<?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/95c5734fcd894e439ea4ff279ece4ae5&quot; frameborder=&quot;0&quot; width=&quot;1920&quot; height=&quot;1440&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1440</height><width>1920</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1440</thumbnail_height><thumbnail_width>1920</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/95c5734fcd894e439ea4ff279ece4ae5-909f4bc54453a8db.gif</thumbnail_url><duration>300.331</duration><title>AI Coach RAG Number Guessing Tips 🚀</title><description>I walk through Rag Coach, a local number guessing game where an AI coach watches your play and gives strategy tips, and you can sign in to see your game history and the tips you received. In hard mode, on your first guess, the Retriever pulls two documents, then a second LLM call critiques the tip for actionability, secrets, and other flags. We have 32 tests passing and 12 out of 12 guardrail blocks. I also share an AB test that cut the response by about 63 percent and made tips safer by staying within two numbers above the secret. No viewer action was requested.</description></oembed>