{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/0fdcf2ca00ba4804b682e36ed4faccae\" frameborder=\"0\" width=\"1490\" height=\"1117\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1117,"width":1490,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1117,"thumbnail_width":1490,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/0fdcf2ca00ba4804b682e36ed4faccae-af59e949493b9dd3.gif","duration":1775.627,"title":"How Five AI Agents Build Websites","description":"This Loom demonstrates how the creator’s five AI agents work together inside Cloud Code to build and update a website from an idea. The agents use a shared memory called the Oracle so they do not guess, and Storm brainstorms while Rune turns the idea into a precise spec, Forge writes code in a real repo with migrations and publishes, and Echo summarizes and updates the Oracle, with Hiro acting as an outward voice to flag risks. The example build targets a climbing services style site named CoreX Farming Services for GreaterBay, runs through stages including building and landing, and is served on localhost with deployment taking longer in auto mode. The finished Next.js site is shown, including swapping an embedded YouTube video link, and the author concludes with the flow from idea to spec to working code to memory."}