<?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/a544db3bfb6c4597bda4e7518a55ca03&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/a544db3bfb6c4597bda4e7518a55ca03-9ad2ff5814051d4e.gif</thumbnail_url><duration>316.16</duration><title>Tribunal: Machine Oversight for Production Fixes</title><description>This Loom explains Tribunal, an agentic platform that handles production pipeline failures by having AI argue before taking action. It uses three agents, including a judge, a prosecution agent that proposes fixes, and a defense agent that warns fixes from last time may worsen the issue, with the judge deciding whether to implement. The decision goes to a human or machine review for about a 10 second SLO window, allowing a veto; otherwise the judge’s decision is executed. Decisions become precedents and are recorded, with confidence scores shown, and the system cites relevant prior cases when similar issues occur.</description></oembed>