Your Claude can now dream! 😨🤯

Añadido: 2026-06-02

[00:00:00]Claude agents can now dream. Yes, you heard it right, and the law firm using them got six times more productive. May 6th, 2026, Anthropic Code with Claude conference, SF CPO Amie Vorah on stage.

[00:00:11]Four features dropped: dreaming, outcomes, multi-agent orchestration, and webhooks. Between jobs, your agent reviews everything it did, finds patterns in its own mistakes, and writes new playbooks for the next session.

[00:00:22]Anthropic compares it to how your brain consolidates memory during sleep. The agent doesn't update model weights. It writes plain text notes future sessions can read. Harvey turned this on. Their legal AI agents kept forgetting file type quirks between sessions. The same drafting jobs failing the same way. With dreaming, the workarounds stuck. Task completion, [music] that's Harvey's own number. Wise Docs cut medical doc review by 50% using outcomes. Netflix runs build log analysis across hundreds of apps in parallel. Dreaming in research preview, outcomes and multi-agent orchestration in public beta. Go check it out.