---
title: "Comment by Kartik N V J K on \"One Agent, One Worktree: Isolation and Concurrency for Coding Agents\" | Hashnode"
description: "Calling parallel coding an isolation problem before a coordination problem is the right ordering. I keep separate worktrees per agent and the thing that still b"
url: "https://hashnode.com/posts/parallel-coding-agents-isolation/6ab4dc91cfcc71d3b3162208/comment/6ab581bfb638151168c2aec3"
---

[Discussion](https://hashnode.com/posts/parallel-coding-agents-isolation/6ab4dc91cfcc71d3b3162208)

Comment [One Agent, One Worktree: Isolation and Concurrency for Coding Agents](https://hashnode.com/posts/parallel-coding-agents-isolation/6ab4dc91cfcc71d3b3162208)

[Kartik N V J K](https://hashnode.com/@kartik-nvj)

[Sep 24](https://hashnode.com/posts/parallel-coding-agents-isolation/6ab4dc91cfcc71d3b3162208/comment/6ab581bfb638151168c2aec3)

Calling parallel coding an isolation problem before a coordination problem is the right ordering. I keep separate worktrees per agent and the thing that still bites me is the merge step, because two agents can each pass their own tests and still conflict at the semantic level, like both renaming the same symbol differently. Do you gate merges on a combined test run, or only per-worktree?
