Agentic UI Exploration
summary
Explains a working prototype that wires an LLM into existing UI actions: the runtime sends enriched context and workflows to the model, the model returns structured tool calls like search hotel inventory with arguments (location, dates, budget), and the page redraws via existing React‑style hooks. Uses the AGUI protocol, server rehydration for private data, and highlights hard specifics: mutation safety, permission boundaries, observability, and where costs and failure recovery bite.
AI-written summary · about
chapters (8)
- 0:00Agentic UI exploration
- 1:50Agentic loop concept
- 4:43Current UIs know context and goals
- 7:12Live demo: agentic UI
- 12:12Challenges in agentic UIs
- 19:23Emerging agentic UI space
- 22:54Key takeaways and future
- 23:57Question and answer session
related talks
23mWebAI: Building the next user experience in the browser
50mUX as an API
28mDynamic UI at the Edge
01h 02mDesigning a Flexible UI Architecture with React and GraphQL
14mUI's Across Multiple Devices
25mAI, GraphQL, and the Rise of Malleable Applications
22mThe agent-ready web: Simplify user actions with WebMCP
26mMerging Design and Developement