Anthropic accidentally shipped Claude Code 2.1.88 with an unminified source map exposing 512,000+ lines of TypeScript. Users found upcoming features including a Tamagotchi-style coding companion and "KAIROS," an always-on background agent. Anthropic confirmed no customer data was exposed, but the code was mirrored to GitHub with 50,000+ forks before the issue was patched.
Products
Claude Code leak exposes a Tamagotchi-style ‘pet’ and an always-on agent
Anthropic's accidental exposure of 512K lines of unminified TypeScript reveals upcoming Tamagotchi-style coding companion and KAIROS always-on agent before official launch.
Tuesday, March 31, 2026 12:00 PM UTC2 MIN READSOURCE: The VergeBY sys://pipeline
Tags
products
/// RELATED
ResearchApr 28
Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines
Researchers benchmark bias mitigation techniques in LLM judges, revealing which strategies actually work against systematic evaluation bias.
ResearchApr 28
Stochastic KV Routing: Enabling Adaptive Depth-Wise Cache Sharing
Stochastic KV routing cuts transformer inference memory overhead by dynamically sharing key-value caches across layers, enabling leaner LLM deployment without sacrificing quality.