Your React app runs smoothly on phones, but the living room is a completely different world. TVs are slower than phones, and adapting for a remote control introduces complex challenges like focus management and replacing touch-and-drag interactions. How can we make the non-deterministic nature of AI coding assistants more predictable to solve these problems effectively?

In this talk, Eric Fahsl, Principal Product Manager at Amazon, shows how Kiro.dev, a new AI assistant from Amazon, helps adapt a mobile React 19 and Three.js app for the TV screen. He'll showcase Kiro's unique "spec mode," a process that turns a single prompt into structured requirements, design documents, and actionable tasks to bring predictability and power to your development workflow.

What you'll discover:

• The spec-driven process: See how Kiro turns a prompt into three distinct phases, Requirements, Design, and Tasks, to create a clear plan of record for AI-driven development.
• Solving TV-specific challenges: Learn techniques for handling performance tiering, focus management, D-pad navigation, and adapting interactive components for the 10-foot UI.
• Making AI predictable: Discover how to use Kiro's "Steering Documents" to enforce project architecture and hooks to automate tasks like running tests, making AI a more reliable partner.
• Iterating on a plan: Watch how an existing spec can be modified mid-project to refine features, like changing a zoom mechanic, without losing context.