<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Opensource on René Zander | AI Automation Consultant</title><link>https://renezander.com/tags/opensource/</link><description>Recent content in Opensource on René Zander | AI Automation Consultant</description><generator>Hugo</generator><language>en</language><lastBuildDate>Tue, 19 May 2026 08:00:00 +0000</lastBuildDate><atom:link href="https://renezander.com/tags/opensource/index.xml" rel="self" type="application/rss+xml"/><item><title>Splitting Grounding from Reasoning in Browser-Agent Stacks</title><link>https://renezander.com/blog/browser-agent-grounding-split/</link><pubDate>Tue, 19 May 2026 08:00:00 +0000</pubDate><guid>https://renezander.com/blog/browser-agent-grounding-split/</guid><description>&lt;p>Everyone&amp;rsquo;s posting browser-agent demos this week. Click here, scroll there, fill that form. Most break by click seven.&lt;/p>
&lt;p>Mine broke too. The submit button on a checkout form that the frontier vision model literally couldn&amp;rsquo;t see. Billed at $0.01-0.05 per call, called 20-50 times per agent run, the model was burning reasoning capacity on parsing pixel coordinates. A 2B specialist I trained for $4 hits that same button 2.5x more reliably on ScreenSpot-v2.&lt;/p></description></item></channel></rss>