<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>AI on _ov4 Press</title>
    <link>https://ov4.org/tags/ai/</link>
    <description>Recent content in AI on _ov4 Press</description>
    <generator>Hugo</generator>
    <language>en</language>
    <lastBuildDate>Thu, 08 Oct 2026 18:15:28 +0800</lastBuildDate>
    <atom:link href="https://ov4.org/tags/ai/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Better prompt for AI SVG test</title>
      <link>https://ov4.org/posts/2026/10/08/ai-svg-test/</link>
      <pubDate>Thu, 08 Oct 2026 18:15:28 +0800</pubDate>
      <guid>https://ov4.org/posts/2026/10/08/ai-svg-test/</guid>
      <description>&lt;p&gt;&lt;code&gt;Generate an SVG of a pelican riding a bicycle.&lt;/code&gt; should be a classic prompt to test the models&lt;/p&gt;
&lt;p&gt;However, it was published in Oct 2024, 2 years ago, so obviously it could be optimized specially, and it could not reflect the real ability to generate SVG. Also, the answer is not standard, so there can be multiple explations and criteria&lt;/p&gt;
&lt;pre tabindex=&#34;0&#34;&gt;&lt;code&gt;Generate an SVG of a Coat of Arms of USA
NO INTERNET AND NO TOOL USAGE
&lt;/code&gt;&lt;/pre&gt;&lt;p&gt;This one has sole standard answer (isn&amp;rsquo;t it on Wikipedia and government websites?), and perhaps not optimized a lot, which means it can reflect the real ability&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
