Frontend after AI: The new UX
summary
Claims that taste is trainable and that measurable UX wins matter now: optimize LCP, CLS and ARIA/WCAG, because a 100ms improves conversions and perceived quality. Demonstrates a Node.js MCP server tunneled via Cloudflare, wired to Claude and ChatGPT as clients, and an MCP app that returns branded, accessible speaker cards. Demo generally works though live model availability sometimes lags; the MCP app preserves brand and accessibility while avoiding cookie banners and messy navigation.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:11Defining taste in AI era
- 10:12Taste for UX and accessibility
- 15:28Frontend standards and UX challenges
- 17:34Tasteful vs hostile web
- 21:49AI agents for user data
- 24:19Model Context Protocol and servers
- 27:27Brand with MCP apps
- 31:39Future of frontend
related talks
23mMCP-UI: Where Intent Becomes Interface
22mThe Critical Advantage With AI
22mMCP UI: Extending the frontier
20mGood design hasn’t changed with AI
38mEvolving AI chat with MCP Apps
17mBeyond Components: Designing Generative UI for MCP Apps
28mCreating fluent Frontend Experiences in an age of AI
26mBuild your First MCP App