Ephemeral UIs That Learn, Adapt, and Vanish
summary
Claims an intent-based architecture that renders UIs from a schema and then prunes or hides components as tasks change, with a prototype called Muse intent platform (to be open sourced). Implements an intent controller, codec libraries (color, Google Fonts), an act validation layer, and schema-to-LLM function calls. Demo is a POC widget that orchestrates state/context/functions and works end-to-end; measured LLM latency around 500 ms response without memory and ~1 s with memory.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:04One-size-fits-all interfaces challenge
- 3:53AI and UI interaction shift
- 6:26Proactively adaptive interfaces
- 8:54Intent-based UI architecture
- 11:40Simplicity and invisibility in design
- 15:56Future of proactive interfaces
- 17:21Flutter for this vision
- 20:30Muse intent platform
- 23:22Demo: intent controller and schemas
- 24:51Codecs for UI components
- 27:52Dynamic UI triggers
- 29:57Next AI-interface paradigm
- 30:19Closing and Q&A
related talks
12mYour UI Has Left the Building: Interfaces That Write Themselves
28mLiquid User Interfaces for the Next Generation of AI Apps
17mBeyond Components: Designing Generative UI for MCP Apps
33mIs Frontend Still a Thing? A UI Reset in the Age of AI Disruption
38m🔥 Generative UI: The Future of Frontend (And Why You Should Care)
23mMind-Reading with Intelligent & Adaptive UIs
37mMind-Reading with adaptive UIs
21mBeyond Components: The Future of AI UI