Thariq Shihipar, who works on Claude Code at Anthropic, has posted an argument that AI agent outputs should shift from Markdown to HTML. He said long Markdown files are hard to read, weak for visual expression and inconvenient to share, while direct editing is increasingly delegated to Claude. He listed four HTML strengths: higher information density, better readability, easier sharing via links, and two-way interaction using controls such as sliders and buttons.

Security concerns sparked by Anthropic\'s AI model Claude Mythos are spreading across government and industry in South Korea, with growing calls to rethink defenses as AI can detect vulnerabilities and generate attack code. A Cloud Security Alliance report says the time from vulnerability discovery to exploitation has shrunk sharply. South Korean ministries have held an emergency meeting, while experts urge stronger governance, standardized information sharing and automated, AI-driven defenses alongside basic asset management.