The HomeView App Dilemma: Why the CarPlay Shell Model Doesn't Work

Apple's newly launched smart home device HomeView is facing a strange problem: it has virtually no app ecosystem of its own. The Vergecast hosts David Pierce and Nilay Patel dug into this in their latest episode, and their conclusion was clear — Apple simply ported CarPlay's app model over to HomeView, but that logic makes no sense at all on a kitchen countertop.

The Current State of HomeView Apps: A Device Without an App Store

HomeView has Apple's first-party apps like FaceTime, Home, Calendar, Messages, and Podcasts, but no App Store. What's even more puzzling is that some apps you'd expect to be there are missing — Apple Music Classical and Apple News, for instance, are both unavailable.

The real problem lies in how Apple handles third-party developers. According to the developer terms, if you've built a CarPlay app, it automatically appears on HomeView, and developers have almost no way to opt out. This means HomeView's app ecosystem was locked into CarPlay's framework from day one.

Pierce was openly baffled: "Apple's greatest strength is apps. They could have challenged developers to think about what a truly multi-user app for a shared kitchen screen looks like. But instead, Apple locked it down to the iPhone."

CarPlay's app distribution mechanism: Apple has a unique set of rules for reviewing and distributing CarPlay apps — CarPlay apps submitted by developers must conform to Apple's "driving scenario" templates (limited categories like navigation, audio, communication, and parking), with the app's core logic running on the iPhone while the car head unit serves only as a display and input terminal. This mechanism extends automatically to HomeView because Apple's developer agreement defines "CarPlay-compatible apps" as renderable across multiple "external display scenarios," and HomeView falls into that category. For developers, this means once you've adapted for CarPlay, your app appears on HomeView and cannot be independently delisted or opted out. This "adapt once, deploy everywhere" approach lowers development costs but also strips developers of the ability to redesign interactions for a shared kitchen screen.

The Fundamental Contradiction of the CarPlay Model

CarPlay is essentially screen projection — it's just a second monitor for the iPhone, with all app logic running on the phone. This approach makes sense in a car: automakers will never be able to compete with the latest iPhone on processors and memory.

But HomeView is a completely different story. It has plenty of processing power, plenty of screen real estate, doesn't need battery power, and the user isn't driving. As Patel pointed out: "There's nothing stopping this device from being a real computer, except Apple won't allow it."

On the show, both hosts agreed that CarPlay's own user experience isn't particularly good. Take Spotify, for example — what you get on CarPlay is a far cry from the full version. If users at home have to go through CarPlay's cumbersome voice-search flow on HomeView, it would be deeply frustrating.

Screen projection vs. native execution: CarPlay is a "projection" approach — apps run on the iPhone, and the car head unit is just an external display that depends on the phone's processing power and network connection. The alternative is "native execution," where apps are installed directly on the device and use its own compute. HomeView has a local chip, continuous power, and a fixed network environment, so in theory it could go the native route entirely — letting apps run independently on the device, support multi-user switching, and stay always-on. Apple chose projection, likely to reuse its existing ecosystem and reduce the developer burden, but the cost is that HomeView can't reach its hardware potential and can't become a truly independent platform. This trade-off is at the heart of the hosts' debate.

Lessons from Apple Watch and the iPad Counter-Question

Patel wrote the cover story for Wired on the first-generation Apple Watch. He noted that the Apple Watch was initially just a remote control for the iPhone, with apps running on the phone. As processor performance improved, apps gradually migrated to run locally on the watch, and the Apple Watch eventually became a fairly independent computer.

But HomeView doesn't appear to be heading down that path — it seems stuck in the first-generation Apple Watch mode.

The show received a flood of listener feedback, and the core sentiment was remarkably consistent: if Apple had simply made an excellent MagSafe stand that an iPad could slot into, that would have been the truly sensible solution. Some even joked about "seeking justice for the Pixel Tablet" — Google's product actually got the direction right.

The Software Test Behind the Hardware Onslaught

Apple is going through an intense hardware release cycle: the iPhone Duo is about to launch, the new Apple TV is already out, HomeView is on the way, and there are rumors about a touchscreen MacBook.

Patel offered a key assessment: "The success or failure of all these devices will be a success or failure of software and ecosystem. HomeView is essentially an LCD screen plus a generation-old chip — that kind of hardware could have been built at any point in the past hundred years."

Both hosts noted a contrast: the developer community is actively adapting apps for the iPhone Duo, and day-one App Store support is expected to be strong. This is precisely Apple's core advantage over Android — convincing developers to support new form factors. But the developer story Apple is offering for HomeView is even weaker than it was for Vision Pro.

The New AI Race Landscape: The Missing Consumer Product

The second half of the show turned to the AI industry. Google released Gemini 4, claiming performance on par with competitors at a much lower price, putting it back at the forefront of the model race. Meta released its Muse series of models.

But Patel argued that the real differentiator isn't model capability — it's product thinking. He had read Kevin Roose's new book "The AGI Chronicles," which points out that Google has long treated AI as a tool to improve existing products rather than as an entirely new product category. "Google is releasing all these models, but they don't know what their product should be."

Meta, by contrast, launched products consumers actually wanted first, then gradually enhanced their capabilities through model upgrades. Pierce observed that Muse proactively pushed him fantasy football suggestions he never asked for — while slightly ridiculous, it perfectly illustrates Meta's strength in consumer products.

OpenAI and Anthropic, meanwhile, are more like enterprise software companies. Patel pointed out: "Everyone demoing these models says they're great at using Blender, but only five people in the world need that capability."

The mismatch between model capability and product thinking: In the AI race, "model capability" refers to technical metrics like benchmarks, inference speed, and context length, while "product thinking" refers to whether you can translate that capability into a product users want to use daily. Google has long treated AI as an enhancement layer for existing products like Search and Workspace, which means its models are powerful but lack an "AI-native" consumer entry point. Meta is the opposite — it first attracts users with tangible products like smart glasses and chat assistants, then iterates on the underlying models. This path difference explains why Google leads on model leaderboards yet lacks an iconic AI product in consumers' minds. Kevin Roose also emphasizes in "The AGI Chronicles" that the real competition isn't about who reaches a certain capability threshold first, but about who defines the everyday form of user-AI interaction.

The X Platform Resurgence and Social Media Fragmentation

The show also discussed the phenomenon of tech journalists returning to X. After MacStories' two founders announced their return to X, it triggered strong backlash within their team. Platformer's Casey Newton and Daring Fireball's John Gruber also announced their returns.

Gruber wrote that "X has marked the spot again" and described the culture on X as "friendly and inclusive." Pierce strongly disagreed: "Gruber and I are both white men, which is a very advantageous identity on Twitter. My experience on X is the exact opposite of inclusive."

Both agreed on one thing: X does have influence in one vertical — AI conversations. All the builders and investors in the AI industry are there. But that doesn't mean X has regained broad cultural influence. Meanwhile, Threads had an opportunity to be an alternative, but its algorithm over-optimizes for engagement, causing the user experience to steadily deteriorate.

Pierce revealed that Threads' product lead recently announced a change — content from people you follow will appear in your feed. His reaction: "What are we even doing? Everyone's lost in the TikTok-style content recommendation sauce, forgetting how people actually want to talk to each other on the internet."

The tension between algorithmic recommendations and the social graph: Social platform feeds are typically driven by two mechanisms — the "social graph" (content from people you follow) and "algorithmic recommendations" (content the platform predicts you'll engage with). Threads initially leaned heavily on algorithmic recommendations to boost time-on-app, leading users to complain they couldn't see content from accounts they followed, with their feeds dominated by viral posts from strangers. This strategy boosted engagement in the short term but eroded users' sense of belonging and conversational continuity on the platform. X's "resurgence" confirms this: AI professionals are returning to X not because its culture has become inclusive, but because the platform still has a dense social graph in certain verticals that efficiently reaches peers. When algorithmic recommendations overwhelm the social graph, a platform loses its function as a "public conversation space."

Other Notable Topics

This episode also covered several noteworthy developments:

Character AI bankruptcy: This early AI companion company filed for bankruptcy, and concerns have been raised about where its user data (including minors' data) will go in the proceedings. Google had previously acquired its founding team through a reverse acquisition, leaving the company essentially a shell.

OpenAI eavesdropping incident: An OpenAI employee hid a listening device behind a conference room TV, wrapped in toilet paper. Even more absurdly, he used his wife's ChatGPT account to search for how to hide the device. He has been arrested.

Smartphone market contraction: A Counterpoint report shows the global smartphone market declined 20% year-over-year, the lowest in 14 years. Samsung and Apple's combined share jumped from 37% to about 45%, and cheap phones are disappearing.

The agent vs. platform battle: Financial institutions like JPMorgan Chase and Bank of America are starting to block AI agents from accessing their platforms. The debate over "whether your computer represents you" is becoming a new industry focal point.

One-line Summary

Apple's HomeView is locked into the CarPlay app framework, exposing deep contradictions in its software ecosystem strategy.

Paragraph Summary

This episode of Vergecast digs into the ecosystem dilemma of Apple's new device HomeView: no App Store, third-party apps forcibly bound to the CarPlay framework, and developers with virtually no way out. The hosts argue that CarPlay's screen-projection logic doesn't make sense on a stationary home device, and that HomeView could have been a real computer. The show also compares Apple Watch's path from remote control to independent computing, noting that the success of Apple's hardware blitz will hinge on software and ecosystem. The second half turns to the AI industry, analyzing Google's lack of consumer product thinking, Meta's product-first-then-model-upgrade approach, and the centralization of AI conversation reflected in tech journalists' return to X.