<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Evaluation on emaat</title><link>https://emaat.dk/tags/evaluation/</link><description>Recent content in Evaluation on emaat</description><generator>Hugo</generator><language>da-DK</language><lastBuildDate>Wed, 04 Nov 2026 20:36:55 +0100</lastBuildDate><atom:link href="https://emaat.dk/tags/evaluation/index.xml" rel="self" type="application/rss+xml"/><item><title>Harness Design for Long Running Apps</title><link>https://emaat.dk/articles/harness-design-long-running-apps/</link><pubDate>Sun, 12 Apr 2026 00:00:00 +0000</pubDate><guid>https://emaat.dk/articles/harness-design-long-running-apps/</guid><description>&lt;p&gt;&lt;strong&gt;Article:&lt;/strong&gt; &lt;a href="https://www.anthropic.com/engineering/harness-design-long-running-apps"&gt;Harness design for long-running application development&lt;/a&gt; by Prithvi Rajasekaran&lt;/p&gt;&#10;&lt;h2 id="summary"&gt;Summary&lt;/h2&gt;&#10;&lt;p&gt;Rajasekaran tackles two problems: getting Claude to produce high-quality frontend designs, and getting it to build complete applications without human intervention. Naive long-running setups fail in two ways. Models lose coherence as the context fills — some show &amp;ldquo;context anxiety&amp;rdquo; and wrap up early — and agents asked to judge their own work confidently praise it regardless of quality. Inspired by the generator/discriminator split in GANs, he separates creation from evaluation: a standalone evaluator is still lenient out of the box, but it is far easier to tune an evaluator to be sceptical than to make a generator self-critical.&lt;/p&gt;</description></item></channel></rss>