Anthropic introduced a capability called “dreaming” on May 6, 2026, at the Code with Claude developers’ conference in San Francisco. It allows Claude Managed Agents to process one to 100 session transcripts in the background, merge duplicate notes, remove contradictions, and identify recurring mistakes and consistently successful workflows. Original inputs remain intact, and users can choose between automatic memory updates and manual approval. The research preview supports claude-opus-4-8, claude-opus-4-7, and claude-sonnet-4-6, but organizations must request access. In early testing with legal AI company Harvey, task completion rates on realistic legal benchmarks reportedly increased sixfold after the feature was applied.
Source: Read the original article
