<?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/460337d003c44718a76f7e5bc66926db&quot; frameborder=&quot;0&quot; width=&quot;1108&quot; height=&quot;831&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>831</height><width>1108</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>831</thumbnail_height><thumbnail_width>1108</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/460337d003c44718a76f7e5bc66926db-3f715df2547916c4.gif</thumbnail_url><duration>290.861</duration><title>How I manage AI agent conflicts with Jujutsu</title><description>Three Claude Code agents working in parallel on the same repo, each in its own Jujutsu workspace, in macOS sandboxes with bypass-permissions. When two of them produce a conflict, I dispatch a fourth agent to resolve it.                                                                                

Code generation got cheap. Review, integration, and judgement are now the work - this setup puts your attention there.

Built on Jujutsu (https://github.com/jj-vcs/jj) and VisualJJ (https://www.visualjj.com).</description></oembed>