Hyperagents

Researchers introduce "hyperagents," self-referential AI systems that combine a task agent and meta-agent in one editable program. Unlike previous self-improving systems with fixed mechanisms, hyperagents enable metacognitive self-adaptation—improving both task performance and the improvement process itself. Built on the Darwin Gödel Machine framework, DGM-Hyperagents eliminate domain-specific assumptions, potentially enabling self-accelerating progress across any computational task.