<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Development on Derald's Workbench</title><link>https://derald.com/tags/development/</link><description>Recent content in Development on Derald's Workbench</description><generator>Hugo -- gohugo.io</generator><language>en-us</language><copyright>© 2026 Derald R. Grimwood Jr.</copyright><lastBuildDate>Thu, 09 Jul 2026 08:00:00 -0700</lastBuildDate><atom:link href="https://derald.com/tags/development/index.xml" rel="self" type="application/rss+xml"/><item><title>Using AI Without Surrendering the Architecture</title><link>https://derald.com/posts/ai-without-surrender/</link><pubDate>Thu, 09 Jul 2026 08:00:00 -0700</pubDate><guid>https://derald.com/posts/ai-without-surrender/</guid><description>&lt;p&gt;AI can write code quickly. That is useful, but speed is not the same as continuity.&lt;/p&gt;
&lt;p&gt;A long-running system contains decisions that may not be obvious from one file: naming rules, lifecycle contracts, compatibility requirements, test surfaces, abandoned approaches, and features that must not quietly disappear.&lt;/p&gt;</description><content>&lt;p&gt;AI can write code quickly. That is useful, but speed is not the same as continuity.&lt;/p&gt;
&lt;p&gt;A long-running system contains decisions that may not be obvious from one file: naming rules, lifecycle contracts, compatibility requirements, test surfaces, abandoned approaches, and features that must not quietly disappear.&lt;/p&gt;
&lt;p&gt;The person directing the work remains responsible for those decisions.&lt;/p&gt;
&lt;h2 id="where-ai-helps"&gt;Where AI helps&lt;/h2&gt;
&lt;p&gt;AI is effective at:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;scanning a large code surface&lt;/li&gt;
&lt;li&gt;drafting repetitive implementation&lt;/li&gt;
&lt;li&gt;comparing approaches&lt;/li&gt;
&lt;li&gt;expanding tests&lt;/li&gt;
&lt;li&gt;organizing documentation&lt;/li&gt;
&lt;li&gt;finding likely omissions&lt;/li&gt;
&lt;li&gt;translating between technical layers&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="where-direction-remains-necessary"&gt;Where direction remains necessary&lt;/h2&gt;
&lt;p&gt;AI needs firm boundaries when:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;a local simplification would remove existing behavior&lt;/li&gt;
&lt;li&gt;two components appear redundant but have different contracts&lt;/li&gt;
&lt;li&gt;a generated answer conflicts with runtime evidence&lt;/li&gt;
&lt;li&gt;a fashionable redesign would erase the project&amp;rsquo;s purpose&lt;/li&gt;
&lt;li&gt;uncertainty is being presented as fact&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="a-practical-relationship"&gt;A practical relationship&lt;/h2&gt;
&lt;p&gt;The useful model is not “AI builds the system.”&lt;/p&gt;
&lt;p&gt;It is:&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;The developer holds the architecture. AI increases the developer&amp;rsquo;s reach.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;That relationship can produce better work, provided the human keeps the long view and the machine remains accountable to evidence.&lt;/p&gt;</content></item></channel></rss>