<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Open Source on Jonathan Franklin</title><link>https://jonnie.github.io/open-source/</link><description>Recent content in Open Source on Jonathan Franklin</description><generator>Hugo</generator><language>en</language><atom:link href="https://jonnie.github.io/open-source/index.xml" rel="self" type="application/rss+xml"/><item><title>agent-bench</title><link>https://jonnie.github.io/open-source/agent-bench/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://jonnie.github.io/open-source/agent-bench/</guid><description>&lt;p&gt;&lt;strong&gt;Compare coding-agent harnesses using the same local model.&lt;/strong&gt;&lt;/p&gt;&#10;&lt;p&gt;&lt;code&gt;agent-bench&lt;/code&gt; is a Python command-line tool that runs one LAN-hosted LLM through three coding-agent harnesses: &lt;strong&gt;pi&lt;/strong&gt;, &lt;strong&gt;oh-my-pi&lt;/strong&gt; and &lt;strong&gt;OpenCode&lt;/strong&gt;. Each attempt starts in a fresh Docker container, and the resulting patch is evaluated independently in a separate grading container using hidden acceptance tests.&lt;/p&gt;&#10;&lt;p&gt;The built-in tasks cover bug fixes, small features, refactoring, security hardening and performance work. Each run produces a machine-readable JSON report and a standalone HTML report you can open offline, with scores, timings, logs and code diffs.&lt;/p&gt;</description></item><item><title>Capture</title><link>https://jonnie.github.io/open-source/capture/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://jonnie.github.io/open-source/capture/</guid><description>&lt;p&gt;&lt;strong&gt;Catch a thought now. Organise it without breaking your flow.&lt;/strong&gt;&lt;/p&gt;&#10;&lt;p&gt;Capture is a local-first web app for thoughts, ideas and tasks. A keyboard shortcut opens a focused capture overlay, and an optional LLM assigns tags so you can save the thought without first deciding where it belongs.&lt;/p&gt;&#10;&lt;p&gt;Captures live in a local SQLite database and can be explored in a tag-grouped board, chronological list or interactive graph. An MCP endpoint also lets compatible AI agents add, search, edit, archive and delete captures.&lt;/p&gt;</description></item></channel></rss>