The paper discusses the integration of Large Language Models (LLMs) in trading systems, highlighting their potential as decision-making agents in financial markets. However, it identifies significant challenges in reproducibility and evaluation protocols that need to be addressed for effective implementation.