AI agents power Shopify's React Native exit and Expo's automated bug fixes

Two agent projects, opposite directions

In the same month, two companies with deep React Native investments turned to AI agents and pointed them at opposite goals. According to a dev.to analysis, Shopify built internal agent tooling, called Helix, to migrate its apps off React Native onto native Swift and Kotlin. Expo, the company behind a widely used React Native toolchain, went the other way: it shipped an agent whose job is to repair React Native bugs automatically.

Despite the opposing directions, the post argues, both moves rest on a single assumption — that AI agents can make large-scale mobile engineering dramatically cheaper.

What Shopify's Helix reportedly does

As the dev.to post describes it, Helix is the internal tooling behind Shopify's move from React Native to native code. The company used it to rebuild its Shop app in 12 weeks and is now working through its main app, which the post says contains more than 300 screens.

Notably, the author writes, Helix does not attempt to generate a complete result from a single specification. It works in small checkpoints, and every checkpoint must clear a series of gates: behaviour tests, a UI review, two adversarial reviews carried out by other agents, and finally a human reviewer.

Expo's agent repairs bugs end to end

Expo's agent, per the post, starts when a label is added to a GitHub issue. From there it reproduces the bug on a cloud simulator, writes a fix, and opens a pull request that includes recordings of the app before and after the change.

The result is an automated pipeline from bug report to reviewable patch — an effort to keep React Native healthy rather than abandon it.

Testability, not frameworks

The author's central claim is that most readers will draw the wrong conclusion from Shopify's story, namely that native platforms have won. Helix's checkpoint system, the post argues, only functions when an application's logic and UI are separable enough to be tested independently. That is a testability problem rather than a React Native problem, and the author says most React Native codebases encountered over ten years of professional work share it.

The post also frames the industry's history in a specific way: earlier native-versus-cross-platform cycles were settled by the shape of teams, not by the quality of the frameworks. The current cycle, in the author's view, is being settled by which codebase an automated agent can navigate safely.

The practical advice follows directly from that framing. Rather than starting a migration immediately, the author suggests measuring how much of an app an agent can change without breaking something, and treats that number as more consequential than the choice of framework.

Why it matters

If the dev.to analysis holds up, the contrast between Shopify and Expo is not really an argument about React Native's merits. It is an early signal of a new selection criterion for mobile teams: how well an agent can work inside the codebase. Companies weighing native against cross-platform may increasingly ask not which framework their engineers prefer, but which one automated agents can modify safely under test gates.

The development also matters for tooling vendors. Expo's agent is an attempt to raise the share of a React Native app that is safe for automated changes, while Shopify's Helix reflects one company concluding that a native codebase gives agents better guardrails. Whichever approach spreads, the underlying wager is the same: that the economics of maintaining large mobile apps — a 12-week rebuild of Shop, bug fixes arriving as automated pull requests — are being reset by agents, and that testability, not framework loyalty, will decide who benefits.