A major source code leak has exposed that Anthropic is evolving Claude from a simple chatbot into a fully autonomous AI agent. The leak, comprising over 500,000 lines of code, reveals features such as UltraPlan, which generates editable plans that run for 10 to 30 minutes, and Kairos, a background daemon that proactively monitors files and retains context even after a session ends. A voice mode and a companion feature with ASCII animations are also included.
This marks a leap from AI as a tool to AI as a teammate. Instead of requiring step-by-step prompts, Claude can now take a high-level goal and execute complex tasks for hours—researching competitors, drafting reports, creating slide decks, and emailing teams autonomously. For developers, the managed agents API enables deploying production-ready agents in days; Rakuten built enterprise agents for spreadsheets and slides in just one week.
Claude also self-evaluates by planning, acting, and reviewing its work, fixing errors before they reach the user. This shift promises a future where computers work for people, not the other way around.