【AI前沿】Aliyun QoderWork Launches Consciousness Function, Achieving AI Self-Memory and Skill Evolution

AI NEWSLatest AI NewsArticleAliyun QoderWork Launches Consciousness Function, Achieving AI Self-Memory and Skill EvolutionPublished in Latest AI NewsTime :Jun 16, 2026Read :4minuteAlibaba Cloud officially launched the new “Consciousness” feature for its desktop AI assistant QoderWork on June 16th, integrating three core modules: memory, reflection, and skill evolution. This upgrade marks the transition of desktop-level AI assistants from simple conversation tools to intelligent agents with autonomous growth capabilities. The core of this feature is that AI can not only actively organize and eliminate redundant information while preserving conversations but also automatically convert users’ frequently performed tasks into reusable skills, significantly improving the efficiency of long-term collaborative work.On the technical architecture, QoderWork’s memory system adopts a hierarchical structure, with daily conversations first stored in short-term memory, and valuable content across sessions promoted to long-term memory. To ensure information security, its reflection mechanism mimics the selective forgetting of the human brain, taking a snapshot backup before each reflection and performing post-validation through hard rules. If any deviation occurs, it automatically reverts to prevent accidental deletion of critical data. All memories are stored locally in an open Markdown text format, granting users full access and editing permissions.In addition, the system uses a multi-factor weighted scoring model to evaluate the value of conversations in real time, and only initiates the evolution process through the Fork Session mechanism when the threshold is reached. This increases the cache hit rate to over 99%, with additional costs controlled within 5% of the total cost of the main conversation.For frequent tasks, the generated skill suggestions are presented to users for selection, ignoring, or rejection. Rejected suggestions will automatically cool down. Currently, QoderWork has formed a diverse product matrix including Chat mode, multiple types of workstations, and industry expert packages. The addition of the “Consciousness” feature not only demonstrates the feasibility of lightweight AI performing autonomous incremental learning on the edge, but also provides a technical demonstration for the industry to explore more cost-effective and privacy-friendly personal AI workflow evolution.