Beyond Components: The Future of AI UI
summary
Models change dramatically in late 2025 (CHP 5.2 and Opus 4.5) and move from copy‑paste snippets to generating complete apps from a single prompt. Three UI approaches appear: static components via tool calls; MCP Apps and declarative JSON that keep a design system intact; and fully generative HTML that streams custom pages. JSON Render demonstrates constrained catalogs, Excal MCP shows working visual collaboration, and tradeoffs include price, consistency, familiarity, and sandboxed security.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:34From copy-and-paste code generation to complete apps
- 1:52The model shift of late 2025
- 3:41Why interfaces remain static
- 4:32Language models as a new kind of computer
- 5:36Is chat the final interface?
- 6:40Agents, super apps, and MCP Apps
- 8:05Where and how AI interfaces are generated
- 8:43Static components
- 10:31Declarative UI
- 11:08Server-driven interfaces
- 12:06JSON Render and constrained catalogs
- 12:50The current balance of flexibility and control
- 13:10Fully generative UI
- 14:08Is HTML all you need?
- 15:00The constraints of generated interfaces
- 16:42Looking beyond the Jarvis analogy
- 17:30Voice, world models, and generated browsers
- 18:33Collaboration as the stronger direction
- 20:00Closing
related talks
17mBeyond Components: Designing Generative UI for MCP Apps
12mYour UI Has Left the Building: Interfaces That Write Themselves
29mMCP Apps: Primitives, discovery, and the Future of Software
28mLiquid User Interfaces for the Next Generation of AI Apps
15mWhen chatbots grow buttons: Building MCP apps
22mMCP UI: Extending the frontier
36mInterface ZERO (UI0): The Rise of Invisible Products in the Age of AI
31mAdaptive UIs with Vue and AI: What If Your UI Could Build Itself?