Shopify Moves Apps from RN Back to Native Using AI
Shopify's engineering blog announced that mobile applications will transition from React Native back to Swift and Kotlin. In 2020, the company had fully invested in cross-platform; Engineering Director Mustafa Ali wrote that large models have changed the premise of "high costs of dual-platform development" from that year, allowing one platform to serve as a reference and generating the other through coding proxies, making cross-platform development acceptable. The company stated that React Native itself is still excellent, but the change is in the cost curve.
The consumer app Shop has completed the migration: with the assistance of artificial intelligence, it took 12 weeks from proof of concept to native app launch. Initially, one engineer spent a week migrating as many RN interfaces to a SwiftUI prototype as possible, proving that functionality could be aligned before rolling out fully. The larger merchant-side Shopify app has over 300 screens, including desktop widgets, watch apps, Siri shortcuts, etc., and migration is ongoing, with plans to release within the year, followed by other applications.
The migration coincided with the planned investment in a new React Native architecture. The team separated business logic from the interface, allowing proxies to drive tests through a desktop command line, reducing some checks from minute-level to millisecond-level to match the speed of code changes by proxies. Evaluation criteria include how much work the proxies can ultimately handle independently. Engineers can also advance in unfamiliar languages, with standards, testing, and reviews used to synchronize dual-platform behavior.
The historical benefit of cross-platform development is writing once and reducing reliance on native manpower. After the proxy made it possible to "produce Android without knowing Kotlin," the repetitive labor saved by cross-platform frameworks was replaced by models, making native performance and system integration economically viable again. Shop has long been a leader in shopping categories, and the native rebuild is seen as a replicable template.
From a market mechanism perspective, this is a repricing of the development tool stack. The buyers are e-commerce application teams that need dual-platform functionality; the sellers are coding proxies that can read reference implementations and write the other side. Funding is shifting from cross-platform frameworks and shared manpower to model invocation and native toolchains. Beneficiaries include the Swift/Kotlin ecosystem and proxy providers; those under pressure are cross-platform framework consultancies that market "write once" as a selling point. The event was driven by the engineering blog's publication.
On a supplementary level, the 12 weeks refers to one product line of Shop, not all applications across the company. The merchant side's 300 screens have not yet been delivered, with risks in deep system integration rather than just the listing page.
Source: Public Information
ABAB AI Insight
The economic rationale for RN in 2020 was that no one can be proficient in two native systems simultaneously, hence the need for a shared layer. The rationale for 2026 is that proxies can read iOS and write Android, turning the shared layer into a constraint. Shopify does not deny the correctness of that year but denies that the premise still holds. Shop proved the migration speed in 12 weeks, while the merchant side's 300 screens are the real stress test. Changing the codebase to be proxy-friendly indicates that the true product is "applications operable by models," not cross-platform abstractions.
The capital path is "reference implementation - proxy generation - native release." Money flows from cross-platform middleware into tokens and native engineer reviews. The motivation is performance and system APIs, as well as eliminating bottlenecks for those who do not know Kotlin. The CEO previously warned about unreviewed proxy code, and the migration document emphasizes testing and standards, which is the same discipline: speed is crucial, and gates cannot be removed.
In contrast to Airbnb's departure from RN and the laboratory's experiments with model-based cross-language migration: both bought frameworks when human resources were scarce and returned to native platforms when models were sufficient. The industry phase is shifting from framework faith to proxy cost accounting. Control lies with those who possess runnable reference applications, as proxies need a reference object.
Structural changes belong to technological substitution. The mechanism is that cross-platform frameworks arbitrage human dual-stack costs; once coding proxies compress dual-stack costs to acceptable levels, the intermediary rents of frameworks disappear, and platform native APIs become the default again. The premium for writing once is replaced by generating once.
ABAB News · Cognitive Laws
Frameworks arbitrage languages that humans cannot master; once models learn the framework, they exit the scene.
What can be rebuilt in twelve weeks is an application with a reference implementation, not an application built from scratch.
The codebase must first be proxy-friendly for migration to outpace testers.