<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Graceful Degradation on Programmer.ie: Modern AI programming</title>
    <link>http://programmer.ie/tags/graceful-degradation/</link>
    <description>Recent content in Graceful Degradation on Programmer.ie: Modern AI programming</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Sun, 09 Aug 2026 11:26:00 +0100</lastBuildDate>
    <atom:link href="http://programmer.ie/tags/graceful-degradation/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Advanced Agents From First Principles 22: What Happens When One Dependency Starts Failing? Add Circuit Breakers, Bulkheads and Graceful Degradation</title>
      <link>http://programmer.ie/post/advanced-agents-from-first-principles-22/</link>
      <pubDate>Sun, 09 Aug 2026 11:26:00 +0100</pubDate>
      <guid>http://programmer.ie/post/advanced-agents-from-first-principles-22/</guid>
      <description>&lt;h1 id=&#34;what-happens-when-one-dependency-starts-failing&#34;&gt;What Happens When One Dependency Starts Failing?&lt;/h1&gt;&#xA;&lt;p&gt;An advanced agent platform can be working perfectly and still collapse.&lt;/p&gt;&#xA;&lt;p&gt;Not because the planner became confused.&lt;/p&gt;&#xA;&lt;p&gt;Not because the router chose the wrong model.&lt;/p&gt;&#xA;&lt;p&gt;Not because MCTS explored the wrong branch.&lt;/p&gt;&#xA;&lt;p&gt;Because one dependency became unhealthy.&lt;/p&gt;&#xA;&lt;p&gt;Maybe:&lt;/p&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;the frontier model starts returning 503s,&lt;/li&gt;&#xA;&lt;li&gt;the browser pool starts timing out,&lt;/li&gt;&#xA;&lt;li&gt;the embedding service slows from 80 ms to 8 seconds,&lt;/li&gt;&#xA;&lt;li&gt;the verifier database starts dropping connections,&lt;/li&gt;&#xA;&lt;li&gt;the search API starts rate limiting,&lt;/li&gt;&#xA;&lt;li&gt;a sandbox cluster stops accepting jobs,&lt;/li&gt;&#xA;&lt;li&gt;or an external provider is technically up but so slow that every request consumes a worker for minutes.&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;p&gt;Then the retries begin.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
