openJiuwen Releases TTSE Dual-Track Self-Evolution Framework

openJiuwen has released the TTSE (Dual-Track Self-Evolution) framework, designed to address the core pain point of AI agents where old experiences become invalid after environmental changes, leading to repeated rework. The framework splits an agent's self-improvement into two independent tracks: environmental cognition and action methodology, each validated and updated separately, drawing on the cognitive science separation of declarative and procedural knowledge. TTSE is now available with WorkSwarm 0.2.8 and has been open-sourced on GitHub and AtomGit, targeting developers of agent memory and skill systems.

Technically, TTSE's key innovation lies in dual-track separation: the environmental cognition track relies on on-site facts for validation, while the action methodology track updates based on conditional effectiveness, avoiding experience conflicts caused by single-track updates. This design borrows from ACT-R theory's distinction between declarative and procedural knowledge, enabling agents to adapt more flexibly to dynamic environments. However, current public information remains primarily methodological, lacking reproducible quantitative evaluation results, and practical effectiveness awaits developer validation.

For the agent ecosystem, TTSE offers a modular self-evolution approach that could reduce maintenance costs during cross-environment migration and promote standardization of memory and skill systems. The open-source initiative will also attract more developers to collaborate, accelerating research and implementation in recursive self-improvement (RSI).