浙大开源 HugAgentOS，智能体可自主沉淀经验并进化
智能体不再每次从零开始，HugAgentOS 让经验自动长成技能，且进化过程可控可审计。
浙大开源 HugAgentOS，解决 AI 智能体无法从过往任务中积累经验的问题。它用记忆、技能、编排三个引擎将执行痕迹转化为可复用能力，并设归因与本体两道关卡，确保每次进化可追溯、可回滚。
正文摘录
 新智元报道  Over the past year, agents have become significantly better at executing tasks. But another, more fundamental problem remains unsolved: agents are incapable of autonomous accumulation and self-evolution. Executing the same type of task ten times, and on the eleventh still planning the path from scratch. The mistakes corrected last time reappear the next time around. The tuning effort users invest during interactions is wiped clean the moment the session ends. Capabilities grow, but experience doesn't accumulate. This is not something a simple long-term memory module can fix. To truly turn experience into capability, at least four …