This project is an inquiry into the nature of musical understanding in artificial intelligence: can a Large Language Model (LLM) go beyond statistical imitation to internalize the rigorous logic of musical structure? To find out, I will be exploring its ability to use a musical technique called counterpoint that flourished in Europe between approximately 1500 to 1750. If we view composition as constructing a building, counterpoint functions as the structural engineering that keeps it standing. Analogously, while modern generative audio tools can convincingly “paint a wall” to resemble a beautiful brick house, they often lack the underlying logic required to construct the building itself. This project seeks to discover if LLMs can be guided to lay the same groundwork that composers like J.S. Bach used to build their greatest works by mastering the strict constraints outlined in Johann Joseph Fux’s 1725 treatise Gradus ad Parnassum.