Code for "Evolutionary System Prompt Learning for Reinforcement Learning in LLMs" (arxiv.org/abs/2602.14697)

Multi-agent orchestration with self-evolution, cognitive memory, and governance.

Endgame-AI: A closed-loop GUI agent that controls Windows like a human. Uses native UI Automation + self-reflection to complete real desktop tasks. Runs fully locally (LM Studio) or with cloud backends. The system literally rewrites its own prompts after each run.