<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Agent Memory on AIBussin — AI applications, systems and books</title><link>https://aibussin.com/tags/agent-memory/</link><description>Recent content in Agent Memory on AIBussin — AI applications, systems and books</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Sat, 08 Aug 2026 17:14:00 +0100</lastBuildDate><atom:link href="https://aibussin.com/tags/agent-memory/index.xml" rel="self" type="application/rss+xml"/><item><title>Memory and Selective Recall</title><link>https://aibussin.com/books/agents-from-first-principles/08-chapter/</link><pubDate>Sat, 08 Aug 2026 17:14:00 +0100</pubDate><guid>https://aibussin.com/books/agents-from-first-principles/08-chapter/</guid><description>&lt;p&gt;The capability boundary gave the agent a defined action space and a rule for which capabilities are eligible at each step. Runtime state gave it an explicit working representation of what this run has established so far and how it reached that point. Together they govern the current execution, but neither gives information from an earlier run a controlled way to influence this one. There is a family of failures that requires exactly that.&lt;/p&gt;</description></item><item><title>Agents From First Principles 07: AI Agent Forgets Previous Work? Add Working, Semantic and Episodic Memory</title><link>https://aibussin.com/post/agents-from-first-principles-07/</link><pubDate>Sat, 08 Aug 2026 17:14:00 +0100</pubDate><guid>https://aibussin.com/post/agents-from-first-principles-07/</guid><description>&lt;h1 id="ai-agent-forgets-previous-work-add-working-semantic-and-episodic-memory"&gt;AI Agent Forgets Previous Work? Add Working, Semantic and Episodic Memory&lt;/h1&gt;&#10;&lt;p&gt;An agent can use the right model, call the right tools, execute the right plan, and still behave as if nothing that happened five minutes ago matters.&lt;/p&gt;&#10;&lt;p&gt;You see the symptoms quickly:&lt;/p&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;it re-reads files it already inspected;&lt;/li&gt;&#10;&lt;li&gt;it repeats research it already completed;&lt;/li&gt;&#10;&lt;li&gt;it asks for information the user already supplied;&lt;/li&gt;&#10;&lt;li&gt;it forgets why a previous approach failed;&lt;/li&gt;&#10;&lt;li&gt;it loses decisions made earlier in a long task;&lt;/li&gt;&#10;&lt;li&gt;it treats every new run as if the system has never seen the problem before;&lt;/li&gt;&#10;&lt;li&gt;it retrieves an old answer and treats it as current truth;&lt;/li&gt;&#10;&lt;li&gt;it fills the prompt with so much history that the useful information is buried.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;&lt;p&gt;The usual response is:&lt;/p&gt;</description></item><item><title>Delta Memory: Cargo-Culting Human Memory with Search</title><link>https://aibussin.com/post/delta/</link><pubDate>Mon, 18 May 2026 11:14:15 +0100</pubDate><guid>https://aibussin.com/post/delta/</guid><description>&lt;blockquote&gt;&#10;&lt;p&gt;AI systems today have no idea why their own memory changes, that’s the problem we are trying to solve in this post.&lt;/p&gt;&#10;&lt;/blockquote&gt;&#10;&lt;h2 id="summary"&gt;Summary&lt;/h2&gt;&#10;&lt;p&gt;Most AI memory systems start from a practical place: retrieval. Retrieval is useful, scalable, and often the right tool for the job. But if we want systems that interact with humans in more human‑like ways, we need a different analogy, not storage, but &lt;strong&gt;thinking&lt;/strong&gt;.&lt;/p&gt;&#10;&lt;p&gt;Humans don’t store perfect records. We don’t retrieve exact text or replay video files. What we call “memory” is a shifting landscape of associations, impressions, weights, and patterns. When you recall something, you’re not pulling a file from disk, you’re running a &lt;strong&gt;search&lt;/strong&gt; across your internal world, shaped by everything you’ve lived through.&lt;/p&gt;</description></item></channel></rss>