Mixed Integer Linear Programming (MILP) solvers are powerful tools for tackling complex optimization problems across various industries. But configuring these solvers, especially selecting the right cutting plane separators, can be incredibly challenging, even for experts. Traditional machine learning approaches often require extensive training data and custom solver interfaces. Now, a new approach is emerging that leverages the power of Large Language Models (LLMs) to configure cutting planes with minimal training data and computational overhead. Imagine configuring a solver based on a simple natural language description of the problem – that's the promise of this LLM-driven approach. By incorporating descriptions of the problem and the solver's available cutting planes, researchers are building prompts that guide LLMs to generate effective configurations. Since LLMs can be unpredictable, a clever ensembling technique clusters and aggregates the generated configurations to create a small portfolio of high-performing options. This innovative method is showing promising results, rivaling the performance of existing approaches but with a fraction of the computational effort. It's a significant leap towards making MILP solvers more accessible and efficient, especially in resource-constrained environments. This opens doors to using LLMs for configuring other aspects of solvers, like heuristic selection and parameter tuning. As LLMs become increasingly sophisticated, we can anticipate even more seamless integration of natural language descriptions and optimization workflows, ushering in a new era of automated algorithm configuration.