A no-op platform implementation is not a platform limit

In a Kotlin Multiplatform project every shared declaration has one implementation per target. When one of them looks like this —

// adapted
actual fun Modifier.frostedSurface(source: SurfaceSource): Modifier = this

— the honest reading is "this target has no implementation yet". The tempting reading is "this target cannot do it". Those are wildly different conclusions and the file cannot tell them apart: a stub written because the API is genuinely absent and a stub written because someone ran out of afternoon are byte-for-byte identical.

The cost of guessing wrong is asymmetric. Saying "the platform can't" ends the conversation and the feature gets dropped; saying "nobody has written it" costs one lookup.

The recorded case