Debates about which AI coding model writes the best React Native code usually rely on anecdotes. A single good or bad experience often shapes strong opinions, but those claims are rarely reproducible. React Native Evals was created to change that by introducing a structured, evidence-based way to measure how well AI models handle real React Native development tasks.

In this live stream, Callstack engineers Kewin Wereszczyński, Artur Morys‑Magiera, Lech Kalinowski, and Piotr Miłkowski will walk through the ideas behind the benchmark and the work that went into building it. The discussion will cover how the evals dataset works, the generation and judging pipeline built with TypeScript and Bun, and why reproducibility matters when evaluating AI coding models.

The team will also explore what the early results tell us about current models and where the benchmark is heading next. Expect insights into categories like animations, async state management, and navigation, along with a broader conversation about AI tooling in the React Native ecosystem and the future direction of developer workflows.

Join us on March 12 at 17:00 CET for a technical deep dive into React Native Evals and a wider discussion about AI in development, including topics from the This Week in React newsletter.