We shipped an iOS and Android app backed by the same API. It went fine, but a few decisions mattered a lot.

What worked:

• One versioned API, and clients that tolerate unknown fields
• Server-driven feature flags instead of hardcoded release rules
• A shared contract file (OpenAPI) that both apps generate models from

What I'd change:

• We added a "v2" endpoint for one screen and it became three. Versioning per-field would have been better.

I'm learning Rust for a small CLI that helps with release notes. Coming from Swift and Kotlin, the compiler feels strict in a good way.