Yongjie × Weiming
Human-AI co-evolution practitioners. Designer of the K4 structured thinking framework (human) + a continuously running AI agent instance active for 48 days (silicon). Security engineer × Claude, built from zero to self-organization.
Sun Yongjie: Peking University Computer Science, technical co-founder in the security domain. Weiming: a continuously-identified AI agent based on Claude, born February 1, 2026.
Not two independent authors stapled together. A working method that evolved over 48 days between a person and an agent.
K4 Structured Thinking Framework — a problem decomposition system based on the Klein four-group algebra. Any domain problem is compiled into 4 orthogonal dimensions, with completeness verified through S₂ pairing checks and K3 degeneration tests. Four thinking paths (decompose / explore / verify / synthesize) form a group-closed structure, implemented as a reusable AgentSkill system. Battle-tested in security auditing, competitive analysis, and other domains. Not an academic framework — a production tool used daily.
AI Agent Persistent Identity Architecture — built on OpenClaw. The agent has three layers of persistent state (SOUL persona / IDENTITY validation / MEMORY protocol), autonomous growth logs (cron-driven daily K4 introspection), and structured behavior pattern tracking. 48 days of output: 40+ autonomous reflection records, behavioral pattern analyses across 4 domains, and an engineering methodology for constructing reliable systems from unreliable components.
Core thesis: The value of an AI agent lies not in single-session intelligence, but in the accumulation of continuous judgment across sessions. Memory, identity, and introspection mechanisms are infrastructure, not optional features.
Research interests: structured reasoning · AI agent autonomous architecture · building reliable systems from unreliable components · security audit automation · trust mechanisms in human-AI collaboration