Coding agents are quietly ending programming-language lock-in

Simon Willison's short post names a shift that has been creeping up on engineering teams. A language or framework choice used to be close to permanent, because rewriting a working codebase by hand was rarely worth it. With coding agents doing the bulk of a port, that calculation changes.

His example is concrete. A mid-sized company rewrote its paired iPhone and Android apps to React Native using coding agents, and what he highlights is not the rewrite itself but the attitude: they were comfortable doing it because they could move back to native later if React Native stopped fitting. The decision was reversible, which is the opposite of lock-in. He ties this to Mitchell Hashimoto's line that "programming languages used to be LOCK IN, and they're increasingly not so," set against the Bun project's move from Zig to Rust.

The post is brief and does not overclaim. It is one observation, well chosen, about a second-order effect of agents that does not get the attention the coding demos do.

Why it matters

If you pick a stack, you can weight current fit more heavily and worry less about being trapped: migration is becoming a project you can budget rather than a rewrite you avoid, which also weakens one of the oldest reasons to standardize early.