Nothing CEO says smartphone apps will disappear as AI agents take their place

Nothing CEO Carl Pei just said out loud what a lot of people in tech have been quietly thinking: the apps on your phone are living on borrowed time. And at SXSW 2026, he explained exactly what’s coming to replace them.

Think about the last time you grabbed coffee with someone. Simple enough, right? Except it’s not. You opened a messaging app to coordinate. Switched to Maps to pick a spot. Jumped to Uber to get there. Checked Calendar to confirm timing. Four apps. Four context switches. Multiple taps through menus that were designed for your fingers, not for getting things done fast.

Carl Pei, co-founder and CEO of Nothing, used that exact example at SXSW 2026 to make his case. And his conclusion was blunt: the smartphone as we know it built around apps, home screens, and app stores is fundamentally broken. It just hasn’t been replaced yet.

“It’s very hard to get things done on a phone,” Pei said during his appearance at the Austin conference. “That’s an intention I want to grab coffee but to execute that intention, we have to go through so many different steps and so many different apps.”

His solution? Stop designing phones for humans to navigate, and start designing them for AI agents to operate.

The App Era Has Lasted 20 Years. Pei Thinks That’s 15 Years Too Long.

Here’s the thing most people don’t stop to think about: the smartphone experience hasn’t fundamentally changed since 2007. Lock screen. Home screen. App icons. An app store to download more. That’s it. That’s the whole model and it’s essentially the same model that Palm Pilots and PDAs were running in the late 1990s.

Pei made exactly this point. “If you think about the user experience, it’s still very similar,” he said. “You have lock screens, home screens, apps. It hasn’t really changed for like, 20 years.”

For a CEO whose brand is built on questioning design conventions Nothing’s transparent-backed Phone (3) being one of the most visually distinct smartphones on the market this kind of critique is on brand. But Pei isn’t just complaining about aesthetics. He’s pointing at something deeper: the entire interaction model of the smartphone is built around humans manually navigating between siloed containers of functionality. That model made sense when software was simpler. In a world where AI agents can understand natural language, infer intent, and execute multi-step tasks autonomously, it makes very little sense at all.

“The current way we use phones is very old-school. It’s pre-iPhone. You have lock screens, home screens, apps. It hasn’t really changed for like, 20 years.” Carl Pei, CEO of Nothing · SXSW 2026

What Actually Replaces Apps The Three Stages

Pei didn’t just drop a hot take and walk off stage. He laid out a specific, three-stage framework for how the transition from app-centric phones to AI-agent-centric devices will actually unfold. It’s worth understanding each step, because we’re arguably already in stage one.

1

Command execution “Do this thing for me”

This is where we are right now. AI features that can execute a single command on your behalf — booking a flight, ordering food, sending a calendar invite. Pei called this stage “super boring,” and he’s not wrong. It’s still very much a human-in-the-loop experience. You’re giving orders; the AI is carrying them out. Useful, but not transformative.

2

Intention learning”I know what you’re trying to become”

This is where things get genuinely interesting. Rather than waiting for commands, the AI begins to learn your long-term intentions and goals. Want to be healthier? The device starts surfacing reminders, routing suggestions, and behavioral nudges without you asking. Pei compared this to a more proactive version of ChatGPT’s memory feature but baked into the operating system itself.

3

Proactive agency “I already did it”

The endgame. A device that acts on your behalf before you think to ask. Not because it’s reading your mind, but because it knows you well enough to anticipate what you’d want. “When the system knows us so well, it will come up with things that we don’t even know we wanted,” Pei explained. This is the version of AI that makes the app grid genuinely irrelevant because there’s no navigation left to do.

The Key Insight: Agents Need Their Own Interface

This is the part of Pei’s vision that most people are missing when they talk about AI on smartphones. The dominant approach right now from Apple Intelligence to Google’s Gemini integration is essentially grafting an AI layer on top of the existing app model. The AI can open apps, tap buttons, navigate menus. It’s mimicking what a human would do.

Pei thinks that’s fundamentally the wrong direction.

“The future is not the agent using a human interface. You need to create an interface for the agent to use. I think that’s the more future-proof way of doing it.” Carl Pei, CEO of Nothing · SXSW 2026

He’s pointing at something real here. When an AI agent has to simulate human touch to navigate a smartphone tap here, scroll there, wait for this screen to load it’s working against massive friction. It’s like building a race car and then having it drive down a hiking trail. The capability is there; the infrastructure isn’t designed for it.

The alternative is building interfaces specifically designed for AI agents to call structured APIs, not visual menus. Think of it as the difference between giving someone a verbal set of instructions and handing them a direct data feed. One requires interpretation and navigation; the other is just information, ready to act on. This is actually where the industry is heading with things like Model Context Protocol (MCP) and agentic API frameworks but most consumer smartphone OSes aren’t there yet.

Nothing’s $200M Bet and What It Actually Means

This isn’t the first time Pei has floated this vision. He used almost identical language when Nothing closed its $200 million Series C funding round last year, led by Tiger Global. At the time, the pitch was an AI-first device built around personalization sophisticated enough that users wouldn’t feel the need to second-guess the AI’s decisions. That’s a high bar and it points to the core challenge of everything Pei is describing.

Trust. That’s the missing ingredient. An AI agent that books your flights and orders your coffee without asking is only useful if you trust it to make the right call. Get that wrong book the wrong flight, schedule coffee during an important meeting and users will immediately revert to doing things manually. The entire value proposition collapses.

Building that trust requires the AI to develop a genuinely accurate model of who you are, what you value, and how you make decisions. It’s a much harder problem than building a good voice assistant. But it’s exactly the problem that models like ChatGPT’s memory and long-context AI systems are starting to work on seriously.

📱 What Has to Change for Pei’s Vision to Work

• Agent-native OS architecture: Operating systems redesigned with structured API layers for AI agents not just AI bolted onto existing app models.
• Deep user modeling: Persistent, accurate understanding of individual goals, preferences, and behavioral patterns far beyond today’s personalization.
• Trust infrastructure: Clear mechanisms for users to understand, verify, and override what their AI agent is doing on their behalf.
• App ecosystem evolution: Developers building backend services designed for agent access, not just human-facing UIs a shift in how software gets built entirely.
• Privacy frameworks: The AI needs to know a lot about you to be genuinely useful. How that data is stored, protected, and controlled is a massive open question.

Apps Aren’t Going Anywhere Tomorrow But the Direction Is Clear

To his credit, Pei isn’t making a reckless “apps are dead by next year” claim. He acknowledged clearly that apps aren’t disappearing in the near term, and pointed out that Nothing’s own OS lets users build mini apps today through vibe coding. The trajectory is what matters.

But “apps are going to disappear” is still a striking thing for a consumer hardware CEO to say publicly. It implies that any startup or company whose core value is locked inside an app interface is sitting on a structural vulnerability. Not an immediate crisis but a long-term one that’s now clearly in motion.

“If you’re a founder or a startup and your app is like where the core value lies, that will be disrupted whether you like it or not,” Pei said. That’s not a gentle nudge. It’s a warning.

The companies that will thrive in an agent-first world aren’t necessarily the ones with the best app interfaces today. They’re the ones building the underlying capabilities the data, the services, the APIs that AI agents will want to call. The interface layer, for decades the primary battleground of consumer tech, may be about to become largely irrelevant.

The Bigger Picture: Everyone’s Moving This Direction

Carl Pei isn’t alone in this thinking. Apple’s latest moves with Apple Intelligence, Google’s Gemini integration across Android, and Samsung’s Galaxy AI features are all incremental steps in the same direction even if none of them have gone nearly as far as Pei is describing. Apple’s push toward on-device AI agents in iOS has been building toward exactly this kind of intent-aware, action-taking model.

And beyond smartphones, the same logic applies to every screen you interact with. Smart TVs, laptops, cars any device that currently relies on humans navigating menus is a candidate for the same transformation. The app grid might be the most visible symbol of the old model, but it’s the model itself that’s changing.

Whether Nothing ends up being the company that actually builds the AI-first device Pei is describing or whether it’s Apple, Google, Samsung, or someone we haven’t heard of yet the destination he’s pointing at is real. The phone of 2030 probably doesn’t have a home screen. It has an AI that already knows what you need.