#4790 Multi-Agent Framework: Recurrent Self Improvement (RSI)
#4790 Multi-Agent Framework: Recurrent Self Improvement (RSI) #4790 Describe a multi-agent framework to improve a given skill A multi-agent framework designed to improve a specific skill operates as a closed-loop deliberate practice system. Rather than relying on a single model to both instruct and evaluate, decomposing the process into specialized, communicating agents separates pedagogical planning, targeted practice, objective critique, and knowledge consolidation. Core Architecture The architecture models Anders Ericsson's principles of deliberate practice: identify granular sub-skills, push beyond comfort zones with focused drills, provide immediate actionable feedback, and iterate until mastery is verified. ┌──────────────────────────────┐ │ Orchestrator Agent │ │ (Curriculum & State Bus) │ └──────┬───────────────▲───────┘ │ │ ┌────────────...