AutoClaw has launched the "self-evolution" mechanism, where an Agent learns from its mistakes, with each lesson becoming a permanent memory. The platform has also introduced a Skill Store for users to enhance their Agents' capabilities simultaneously.
According to Perceive Beating monitoring, within the Spectrum AI umbrella, the Agent framework AutoClaw has officially launched its self-evolution mechanism. This mechanism aims to address the pain point of Agent "amnesia" by identifying user corrections, new methods, or expression preferences in conversations and converting them into the Agent's permanent memory. Users no longer need to repeatedly remind the Agent to "be more concise" or "avoid using hyphens," as the Agent can autonomously evolve after being taught once.
The self-evolution mechanism is triggered by keywords (such as "in the future" or "remember") or automatic detection (experience refinement after complex tasks). After each conversation round, AutoClaw will scan the dialogue and present an "evolution request" card, which can be written into memory only after user approval. To avoid information noise, Spectrum has established high-quality evolution principles, preferring to generate 1-3 deep evolutions per week rather than massive daily fragmented records. This mechanism allows the Agent to shift from "passive instruction execution" to "continuous growth through usage," where the more users teach, the more the Agent aligns with individual habits.
Simultaneously launched is the AutoClaw Skill Store, featuring the initial set of Spectrum's in-house developed GLM Office Skills suite (PPT, DOCX, XLSX, PDF, Charts). Powered by the office-scenario-optimized GLM-5.1 model at its core, it supports intelligent self-layout formatting, cross-format conversion (such as Word to PDF), and multi-file synchronous generation. Furthermore, the store has introduced expert co-creation Skills, including the "Nüwa Skill" that can replicate human thinking and the "Horizontal and Vertical Analysis Method" for in-depth research, packaging industry-leading methodologies as plug-and-play Agent capabilities.
