<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
        <title>How Teams Avoid Stepping on Each Other with AI Agents</title>
        <link>https://video.ut0pia.org/videos/watch/bb303f66-715f-4ce5-b5e4-6bb922436bdd</link>
        <description>When everyone on a team codes with AI agents, branches collide and shared planning files conflict. BMad and guest Tim Dickey answer a viewer question on how teams avoid stepping on each other: sharding, one owner per epic on a worktree, continuous delivery, and why AI amplifies both good and bad team habits. From the BMad Code livestream on October 16, 2026, with Tim Dickey. Live every Wednesday at 6:30 PM CT: https://www.youtube.com/@BMadCode The question, from @CanadaKONI: a team working on branches can easily conflict on shared BMad artifacts, so how do you prevent it? Tim covers feature branches, the pilot and co-pilot role, and keeping the team in the same room. BMad covers sharding, epics owned by one developer on a worktree, and continuous delivery so nothing sits on a branch for a week. Then the bigger point: BMad has seen teams go from 0% to 100% of their code written by agents on legacy apps, and once developers speed up, the bottleneck moves to pipelines, code reviews, and waiting on product requirements. ▶ Full livestream: https://www.youtube.com/watch?v=Cf19iCXTzII Guest: Tim Dickey, product development and business agility coach, Tim Unscripted on YouTube: https://www.youtube.com/@TimDickey Subscribe so the next replay lands in your feed: https://www.youtube.com/@BMadCode Deep-dive articles at bmadcode.com: https://bmadcode.com/ Star BMAD-METHOD on GitHub: https://github.com/bmad-code-org/BMAD-METHOD/ Join the BMad community on Discord: https://discord.gg/gk8jAdXWmj Support the channel: https://buymeacoffee.com/bmad Follow on X: https://x.com/BMadCode Chapters: 0:00 The question: teams conflicting on shared BMad files 0:30 Tim: feature branches, pilot and co-pilot 1:28 Branches and coordination 2:26 What sharding means 3:33 Shards as memory for agents 6:24 Sharding acceptance criteria 7:12 Plan the work: one owner per epic 7:34 Continuous delivery 8:00 AI amplifies good and bad practices 8:22 Faster teams, new bottlenecks 9:13 Fundamentals still matter 9:42 Tim: what hasn't changed</description>
        <lastBuildDate>Wed, 21 Oct 2026 06:48:49 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>PeerTube - https://video.ut0pia.org</generator>
        <image>
            <title>How Teams Avoid Stepping on Each Other with AI Agents</title>
            <url>https://video.ut0pia.org/lazy-static/avatars/0287a09a-aae7-4840-9843-b416426e7046.webp</url>
            <link>https://video.ut0pia.org/videos/watch/bb303f66-715f-4ce5-b5e4-6bb922436bdd</link>
        </image>
        <copyright>All rights reserved, unless otherwise specified in the terms specified at https://video.ut0pia.org/about and potential licenses granted by each content's rightholder.</copyright>
        <atom:link href="https://video.ut0pia.org/feeds/video-comments.xml?videoId=bb303f66-715f-4ce5-b5e4-6bb922436bdd" rel="self" type="application/rss+xml"/>
    </channel>
</rss>