DAVID KHOURSHID Making state management intelligent
summary
Claims that combining state machines + reinforcement learning + GPT-4 makes UI agents that can plan and act, not just chat. Uses XState (xstate v5 beta and xstate/graph) to compute shortest paths through ~105 possible states, sends that to OpenAI to pick a target, and executes events; an espresso machine demo actually produces a cortado. LLMs are useful but unpredictable, need guard rails and server-side API keys; prospects: state-machine synthesis and adaptive UIs.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:52Declarative state in React
- 6:04State machines and AI
- 8:59XState crash course
- 10:25Reinforcement learning
- 16:46Large language models fundamentals
- 18:54Demo: Intelligent espresso machine
- 25:18Demo learnings
- 28:07Future AI in XState
- 30:32Summary
related talks
42mMaking state management intelligent
34mMaking state management intelligent
23mMind-Reading with Intelligent & Adaptive UIs
27mXState: the Visual Future of State Management
38mInto the Visual World of State Machines
31mXState: the Visual Future of State Management
39mMind-Reading with Adaptive and Intelligent User Interfaces
24mMind-Reading with Adaptive and Intelligent User Interfaces