<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/d78c46fe8b2d4f2899bcb70a22d2ae96&quot; frameborder=&quot;0&quot; width=&quot;1658&quot; height=&quot;1243&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1243</height><width>1658</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1243</thumbnail_height><thumbnail_width>1658</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/d78c46fe8b2d4f2899bcb70a22d2ae96-257ff9fc882957e1.gif</thumbnail_url><duration>199.303</duration><title>Monitoring and Replacing Hallucinating AI Agents</title><description>This Loom demonstrates an AI engine that documents and monitors multiple agents in a workflow, including hallucination detection and agent replacement. It describes Layer 1 generating an MD file log of each agent’s actions, tracked files, prompts, models, and hallucination checking, while Layer 2 prompts the user to swap or spin up a different agent when hallucinations are found. In the demo, the system splits the work for a prompt to create a simple HTML and CSS flowers website, spins up multiple Codesmith agents, and shows agent logs and a prompt graph. It also simulates a hallucination where Codesmith is replaced with Mood Maven to continue the work, with further switching to Cynthia, and reviews the generated HTML and CSS in various workspace runs.</description></oembed>