@menegatti in the following example

x?.let { y ->
   doSomeNetworkCall(
       { 
           response -> x?.doSomething() // x could be null here, but y couldn't
       }
   )
}

The outermost lambda is inlined, because

let

is inline function, but the innermost is not, as

doSomeNetworkCall

is not inline (as I suppose). So the innermost lambda is represented as an anonymous class instance, which holds references captured in the closure. If

x

is a nullable property of some object, the closure captures instance of that object, so you could get it's

x

property value later, and the value of that property could become

null

at that point. And

y

is a local non-nullable val, and after it's captured in closure, you can still be sure that it remains not null in that closure.