The End of Apps? Nothing CEO Carl Pei Envisions a World Where AI Agents Replace Your Smartphone’s Home Screen

The smartphone as we know it is a graveyard of icons. Every morning, millions of users wake up to a grid of colorful squares—each one a portal to a specific function, a siloed experience, a walled garden. But what if that grid vanished? What if, instead of tapping on a weather app, you simply asked your phone, “Should I bring an umbrella today?” and it answered, proactively, before you even finished the thought?

This is the future that Nothing CEO Carl Pei is betting on. In a bold statement that has sent ripples through the tech community, Pei predicted that smartphone apps will eventually disappear, replaced entirely by intelligent AI agents [1]. Speaking at an event earlier this month, he painted a picture of a world where users interact with their devices through fluid, intent-driven conversations rather than manual taps and swipes. It’s a vision that feels both radical and inevitable—a logical endpoint for an industry that has spent the last two years obsessing over generative AI, large language models, and the promise of truly adaptive computing.

But what does this mean for the 1.5 billion apps currently sitting in the App Store? And more importantly, is Pei’s prediction a glimpse of the future, or just another piece of tech utopianism that overlooks the messy realities of human behavior?

The Death of the Grid: Why Apps Are Already Showing Their Age

To understand Pei’s argument, you have to go back to 2007. The iPhone didn’t just invent the smartphone; it invented the app economy. Steve Jobs famously called the App Store a “revolutionary” platform, and it was. For the first time, users could customize their devices with third-party software, turning a phone into a camera, a GPS, a gaming console, or a music player with a single download. The app became the atomic unit of mobile computing.

But that model is showing its age. The walled garden approach—exemplified by Apple’s strict App Store policies—has created a system where innovation is often stifled by gatekeeping. Consider the case of Musi, a free music streaming app that was recently delisted by Apple despite having millions of loyal users [2]. The legal battle over Musi highlights a fundamental tension: platforms control the distribution of software, and they can pull the plug on a product at any time, for any reason. This is not a bug; it’s a feature of the current ecosystem. But it’s a feature that users are increasingly frustrated with.

Pei’s AI agent vision offers an alternative. Instead of downloading a discrete app for music, weather, or navigation, users would interact with a single, omnipresent AI system that understands context, intent, and history. This isn’t just about convenience; it’s about breaking down the silos that define modern computing. An AI agent doesn’t need to open a separate app to check your calendar, send a message, or book a ride. It can do all of those things in a single, fluid interaction, pulling data from multiple sources in real-time.

This shift aligns with broader trends in the industry. Google, for instance, has been aggressively integrating its generative AI capabilities into its ecosystem through the Gemini project. Apple, meanwhile, has been quietly embedding AI into its Vision Pro headset, using it to enhance spatial computing and user interaction [4]. The pieces are already in place; Pei is simply stating the logical conclusion.

The Agent Paradigm: How AI Will Reimagine User Intent

The technical leap from apps to agents is more profound than it might seem at first glance. Traditional apps are reactive: they wait for a user to open them, provide input, and then execute a task. AI agents, by contrast, are proactive. They don’t just respond to commands; they anticipate needs.

Imagine a scenario where you’re running late for a meeting. A traditional smartphone setup requires you to open your calendar, check the time, open a maps app to estimate travel time, and then send a message to the other party. An AI agent, on the other hand, could monitor your calendar, detect the conflict, check traffic data, and draft an apology message—all before you even look at your phone. This is the promise of intent-driven computing, and it’s a paradigm shift that requires a fundamentally different architecture than the app-based model.

From a technical perspective, this shift relies heavily on advancements in open-source LLMs and vector databases. AI agents need to understand not just the words you say, but the context behind them. They need to remember your preferences, your habits, and your relationships. This requires a persistent, personalized knowledge base—something that vector databases are uniquely suited to provide. By storing user data as high-dimensional embeddings, these systems can retrieve relevant information in milliseconds, enabling agents to act with near-human intuition.

This is not science fiction. Companies like Google and Apple are already investing heavily in these technologies. Google’s Gemini project, for example, is designed to be a “universal” AI model that can understand and generate text, images, audio, and code. Apple’s Vision Pro headset uses AI to map your environment, track your gaze, and interpret your gestures. The infrastructure for an agent-driven future is being built right now.

The Developer Dilemma: Building for a Post-App World

For the millions of developers who have built careers on the app economy, Pei’s prediction is both an opportunity and a threat. On one hand, AI agents offer a more flexible and powerful platform for building innovative solutions. Instead of being constrained by the limitations of a mobile operating system, developers can create services that integrate seamlessly into a user’s daily life. An AI agent could, for example, book a restaurant reservation, order groceries, and adjust your thermostat—all without requiring a dedicated app for each task.

But this flexibility comes at a cost. The skills required to build for an AI-driven ecosystem are fundamentally different from those needed to build a traditional app. Developers will need to understand natural language processing, context-aware computing, and API orchestration. They will need to design for conversation rather than navigation. This is a steep learning curve, and it risks displacing those who cannot adapt.

Moreover, the shift to AI agents could exacerbate existing power imbalances in the tech industry. Large corporations like Apple, Google, and Samsung have the resources to build and maintain sophisticated AI systems. Smaller startups and independent developers, by contrast, may struggle to compete. This could lead to a new era of technological haves and have-nots, where access to advanced AI systems becomes a key determinant of success.

The legal landscape is also shifting. Apple recently rolled out its first-ever "background security improvement" update for iPhones, iPads, and Macs, fixing a vulnerability in its Safari browser [3]. While this update is unrelated to Pei’s comments, it highlights the growing importance of security in AI-driven systems. As agents become more powerful, they will also become more attractive targets for malicious actors. Ensuring that these systems are secure, transparent, and fair will be one of the defining challenges of the next decade.

The Security Paradox: Trusting Your Digital Butler

One of the most overlooked aspects of the AI agent revolution is the question of trust. When you open an app, you know exactly what it does: it accesses your camera, your location, or your contacts, and you can revoke those permissions at any time. An AI agent, by contrast, operates in the background, making decisions on your behalf based on a vast array of data points. This creates a security paradox: the more capable the agent, the more access it needs, and the more vulnerable you become.

This is not a hypothetical concern. The cybersecurity landscape is evolving rapidly, with agencies like CISA playing a critical role in ensuring that new technologies are secure [3]. As AI becomes more pervasive, the importance of robust security frameworks will only increase. Users will need to trust that their agents are not only competent but also ethical—that they won’t share sensitive data, make biased decisions, or act in ways that harm the user.

Pei’s vision assumes a level of trust that the current ecosystem has not yet earned. The same companies that are building these AI agents are also the ones that have been criticized for data breaches, privacy violations, and monopolistic practices. Until these issues are addressed, the transition from apps to agents will remain a work in progress.

The Bigger Picture: A New Kind of Digital Divide

Pei’s statement is part of a broader trend in the tech industry toward AI-driven ecosystems. Over the past year, major companies have made significant strides in integrating AI into their products and services. Google has been actively promoting its generative AI capabilities through tools like Gemini, while Apple continues to enhance its Vision Pro headset with advanced AI features [4]. NVIDIA, meanwhile, announced a collaboration with Apple to integrate its CloudXR 6.0 platform with the Vision Pro headset, further solidifying the role of AI in enhancing user experiences [4].

But this shift also raises uncomfortable questions about equity. As large corporations invest heavily in AI research and development, smaller players and under-resourced communities may be left behind. This could lead to a new era of technological haves and have-nots, where access to advanced AI systems becomes a key determinant of success. The same dynamics that have shaped the app economy—winner-take-all markets, platform lock-in, and data monopolies—could be replicated, or even amplified, in an agent-driven world.

There is also the risk of over-reliance. While AI agents offer significant benefits in terms of efficiency and convenience, they also introduce risks. Users may become overly dependent on AI-driven interfaces, losing the ability to perform tasks independently or troubleshoot issues when things go wrong. This is not just a technical problem; it’s a cognitive one. The more we outsource our decision-making to machines, the less capable we become of making decisions on our own.

What Comes Next: Coexistence or Conquest?

So, what does the future hold for traditional apps? Will they fade into obscurity as AI agents take over, or will they continue to coexist alongside these new technologies? The answer, as with most things in tech, is likely somewhere in between.

For the foreseeable future, apps will remain a critical part of the mobile experience. They are familiar, reliable, and well-understood. But as AI agents become more sophisticated, they will increasingly handle the tasks that apps currently perform. The home screen of tomorrow may not be a grid of icons; it may be a single, conversational interface that adapts to your needs in real-time.

Pei’s prediction is a bold one, but it reflects a growing consensus within the tech industry. The era of the app is not over, but its dominance is waning. The question is not whether AI agents will replace apps, but how quickly the transition will happen—and who will be left behind when it does.

For developers, enterprises, and startups, the message is clear: adapt or risk obsolescence. The tools for building an agent-driven future are already here. The only question is whether we have the courage to use them.

---

References

[1] Editorial_board — Original article — https://techcrunch.com/2026/03/18/nothing-ceo-carl-pei-says-smartphone-apps-will-disappear-as-ai-agents-take-their-place/

[2] Ars Technica — Apple can delist apps "with or without cause," judge says in loss for Musi app — https://arstechnica.com/tech-policy/2026/03/judge-upholds-apple-delisting-of-free-musi-app-that-streams-songs-from-youtube/

[3] TechCrunch — Apple rolls out first ‘background security’ update for iPhones, iPads, and Macs to fix Safari bug — https://techcrunch.com/2026/03/17/apple-rolls-out-first-background-security-update-for-iphones-ipads-and-macs-to-fix-safari-bug/

[4] NVIDIA Blog — More Than Meets the Eye: NVIDIA RTX-Accelerated Computers Now Connect Directly to Apple Vision Pro — https://blogs.nvidia.com/blog/nvidia-cloudxr-apple-vision-pro/