<?xml version="1.0" encoding="UTF-8"?>


<feed xmlns="http://www.w3.org/2005/Atom">
    <id>https://www.julianwyngaard.com/tags/llm-knowledge-base/feed.xml</id>
    <title>Loosely Coupled Thoughts &#x2014; tagged llm-knowledge-base</title>
    <subtitle>A blog about software engineering, AI, and development.</subtitle>
    <link rel="self" type="application/atom+xml" href="https://www.julianwyngaard.com/tags/llm-knowledge-base/feed.xml"/>
    <link rel="alternate" type="text/html" href="https://www.julianwyngaard.com/tags/llm-knowledge-base"/>
    <updated>2026-04-12T20:51:56.111754+12:00</updated>
    <author>
        <name>Julian Wyngaard</name>
    </author>
    <entry>
        <id>urn:uuid:019d80c7-8de1-74aa-9a0c-a987f6edf671</id>
        <title>LLM Knowledge Bases</title>
        <link rel="alternate" type="text/html" href="https://www.julianwyngaard.com/2026/04/06/llm-knowledge-bases"/>
        <published>2026-04-06T20:51:56.111754+12:00</published>
        <updated>2026-04-12T20:51:56.111754+12:00</updated>
        <category term="ai" label="ai"/>
        <category term="llm-knowledge-base" label="llm-knowledge-base"/>
        <content type="html">&lt;p&gt;&lt;a href=&quot;https://x.com/karpathy/status/2039805659525644595&quot;&gt;LLM Knowledge Bases&lt;/a&gt;&lt;/p&gt;&lt;p&gt;Andrej Karpathy recently shared his approach to building LLM-managed knowledge bases, and it resonates with me.&lt;/p&gt;
&lt;p&gt;I&apos;ve never been great at organizing notes. I suspect most people are the same &#x2014; they want things organized, they just don&apos;t want to be the one doing it. Which makes this a perfect job for an LLM.&lt;/p&gt;
&lt;p&gt;The core idea:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&amp;amp;A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it&apos;s the domain of the LLM.&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;You provide raw information, the LLM organizes and makes it discoverable, and then it&apos;s accessible by you directly, and also made available to the LLM that you work with for looking up information, Q&amp;amp;A and outputting it in different formats.&lt;/p&gt;
&lt;p&gt;In a way it&apos;s similar to RAG but less complicated, and probably a bit more like the progressive disclosure approach with &lt;a href=&quot;https://agentskills.io&quot;&gt;agent skills&lt;/a&gt; where information is discoverable when needed in the context of the conversation but without blowing out the context with wiki information that&apos;s not relevant to the conversation.&lt;/p&gt;
&lt;p&gt;The real payoff comes with scale:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc.&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;I can see this extending further &#x2014; multiple wikis, all accessible to the LLM, with controlled access say by different agents. A personal wiki, a business wiki, and a general one for collected articles and research interests. Different agents, different scopes, same underlying approach. Or a single agent with access to all of them.&lt;/p&gt;
&lt;p&gt;This feels like something we are going to see more of as a new product or enhancement to existing products.&lt;/p&gt;</content>
    </entry>
</feed>
