Shopify is moving from React Native back to Swift and Kotlin
summary
Shopify is migrating its mobile apps from React Native back to Swift and Kotlin because AI coding agents have drastically reduced the cost of maintaining two separate codebases. While cross-platform frameworks were essential in 2020 to avoid duplicating effort, modern LLMs can now handle implementation, translation, and testing across platforms with high efficiency. This shift allows Shopify to leverage first-party platform capabilities and tooling without the overhead of a middle-layer framework. The company is using a greenfield approach powered by custom AI tooling to rebuild its apps faster and with higher quality.
key points
- AI agents now enable developers to implement features on one platform using the other as a reference, effectively removing the 'double work' penalty of native development. This allows teams to maintain parity through shared specifications and automated reviews rather than a shared codebase.
- To prevent low-quality AI-generated code, Shopify developed Helix, a system that breaks migrations into small, verifiable checkpoints that must pass tests and human reviews before proceeding. This ensures that the rebuilt apps are maintainable and meet high quality standards.
- Shopify is redesigning its app architecture to decouple business logic from the UI, allowing AI agents to test and iterate on logic via a CLI in milliseconds. This bypasses the slow process of using mobile simulators, which previously acted as a bottleneck for agentic development.
who should read it
Mobile engineers and technical leads interested in the impact of AI on software architecture and the trade-offs between native and cross-platform development.
discussion mixed
consensus
Commenters are divided between 'vibe coders' who believe AI can generate entire native apps that only require functional testing, and professional engineers who argue this approach ignores critical maintenance, security, and edge-case risks. While some successfully use AI to port apps to languages they don't know, critics maintain that without domain expertise, developers cannot verify the correctness or security of the generated code. Ultimately, a middle ground emerges suggesting that the most effective workflow combines human expertise with AI to increase both speed and quality.
top insight
The debate highlights a fundamental shift in the definition of software engineering, where some argue that AI replaces the need for deep language knowledge if the output is functionally verified. However, the counter-insight is that 'functional' does not equal 'correct,' as silent failures, race conditions, and security vulnerabilities often bypass manual QA and only a professional review can prevent long-term technical debt.
Generated by an LLM from the article and thread. Check the sources before relying on it.