<?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/8c560fabd0304f63aacf4543a9856b54&quot; frameborder=&quot;0&quot; width=&quot;1232&quot; height=&quot;924&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>924</height><width>1232</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>924</thumbnail_height><thumbnail_width>1232</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/8c560fabd0304f63aacf4543a9856b54-f7a7658084c0f87b.gif</thumbnail_url><duration>137.973</duration><title>Agent Memory Palaces to Reduce Hallucinations</title><description>This Loom explains an Agent Palace approach to reduce hallucinations by using a memory palace concept for agent information retrieval. Bert says Agent Palace stores company knowledge as markdown files in G-Brain and lets different agents “walk through” to fetch the correct details. He demonstrates with an Acme Robotics example using QM as nufo.ai to generate a new engineering onboarding page by coordinating with engineering and legal for items like an NDA and I-9, tracing results into a memorable log. Runtime improved from 46 seconds with 32 tool calls to 34 seconds with 21 tool calls after repeated runs.</description></oembed>