Imagine a factory where the software controlling the production line isn't painstakingly coded by human engineers but is instead automatically generated by artificial intelligence. This futuristic scenario is closer than you might think. Researchers are exploring how Large Language Models (LLMs), the technology behind chatbots like ChatGPT, can be used to automate the complex task of creating scheduling software for factories. Traditionally, creating scheduling software for dynamic job shop environments, where jobs arrive unpredictably and machines can break down, requires expert knowledge and extensive manual tuning. Existing solutions like Genetic Programming (GP) and Gene Expression Programming (GEP) help, but they have limitations in how well they adapt to new situations. Even cutting-edge Deep Reinforcement Learning (DRL) approaches can fall short. The new approach uses a method called "population self-evolution" (SeEvo). SeEvo works by having the LLM generate multiple versions of scheduling rules (like small programs). These rules are tested in a simulated factory environment, and the results are fed back to the LLM. The LLM then uses this feedback to refine the rules, evolving them over multiple generations to become increasingly efficient. Like a digital Darwinism, the best-performing rules survive and reproduce, combining their strengths to create even better offspring. This cycle continues until a highly optimized scheduling solution emerges. In experiments, this LLM-powered approach outperformed existing techniques like GP, GEP, and DRL, especially when tested on new factory scenarios. This is a significant leap forward, suggesting that LLMs could dramatically simplify the development of manufacturing software. However, the technology isn’t without its limitations. Current research relies on APIs and doesn't directly fine-tune LLMs for specific factory knowledge. Furthermore, using a single scheduling rule in dynamic scenarios can restrict the effectiveness. Future research will focus on improving how these AI systems learn about specific factories and on combining multiple scheduling rules to handle real-world complexity. If successful, this could lead to a revolution in manufacturing, enabling factories to adapt to change and optimize production with unprecedented speed and efficiency.