<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Opinion on Afraz Ahmed</title><link>https://afraz.dev/tags/opinion/</link><description>Recent content in Opinion on Afraz Ahmed</description><generator>Hugo</generator><language>en-us</language><copyright>© Afraz Ahmed</copyright><lastBuildDate>Wed, 13 May 2026 12:00:00 +0500</lastBuildDate><atom:link href="https://afraz.dev/tags/opinion/index.xml" rel="self" type="application/rss+xml"/><item><title>Building AI agents for production: the four walls</title><link>https://afraz.dev/blog/four-walls-of-agentic-ai/</link><pubDate>Wed, 13 May 2026 12:00:00 +0500</pubDate><guid>https://afraz.dev/blog/four-walls-of-agentic-ai/</guid><description>&lt;p>I build AI agents that run in production: a server diagnostics agent with SSH access, a conversation QA evaluator, and a signup verification system that collects threat intelligence. They handle real data every day.&lt;/p>
&lt;p>The longer I do this, the more I notice a pattern. The problems have nothing to do with model capability. The models are great and getting better every quarter. My problems are all in the space between the model and the real world. And I do not think they go away with GPT-6 or Gemini 4 or whatever comes next.&lt;/p></description></item></channel></rss>