The previous implementation collected all node values into a vector via a full inorder traversal before indexing into it, resulting in O(n) time and O(n) space regardless of where the kth smallest element sits in the tree.

The new implementation passes k by reference as a countdown and short-circuits as soon as it reaches zero. When the kth node is found, the return prevents any further right subtree traversal in that frame. For ancestors that were waiting on a left call, the if (k == 0) return guard intercepts them before they descend right, ensuring no unnecessary nodes are visited after the answer is found.

This brings time complexity down to O(k + h) and space complexity down to O(h), where h is the height of the tree. In the worst case (skewed tree) h = n, but for balanced trees this is a significant improvement over the previous approach.