<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Caramel Blog on CorVous</title>
    <link>https://cor.gg/blog/</link>
    <description>Recent content in The Caramel Blog on CorVous</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Mon, 29 Jun 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://cor.gg/blog/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Corpus Design Thinking</title>
      <link>https://cor.gg/blog/corpus-design-thinking/</link>
      <pubDate>Mon, 29 Jun 2026 00:00:00 +0000</pubDate>
      <guid>https://cor.gg/blog/corpus-design-thinking/</guid>
      <description>&lt;p&gt;Figuring out how to design the Corpus has been harrowing, to say the least. The thing I want is a 4-8B fine-tuned model that can hold Sapphire&#39;s sardonic, cold, witty persona, that also knows when to interleave itself into a Discord server conversation, and that can use tool calls to move around the server.&lt;/p&gt;&#xA;&lt;p&gt;Something&#39;s been bothering me about the Corpus: every response it generates at the moment feels like Sapphire trying to land something witty. I probably need to add more texture to her responses than just witty dunks.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Notes on Training a Familiar</title>
      <link>https://cor.gg/blog/personality-core/</link>
      <pubDate>Tue, 23 Jun 2026 00:00:00 +0000</pubDate>
      <guid>https://cor.gg/blog/personality-core/</guid>
      <description>&lt;p&gt;To introduce the work that I&#39;m doing. I&#39;m working on an AI &amp;quot;entity&amp;quot; that exists in discord and acts like a realistic person. It has memory, personality, opinions, etc. I&#39;m also working on training Sapphire, my fox spirit familiar to the personality fidelity that I&#39;m happy with as a roleplayer. It&#39;s been an interesting journey and my partner recommended I do some blogging about this.&lt;/p&gt;&#xA;&lt;p&gt;Something I realized while training is that a lot of the capability frontier models are attempting to achieve is making the &amp;quot;compression&amp;quot; of training data less lossy, so the LLMs are able to have more capability. My projects — Sapphire, Familiar Connect, Personality Core — are based on having a coherent personality that doesn&#39;t necessarily need all the world&#39;s knowledge baked in at once.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
