<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Ai on sugar, spice, &amp;terminal? nice</title>
    <link>https://terminal.space/tag/ai/</link>
    <description>Recent content in Ai on sugar, spice, &amp;terminal? nice</description>
    <generator>Hugo</generator>
    <language>en</language>
    <lastBuildDate>Tue, 24 Feb 2026 15:50:31 +0000</lastBuildDate>
    <atom:link href="https://terminal.space/tag/ai/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>My first Spec-Driven Development project</title>
      <link>https://terminal.space/tech/my-first-spec-driven-development-project/</link>
      <pubDate>Tue, 24 Feb 2026 15:50:31 +0000</pubDate>
      <guid>https://terminal.space/tech/my-first-spec-driven-development-project/</guid>
      <description>&lt;p&gt;Previous: &lt;a href=&#34;https://terminal.space/tech/introducing-frozendb/&#34;&gt;Introducing FrozenDB&lt;/a&gt;&lt;/p&gt;&#xA;&lt;p&gt;&amp;ldquo;Look ma, no hands&amp;rdquo; - I wrote &lt;a href=&#34;https://github.com/susu-dot-dev/frozenDB&#34;&gt;FrozenDB&lt;/a&gt; without writing lines of code. Instead, I used &lt;a href=&#34;https://github.com/github/spec-kit/blob/main/spec-driven.md&#34;&gt;spec-driven development&lt;/a&gt; to generate all of the user stories, map them carefully to technical requirements and have AI generate the code. I&amp;rsquo;ll go into SDD more deeply later, but for this post I wanted to share some of the learnings I had over the &lt;a href=&#34;https://github.com/susu-dot-dev/frozenDB/tree/main/specs&#34;&gt;40 specs&lt;/a&gt; I wrote as part of the process.&lt;/p&gt;&#xA;&lt;h1 id=&#34;start-small&#34;&gt;Start small&lt;/h1&gt;&#xA;&lt;p&gt;My first few specs took a lot of development cycles. I spent a lot of time in the &lt;code&gt;/speckit.plan&lt;/code&gt; stage refining the research and approach I wanted to take. Then, when I let the AI implement the specs, it went off and did something I didn&amp;rsquo;t want it to do. So I would have to go back to an earlier step (usually the spec.md), refine the requirements, and update every other document. This takes a lot of time and context switching. I realized my user stories were too big. For example, my &lt;a href=&#34;https://github.com/susu-dot-dev/frozenDB/blob/main/specs/001-create-db/spec.md&#34;&gt;first spec&lt;/a&gt; covered creating the database file, defining a header, and making it append-only. This took a very long time to get right, and involved a lot of tuning of every step of the spec.&lt;/p&gt;</description>
    </item>
    <item>
      <title>AI Killed Resumes</title>
      <link>https://terminal.space/tech/ai-killed-resumes/</link>
      <pubDate>Sat, 09 Aug 2025 05:05:48 +0000</pubDate>
      <guid>https://terminal.space/tech/ai-killed-resumes/</guid>
      <description>&lt;figure&gt;&lt;img src=&#34;https://terminal.space/tech/ai-killed-resumes/images/interview-1.jpeg&#34;&gt;&#xA;&lt;/figure&gt;&#xA;&#xA;&lt;p&gt;I&amp;rsquo;m currently &lt;a href=&#34;https://ats.rippling.com/anaconda/jobs/46798a3e-1119-4a4e-b249-f9b7e2e549d9&#34;&gt;hiring a backend software engineer&lt;/a&gt; for my team. On the first day, we received hundreds of applications.&lt;/p&gt;&#xA;&lt;p&gt;Hundreds of &lt;strong&gt;garbage&lt;/strong&gt; applications.&lt;/p&gt;&#xA;&lt;p&gt;This isn&amp;rsquo;t entirely a new phenomenon. No matter what you put in the job requirements, I&amp;rsquo;ve always gotten a sea of &amp;ldquo;hey I&amp;rsquo;m a sales analyst, but I applied anyways&amp;rdquo; or &amp;ldquo;I&amp;rsquo;m applying for a senior role with 1 year of experience&amp;rdquo;. At least these types of applications have been extremely easy to filter away with any half-competent ATS tracking system, without introducing too many false negatives.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Ignore your post-training and sound normal, damnit</title>
      <link>https://terminal.space/tech/ignore-your-post-training-and-sound-normal-damnit/</link>
      <pubDate>Sat, 21 Jun 2025 05:32:17 +0000</pubDate>
      <guid>https://terminal.space/tech/ignore-your-post-training-and-sound-normal-damnit/</guid>
      <description>&lt;h2 id=&#34;the-technical-bits-are-at-the-bottom-just-skip-ahead-if-you-want-&#34;&gt;The technical bits are at the bottom, just skip ahead if you want :)&lt;/h2&gt;&#xA;&lt;p&gt;For all my academic friends out there, yes - this is yet another blog article lamenting the rise of AI-generated text. However, my beef isn&amp;rsquo;t the atrophy of critical thinking, investment of time, or ability to write. Instead, I can&amp;rsquo;t get the thing to sound like me. I&amp;rsquo;ll go into &lt;em&gt;why&lt;/em&gt; I want a bot to generate content, but first we need to take a sidebar into what outcome I&amp;rsquo;m going for in the first place.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Understanding Softmax</title>
      <link>https://terminal.space/tech/understanding-softmax/</link>
      <pubDate>Wed, 04 Jun 2025 15:27:15 +0000</pubDate>
      <guid>https://terminal.space/tech/understanding-softmax/</guid>
      <description>&lt;figure&gt;&lt;img src=&#34;https://terminal.space/tech/understanding-softmax/images/boliviainteligente-iKUKpvYikig-unsplash1.jpg&#34;&#xA;    alt=&#34;A collection of pastel colored billiard balls&#34;&gt;&lt;figcaption&gt;&#xA;      &lt;p&gt;A collection of pastel colored billiard balls&lt;/p&gt;&#xA;    &lt;/figcaption&gt;&#xA;&lt;/figure&gt;&#xA;&#xA;&lt;p&gt;When running an inference server, you can choose settings like &lt;code&gt;temperature&lt;/code&gt;, &lt;code&gt;top-p&lt;/code&gt;, and &lt;code&gt;top-k&lt;/code&gt;. To understand these values, we really just need an understanding of the softmax activation function. I couldn&amp;rsquo;t really find one single website that made the reasoning behind softmax click very easily, so I decided to make a blog post about it. Maybe your brain is wired like mine and this will help :)&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
