A recent paper published on arXiv introduces a novel approach to machine unlearning by formulating it as a constrained optimization problem (F1). The authors propose feasible model updates that improve the unlearning process while maintaining the model's original performance (F2). This method offers a promising direction for enhancing machine unlearning capabilities, as these updates are designed to align with gradient information, thereby supporting more effective forgetting of specific data (A1). The approach balances the need to remove learned information without degrading overall model quality, addressing a key challenge in the field. This contribution aligns with ongoing research efforts that explore optimization-based techniques for model refinement and data privacy. By framing unlearning within a constrained optimization framework, the study provides a structured methodology that could inform future developments in AI model management. These findings complement related work recently discussed in the arXiv community, emphasizing optimization strategies in machine learning.