Shopify moves mobile apps back to native Swift and Kotlin as AI agents cut rewrite costs

Shopify reverses its mobile strategy

In 2020, Shopify declared React Native the future of its mobile development. A January 2025 retrospective from the same team said the framework had delivered and that Shopify planned to keep investing in it. Then, on September 10, 2026, the team published a post with the opposite conclusion: every Shopify mobile app is moving back to native Swift and Kotlin, according to a dev.to write-up of the announcement.

The Shop app has already completed the switch. Per the dev.to article, six engineers took it from proof of concept to a fully native app published in the stores in 12 weeks. The main Shopify app is next, a far larger job covering more than 300 screens plus home and lockscreen widgets, an Apple Watch app and Siri Shortcuts. The rest of the fleet follows.

React Native did not fail

The dev.to piece argues against the obvious headline that React Native simply lost. The framework worked as advertised: the 2025 retrospective cited P75 screen load times under 500 milliseconds in the Shopify app, crash-free session rates above 99.9 percent, and feature parity the team no longer treated as a problem. Their stated view was that native does not guarantee speed and React Native does not guarantee slowness.

What changed is the price of the alternative. Shopify originally adopted React Native to stop building every feature twice, to let engineers work across platforms, and to spend less effort chasing iOS and Android parity. Those were economic arguments, and coding agents weakened them.

How agents changed the economics

Shopify has used LLMs since 2021, but by late 2025, the dev.to article says, the team concluded that agents no longer merely sped up writing code — they called into question whether supporting two platforms still meant doing twice the work. Three shifts stood out:

• Platform translation got cheap. In prototypes, agents implemented a feature on Android using the iOS version as the reference, and the reverse. One engineer spent a week of agent-assisted prototyping moving most of the Shop app to SwiftUI from the React Native codebase. The result was not production-ready, but it proved the point.
• Cross-stack ramp-up got cheap. Agents let specialists contribute outside their primary platform, so a strong iOS engineer can meaningfully drive a Kotlin migration.
• Parity maintenance got cheap. Shared specs, shared tests and review checkpoints let agents carry much of the burden a shared codebase used to absorb.

The dev.to post stresses that Shopify is not claiming native is free: two platforms still mean two implementations. That cost simply stopped being the deciding factor. Meanwhile, the costs of React Native — heavy investment in performance work, framework fundamentals, and keeping up with updates and dependencies — remained, while native offers direct access to platform capabilities and first-party tooling with fewer layers in between.

A greenfield rebuild, gated by Helix

Although millions of merchants rely on these apps, Shopify chose a full rebuild rather than the incremental migration it used in 2020. The dev.to article gives three reasons: agents can build native features using the React Native version as a working reference, a clean slate avoids carrying old constraints forward, and prototypes showed a rebuild was substantially faster than pre-agent estimates. During the Shop migration the priorities were preserving feature behavior and analytics events, keeping users signed in, and keeping push notifications working, so the change feels like a routine update.

Shopify first tried the naive approach, pointing an LLM at the codebase for a one-shot port, and judged the result unmaintainable and unshippable. It then built a system called Helix, which decomposes a screen's migration into small ordered checkpoints. Each checkpoint must pass behavior tests, match the running app in visual review, survive two adversarial AI reviewers, and get human approval before the next slice starts. Review feedback is remembered, so the loop grows more autonomous as the migration progresses.

The bottleneck was feedback speed, not model quality

Perhaps the most transferable detail in the dev.to summary concerns what actually limited the agents. The problem was not the models but verification: agents edited code in seconds yet needed minutes to test changes through simulators, which depended on brittle accessibility-tree and screenshot parsing. Shopify's fix was architectural. Business logic was decoupled from the UI and runs headlessly on desktop, and a CLI lets agents inspect app state, navigate and perform actions in milliseconds without a simulator. When real UI interaction is needed, the CLI drives simulators in remote mode using commands rather than layout parsing. The outcome is agents able to work autonomously for hours.

On open source, the dev.to post notes Shopify is not abandoning the ecosystem abruptly and will sponsor React Native Skia through the end of 2026.

Why it matters

The founding economic case for cross-platform frameworks was avoiding the cost of building everything twice. If agents can translate between Swift and Kotlin cheaply, a shared codebase stops being a saving and becomes a constraint. Shopify, one of the most prominent React Native adopters, is now the clearest public example of AI coding tools changing a fundamental architecture decision rather than just developer productivity. The engineering lessons generalize too: gated increments with tests and adversarial review beat one-shot generation, and fast, agent-drivable verification loops matter more than smarter models.