Apple blocks Replit updates and pulls Anything app, citing rules against agent‑style apps
- Apple has blocked updates to Replit and removed the Anything app, enforcing rules against apps that change code after App Store review. - The policy fight is timely: the App Store saw more than 550,000 AI‑driven app submissions in 2025, an 84% year‑over‑year jump, developers say. - Developers argue Apple’s posture acts as gatekeeping tied to App Store services revenue, while Apple tightens how AI is integrated in Xcode. (x.com) (x.com)
Apple’s App Store fight with AI coding apps looks small on the surface — one blocked update here, one pulled app there. But the real issue is bigger. iPhone apps that can generate, preview, and ship other apps are running straight into rules Apple wrote for a pre-agent era, and Apple is enforcing them unevenly enough that founders now see the store itself as the product bottleneck. ### What actually happened? Apple blocked updates to Replit’s iOS app and held up Vibecode, while it removed the app builder Anything from the App Store, briefly restored it on April 3, then removed it again. The companies say the trouble started in late 2025 and intensified this spring, with Replit still stuck in what its CEO Amjad Masad called App Store “purgatory” as of May 1, 2026. ### What rule is Apple leaning on? The center of gravity is Guideline 2.5.2. In plain English, Apple does not want an approved app downloading or executing new code that changes what the app does after review. That makes sense if you picture malware or hidden features sneaking in later. The problem is that modern “vibe coding” apps are built around exactly that loop — generate code, run it, preview it, tweak it, repeat. ### Why does that hit these apps so directly? Because these products are not just chatbots that talk about code. They are app factories. Replit’s own docs say users can build native iOS apps from a prompt and publish to the App Store with one click. Anything was pitching itself as a mobile app builder with code export, source-code editing, and app previews on the device itself. From Apple’s perspective, that starts to look less like a normal app and more like a runtime for unreviewed software. ### Why are founders so angry? Mostly because the enforcement looks inconsistent. Anything says it rewrote the product multiple times, got conflicting feedback, and after removing the preview feature was told the app now had “minimum functionality.” Replit says it had lived on the platform for years before updates suddenly got blocked. And Lovable — another AI app builder — managed to launch mobile apps on iOS and Android last week. Same broad category, very different outcomes. ### Is this really about safety? Partly, yes. If an app can generate and run fresh code on-device, Apple loses the clean review boundary it has always relied on. A bad actor could use an approved shell app to create something harmful later. Apple also told Anything that users could sideload an app built this way and then point to the original App Store approval as if the generated app had been vetted too. That is a real platform-risk argument, not a made-up one. ### So why does this feel different now? Because AI has changed the scale. The old rule was built for edge cases. Agent-style coding apps turn it into a mainstream workflow. Financial Times’ reporting says Apple is struggling to map existing rules onto tools that can generate and launch software almost instantly, while app releases overall accelerated sharply in 2025. Basically, the review model was designed for static apps, but these products behave more like live software workshops. ### Could these companies just work around Apple? Some already are. Anything has been pushing users toward desktop tools and even iMessage-based flows, and its founder has openly pointed to Android as the more permissive platform. Apple’s own guidelines also make its view pretty explicit — if the App Store model does not fit, there is the web, and in some markets there are alternative iOS distribution paths. But that is still a downgrade if your whole pitch is “build on your phone, ship from your phone.” ### What matters from here? This is turning into a test of whether Apple can adapt App Store rules without breaking its safety model. If it cannot, the next generation of agent-style software may end up web-first, desktop-first, or Android-first by default. The immediate fight is about Replit and Anything. The bigger one is about whether iPhones are allowed to be places where software gets made, not just installed.