Shopify Leaves React Native for Swift and Kotlin, Crediting LLMs for a 12-Week Native Rewrite
The economic case for cross-platform frameworks rests on "don't write twice". When LLMs can translate a React Native feature into Swift and Kotlin with acceptable quality, that case weakens; Shopify's move shows the cost gap is closing. Teams relying on Shopify's open source libraries face reduced maintenance, and cross-platform frameworks generally must now answer why their abstraction layer beats a cheaper native build.
In 2020, Shopify bet on React Native to avoid writing the same features twice and to let non-mobile developers build apps. By 2026, after using LLMs since 2021, the cost of building and maintaining separate Swift and Kotlin codebases has dropped enough that the write-once advantage no longer outweighs native's cleaner platform access. Shopify rebuilt the Shop app from scratch in 12 weeks using AI, and plans to migrate all its mobile apps to native.
The migration is not a blind "let AI rewrite everything" exercise. Shopify built Helix, a system that breaks each page into sequential checkpoints, each requiring passing tests, visual review against the running app, two adversarial reviewers, and human sign-off. They also decoupled business logic from UI so agents can verify changes at millisecond speed through a CLI instead of a sluggish simulator. The open source libraries React Native Skia and Restyle are being wound down or archived, while FlashList will get critical fixes and a new maintainer.
The reversal is a cost decision, not a performance one; Shopify says React Native apps can be fast, but shared implementation no longer saves enough time when LLMs can generate and verify native code quickly.
The success of AI-driven native rewrites depends on having an existing cross-platform codebase to serve as reference; a greenfield project without that would not get the same speed.
Verification, not code generation, is the new bottleneck; Shopify's headless architecture and CLI tools show that making agents fast to validate is a prerequisite for working at scale.
The author's side story about Uni App reinforces a broader point: as AI reduces the cost of building native, cross-platform frameworks lose their default-first position, especially for small teams focused on cost.
AI solving cross-platform is indeed a good point. But cross-platform frameworks also tend to support hot updates, which is something native doesn't handle well.
Mm, actually each has its own merits; my title was a bit clickbait-y. But reading the full article carefully, everyone is still in the exploration phase right now.