Microsoft's AutoGen has entered maintenance mode as its team merged with Semantic Kernel to form the Microsoft Agent Framework. While no new features are being developed, critical bug fixes and security patches will continue to be released. Existing projects using AutoGen can continue to operate, and Microsoft provides a migration guide. This transition is significant because AutoGen served as a foundational tool for understanding multi-agent systems. Four key concepts from AutoGen remain relevant: conversation history as shared state, the centrality of speaker selection in design, the cost-saving benefits of using different models per agent, and the essential separation of code writing from code execution. Conversation history acts as the shared state, simplifying debugging and agent integration but requiring careful management to control context length. The system's design hinges on how the "next speaker" is determined, with a preference for coded rules over inference calls when possible. Utilizing different AI models for different agent tasks can drastically reduce costs by employing cheaper models for simpler operations. Finally, separating agents that reason and write code from those that execute it in a sandbox provides crucial security and control. Users with AutoGen in production should plan their migration, while new projects should start with the Microsoft Agent Framework. The enduring value lies in understanding the underlying patterns of multi-agent systems rather than specific library implementations.