Anthropic, an AI research company, has introduced a new feature for its Claude AI called "dreaming," currently in research preview. This functionality is integrated into its agent management software and aims to enable AI agents to self-improve by reviewing their work between sessions, identifying patterns, and updating contextual files such as user preferences. Launched in conjunction with its San Francisco developer conference, "dreaming" is part of Anthropic's strategy to attract business clients, following increased demand for its AI coding agent. This development comes after Anthropic also unveiled ten finance-focused AI agents in New York and announced broader access to other features, like task delegation capabilities for its AI agents.