Google tests the app market version of the SaaSpocalypse

Google AI Studio can generate Android apps from a prompt. That fits a bigger shift in the software market: some software won't be searched for, bought, or subscribed to anymore. It'll just be created on the spot.

Google used its I/O 2025 to show off new AI Studio features that let users build native Android apps right in the browser using a prompt. The apps are built in Kotlin with Jetpack Compose, can tap into sensors like GPS, Bluetooth, and NFC, and run in an Android emulator inside the browser for testing. For now, they're meant for personal use. Sharing with family and friends is on Google's roadmap.

At first glance, this looks like a developer tool. The market logic behind it is more interesting: if users can spin up a simple to-do list, a water tracker, a GPS logger, or a packing list in minutes, there's less reason to open the Play Store for those one-off needs.

Personal software becomes a new layer below the app market

This could create a software category that sits below the traditional app market. These apps are too small, too personal, or too short-lived to make sense as public products. Until now, software had to be general enough to be found, downloaded, and maintained. With vibe coding, it can get extremely specific, like an app for your next vacation, your shared apartment, a hobby project, or an internal workflow.

That's bad news for makers of simple utility apps. Anyone selling a generic checklist app or a basic tracker today won't just compete with other developers. They'll compete with users themselves. For many everyday needs, the question soon won't be which app to install at all, but whether a ready-made app is even needed.

The app-market version of the "SaaSpocalypse"

This fits a debate that's been building in the enterprise software market since 2024 and reached its height in 2026 with the "SaaSpocalypse." Wall Street worries that AI agents could weaken traditional enterprise software by handling tasks companies currently pay for on a per-user, per-app basis. On top of that, there's the fear that companies will just vibe-code their own tools.

OpenAI and Anthropic are already pushing the market in this direction. OpenAI launched Frontier, an enterprise platform that gives AI agents their own identity, permissions, and shared business context across CRM systems, ticketing tools, and internal apps. Anthropic is plugging Claude into tools like QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365 through "Claude for Small Business," offering workflows for finance, sales, marketing, HR, and customer service.

The direction is the same: individual software interfaces are losing their grip. In the enterprise, an agent handles tasks across multiple SaaS products. On Android, a user could generate simple apps for their own needs instead of searching the Store.

Software isn't dying, but some packaging is

Still, the blunt take that "SaaS is dead" goes too far. Salesforce CEO Marc Benioff pushes back hard, arguing that AI makes Salesforce more valuable, not less. Self-built solutions carry real risks around data security and compliance. And so far, today's agents still hit walls on complex tasks.

The same goes for apps. Generating an app in minutes is one thing. Keeping it running is another. Apps stay valuable when users expect security, syncing, support, payments, data quality, or reliability. What's under pressure right now are interchangeable interfaces for simple tasks.

Why Google can afford this risk more than Apple

Google can run this experiment more easily than Apple because AI Studio doesn't flood the public Play Store with generated apps. They run locally, can be sideloaded to a device via USB, or pushed to internal test channels in the Play Console. A poorly built personal app isn't a publicly listed product with ratings, payments, and mass distribution.

At the same time, Google is boosting the traditional app market through a new channel. In the coming weeks, Gemini on Android and the web will start suggesting existing apps directly in conversations with users. Later this year, the assistant will also surface over 450,000 movies, shows, and live sports streams, linking users straight into the provider's Android app. Google is building at both ends: simple needs will increasingly be solvable with a prompt, while professional apps get new reach through Gemini.

Android's more open platform model helps here. Google needs to protect the Play Store, but Android has always thrived on software existing outside a tightly controlled storefront. Strategically, Google can stomach some cannibalization of small Store apps if it makes Android more useful and turns Gemini into the layer where users find or create software.

Apple draws a tighter line. Apple blocked updates from vibe-coding apps like Replit and Vibecode, citing Guideline 2.5.2. That rule bars apps from downloading, installing, or running code that changes the functionality of the app or other apps. TechCrunch also reported that the vibe-coding app Anything was pulled from the App Store twice.

Apple is defending a security model, not just App Store revenue: whatever runs on the iPhone should be reviewed first. Google is testing more local autonomy instead. Whether that leads to less app slop, more app slop, or a genuinely new layer of personal software remains an open question. What's clear: for small, simple use cases, the app as a product is no longer a given.