<?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>Physics on The Practical AI Builder</title>
    <link>https://aitechy226.github.io/tags/physics/</link>
    <description>Recent content in Physics on The Practical AI Builder</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Sun, 01 Nov 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://aitechy226.github.io/tags/physics/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Cost-Per-Token: The Physics Behind the Number</title>
      <link>https://aitechy226.github.io/posts/cost-per-token-physics/</link>
      <pubDate>Sun, 01 Nov 2026 00:00:00 +0000</pubDate>
      <guid>https://aitechy226.github.io/posts/cost-per-token-physics/</guid>
      <description>I built a cost-per-token (CPT) engine for an inference GPU recommender and rewrote it twice — first because the formula ignored the KV cache (the conversation history the GPU re-reads for every token), then because the UI described the same number three different ways. Here is how that changed the way I read a CPT number.</description>
    </item>
  </channel>
</rss>
