<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Software-Engineering on hippotion</title><link>https://blog.hippotion.com/tags/software-engineering/</link><description>Recent content in Software-Engineering on hippotion</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 01 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://blog.hippotion.com/tags/software-engineering/index.xml" rel="self" type="application/rss+xml"/><item><title>The Coding Skill That Actually Degrades Isn't Syntax — It's Judgment</title><link>https://blog.hippotion.com/posts/not-syntax-judgment/</link><pubDate>Wed, 01 Jul 2026 00:00:00 +0000</pubDate><guid>https://blog.hippotion.com/posts/not-syntax-judgment/</guid><description>Everyone worries that coding with an LLM makes them forget how to code. That&amp;rsquo;s the wrong thing to worry about. The skill that really degrades is judgment — knowing when to trust the model and when to take the keyboard back.</description><content:encoded><![CDATA[<p>Most developers I know treat manual coding as a muscle — skip it too long and it
degrades. Fair enough. But that&rsquo;s not the skill worth worrying about.</p>
<p>When you drive instead of riding a bike for a month, you don&rsquo;t forget how to
pedal. You adapt. The fear of &ldquo;forgetting syntax&rdquo; because you&rsquo;re not writing it
by hand is the same category of anxiety — and about as useful.</p>
<p>The skill that&rsquo;s actually degrading is quieter: <strong>knowing how to think with an
LLM.</strong> When to trust it. When to verify. When to throw the output away and
rethink the prompt instead of patching what it gave you. When the abstraction is
helping, versus hiding a design mistake you&rsquo;ll pay for in six months.</p>
<p>That&rsquo;s not a passive skill. It degrades if you don&rsquo;t use it.</p>
<p>What I try to stay sharp on:</p>
<ul>
<li><strong>Architecture.</strong> The model still needs someone to make the hard calls.</li>
<li><strong>Catching wrong-but-confident output</strong> before it lands in prod. The model is
never more convincing than when it&rsquo;s wrong.</li>
<li><strong>Writing prompts that constrain the problem space</strong>, not just describe it.</li>
<li><strong>Knowing when to take over the keyboard entirely</strong> — some problems are faster
to just do yourself.</li>
</ul>
<p>The real danger isn&rsquo;t forgetting syntax. It&rsquo;s outsourcing judgment without
noticing.</p>
<p>And if your workplace hasn&rsquo;t blessed LLMs yet — spin up something at home and
train yourself there. The skill gap won&rsquo;t wait for the policy.</p>
]]></content:encoded></item></channel></rss>