Bedtime School Bridge
Team led by a former AWS Principal PM who built Amazon Q Developer, featuring deep expertise in multi-agent architectures, LLMs, and code modernization.
YouTube Video
Project Description
A voice-activated iPad bedtime UI that the agent assembles, page by page, from one parent sentence about today.
The parent taps a microphone and says one sentence (“Move It Week 2, garbage trucks, basketball, calmer tonight”). The agent generates the bedtime interface their 2.5-year-old will use: tap-through illustrated picture book with auto-narration, tap-to-do movement game, food tiles that speak when tapped, voice mascot, and (when the parent’s sentence implies stress) an invented “feelings reset” breathing circle. The UI re-organizes when tone changes — calmer collapses to a dark night-sky listen layout; sillier brings up a bright sun, kangaroo-led movement scene. Same context, fundamentally different interface.
Why it’s generative UI not chatbot: voice in, UI out; variable module sets per tone (2-5 modules); invented module kinds for parent context (e.g. “feelings-reset”); per-module interactivity (tap pages, tick movement steps, breathe with the circle) for a non-reader.
Prior Work
Built on the Generative UI Global Hackathon Starter Kit (https://go.copilotkit.ai/GenUITemplate) which provided Next.js + CopilotKit scaffolding. The bedtime product, generated UI manifest schema, multi-provider LLM cascade router, voice flow, and toddler-first interactive renderers (tap-through illustrated story, food flip cards, movement checklist, breathing reset) are original to this build.