<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Agents on Git Push and Run</title><link>https://manuelfedele.github.io/tags/agents/</link><description>Recent content in Agents on Git Push and Run</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>© 2026 Manuel Fedele</copyright><lastBuildDate>Fri, 07 Aug 2026 10:00:00 +0200</lastBuildDate><atom:link href="https://manuelfedele.github.io/tags/agents/index.xml" rel="self" type="application/rss+xml"/><item><title>Engineering a Semi-Deterministic AI Dark Factory</title><link>https://manuelfedele.github.io/posts/from-requirements-agent-to-dark-factory/</link><pubDate>Fri, 07 Aug 2026 10:00:00 +0200</pubDate><guid>https://manuelfedele.github.io/posts/from-requirements-agent-to-dark-factory/</guid><description>&lt;div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl">
 What if a business requirement could inspect the system it is about, challenge its own ambiguities, become an executable specification, and then move through implementation, review, deployment, and acceptance testing without losing human control? This is how we built a semi-deterministic production line around probabilistic coding agents.
&lt;/div>

&lt;p>Most discussions about AI-assisted development start too late. They start with a coding agent and a prompt.&lt;/p>
&lt;p>The difficult part is not producing code. It is converting an imprecise business request into a requirement that is consistent with the existing system, bounded enough to implement, and precise enough to test. If that input is weak, adding more agents only automates ambiguity.&lt;/p></description></item></channel></rss>