doc: update for next release

src/node.rs

@@ -27,8 +27,6 @@ type ExecuteFuture =
 ///
 /// This is the ONLY place where type erasure occurs - tasks are created with full type
 /// information and only erased at this trait boundary.
-///
-/// This is an implementation detail and not part of the public API.
 pub(crate) trait ExecutableNode: Send {
     /// Execute the task with type-erased inputs and output
     ///
@@ -43,14 +41,14 @@ pub(crate) trait ExecutableNode: Send {
 /// Fully-typed node storage for a single task.
 ///
 /// Stores a task of type T. Implements ExecutableNode for type erasure,
-/// allowing heterogeneous tasks to be stored in a single Vec.
+/// allowing heterogeneous tasks to be stored together.
 ///
 /// # Ownership Model
 ///
 /// The task is owned directly (no Mutex needed) and consumed during execution.
 /// Each task executes exactly once, taking ownership and producing an output.
 ///
-/// The output is also returned from execute_with_deps for central storage.
+/// The output is also returned from execute_with_deps to be passed to dependents and returned as an output
 pub(crate) struct TypedNode<Input, T>
 where
     Input: Send + Sync + 'static,
@@ -84,14 +82,11 @@ where
     ) -> ExecuteFuture {
         Box::pin(async move {
             // Use the task's run method which has inline extraction logic
-            // generated by the #[task] macro. This allows ANY type to work without
-            // requiring ExtractInput trait implementations.
+            // generated by the #[task] macro.
             let output = self.task.run(TaskInput::new(dependencies.iter())).await;
 
             // Arc Wrapping Strategy:
             // Wrap output in Arc ONCE for efficient fan-out sharing.
-            // For 1:N dependencies, this enables O(1) sharing (Arc clone = atomic increment)
-            // instead of O(N) cloning (N deep copies of output data).
             // Downstream tasks receive &T after extraction from Arc.
             let arc_output = Arc::new(output);