Antigravity feels heavy and Claude Skills are light

A hands-on comparison of Google Antigravity IDE and Claude Code for AI-assisted development workflows. The author tested both tools by creating a Slidev presentation with AI-generated images, finding Antigravity slow, underpolished, and plagued by token limits and data risks. Claude Code proved more reliable, especially when enhanced with custom Claude Skills—reusable scripts that extend AI capabilities and allow swapping underlying models without changing editors. The piece argues that Skills may be more transformative than the choice of base model or IDE.