Apple’s App Store submission is one of the more demanding gatekeeping mechanisms in consumer software. It requires accurate privacy disclosures, published security standards, measurable performance and accessibility thresholds, and design compliance reviewed by human reviewers. With Artificial General Intelligence (AGI) claims refusing to die, I decided to take a look at whether Claude Code would fit the bill and what would happen if I brought an app from Claude Code and introduced it to the App Store.

Claude Code moves fast at ideation, screen mapping, scaffolding, and boilerplate generation, and developers have shipped real apps this way. But at the back of the development life cycle, where compliance, privacy disclosure, security architecture, and App Store submission live, the human cost reasserts itself. Studies of AI-generated code indicate a significant share requires refactoring to meet Apple’s accessibility and performance standards, and a meaningful fraction of AI-driven apps fail review due to privacy or design violations.

A little more than three years ago, I began developing the AI Life Cycle Core Principles (AILCCP). This is a framework—and now an app—that organizes AI development and deployment obligations across 37 principles, 10 development phases, and 48 controls, mapped to international standards and regulatory enforcement contexts. It gives developers, deployers, lawyers, and policymakers a shared vocabulary and methodology for assessing where an AI system meets its obligations and where it falls short across the development and deployment life cycle. I use it to granularly analyze things like AI legislation, policies, AI vendor agreements, AI governance documents, and questions such as whether Claude Code is AGI. The AILCCP contains 37 principles, each with multiple requirements. Three principles apply most directly here: Wherewithal, Human-Centered, and Workforce Compatible. For each, I focus on the requirements most relevant to what the App Store test exposes, then apply it to Claude Code.

Wherewithal asks whether the capability matches what is being claimed. The enthusiasm around AI coding tools has generated claims that Claude Code can take a developer from idea to shipped app with minimal effort. That framing describes the front of the life cycle accurately and the back poorly, and developers who plan around it will discover the gap at exactly the point where it costs the most to close.

Human-Centered requires human-in-the-loop oversight at the pre-deployment review and deployment phases. Those are the phases where an iOS app is tested against Apple’s privacy guidelines, where data handling disclosures are drafted and verified, where security architecture is stress-tested, and where the submission package is assembled and submitted for review. Claude Code does not do those things independently. A developer who has moved quickly through scaffolding and code generation arrives at those phases with the tool’s momentum behind them and its limitations fully exposed.

Workforce Compatible asks whether an AI tool builds human capability or displaces it. A developer who uses Claude Code to generate iOS code throughout a project never learns iOS development. They learn to prompt. When the tool produces architecturally flawed code, which it does with some regularity, the developer has no independent basis for catching the error. They are dependent on the tool to identify problems that the tool created. That is not augmentation. It is a different kind of dependency, and it grows less visible the more the tool appears to be working.

Claude Code is powerful, without a doubt. But a phase-specific competence at a high level is not what “G” in AGI means.