Shopify's Native Rebuild Proves a Narrower Claim Than It Sounds

Shopify's decision to rebuild its mobile apps in native Swift and Kotlin, six years after standardizing on React Native, is being read as proof that AI coding agents have crossed a real threshold: that maintaining two codebases instead of one is no longer the tax it used to be. The evidence Shopify itself published supports a narrower claim than that.

The evidence Shopify published

In a Sept. 10 engineering post, Shopify said one engineer spent a week on a proof-of-concept iOS migration of its consumer Shop app, before a core team of six engineers took it to a published App Store release in 12 weeks total, with feature teams joining midway to check specific areas. A companion post on the Shop app migration lists the results: iOS startup time fell from 3,200 milliseconds to 2,466, a 23 percent drop, and Android startup time fell from 4,433 milliseconds to 2,233, a 50 percent drop. Session stability rose from 99.5 percent to 99.95 percent, which Shopify's engineers call a tenfold reduction in crashing sessions. Android app size fell 37 percent, to 184 megabytes from 293; iOS grew slightly, by 1 megabyte, to 68.

Metric	React Native	Native	Change
iOS startup time	3,200 ms	2,466 ms	-23%
Android startup time	4,433 ms	2,233 ms	-50%
Session stability	99.5%+	99.95%+	~10x fewer crashes
Android app size	293 MB	184 MB	-37%

Mustafa Ali, an engineering director at Shopify, told DigitalToday that "LLMs changed one of the core assumptions behind the 2020 decision," adding that AI agents began handling conversion between Android and iOS code well enough, in the second half of 2025, that engineers could work on platforms they did not specialize in.

Why the 2020 logic held for six years

Shopify's own post is specific about what changed and what did not. In 2020, the company picked React Native for three reasons it gave at the time: to stop building the same features twice, to let developers move across the stack, and to cut the time spent chasing feature parity between iOS and Android. None of those reasons disappeared. Shopify's post says plainly: "Native still means building and maintaining software on two platforms, that cost has not disappeared." What changed is a variable, not the equation: agents now absorb enough of the implementation, translation, testing and review work that the duplicate-codebase cost stopped being the deciding factor it was six years ago.

That distinction matters, because it means Shopify's move is not evidence that native development itself got cheaper. It is evidence that the specific cost React Native was hired to solve, in a company able to build an internal system, called Helix, to run checkpoint-based agent review and a CLI that lets agents iterate on business logic in milliseconds, got cheaper for Shopify.

What the switch costs the people who stayed

Shopify is not walking away from React Native cleanly. The company maintains three widely used libraries built on top of it. FlashList, a list-rendering library with roughly 2 million downloads a week, will keep getting Shopify's attention while the company looks for a long-term steward. React Native Skia, created by developer William Candillon, will keep Shopify's sponsorship through the end of 2026, after which Candillon plans to fork and republish it under a new name. Restyle, the smallest of the three by user base, will be archived once 2026 ends. Shopify built for itself; the React Native ecosystem inherits the maintenance gap.

What would falsify the broader claim

The Shop app is Shopify's consumer-facing storefront app, a comparatively contained surface. The company's main Shopify app, used by merchants, is the harder case: more than 300 screens, plus home-screen and lock-screen widgets, an Apple Watch app, watch complications and Siri Shortcuts, according to Shopify's own migration post. If that migration takes meaningfully longer than 12 weeks, or needs a team much larger than six engineers relative to its size, the Shop app result looks like a favorable data point rather than a threshold every company can now expect to clear. If it lands on a similar timeline, the claim gets much harder to dismiss as Shopify-specific.

The comparison worth watching is not Shopify against its own 2020 decision. It is Shopify's 12-week, six-engineer rebuild of a single consumer app against whatever GitHub's own numbers on agent-driven engineering cost turn out to say about teams without an internal system as elaborate as Helix, and whether multi-model orchestration efforts like GitHub's HydraFusion can get smaller teams even part of the way to what Shopify built for itself.

What Shopify is not claiming

Nothing in either Shopify post argues that React Native itself is deficient, and Shopify says its React Native apps were fast and stable before the rebuild began. The argument is narrower and more specific to Shopify's own position: a company large enough to build Helix, staff a six-engineer migration team, and sponsor two open-source libraries through a wind-down period found that the arithmetic behind a six-year-old platform decision had changed. A smaller team without Shopify's tooling budget, or without an app as contained as Shop, is the group this piece cannot yet make a claim about, because Shopify has not published numbers for one.