<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>TyKolt</title>
    <link>https://tykolt.pages.dev</link>
    <atom:link href="https://tykolt.pages.dev/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Minimal but powerful.</description>
    <lastBuildDate>Sun, 10 May 2026 00:00:00 GMT</lastBuildDate>
    <item>
      <title>I spent months trying to stop LLM hallucinations. So I wrote a graph engine in Rust.</title>
      <link>https://tykolt.pages.dev/writing/llm-hallucinations-rust-graph-engine/</link>
      <guid isPermaLink="true">https://tykolt.pages.dev/writing/llm-hallucinations-rust-graph-engine/</guid>
      <pubDate>Mon, 30 Mar 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Why prompt engineering, RAG, and vector DBs weren't enough — and how Kremis became a deterministic verification layer instead of another generator.]]></description>
    </item>
    <item>
      <title>I built an MCP server for a knowledge graph. It doesn't call any LLM.</title>
      <link>https://tykolt.pages.dev/writing/mcp-server-no-llm/</link>
      <guid isPermaLink="true">https://tykolt.pages.dev/writing/mcp-server-no-llm/</guid>
      <pubDate>Sun, 10 May 2026 00:00:00 GMT</pubDate>
      <description><![CDATA[Most MCP servers assume an LLM in the pipeline. This one is a deterministic memory layer: grounding, not retrieval.]]></description>
    </item>
  </channel>
</rss>
