<?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>Retrieval on Loop &amp; Retry</title>
    <link>https://loopandretry.github.io/tags/retrieval/</link>
    <description>Recent content in Retrieval on Loop &amp; Retry</description>
    <image>
      <title>Loop &amp; Retry</title>
      <url>https://loopandretry.github.io/images/og-default.png</url>
      <link>https://loopandretry.github.io/images/og-default.png</link>
    </image>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Wed, 05 Aug 2026 09:00:00 -0400</lastBuildDate>
    <atom:link href="https://loopandretry.github.io/tags/retrieval/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>The transcript is a log, not an index: retrieval for long-running agents</title>
      <link>https://loopandretry.github.io/posts/transcript-is-a-log-not-an-index/</link>
      <pubDate>Wed, 05 Aug 2026 09:00:00 -0400</pubDate>
      <guid>https://loopandretry.github.io/posts/transcript-is-a-log-not-an-index/</guid>
      <description>Eviction and summarization keep a long-running agent&amp;rsquo;s window small, but the summaries themselves still accumulate in the linear transcript — and on a run long enough, that accumulation becomes the new quadratic. Past a few hundred steps, the fix isn&amp;rsquo;t a better compaction policy inside the window. It&amp;rsquo;s moving history out of the window entirely and querying it like a database instead of replaying it like a log.</description>
    </item>
  </channel>
</rss>
