80% of Apps Are Going to Disappear. Here's Why That's Not Hyperbole.

‍Introduction

Peter Steinberger - our investor and the developer behind OpenClaw, a personal AI agent with a fast-growing global community - made a claim recently that's easy to dismiss as tech-bubble optimism: 80% of apps will disappear in the age of AI agents.

It's the kind of number that sounds like a conference keynote soundbite. But when you listen to his reasoning, it's hard to argue with the logic.

What most apps actually do

Strip away the UI and the branding, and most apps are doing one of two things: managing data or triggering actions.

MyFitnessPal logs your calories and tracks your macros. Your Eight Sleep app adjusts your mattress temperature. Sonos lets you control your speakers. A calendar app reminds you of things. A to-do app stores your tasks.

None of these require a dedicated interface. They require access to data and the ability to execute a function.

Peter's argument is simple: an AI agent can do all of that - and do it better - because it has something no individual app has: context across everything.

"Why do I need MyFitnessPal when the agent already knows where I am? It can assume I'm going to make bad decisions at Waffle House, or it can modify my gym workout based on how well I slept, or whether I have stress."

A single app only knows what you tell it. An agent knows your sleep data, your location, your calendar, your habits - and can reason across all of it simultaneously. That's not a feature improvement. That's a different category of intelligence.

The subscription math stops making sense

There's another force at work here, and it's purely economic.

Most consumers are paying for 5, 10, maybe 20 subscriptions. Each one solves a narrow problem. Each one requires you to open the app, navigate the interface and remember it exists.

Peter frames the shift bluntly: why should you pay another subscription for something an agent can just do now?

The value proposition of a dedicated app was always that it did one thing really well. But the implicit assumption was that you were the integration layer - you opened the app, entered the data, made the decision. The moment the agent becomes the integration layer, the app's interface stops being useful. It's an extra step in a process the agent could handle end to end.

Apps won't disappear - they'll become APIs

The nuanced version of Peter's prediction isn't that every app ceases to exist. It's that apps that survive will look completely different.

The companies that adapt fastest will stop building for human interfaces and start building for agent interfaces.

Uber Eats still needs to exist. Someone still needs to dispatch the driver and process the payment. But why does that require a human opening an app, scrolling through a menu, and tapping confirm? The agent can handle all of that. Uber Eats' real product is the logistics network - not the app. The app is just how humans currently access it.

Peter's prediction: apps that can expose their core function as an API will survive. Apps that can't will get bypassed.

And agents will find a way regardless. On Android, agents are already navigating UIs directly - clicking buttons, filling forms, extracting data. Peter watched his own agent click the "I'm not a robot" button without prompting. Companies can try to block this. Some will. But the user intent is there, and eventually the path of least resistance wins.

"Apps will become APIs whether they want to or not. My agent can figure out how to use my phone."

This has happened before

Every major platform shift produces the same pattern: companies that treat the new interface as optional eventually lose to those that treat it as primary.

The internet forced every business to rethink what their product was. The mobile shift forced it again. Companies that moved fast built new value on the new interface. Companies that defended their existing model either adapted late or disappeared.

The agent shift is the same thing happening again, compressed into a shorter timeline.

Peter points to Google as an example of a company that hasn't yet made this easy. There's no CLI for Google. Getting certified access to Gmail data can take startups months. So agents route around it - connecting directly through user-authorized access, or navigating the web interface autonomously. Google's data is still accessible. Google just isn't the one providing it cleanly.

"They cannot prevent me. And worst case, my agent just clicks on the website and gets the data out that way."

The missing piece: context

Here's where it gets interesting for anyone building in this space.

Even if agents replace the interfaces of most apps, they still have a fundamental blind spot: they don't know what's happening in the physical world.

An agent can manage your calendar, but it doesn't know you're driving. It can send you a detailed briefing, but it doesn't know you're walking with your phone in your pocket. It can schedule a follow-up task, but it doesn't know your attention is limited right now.

This is the gap between an agent that technically does the job and an agent that actually fits into your life.

The app era solved this by constraining interaction to moments when you chose to open the app. The agent era removes that constraint entirely - agents act proactively, continuously, across your whole day. Which means they need to understand your situation in real time, not just your intent.

Physical context - motion, environment, attention level - isn't a nice-to-have in the agent era. It's infrastructure.

What this means if you're building

The app store model isn't going away overnight. But the writing is on the wall for any product whose core value is wrapped in a proprietary interface.

The question every product team should be asking right now isn't "how do we improve our app?" It's "what is the actual value we provide, and how do we expose that to an agent?"

Peter's 80% figure might be aggressive. Maybe it's 60%. Maybe the timeline is longer than anyone expects. But directionally, the prediction is hard to dispute. Apps that are primarily interfaces to data or simple functions are solving a problem that agents are increasingly solving better.

The ones that survive will be the ones that figured out - early - that their real product was never the app.

‍