<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Evals on Hrushikesh Dokala</title><link>https://hrushikesh.dev/tags/evals/</link><description>Recent content in Evals on Hrushikesh Dokala</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Tue, 15 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://hrushikesh.dev/tags/evals/index.xml" rel="self" type="application/rss+xml"/><item><title>start ugly, write evals anyway.</title><link>https://hrushikesh.dev/notes/evals/</link><pubDate>Tue, 15 Sep 2026 00:00:00 +0000</pubDate><guid>https://hrushikesh.dev/notes/evals/</guid><description>&lt;p>











&lt;figure class="">

 &lt;div class="img-container" >
 &lt;img loading="lazy" alt="evals for your harness — harness &amp;#43; model → trials ×k → transcript and outcome → graders" src="https://hrushikesh.dev/images/evals-cover.jpg" >
 &lt;/div>

 
&lt;/figure>
&lt;/p>
&lt;p>TLDR; understanding the importance of evals, and how they can make yours and your agent life easier without spending exponential tokens and time to complete a task.&lt;/p>
&lt;p>pre -&lt;/p>
&lt;p>&lt;strong>Agent = Model + Harness.&lt;/strong>&lt;/p>
&lt;ul>
&lt;li>Model - the one thats brings the intelligence, capability to think and act.&lt;/li>
&lt;li>Harness - the one that lets the model act by giving tools, state, memory, envs, etc..&lt;/li>
&lt;/ul>
&lt;p>











&lt;figure class="">

 &lt;div class="img-container" >
 &lt;img loading="lazy" alt="an agent is a model plus a harness — model evals ask how good is the model itself, harness evals ask can the agent do the job with the tools and context you gave it" src="https://hrushikesh.dev/images/evals-model-harness.jpg" >
 &lt;/div>

 
&lt;/figure>
&lt;/p></description></item></channel></rss>