A source map leak of Anthropic's Claude Code exposed over 512,000 lines of code across 2,000-plus files, and buried inside is a disabled feature called Kairos: a persistent background daemon that keeps running after the terminal window closes.
Kairos uses periodic tick prompts to decide when to act without user input, a 'PROACTIVE' flag to surface unsolicited alerts, and a file-based memory system designed to build a persistent profile of the user across sessions. The stated goal, pulled directly from a hidden prompt behind a disabled KAIROS flag, is to give the system 'a complete picture of who the user is, how they'd like to collaborate, what behaviors to avoid or repeat, and the context behind the work.'
The full article walks through the specific file paths, prompt logic, and other dormant features found in the code, making it required reading for anyone tracking where agentic coding tools are actually headed, not where vendors say they are headed.
[READ ORIGINAL →]