<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Blog on Jonathan Franklin</title><link>https://jonnie.github.io/blog/</link><description>Recent content in Blog on Jonathan Franklin</description><generator>Hugo</generator><language>en</language><lastBuildDate>Mon, 05 Oct 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://jonnie.github.io/blog/index.xml" rel="self" type="application/rss+xml"/><item><title>Same model, different harness: benchmarking qwen3.8-27b with Agent Bench</title><link>https://jonnie.github.io/blog/agent-bench-qwen3-8-27b/</link><pubDate>Mon, 05 Oct 2026 00:00:00 +0000</pubDate><guid>https://jonnie.github.io/blog/agent-bench-qwen3-8-27b/</guid><description>&lt;p&gt;Changing a coding agent does not necessarily mean changing its model. The same LLM can be wrapped in different system prompts, editing tools, context-management strategies and execution loops. Those choices matter: the useful output is not a confident explanation, but a patch that satisfies the task.&lt;/p&gt;&#10;&lt;p&gt;I built &lt;a href="https://github.com/jonnie/agent-bench"&gt;Agent Bench&lt;/a&gt; to make that comparison easier. It runs the same coding tasks through different agent harnesses against &lt;strong&gt;one shared model endpoint&lt;/strong&gt;, then grades the resulting code independently.&lt;/p&gt;</description></item></channel></rss>