Will apps disappear or multiply? Both, actually.

Why Pete Steinberger is both right and wrong about the future of apps.

Peter Steinberger, creator of OpenClaw, made a bold claim recently: 80% of current apps will disappear over the next couple of years as autonomous AI agents take over.

And yet, everywhere you look, someone is shipping a new app using the latest AI tools.

So what gives?

Let’s clear out the second part first:

The proliferation of new apps that you see is largely driven by the infused superpowers of AI revolution rather than a grounding in SCALABLE real user problem. One look at Lovable and Bolt.new homepage will substantiate more than I need to mention. The pages are full of enthusiastic projects borne out of a long nurtured desire to build but a lack of technical ability to support.

However, this enthusiasm mostly deflates when the maintenance and distribution rubber hits the road.

(Honest admission: I have been guilty of falling for it more times than I’d like to admit here 😅)

Now the first part:

The argument makes sense on the surface: many apps are indeed just convenient UIs sitting on top of simple data layers - weather apps, tracking apps, to-do lists, among countless other examples. Steinberger believes “an entire layer of apps will slowly disappear because if they have APIs, they are just services that your AI will call.”

This is playing out already. Users (using agents) are accessing multiple enterprise applications (the ones that never paid attention to a clumsy UI - more on why a decent UI matters later) via the API rather than the GUI. And they seem to be much happier with the trade-off.

But the argument also completely overlooks something very human.

Just because you can cook doesn’t mean restaurants go out of business.

Restaurants survive not just on food, but on experience, habit, and the mental relief of not having to decide - which is exactly what great apps do.

Here’s why I believe the case for apps disappearing isn’t as straightforward and will instead lead to more interesting apps with a higher bar:

🧠 The human mind compartmentalises through distinct UIs

Users judge apps in under 50 milliseconds, and the most effective icons invoke emotion while being instantly recognisable. The yellow icon for groceries. The green one for banking. The white one for notes. Those little icons on the home screen quietly drive you to call them in the moment of need. Agents are expected to be uniform and abstracted - and that’s precisely where they will lose the human mind.

⏳ Convenience compounds over time

That’s why businesses exist at scale. The average smartphone owner uses 30 different apps each month, and 88% of mobile time is spent inside apps. Would you rather spin up data layers, feed them context, maintain memory, and deploy agents securely

OR

tap an icon you’ve trusted for years? For most people, that’s not even a question.

While the enthusiasts would argue in favour of building more personal (sub)agents, the app market runs on the average consumer’s time and not a few power users’.

🏦 Trust is category-specific

Would you let an agent handle your banking, your medical records, your child’s screen time? Certain categories require the accountability of a named, regulated, auditable product, not a general-purpose agent running locally on your machine.

While the recent releases from the big players like Anthropic and OpenAI do hint at making a move towards category-specific agents (think Finance and Health), there are still way too many categories you wouldn’t trust an AI agent with for a long time.

🌍 Accessibility is overstated

Not everyone has the technical literacy or the reliable connectivity to run a personal AI agent. This is probably the most understated fact about the world we live in. Because everyone seems to be shouting AI, it appears as if everyone is running tens of personal agents everyday. The reality is far from it.

On the other hand, apps are designed for the lowest common denominator on purpose. Billions of users in emerging markets aren’t spinning up local agents anytime soon.

📈 Recent numbers - apps are booming, not dying

The iOS App Store alone has over 2 million apps, with around 2,000 new ones being added every single day. Now this might be a small function of the enthusiasm I alluded to earlier. However, mobile apps also generated over $935 billion in revenue in 2025 - a 52% increase from just two years prior. That doesn’t sound like a channel on its deathbed yet.

Google’s Play Store also has a similar story.

BUT

The nature of apps will change.

Having argued that apps are here to stay, I do believe majority of them will be of a different type to the current ones.

✅ The winners: apps that are true companions and earn your attention every day

AI tools have dramatically lowered the cost of building apps. This means more niche, hyper-personalised apps - not fewer.

An app for left-handed guitarists. An app for parents of premature babies. The long tail of apps may actually grow as the cost to build approaches zero, even as the middle of the market hollows out.

Some stats that signal the growth of hyper-personal and interesting apps:

Productivity apps recorded nearly 80% revenue growth year-over-year in 2025 - the strongest among all major categories.

Finance apps grew 71%.

Utility & Productivity apps 47%.

Mental health apps saw 67% of entries grow by more than 50% (as users seek wholesome wellness, not just step counters).

❌ The losers: joyless/thin data layer apps (Pete is spot on here)

Think about how many apps you've downloaded to do one simple thing - convert a file, check a currency, track a package, split a bill. These single-utility apps are genuinely dead. Any agent handles them without a second thought, and no user minds that.

A huge chunk of the app economy is also B2B - tools people use not out of love but obligation. Expense reporting, HR portals, project trackers - these are exactly the kind of API-accessible, joyless data layers agents will absorb first. And unlike consumers, companies will make the switch if it saves money.

Data from 2025 on other major categories:

Photography apps dropped 7.4% (thank you, Nano Banana!).

News & Magazines fell 1.4% (thank you, Perplexity!) - categories where agents have become sufficiently good at aggregating, curating, and delivering on command already.

In fact, in 2025, the AI sector itself entered an “elimination phase” - multiple reports suggest many small and mid-sized products exited due to weak monetisation and severe product homogeneity (think Vibe Coded but more professional), causing the total number of advertisers to drop by nearly half. No surprises there - typical of an early phase with any technology - people build, tinker and exit when things don’t seem to move.

Even Steinberger himself, after OpenClaw exploded with ~150,000+ GitHub stars, ended up joining OpenAI - with a stated mission to “build an agent that even my mum can use.”

That sounds less like apps disappearing and more like the bar for what a great app needs to be just got a lot higher.