Apple’s Strict Stance on Vibe Coding

Dhruv Amin, co-founder of Anything, recently received an unwelcome notification: Apple removed their app from the App Store.

The reason cited was a violation of App Store Review Guideline 2.5.2, which states that apps must be self-contained and cannot read or write data outside their designated container or download, install, or execute code that changes the app’s features or functions, except in rare educational scenarios where the source code is visible and editable.

Anything is a tool that supports Vibe Coding. It launched on the web in August last year and released its iPhone version in November, passing Apple’s review without issues. However, after several successful updates, Apple began rejecting their submissions in mid-December, citing Guideline 2.5.2. Recently, the Anything official account tweeted a sarcastic yet poignant message: “Breaking news: Apple is panicking over Vibe Coding and has removed Anything from the App Store, so we are moving our app development to iMessage.”

Vibe Coding Hits a Wall with the App Store

To understand this removal controversy, one must grasp what Apple’s Guideline 2.5.2 regulates and why Anything crossed the line.

Imagine ordering a meal at a restaurant; the food served must be pre-approved by the kitchen. Once the app is approved, the version users download must be the same as the one reviewed. This rule has long targeted developers who use “hot updates” to sneak new code into apps after approval, bypassing the review process.

Vibe Coding tools like Anything allow users to describe requirements in natural language, with AI generating code that runs and previews directly on the device. Each new requirement leads to new code being generated and executed on the device.

From Apple’s perspective, this is structurally indistinguishable from “hot updates.” The approved shell of Anything runs code that changes dynamically each time a user interacts with it, which Apple has not reviewed.

Earlier this year, after reports of Apple freezing updates for a batch of Vibe Coding apps, Anything attempted to compromise by moving code previews to a web browser. However, Apple still did not approve it and removed the entire app.

This indicates that Apple’s judgment may not be about a specific feature but rather a fundamental conflict between the operational logic of such products and the App Store’s review model. As long as Anything acts as an entry point for “generating and distributing dynamic code,” Apple believes it exceeds the review scope.

Other apps have also been affected. Since around December, Apple has been blocking updates for several AI coding applications, with some teams, like Vibecode, abandoning mobile development for pure web products.

The Impact of Vibe Coding

To understand why Apple is suddenly so concerned, one must recognize the significant wave Vibe Coding has created. In March 2008, Apple opened the iPhone SDK for free, leading to a rapid explosion in mobile development. Eighteen years later, a similar phenomenon is occurring.

The catalyst for this shift was a post by Andrej Karpathy in February 2025, where he humorously described his coding experience as being immersed in the atmosphere, issuing commands without delving into code details, essentially forgetting about the code itself. He termed this state “Vibe Coding.”

Vibe Coding has gained traction in the AI programming community, with influential figures like Ryan Dahl declaring the end of the era of handwritten code. The release of Claude Sonnet 4 and GPT-5-Codex in 2025 marked the emergence of practical AI models capable of understanding entire projects and autonomously completing tasks.

The lowering of barriers has directly impacted the volume of app submissions to the App Store. Statistics show a 24% year-over-year increase in new app submissions for 2025, with a record of approximately 557,000 new applications.

However, the surge in submissions has led to longer review times, with developers expressing frustration over delays. Apple claims to process over 200,000 submissions weekly, with an average review time of 1.5 days.

Importantly, Apple has also invested in Vibe Coding. The recent release of Xcode 26.3 integrates Claude Agent and OpenAI Codex, allowing developers to leverage AI within Apple’s own development environment.

The Dangers of Vibe Coding

Apple’s tightening of guidelines is not without reason. For instance, after the launch of the Sora app, numerous counterfeit Sora applications appeared on the App Store, some even using OpenAI’s official logo in their icons.

In response to the surge of low-quality apps, Apple revised its review guidelines to emphasize originality and discourage plagiarism.

The issues arising from Vibe Coding are real. Karpathy himself noted that his approach to projects often lacks deep understanding of the code, which can lead to significant problems later on. AI-generated code frequently lacks proper error handling and can lead to system crashes.

A group of self-proclaimed “AI archaeologists” has emerged, hired by companies to clean up the mess left by Vibe Coding, sifting through millions of lines of AI-generated code to find critical bugs.

Despite Apple’s efforts to control the situation, the trend toward democratizing software development is irreversible. The removal of Anything is merely a friction point between a startup and platform rules.

Vibe Coding represents a restructuring of the entire software production relationship. The era of the “1 founder + 1 AI = 1 million-dollar app” is on the horizon, and the App Store may not be able to hold back this wave for long.

This is both the best and worst of times, a sentiment that resonates with Dickens’ famous quote, now fittingly applicable to the App Store’s review process.