Shopify drops React Native for Swift and Kotlin, agents in hand

Six years, two U-turns

In 2020, Shopify moved its mobile apps to React Native for three stated reasons: stop writing the same feature twice, open mobile development to people who hadn’t come from mobile, and stop chasing parity between iOS and Android. In January 2025, that same team was still writing that the framework’s future was bright and that the investment would continue. On 10 September 2026, the company’s engineering blog announced the opposite route: back to Swift and Kotlin.

The post takes nothing back. It calls the 2020 bet extremely successful and describes React Native as an excellent framework. What moved, it says, is a starting assumption — not the framework.

The argument fits in one word: agents

Shopify says it has been using LLMs to build software since 2021, a year before ChatGPT — a detail you can practically hear being typed with a smirk. First to implement features, hunt bugs and review code. Then, in late 2025, to poke at a far more awkward question: does building twice still mean working twice?

The team says it rebuilt several core parts of its biggest apps in Swift and Kotlin with the help of agents, and was surprised by the result. Three concrete observations come out of it: an agent implements a feature on Android using the iOS version as its reference, and the other way round; developers contribute effectively outside the stack they came from; and the cost of holding parity drops sharply, through specs, tests and review points shared across platforms.

The post doesn’t claim that cost has vanished. Two platforms are still two platforms. It’s simply, it says, no longer the factor that decides.

The ceiling isn’t the one you’re picturing

If you came for the usual “it’s slow” verdict, you’ll leave empty-handed: Shopify states in black and white that React Native apps can be fast, and that its own are. The technical grievance sits elsewhere. The framework demanded continuous investment — performance tuning, improving fundamental pieces of React Native itself, keeping up with updates and third-party dependencies. Trade-offs the team found acceptable for as long as code sharing paid for them.

That’s the real lesson, and it flatters native less than it looks. Cross-platform didn’t hit a technical ceiling; it lost its side of the bargain. Native keeps its advantages — closer to platform capabilities, first-party tooling, fewer framework layers and dependencies between your code and the system — while its main drawback quietly gets absorbed by agents. Shopify’s decision says nothing about React Native’s ceiling. It says the price of the ticket changed.

The ecosystem picks up the tab

That leaves the libraries Shopify published, several of which became defaults. React Native Skia is sponsored through the end of 2026; William Candillon will keep working on it beyond that, will fork the repo under a new name, and the original will be archived once the transition is done. FlashList — roughly two million downloads a week, and the default way to render performant lists — will still get its critical breaking changes fixed while Shopify looks for a long-term steward among several companies it has approached. Restyle, smaller audience, is archived outright. Translation, if you depend on any of these: your critical building block just became a sponsorship appeal, which still beats waking up to an archive commit nobody warned you about.

And notice the mechanism on the way past, which is more interesting than the choice itself: a team reopened a decision that was working, because an assumption had moved. In this industry, that’s rarer than a framework migration.