[Tencent Hy Memory 2.0 Open Source: 10 AI Tools Share Memory, Even Rolled-Back Code Can Be Traced] Beating AI Newsflash: Tencent's HunYuan has upgraded Hy Memory to version 2.0, open-sourcing its memory framework, dedicated models, and MemHub application all at once. The 1.0 version, launched in May, primarily handled chat memory. It could track changes in the same matter over time, identify and resolve contradictions, and summarize user habits that were not explicitly stated across multiple conversations. When users change preferences, the system can retain old records while remembering the latest decisions. Version 2.0 introduces working memory and code memory. It records the progress of a task, the approaches used, and the final outcomes. For similar tasks, it checks whether past experiences are applicable. Code memory goes a step further, combining Git records to track the modification history of files and functions. Even if a piece of code is later rolled back, it can retrieve the changes made at the time and clarify which methods were tried previously. The accompanying MemHub integrates memory across different AI tools. Ten tools, including Cursor, Codex, and Claude Code, can share a unified local memory. With user authorization, MemHub can import historical conversations and continuously sync new records. For example, if project specifications are finalized in Cursor and development continues in Codex, the memory can be directly retrieved without needing to re-explain. Tencent has also trained a dedicated model, Hy-Memory-30B-A3B, with a total of 30 billion parameters, activating only about 3 billion at a time. This model specializes in filtering, organizing, and updating memory and can also connect to existing frameworks like Mem0 and MemOS. Tencent tested four memory systems, achieving an average score of 55.49, slightly higher than Opus 5's 54.62. [Original Link]