<?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/e208723b422e42fda8b1ed36811b58bd&quot; frameborder=&quot;0&quot; width=&quot;1664&quot; height=&quot;1248&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1248</height><width>1664</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1248</thumbnail_height><thumbnail_width>1664</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/e208723b422e42fda8b1ed36811b58bd-98e48637854ce61b.gif</thumbnail_url><duration>202.216</duration><title>Exploring Tool Use in Browser AI: A Demo of On-Device Models 🚀</title><description>In this video, I demonstrate the capabilities of Browser AI, a prototype that allows developers to connect to on-device models through a simple API in the window object. Using a cafe website as an example, I showcase how the AI can assist with actions like adding and removing items from a cart, filtering by categories, and searching for products. The model supporting this demo is Llama 3.2, a lightweight 3 billion parameter version, which enables secure and low-latency interactions. I encourage viewers to explore the potential of running models on users&apos; devices for enhanced privacy and functionality. Please consider how these advancements could be applied in your own projects.</description></oembed>