Shopify leaves React Native, and publishes the bill it had been paying
The Shop app rebuilt native in twelve weeks cut Android cold start from 4,433 ms to 2,233 ms and the Android build by 109 MB.

Shopify said on September 10 that it is moving its mobile apps off React Native and back to Swift and Kotlin, twenty months after publishing a post titled "Five years of React Native at Shopify" that called the framework a success and promised more investment. The Shop app has already shipped fully native, rebuilt from proof of concept to store release in twelve weeks by a core team of six engineers. The Shopify app, 300-plus screens with widgets, an Apple Watch app and Siri shortcuts, is underway.
The reason given is agents, and the wording is careful enough to quote exactly.
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's no longer the deciding factor it was in 2020.
Read that twice, because almost every write-up of this has flattened it into "AI killed React Native". Shopify is not saying native got cheaper. It is saying the duplication tax is now payable by machines, and so it stopped being the thing that decides.
But the sibling post is where the receipts are, and hardly anyone has opened it.
Shopify published the Shop app migration numbers the same day. Android cold start to first home-feed content fell from 4,433 ms to 2,233 ms. iOS went from 3,200 ms to 2,466 ms. The Android release build shrank by 109 MB, from 293 MB to 184 MB — a 37% cut. Session stability moved from a historical 99.5%-plus to 99.95%-plus, which the team describes as a tenfold reduction in crashing sessions. Android release build time fell roughly 75%.
None of that is an agent win. That is the bill React Native had been running up, itemised at last, and Shopify had been paying it on purpose for six years because the alternative cost more engineers.
Now set the two posts side by side. In January 2025 Shopify wrote that its React Native apps were "blazing fast (<500ms screen loads)" and "stable (>99.9% crash-free sessions)". Twenty months later the same company reports a React Native Android cold start of 4.4 seconds and historical session stability of 99.5%. Those are different measurements, and we are not accusing anyone of anything — screen loads are not cold starts, and crash-free sessions are not session stability. But if you are choosing a mobile stack on a vendor's published numbers, that is a useful lesson about which numbers get published.
The mechanism is worth your attention even if you never write Swift. Shopify built a system called Helix that refuses to let an imperfect attempt move forward: it reads the React Native screen, proposes small ordered checkpoints, and each one must pass tests, match the running app in a visual review, survive two adversarial code reviewers and get a human's approval before the next begins. They also decoupled business logic from UI so it runs headlessly on a desktop, reachable by a CLI, because agents could change code in seconds but needed several minutes to test it through a simulator. And for the Shop app they tied plan acceptance to a hash of the plan's contents, so editing a plan invalidates the approval it already had.
That last detail is the one to steal. Approval attached to a document, not to a promise.
Our read: this is a story about large apps rather than about React Native, and Meta is not who gets hurt. The pain lands on every team of eight that adopted React Native because they could not afford two codebases and still cannot afford two codebases, and now gets to maintain libraries whose corporate steward is leaving. Skia gets forked and renamed. Restyle is archived at the end of 2026. FlashList — which the post puts at roughly 2 million downloads a week, though the npm registry counted 1,082,601 in the week to September 9 — is looking for a long-term owner Shopify has not yet found.
What would we expect? At least one more company with a top-100 app-store title announcing the same reversal before the end of the first quarter of 2027, and react-native's own npm downloads (7.5 million in the same week) still growing while it happens. Frameworks do not die of defections at the top. They just stop being where the interesting apps are.
If a mid-sized team tries this and publishes a twelve-week number without six engineers and a bespoke agent harness behind it, we will want to see the diff.

