Large Language Models (LLMs) are changing how we write code, and GitHub, the world’s largest host of source code, offers a fascinating glimpse into this transformation. A recent study explored how LLMs like ChatGPT and Copilot are actually being used by developers on real-world projects. Surprisingly, while LLMs are known to struggle with complex coding challenges in controlled tests, the study found that the LLM-generated code on GitHub is often relatively simple and contains fewer bugs than one might expect. The majority of this AI-generated code appears in smaller, lesser-known projects, primarily focused on Python and JavaScript, and is often used for data processing and user interface development. This reveals that while LLM use is growing, developers currently leverage them mostly for smaller, less complex tasks. Interestingly, even within larger, more popular projects, LLM code tends to be short and straightforward. Furthermore, AI-generated code often undergoes minimal changes after its creation. This suggests developers are finding the code quite effective for these specific use cases. The study’s findings also shed light on how developers annotate AI-generated code. While some provide helpful context like the prompt used or whether the code has been tested, many simply mark the code as ‘generated by ChatGPT/Copilot,’ leaving future maintainers with little extra information. This lack of standardized commenting practices raises questions about how best to document and manage AI-generated code in collaborative environments. As LLM capabilities evolve, the need for best practices in integrating and documenting AI-generated code becomes increasingly important.