<?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>#AICost on Home</title>
    <link>https://yakinin.com/en/tags/%23aicost/</link>
    <description>Recent content in #AICost on Home</description>
    <generator>Hugo -- 0.148.2</generator>
    <language>en</language>
    <lastBuildDate>Wed, 13 Aug 2025 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://yakinin.com/en/tags/%23aicost/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Quality AI Isn&#39;t Free: The Real Lesson from the GPT-5 Launch</title>
      <link>https://yakinin.com/en/posts/20250813-openai-gpt-5-rollout-issues-and-psychosis/</link>
      <pubDate>Wed, 13 Aug 2025 00:00:00 +0000</pubDate>
      <guid>https://yakinin.com/en/posts/20250813-openai-gpt-5-rollout-issues-and-psychosis/</guid>
      <description>&lt;p&gt;My take on the recent discourse around GPT-5&amp;rsquo;s instability is that it&amp;rsquo;s less about a technical stumble and more about a classic release management problem, a challenge familiar to any large IT company.&lt;/p&gt;
&lt;p&gt;The more telling issue, however, is the reports that the GPT-5 model is performing worse than its predecessor. This isn&amp;rsquo;t a bug; it&amp;rsquo;s a feature of a business decision. It indicates a strategic push to make the models more cost-effective. When you&amp;rsquo;re serving 700 million users, the priority shifts from peak performance to scalable, affordable operations. The casualty in this equation is often quality.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
