Shopify left React Native. Should you?
Shopify is rebuilding its apps in Swift and Kotlin because agents made building twice cheap. What it said, where Theo is right, what the 2026 data says.
· 7 min read
In 2020 Shopify went all in on React Native and became its best reference customer. On 10 September 2026 it announced the way back to Swift and Kotlin. Theo Browne spent an hour on it and half the internet declared React Native dead. Shopify's reasons are real, most of the hot takes are not, and the numbers say React Native is fine. Sources below, with Theo's opinions labelled as Theo's.
What Shopify actually said
The announcement is Native is now the future of mobile at Shopify by Mustafa Ali, Shopify's head of mobile, the same man who wrote Five years of React Native at Shopify in January 2025 with the line "the future of React Native is bright". He does not pretend otherwise: React Native "remains an excellent framework".
One assumption changed. In 2020, building a feature twice cost twice. By late 2025 their coding agents could implement an Android screen from the iOS version as a reference, and the reverse. In Ali's words: "Native still means building and maintaining software on two platforms, that cost has not disappeared. What changed is that agents can now do enough of the implementation, translation, testing, and review work" that it stopped being the deciding factor. That is the whole argument. Not performance, not bugs, not "React Native failed".
The proof point is the Shop app, written up in Migrating Shop app from React Native to native: a one-week proof of concept by one engineer, six engineers on the foundations, a rebuilt app in both stores twelve weeks later. The published results:
| Metric (Shop app) | React Native | Native |
|---|---|---|
| iOS release size | 67 MB | 68 MB |
| Android release size | 293 MB | 184 MB |
| Cold start to home feed, iOS | 3,200 ms | 2,466 ms |
| Cold start to home feed, Android | 4,433 ms | 2,233 ms |
| Crash-free sessions | 99.5%+ | 99.95%+ |
The Shopify app itself, 300+ screens plus widgets and a Watch app, is due later in 2026. For it they built Helix: it splits a screen into small checkpoints and refuses to advance until each passes behaviour tests, a visual comparison with the running app, two context-isolated reviewer agents and a human. The gated loop from my note on multi-agent workflows, applied to a rewrite.
The open-source part touches you if you ship React Native. Skia is sponsored until the end of 2026, then William Candillon carries on under a new name; as of October 2026 github.com/Shopify/react-native-skia already redirects to his fork and v3 is on npm as react-native-skia. FlashList, at roughly 1.5 million npm downloads a week in mid-October 2026, gets critical fixes while Shopify finds a steward. Restyle is archived at the end of the year.
What Theo adds, and where it holds up
Theo's video is Did AI Kill React Native? on t3.gg. Everything here is his view, not Shopify's and not a measured fact, unless I say so.
Shopify never mentions over-the-air updates. His strongest point. React Native's JavaScript bundle can be swapped without a store release, so a hotfix lands in an hour instead of waiting on review and on users with 54 pending updates (his phone, on stream). It holds: neither Shopify post mentions OTA, and native loses it.
Shopify was not on Expo. Jamon Holmgren of Infinite Red said it in Theo's chat; Theo's conclusion is "if you're using React Native without Expo, you're probably rebuilding Expo." Shopify's posts do not mention Expo, so: plausible, unverified.
"They shaved a whopping 1 MB." Half right. Theo reads the iOS row (67 to 68 MB, native is bigger) and skips Android: 293 to 184 MB, 37% smaller. His point that the React Native runtime is not what bloats apps is fair; the Android number says there was more than a runtime in that bundle.
The comparison is old architecture versus native. This checks out. The Shop post says adopting the New Architecture "would have required us to revisit native module integrations, rendering, and the boundaries between shared and platform-specific code", so they skipped it. The startup numbers above compare a pre-New-Architecture app with a greenfield rewrite, and every rewrite gets second-system wins. The New Architecture has been the default since 0.76 (October 2024) and the only one since 0.82 (October 2025); an app started today is not the app Shopify measured.
"The average React Native dev is worse than the average native dev." His opinion, stated as fact, no data, not one I am repeating. An easier tool attracting more beginners is a hiring problem, not a framework problem.
"Meta is disbanding the React and React Native teams." No public statement from Meta says so. Verifiable: Meta cut roughly 10% of staff in May 2026 (Reuters, via The Pragmatic Engineer), and React Native still ships every two months, 0.88 on 19 October 2026. An insider's worry, not a fact.
The Tobi story. Theo relays an anonymous account that Shopify's CEO forces calls like this over a weekend, then reads Tobi Lütke's reply: the example cited was an internal tool in "an architecture nightmare of their own doing", and he has explained such decisions on an internal podcast for ten years. I found only fragments of that reply on X, so treat the quotes as Theo's reading. His verdict: "this one flipped me". A CEO who reviews his own decisions on tape is not the villain here.
React Native popularity in 2026, with dates
Every number is from the primary source, dated, so you can check it.
- npm registry:
react-nativewas downloaded 55.3 million times in September 2026, against 18.2 million in October 2025;expodid 38.2 million in September 2026 (npm download API, queried 23 October 2026). Registry counts include CI and mirrors: a trend, not a headcount. - State of React 2025: of 2,341 people who answered the mobile question, 1,064 had used React Native (45%) and 971 Expo (41%).
- State of React Native 2025 (Software Mansion, unofficial): New Architecture adoption "at 80%".
- Appfigures, 19 December 2025: of the top 10,000 non-game iOS apps by downloads, 1,350 ship React Native, 1,184 Flutter, 218 Kotlin Multiplatform; React Native apps took 47% of those three groups' downloads. "Ship" means the framework is in the bundle.
- Stack Overflow Developer Survey: the last edition listing both was 2024, React Native 8.4% and Flutter 9.4% of all respondents (n = 45,841). The 2026 edition lists neither, so there is no newer like-for-like figure.
Nothing there looks like a framework in decline. It looks like a framework whose loudest advocate left, which is a different thing.
What I'd actually do
If you write the code. Shopify's argument is about Shopify: six engineers on one rebuild, plus a tooling team building Helix, a debugging bridge and a headless CLI before the migration could start. Copy the conclusion only with the preconditions. What you can copy today is cheaper: use Expo so OTA and dev builds are a config file rather than a project, move to the react-native-skia package before the old repo is archived, watch who ends up stewarding FlashList. And learn one native platform properly; the person who can read a Swift stack trace is the one who ships the fix.
If you commission the app. Ask the agency three questions. Who keeps two codebases in sync after launch, and who pays for that in month seven. How does a critical bug reach users: two store reviews, or an over-the-air push. What happens when a library changes hands, as Skia and FlashList just did. A React Native app with a small team and a boring stack is still the cheapest way to be on both stores with one product; a native pair is right when platform features, background work or raw startup time are the product. The agents made the implementation cheap; they did not make the coordination free.
Shopify made a big bet in 2020 and another one now, and was honest about the reasoning both times. That is the part worth copying. Choosing a mobile stack and want a second opinion before the contract is signed? Write to me.