<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>issue-management on Arrans Technical blog</title><link>https://arran4.github.io/blog/tags/issue-management/</link><description>Recent content in issue-management on Arrans Technical blog</description><generator>Hugo -- gohugo.io</generator><language>en</language><managingEditor>technicalblog123@arran4.com (Arran Ubels)</managingEditor><webMaster>technicalblog123@arran4.com (Arran Ubels)</webMaster><copyright>Arran Ubels. This work is licensed under a &lt;a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/">Creative Commons Attribution-ShareAlike 4.0 International License&lt;/a>.</copyright><lastBuildDate>Wed, 09 Sep 2026 12:26:55 +1000</lastBuildDate><atom:link href="https://arran4.github.io/blog/tags/issue-management/index.xml" rel="self" type="application/rss+xml"/><item><title>Managing Jules with a Management LLM</title><link>https://arran4.github.io/blog/post/2026/044-jules-management/</link><pubDate>Wed, 09 Sep 2026 12:26:55 +1000</pubDate><author>technicalblog123@arran4.com (Arran Ubels)</author><guid>https://arran4.github.io/blog/post/2026/044-jules-management/</guid><description>I use Jules as an asynchronous implementation worker, but the useful workflow is larger than Jules itself. The part that makes it practical is a separate management LLM that sits between me, GitHub, the implementation agent, and the durable issue history.
This post describes how I currently work, why it works for me, where the human attention cost comes from, and how I want a management LLM to behave when it is dropped into a fresh environment with little or no prior conversational context.</description></item></channel></rss>