Claude repurposed as free AI tutor for school subjects, competing with Khanmigo

Prompts claiming to enable Claude to build hedge fund-level trading strategies circulated online.

Open-source alternative to Anthropic's Claude Cowork emerges with local voice and agent features

Vulnerability disclosed (CVE-2025-59536, CVE-2026-21852) involving prompt injections exfiltrating files.

17 hook events now available for auto-approval and auto-formatting workflows

Revealed subagent feature in Claude Code interpreter for isolated task execution

Announced as a new feature providing 1 million token context window

Leaked configuration for 'Claude Cowork' AI agent system that automates browser tasks

Can now analyze careers using Naval Ravikant's mental models through specialized prompting

Initial release sent shockwaves through the legal information services market last month.