<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Agent-Infrastructure on Engineering Notes</title><link>https://notes.muthu.co/tags/agent-infrastructure/</link><description>Recent content in Agent-Infrastructure on Engineering Notes</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 25 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://notes.muthu.co/tags/agent-infrastructure/index.xml" rel="self" type="application/rss+xml"/><item><title>The Unsolved Engineering Problems of AI Agents</title><link>https://notes.muthu.co/2026/05/the-unsolved-engineering-problems-of-ai-agents/</link><pubDate>Mon, 25 May 2026 00:00:00 +0000</pubDate><guid>https://notes.muthu.co/2026/05/the-unsolved-engineering-problems-of-ai-agents/</guid><description>&lt;p>
 &lt;img src="https://notes.muthu.co/agents.png" alt="">

&lt;/p>
&lt;p>AI agents engineering is still in its &amp;ldquo;distributed systems in the 1990s&amp;rdquo; phase.&lt;/p>
&lt;p>Most demos work. Very few systems are reliable, scalable, explainable, secure, cheap, and autonomous at the same time.&lt;/p>
&lt;p>That gap matters. A demo can succeed once under ideal conditions. A production agent has to survive ambiguous inputs, partial failures, bad data, tool errors, cost constraints, security attacks, long-running state, and human accountability. It has to act like software while being powered by probabilistic reasoning.&lt;/p></description></item></channel></rss>