JevRSI framework for cost-effective recursive self-improvement
November 18, 2026
JevRSI is a collaboration framework that uses a fast System-1 model for low-latency actions and an asynchronous System-2 LLM to develop a modular harness. This approach enables recursive self-improvement of decision modules based on environmental feedback.
HOW THIS AFFECTS YOU
●
builderYou can use a fast model for immediate actions while a larger model builds better logic in the background.
●
researcherThis introduces a new way to bridge the gap between fast response and deep reasoning.