<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Design Notes on Elarity Systems</title><link>https://elaritysystems.com/es-archive/notes/</link><description>Recent content in Design Notes on Elarity Systems</description><generator>Hugo</generator><language>en-US</language><atom:link href="https://elaritysystems.com/es-archive/notes/index.xml" rel="self" type="application/rss+xml"/><item><title>UUIDs Are Teargas for AI</title><link>https://elaritysystems.com/es-archive/notes/uuids-are-teargas-for-ai/</link><pubDate>Thu, 01 Oct 2026 11:00:00 -0700</pubDate><guid>https://elaritysystems.com/es-archive/notes/uuids-are-teargas-for-ai/</guid><description>&lt;p&gt;An early version of my ES Archive memory system used UUIDs as identifiers for memories.&lt;/p&gt;&#10;&lt;p&gt;It seemed like the obvious choice. UUIDs are everywhere. They are robust, globally unique, wonderfully boring database identifiers.&lt;/p&gt;&#10;&lt;p&gt;And then something strange happened.&lt;/p&gt;&#10;&lt;p&gt;The AI became &lt;strong&gt;taciturn&lt;/strong&gt;.&lt;/p&gt;&#10;&lt;p&gt;The more memories I retrieved, the less communicative the responses seemed to become. This puzzled me because loading more actual text into the context usually had the opposite effect. More context produced richer, more expansive answers.&lt;/p&gt;</description></item></channel></rss>