<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Олександр</title>
    <description>The latest articles on DEV Community by Олександр (@hatabish).</description>
    <link>https://dev.to/hatabish</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4005960%2F45d8a0ed-41c7-455a-b8de-1b3753369164.jpg</url>
      <title>DEV Community: Олександр</title>
      <link>https://dev.to/hatabish</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/hatabish"/>
    <language>en</language>
    <item>
      <title>I built an LLM router that picks a different model per article type</title>
      <dc:creator>Олександр</dc:creator>
      <pubDate>Sun, 28 Jun 2026 01:37:26 +0000</pubDate>
      <link>https://dev.to/hatabish/i-built-an-llm-router-that-picks-a-different-model-per-article-type-48lf</link>
      <guid>https://dev.to/hatabish/i-built-an-llm-router-that-picks-a-different-model-per-article-type-48lf</guid>
      <description>&lt;p&gt;I've been quietly running a small network of niche content sites for a few weeks. Two are live on real domains:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://afixu.com" rel="noopener noreferrer"&gt;afixu.com&lt;/a&gt; - home improvement and DIY tool guides&lt;/li&gt;
&lt;li&gt;
&lt;a href="https://aceju.com" rel="noopener noreferrer"&gt;aceju.com&lt;/a&gt; - weekly AI tool awards, editorial format scored across 4 criteria&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The niche isn't the interesting part. The pipeline is.&lt;/p&gt;

&lt;h2&gt;
  
  
  The LLM router
&lt;/h2&gt;

&lt;p&gt;Instead of one model for everything, a router picks per article type:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;premium/commercial pieces go to a stronger model&lt;/li&gt;
&lt;li&gt;high-volume informational drafts go to a cheaper, faster one&lt;/li&gt;
&lt;li&gt;if a provider is rate-limited, it falls through a cascade so generation never stalls&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  The QA loop
&lt;/h2&gt;

&lt;p&gt;Every draft, before it ships, passes through:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;an anti-AI-tells cleanup (80+ regex replacements - "delve", "leverage", "robust", etc.)&lt;/li&gt;
&lt;li&gt;a rankability judge scoring depth, specificity, intent-match and trust on an 8-point scale&lt;/li&gt;
&lt;li&gt;a structure/schema check (FAQ block, JSON-LD, internal links)&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;A daily job re-scores everything published in the last 24h and auto-regenerates the failures. The whole thing is Next.js 16 + MDX, one git repo per site.&lt;/p&gt;

&lt;p&gt;The honest target: a $30k/mo run-rate in ~24 months across 30 mature sites, monetized through guest-link sales, not ads. Slow on purpose.&lt;/p&gt;

&lt;p&gt;Happy to go deeper on the router logic or the QA scoring if anyone's interested.&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>ai</category>
      <category>seo</category>
      <category>nextjs</category>
    </item>
  </channel>
</rss>
