<?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: Yehuda Frankel</title>
    <description>The latest articles on DEV Community by Yehuda Frankel (@yehuda_frankel_0157cf0a82).</description>
    <link>https://dev.to/yehuda_frankel_0157cf0a82</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.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3842337%2Fa1a50f63-94f5-48e5-8f93-ee4200faf0d8.jpg</url>
      <title>DEV Community: Yehuda Frankel</title>
      <link>https://dev.to/yehuda_frankel_0157cf0a82</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/yehuda_frankel_0157cf0a82"/>
    <language>en</language>
    <item>
      <title>I got tired of re-explaining my codebase to Claude every session — so I built a memory system that actually stays accurate</title>
      <dc:creator>Yehuda Frankel</dc:creator>
      <pubDate>Wed, 25 Mar 2026 00:31:36 +0000</pubDate>
      <link>https://dev.to/yehuda_frankel_0157cf0a82/i-got-tired-of-re-explaining-my-codebase-to-claude-every-session-so-i-built-a-memory-system-that-53be</link>
      <guid>https://dev.to/yehuda_frankel_0157cf0a82/i-got-tired-of-re-explaining-my-codebase-to-claude-every-session-so-i-built-a-memory-system-that-53be</guid>
      <description>&lt;p&gt;very time I started a new Claude Code session I went through the same ritual.&lt;/p&gt;

&lt;p&gt;What's my stack? What files matter? What approach did we decide against last week?&lt;br&gt;
Claude had no idea. I'd spend 10-15 minutes just getting it back up to speed.&lt;br&gt;
Every. Single. Session.&lt;/p&gt;

&lt;p&gt;So I built Claude Recall.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it actually does
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Memory that stays accurate&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Most memory tools go stale — you document once, code moves on. Claude Recall &lt;br&gt;
runs a drift detector after every file edit. It compares live code against what &lt;br&gt;
Claude has in memory and flags undocumented changes automatically.&lt;/p&gt;

&lt;p&gt;First run on my project caught 21 undocumented functions Claude had no idea existed.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Skills that patch themselves&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Every built-in skill scores itself — did it need correction? Run /evolve every &lt;br&gt;
few sessions and it rewrites the failing steps using your actual failure data. &lt;br&gt;
The same mistake becomes architecturally impossible after /evolve runs.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Autonomous operation&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Skill chaining, self-healing on verify failure, session journal auto-captured &lt;br&gt;
on every response. Claude works through multi-step tasks without human &lt;br&gt;
checkpoints between each step.&lt;/p&gt;

&lt;h2&gt;
  
  
  Real numbers
&lt;/h2&gt;

&lt;p&gt;Tested across 119 real sessions on a production codebase — legacy Java backend, &lt;br&gt;
100+ functions, scheduler, email system. Not a demo project.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Sessions crashed mid-task — Start Session recovered every time&lt;/li&gt;
&lt;li&gt;Skills patched themselves — the same failure never happened twice
&lt;/li&gt;
&lt;li&gt;21 undocumented functions caught on first drift detection run&lt;/li&gt;
&lt;li&gt;velocity.md reached 30+ entries — estimates now reflect real track record&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  No API keys. No database. No background service.
&lt;/h2&gt;

&lt;p&gt;Plain markdown files + Python scripts. MIT license.&lt;/p&gt;

&lt;p&gt;GitHub: github.com/YehudaFrankel/claude-recall&lt;/p&gt;

</description>
      <category>claudecode</category>
      <category>ai</category>
      <category>productivity</category>
      <category>tooling</category>
    </item>
  </channel>
</rss>
