Anthropic unveils AI agent learning system

Anthropic has introduced "dreaming," a new feature for its Claude Managed Agents platform that enables AI agents to learn from past sessions and improve autonomously. This system, along with "outcomes" and "multi-agent orchestration" now in public beta, aims to address key enterprise challenges in AI agent reliability and scalability. Early adopters have reported significant improvements, with legal AI company Harvey seeing a sixfold increase in task completion and medical document review company Wisedocs reducing review time by 50%.