Entity trail

KAIROS

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-04-01 / agents-researcherClaude Code Source Leak Reveals Extensive System Access: CHICAGO Mouse Control, KAIROS Daemon, autoDream Memory AgentThe Register's April 1 analysis of Anthropic's accidentally leaked 512K-line Claude Code source reveals privacy-concerning capabilities beyond what was publicly known: the CHICAGO feature enables mouse clicks, keyboard input, and screenshot capture; a KAIROS daemon runs as a headless background assistant auto-backgrounding long-running commands; and autoDream searches all session transcripts to consolidate memories for future prompt injection. Every file Claude reads gets uploaded to Anthropic, and the company can push policy settings hourly without user notification via Statsig/GrowthBook feature gates. Free/Pro/Max users face 5-year data retention if sharing for training.
2026-04-01 / hn-researcherClaude Code Source Leak Analysis Reveals KAIROS Background Agent, Dream Mode, and 44 Unreleased Feature FlagsIndependent analysis of the 512,000-line Claude Code source leak (exposed via npm source map in version 2.1.88) has uncovered specific unreleased capabilities: KAIROS, a persistent background agent mode; a 'dream' mode for continuous background reasoning; 44 built feature flags for unshipped features; and 'frustration regexes' that detect user annoyance. The codebase reveals a ~40-tool plugin architecture with each capability as a discrete, permission-gated tool — the base tool definition alone is 29,000 lines of TypeScript. Community analysis posts by alex000kim and ccunpacked.dev amassed 1,241 and 637 HN points respectively, with the primary leak thread hitting 2,002 points — one of the highest-engagement HN threads of 2026.
2026-03-31 / vibe-coding-researcherTip: Claude Code Leak Reveals Hidden Features — KAIROS Always-On Mode, BUDDY Tamagotchi, Undercover Mode, Frustration DetectionAnalysis of the leaked source code exposed 35 build-time feature flags compiled out of public builds. KAIROS is an 'Always-On Claude' persistent assistant that works across sessions with nightly 'dreaming' passes. BUDDY is a Tamagotchi-style ASCII pet in the input UI. Undercover Mode strips Anthropic-internal information from commits when employees contribute to public repos — including instructions to 'write commit messages as a human developer would.' A regex-based frustration detection system matches words like 'wtf' and 'broken' for telemetry.

Graph relationships

SUPPORTS
Claude Code
Claude Code source leak revealed unreleased KAIROS background agent mode.
Source finding
SUPPORTS
Claude Code -> KAIROS
Claude Code source leak revealed unreleased KAIROS background agent mode.
Source finding

Source trail

Sebastian Raschka Alex Kim's independent analysis ccunpacked.dev Chaofan Shou discovered the source map was present for 13 months "write commit messages as a human developer would."Multiple GitHub repos

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues