Proof of concept to published native Shop app: 12 weeks (Source: Shopify Engineering, back-to-native post (as cited by The Pragmatic Engineer, 2026-09-29))
Every architecture choice is a bet on a price
Shopify is moving its mobile apps from React Native back to Swift and Kotlin, the native languages for iPhone and Android. The company gives a plain reason. Coding agents changed what it costs to build the same app twice.
The lesson here is that a technology choice is a bet on a price. When the price moves, the choice can expire, even if it worked. Shopify's engineers say React Native "remains an excellent framework" and that their apps are fast. Their assumption about cost changed, so they reopened the decision.
Executives should notice how rarely organisations do this. Most teams defend a platform choice because it once succeeded.
What Shopify says changed
Shopify chose React Native in 2020 for three reasons. It let developers work across the whole stack. It let them build a feature once. And it cut the effort of keeping iOS and Android in step. The company says those benefits held for years.
By late 2025, Shopify's prototypes suggested the second and third reasons were weaker. The company rebuilt core parts of its biggest apps in Swift and Kotlin using large language models. It says the results surprised the team. Agents could use the React Native version as a reference and produce the native version.
Shopify does not claim the cost is gone. Native still means two codebases. Its argument is that agents now handle enough of the translation, testing and review work that the cost no longer decides the question.
How the rebuild works
Shopify says pointing an AI model at old code and asking for a copy does not work. It produced "a huge amount of unmaintainable code that can't be shipped."
Instead, Shopify built a system called Helix. A developer points it at one screen. Helix reads the React Native code and proposes small, ordered checkpoints. Each checkpoint must pass four gates before the next begins. It must prove its behaviour with tests. It must match the running app in a visual review. It must survive two adversarial code reviewers. And a human must approve it.
Shopify says feedback from each review is remembered, so the loop needs less supervision over time. The design assumes the first attempt is flawed. It blocks flawed work from moving forward. Companies weighing AI-assisted rewrites can ask whether their own process has gates like these.
The bottleneck is testing, not writing
Shopify reports that agents change code in seconds but need several minutes to test it on a phone simulator. The agent has to read the screen through screenshots or the accessibility tree. Shopify says this made iteration slow and manual.
Its fix is architectural. Business logic is separated from the screen so it can run without a phone. Agents drive it through a command-line tool and get feedback in milliseconds instead of minutes. Shopify says this lets agents work autonomously for hours.
This is the deeper shift. Shopify is redesigning its apps to be readable and testable by machines as well as people. Its stated aim is that humans and agents can both "understand, test, and change them quickly."
The cost lands on people outside Shopify
Shopify built and sponsored several open-source React Native libraries. It says one, FlashList, gets about 2 million downloads a week. Shopify will keep fixing critical compatibility issues and is talking to several companies about long-term stewardship. Another library, Restyle, will be archived after 2026, though anyone may fork it.
For any team that depends on such a library, the risk sits with the dependency, not the app. A company that maintains an open-source project can change strategy. Shopify has given notice and a timeline. Not every sponsor does.
Limits of the evidence
This is Shopify's account of its own work. The source gives no cost figures, so the size of the savings is unverified. It is also one company's decision, not a measured industry trend.
The Pragmatic Engineer notes that Airbnb adopted React Native in 2016 and returned to native two years later, citing performance. Shopify says performance is not its reason. Similar moves can have different causes.
Questions to put to your team
Which of our platform choices were made because building twice was too expensive? Has anyone re-priced that assumption since 2025?
Can an AI agent test our software quickly, or does every check wait on a slow manual step?
Which open-source libraries do we rely on that depend on a single company's sponsorship?
Shopify says success will be measured by product velocity, app quality and how much work agents complete on their own. Ask for measures like those before approving any rewrite.
Shopify's own summary is worth remembering: native is right now, and React Native was right in 2020. A decision can be correct for its time and still be replaced.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Shopify Engineering.
