<?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/5c622b884c8d4b22bc195ce8ccee8b9b&quot; frameborder=&quot;0&quot; width=&quot;1670&quot; height=&quot;1252&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1252</height><width>1670</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1252</thumbnail_height><thumbnail_width>1670</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/5c622b884c8d4b22bc195ce8ccee8b9b-97be1a2fd1ae4a8d.gif</thumbnail_url><duration>141.845</duration><title>Enhancing Agent Performance with GoodNight&apos;s Reflection Cycle 💤</title><description>In this video, I introduced GoodNight, a system designed to enhance our coding agents by allowing them to reflect on their experiences and identify recurring issues. I demonstrated how our multistage identity pipeline works, processing agent conversations and producing artifacts like CloudSkills or Cloud.MD edits for approval. I also highlighted the importance of using BendyBeave to ensure that our outputs are safe and do not expose private information. I made some changes to the clod.md file during the demo, adding entries like core workflow preferences and confirming before bulk operations. I encourage you to review these updates and consider how they can improve our processes.</description></oembed>