The recent developments in the research area of tree-based algorithms and treewidth approximation have shown significant advancements in both theoretical efficiency and practical performance. Researchers are focusing on optimizing tree structures to enhance their accuracy and computational efficiency, challenging traditional assumptions about the superiority of ensemble methods over single trees. Notably, there is a shift towards reformulating tree training as a differentiable optimization problem, which allows for gradient-based optimization techniques to be applied, significantly improving the accuracy of single trees. Additionally, advancements in root recovery algorithms for random trees have led to optimal algorithms with sharp bounds on the size of the output set, addressing long-standing questions in the field. These innovations collectively push the boundaries of what can be achieved with tree-based models, making them more competitive and versatile in various applications.