Using ChatGPT for modular design in high-load systems presents significant challenges, primarily due to its limited context window, which hinders comprehensive understanding of complex module interdependencies and system-wide implications. It often struggles with effectively prioritizing scalability and performance requirements, potentially leading to design suggestions that overlook critical aspects like concurrency, resource optimization, or latency in high-demand environments. A notable drawback is the lack of real-time system state awareness, preventing ChatGPT from dynamically adapting modular designs based on live operational metrics or predicting bottlenecks under actual load conditions. Furthermore, ensuring security, reliability, and fault tolerance in AI-generated modular components is highly problematic, as ChatGPT lacks the inherent capability for formal verification or deep domain-specific engineering expertise required for robust high-load systems. The potential for generating suboptimal or biased architectural patterns, coupled with the difficulties in debugging and maintaining complex AI-derived modules, introduces considerable operational risks and increased long-term costs. More details: https://www.ucg.ac.me/include/promjena_pisma.php?url=https%3A%2F%2Fabcname.com.ua/