Shopify Abandons React Native, Rebuilds Shop App in 12 Weeks Using AI Agents
After five years with React Native, Shopify is switching to native mobile development and leveraging AI agents to accelerate the transition. The company rebuilt its consumer Shop app from proof of concept to production in just 12 weeks.
Shopify has decided to move away from React Native and return to native mobile applications, relying on AI agents to power the transition and fundamentally reshape how it thinks about building codebases for machine assistance.
Back in 2020, Shopify made a strategic choice that gained traction throughout the developer community: use React Native to write mobile code once rather than maintain separate implementations in Swift and Kotlin. This week, the company announced it is reversing course.
The shift represents a dramatic pivot. Shopify is abandoning the cross-platform approach in favor of native development, with AI agents handling much of the conversion work. The consumer-facing Shop application transitioned from initial concept to a fully native production version in just 12 weeks. The merchant application, a significantly more complex undertaking with over 300 screens and extensive iOS platform dependencies, comes next.
Notably, Shopify views its React Native investment not as a misstep but as a framework that served the company's needs effectively for years. The shift caught many by surprise, given that as recently as January 2025, Shopify was still publicly discussing its continued commitment to React Native.
Agents replaced cross-platform tradeoffs
The calculus changed dramatically as AI coding agents improved substantially by late 2025. Shopify discovered that agents could examine how a feature functioned on iOS and generate the corresponding Android implementation, or perform the reverse operation. This capability allowed engineers to work effectively on platforms outside their expertise, since agents could manage much of the platform-specific complexity.
LLMs changed one of the core assumptions behind our 2020 decision
Mustafa Ali, Shopify's director of engineering
The Rewrite Economics Keep Changing
Shopify is far from alone in deploying agents for this type of work. Jarred Sumner, who created Bun and currently serves on Anthropic's technical team, deployed 64 parallel Claude Fable 5 instances to convert the JavaScript runtime from Zig to Rust—approximately one million lines of code completed in 11 days with an estimated API expense of $165,000. The existing test suite passed across all six supported platforms following the migration.
A year ago, that would have taken a small team multiple quarters but today it's an 11-day sprint supervised by one person.
These productivity improvements are not uniformly distributed across organizations. As recently documented, AI agents can generate additional work elsewhere within engineering teams even while accelerating certain development activities. However, they are reshaping the economics of decisions that were previously difficult and costly to reverse.
In May, developer Simon Willison described encountering an engineer whose organization had employed coding agents to merge its separate iPhone and Android applications into a single React Native codebase. When Willison questioned why consolidation made sense given that agents were simplifying separate codebase maintenance, the engineer expressed no concern about future constraints. React Native was meeting the company's requirements, and if circumstances shifted, returning to native development remained feasible. Shopify is now executing this identical strategy, though with a substantially larger application and greater stakes involved.
Zero-shot prompting produced junk
Shopify's initial approach involved providing an LLM with the existing React Native codebase and requesting a native rewrite. The outcome was disappointing. Ali characterized the generated code as slop, and introducing additional process steps failed to resolve the fundamental issue.
Even when engineers directed the model to generate specifications and task documentation before writing any code, the output still contained excessive amounts of code that development teams would find unmaintainable.
Helix breaks migrations into pieces
Shopify introduced Helix, an internal platform designed to coordinate agents performing the rewrite. Rather than attempting to reconstruct the entire application simultaneously, Helix processes the migration on a screen-by-screen basis, subdividing each screen into manageable units. Shopify continuously compares the output against the original application, while dedicated agents identify potential issues in the generated code before human reviewers approve it. The system preserves review feedback, allowing insights from earlier problem detection to shape subsequent work.
The approach appears connected to an earlier incident when Shopify's CEO publicly warned of restricting Claude Code after engineers deployed agent-generated code with insufficient human oversight. Helix adopts a similarly conservative stance, treating agent output as requiring validation before it enters production systems.
Codebases built for agents
The migration revealed an additional constraint. Agents can modify code in seconds, but validating changes through a mobile simulator requires approximately twice that duration.
Shopify addressed this by decoupling business logic from user interface components and enabling agents to interact with the application via a command-line interface running on a desktop machine. Verification procedures that previously consumed minutes now complete in milliseconds, while the same interface can launch a simulator when necessary.
Most existing codebases were not designed with AI agents in mind. Shopify is intentionally architecting its systems around agent capabilities and plans to evaluate the native applications partly based on how much work agents can eventually manage independently.
Some checks that once took minutes now happen in milliseconds, while the same interface can control a simulator when needed.
Source: The New Stack