A method reportedly lets language models explain why one chess move is better than another

A user says his brother developed a method that lets language models explain why one chess move is better than another. The model writes code to analyze the board, he says, and can do so because it gets access to 200,000 board states from real games alongside engine analysis.