headset is great but not perfect. There is some warping of reality on the edges, and that was just enough to cause a little motion sickness.

Second, when I’m sitting at my desk, I’m used to working with dual monitors. I usually have email or comms software on the smaller laptop screen while I work in Figma, Illustrator, or Photoshop on my larger 5K Apple Studio Display. If I sit at my desk and turn on Mac Virtual Display, I also lose my Studio Display. Only one virtual display shows up in Vision Pro. 

I tried to mitigate the lost space by opening Messages, Spark Email (the iPad version), and Fantastical in Vision Pro and placing those apps around me. But I found switching from my Mac to these other apps cumbersome. I’d have to stop using my mouse and use my fingers instead when I looked at Spark. I found that keyboard focus depended on where my eyes were looking. For example, if I were reading an email in Spark but needed to look at my keyboard to find the “E” key to archive that email, if I pressed the key before my eyes were back in the Spark window, that E would go to whatever app my eyes happened to cross. In other words, my eyes are my cursor, which takes a while to get used to.

Spatial Computing 1.0

It is only the first version of visionOS (currently 1.1). I expect many of these issues, like window management, eye tracking and input confusion, and contrast, to improve in the coming years. 

Native visionOS Apps

In many ways, Apple has been telegraphing what they want to achieve with Vision Pro for years. Apple’s API for augmented reality, ARKit, was released way back in June 2017, a full six years before Vision Pro was unveiled. Some of the early AR apps for Vision Pro are cool tech demos.

There’s a jet engine in my living room!

The JigSpace app plunks real-world objects into your living room. I pulled up a working jet engine and was able to peel away the layers to see how it worked. There’s even a Formula 1 race car that you can load into your environment.

The Super Fruit Ninja game was fun. I turned my living room into a fruit-splattered dojo. I could even launch throwing stars from my hands that would get stuck on my walls.

That’s half a floor plan on top of a low-resolution 360° photo.

Some Vision Pro apps were rushed out the door and are just awful. The Zillow Immerse app is one of them. I found the app glitchy and all the immersive house tours very low-quality. The problem is that the environments that ship with Vision Pro are so high-resolution and detailed that anything short of that is jarringly inferior. 

UX Considerations in Vision Pro

Apple Vision Pro can run iPad apps, at least the ones where the developer has enabled the capability. However, I found that many of the touch targets in iPad apps were not sufficient. Apple’s Human Interface Guidelines specify that hit targets should be at least 44x44 pts. But if opened in Vision Pro, that’s not enough. For visionOS, Apple recommends controls’ centers be at least 60 pts apart. 

I would further recommend that controls for visionOS apps should have large targets. In Apple’s own Photos app, in the left sidebar, only the accordion arrow is a control. Looking at and selecting the accordion label like “Spatial” or “Selfies” does not work. I had to look to the right of the label, to the arrow in order to select the item. Not great.

Eye and hand tracking in Vision Pro are excellent, although not perfect. There were many times when I couldn’t get the device to register my pinch gesture or get my eyes to a point in a window to resize it.

Some apps take advantage of additional gestures like pinching with both hands and then pulling them apart to resize something. I do believe that more standard gestures need to be introduced in the future for visionOS.

Steve Jobs famously once said, “God gave us ten styluses. Let’s not invent another.” Apple eventually introduced the Pencil for iPad. I think for many applications and for users to be productive with them, Apple will have to introduce a controller.

IMAX in My Bedroom

The single most compelling use case for Apple Vision Pro right now is consuming video content, specifically movies and TV shows. The built-in speakers, which Apple calls audio pods, sound fantastic. Apple has been doing a lot of work in Spatial Audio over the years and I experienced really great surround sound in the Vision Pro. The three apps that currently stand out for video entertainment are IMAX, Disney Plus, and Apple TV. 

Watching content in the IMAX —only a couple of trailers were free—reminded me of the best IMAX screen I’ve ever been to, which is the one in the Metreon in San Francisco. The screen is floor-to-ceiling high with a curved railing in front of it. On either side is a backlit IMAX logo, and I could choose from a few different positions in the theater!

Watching a Star Wars movie on Tatooine.

Disney leverages its IP very well by giving us various sets to watch their content. I could watch Avengers: End Game from Avengers Tower, Monsters, Inc. from the scare floor, or The Empire Strikes Back from Luke’s land speeder on Tatooine. 

With Apple TV, I could watch Masters of the Air in a window in my space or go into an immersive environment. Whether it’s lakeside looking towards Mount Hood, on the surface of the moon, or in a discrete movie theater, the content was the star. My wife goes to sleep before me, and I usually put on my AirPods and watch something on my iPad. With Vision Pro, I could be much more engrossed in the show because the screen is as big as my room.

From the Apple commercial “First Timer”

I rewatched Dune from 2021 and was blown away by the audio quality of my AirPods Pro. The movie has incredible sound and uses bass and sub-bass frequencies a lot, so I was surprised at how well the AirPods performed. Of course, I didn’t feel the bass rumble in my chest, but I could certainly hear it in my ears.

Vision Pro Industrial Design

The Vision Pro hardware is gorgeous.

As many others have pointed out, the hardware is incredible. It feels very premium and is a technological marvel. The cool-looking Solo Knit Band works pretty well for me, but everyone’s heads are so different that your mileage may vary. Everyone’s face is also very different, and Apple uses the Face ID scanner on the iPhone to scan your face when you order it. This determines the exact light seal they’ll include with your Vision Pro.

There are 28 different models of light seals. Finding the right light seal to fit my face wasn’t as easy as taking the recommendation from the scan. When I went to pick it up, I opted for a fitting, but the 21W that was suggested didn’t feel comfortable. I tried a couple of other light seal sizes and settled on the most comfortable one. But at home, the device was still very uncomfortable. I couldn’t wear it for more than 10 minutes without feeling a lot of pressure on my cheeks.

The next day, I returned to the Apple Store and tried three or four more light seal and headband combinations. But once dialed in, the headset was comfortable enough for me to watch an hour-long TV show.

I wonder why Apple didn’t try to develop a method that requires less variation. Wouldn’t some memory foam cushioned light seal work?

Apple’s Ambitions

The Apple Vision Pro is an audacious device, and I can tell where they want to go, but they don’t yet have the technology to get there. They want to make AR glasses with crystal-clear, super-sharp graphics that can then be converted to immersive VR with the flick of a dial.

That’s why EyeSight, the screen on the front of the headset, allows people in the surrounding area to see the user’s eyes. The device also has a passthrough camera, allowing the user to see out. Together, these two features allow Vision Pro to act as a clear two-way lens.

But Apple seems to want both AR and VR in the same device. I would argue that it might be physically impossible. Imagine an Apple device more like the HoloLens, where they are truly glasses with imagery projected onto them. That eliminates the smaller-than-their-competitors’ field of vision, or FOV. That would eliminate the ridiculous fitting conundrum as the glasses could float in front of your eyes. And that would probably reduce the device’s weight, which has been discussed at length in many reviews.

And then, for VR, maybe there’s a conversion that could happen with the AR glasses. A dial could turn the glasses from transparent to opaque. Then, the user would snap on a light-blocking attachment (a light seal). I believe that would be a perfectly acceptable tradeoff.

What $3,500 Buys You

In 1985, when I was 12 years old, I badgered my father daily to buy me a Macintosh computer. I had seen it at ComputerLand, a computer shop on Van Ness Avenue. I would go multiple times per week after school just to mess around with the display unit. I was enamored with MacPaint.

After I don’t know how many months, my dad relented and bought me a Macintosh 512K. The retail cost of the machine in 1985 was $2,795, equivalent to $8,000 in 2024 dollars. That’s a considerable investment for a working-class immigrant family. But my wise father knew then that computers were the future. And he was right.

With my Mac, I drew illustrations in MacPaint, wrote all my school essays in MacWrite, and made my first program in HyperCard. Eventually, I upgraded to other Macs and got exposed to and honed my skills in Photoshop and Illustrator, which would help my graphic design career. I designed my first application icon when I was a senior in high school.

Of course, computers are much cheaper today. The $999 entry model MacBook Air is able to do what my Mac 512K did and so much more. A kid today armed with a MacBook Air could learn so much!

Which brings us to the price tag of the Apple Vision Pro. It starts at $3,499. For a device where you can’t—at least for now—do much but consume. This was an argument against iPad for the longest time: it is primarily a consumption device. Apple went so far as to create a TV spot showing how a group of students use an iPad to complete a school project. With an iPad, there is a lot of creation that can happen. There are apps for drawing, 3D sculpting, video editing, writing, brainstorming, and more. It is more than a consumption device.

More than a Consumption Device? Not So Fast.

For Vision Pro, today, I’m not so sure. The obvious use case is 3D modeling and animation. Already, someone is figuring out how to visualize 3D models from Blender in AVP space. It’s tied to the instance of Blender running on his Mac, though, isn’t it? 3D modeling and animation software is notoriously complicated. The UI for Cinema 4D, the 3D software that I know best, has so many options and commands and so many keyboard shortcuts and combinations that it would be impossible to replicate in visionOS. Or take simpler apps like Final Cut Pro or Photoshop. Both have iPad apps, but a combination of the keyboard and mouse can make a user so much more productive. Imagine having to look at precisely the right UI element in Vision Pro, then pinch at exactly the right thing in a dense interface like Final Cut Pro. It would be a nightmare.

Being creative with djay in Apple Vision Pro

I do think that creative apps will eventually find their way to the platform. One of the launch apps is djay, the DJing app, of course. But it will take some time to figure out.

Beyond that, could a developer use Vision Pro to program in? If we look to the iPadOS ecosystem there are a handful of apps to write code. But there is no way to check your code, at least not natively. Erik Bledsoe from Coder writes, “The biggest hurdle to using an iPad for coding is its lack of a runtime environment for most languages, forcing you to move your files to a server for compiling and testing.” The workaround is to use a cloud-based IDE in the browser like Coder. I imagine that the same limitations will apply to Vision Pro.

The Bottom Line

For $3,500, you could buy a 16-inch MacBook Pro with an M3 Pro chip and an iPhone 15 Pro. Arguably, this would be a much more productive setup. With the Mac, you’d have access to tens of thousands of apps, many for professional applications. With the iPhone, there are nearly five million apps in the App Store.

In other words, I don’t believe buying an Apple Vision Pro today would open a new world up for a teenager. It might be cool and a little inspirational, but it won’t help the creator inside them. It won’t do what the Mac 512K did for me back in 1985.

Vision Pro’s Future

Clearly, the Apple Vision Pro released in 2024 is a first generation product. Just like the first-gen Apple Watch, Apple and its customers will need to feel their collective way and figure out all the right use cases. We can look to the Meta Quest 3 and Microsoft HoloLens 2 to give us a glimpse.

As much as people were marveling at the AR vacuum cleaning game for Vision Pro, AR and VR apps have existed for a while. PianoVision for Meta Quest 3 combines your real piano or keyboard with a Guitar Hero-like game to teach you how to play. The industrial applications for HoloLens make a lot of sense.

Now that Apple is overtly out of the closet in the AR/VR game, developers will show great enthusiasm and investment in the space. At least on Reddit, there’s a lot of excitement from users and developers. We will have to see if the momentum lasts. The key for the developers will be the size of the market. Will there be enough Vision Pro users to sustain a thriving app ecosystem?

As for me, I decided to return my Vision Pro within the 14-day return window. The only real use case for me was the consumption of media, which I couldn’t justify spending $3,500 for a room-sized TV that only I could watch. Sign me up for version 2, though.