<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Recovery on Programmer.ie: Modern AI programming</title>
    <link>http://programmer.ie/tags/recovery/</link>
    <description>Recent content in Recovery on Programmer.ie: Modern AI programming</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Mon, 14 Sep 2026 05:00:22 +0100</lastBuildDate>
    <atom:link href="http://programmer.ie/tags/recovery/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Externalize Working Memory</title>
      <link>http://programmer.ie/books/applied-ai/16-chapter/</link>
      <pubDate>Mon, 14 Sep 2026 05:00:16 +0100</pubDate>
      <guid>http://programmer.ie/books/applied-ai/16-chapter/</guid>
      <description>&lt;p&gt;&lt;em&gt;Part 3 — Give Intelligence a Runtime&lt;/em&gt;&lt;/p&gt;&#xA;&lt;h2 id=&#34;the-step-that-didnt-finish&#34;&gt;The step that didn&amp;rsquo;t finish&lt;/h2&gt;&#xA;&lt;p&gt;A process was reviewing a paragraph. It had compiled its context, recorded the call&amp;rsquo;s manifest, and sent the request. While the provider was still working on it, the process died.&lt;/p&gt;&#xA;&lt;p&gt;The standard recovery is familiar: restart, find the step that didn&amp;rsquo;t finish, and run it again. Someone did exactly that in this chapter&amp;rsquo;s experiment, on a copy of the dead process&amp;rsquo;s files, and the step re-ran to completion.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Retries Are Side Effects Too</title>
      <link>http://programmer.ie/books/applied-ai/22-chapter/</link>
      <pubDate>Mon, 14 Sep 2026 05:00:22 +0100</pubDate>
      <guid>http://programmer.ie/books/applied-ai/22-chapter/</guid>
      <description>&lt;p&gt;&lt;em&gt;Part 4 — Make It Safe and Verifiable&lt;/em&gt;&lt;/p&gt;&#xA;&lt;h2 id=&#34;the-line-that-may-already-be-there&#34;&gt;The line that may already be there&lt;/h2&gt;&#xA;&lt;p&gt;&lt;strong&gt;Constructed scene.&lt;/strong&gt; An adapter appends one line to a file. The process dies before the completion is recorded. On restart the ledger shows no completed result — but the file has changed.&lt;/p&gt;&#xA;&lt;p&gt;Three responses are available, and two of them lie. Appending again may write the line twice. Writing a success record by hand manufactures evidence for an effect nobody observed. Saying “unknown” is the only honest answer, and it is also the least actionable one. The question that matters is narrower than “what do we do?” It is: &lt;strong&gt;what evidence would justify another append?&lt;/strong&gt;&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
