<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Blog — Victor Demin</title>
    <link>https://sg4.tech/blog/</link>
    <description>Long-form essays on engineering delivery systems, metrics, bottlenecks, and the operator practices behind predictable shipping.</description>
    <language>en</language>
    <lastBuildDate>Thu, 08 Oct 2026 00:00:00 GMT</lastBuildDate>
    <atom:link href="https://sg4.tech/blog/feed.xml" rel="self" type="application/rss+xml"/>
    <item>
      <title>Code entropy: how CI checks keep AI from piling up legacy</title>
      <link>https://sg4.tech/blog/code-entropy-ci-checks-ai-legacy/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/code-entropy-ci-checks-ai-legacy/</guid>
      <pubDate>Thu, 08 Oct 2026 00:00:00 GMT</pubDate>
      <description>Code entropy grows on its own, and AI coding agents speed it up. Seven levels of CI checks that stop legacy piling up, with tools for Python, PHP, and JS/TS.</description>
    </item>
    <item>
      <title>AI coding agents: give them the periphery, keep humans on the core</title>
      <link>https://sg4.tech/blog/ai-coding-agent-periphery-core/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/ai-coding-agent-periphery-core/</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 GMT</pubDate>
      <description>AI coding agents are brilliant at some work and dangerous at other work. What to delegate: give them the periphery, keep humans on the core.</description>
    </item>
    <item>
      <title>AI made starting work nearly free. Finishing is still expensive.</title>
      <link>https://sg4.tech/blog/ai-made-starting-free-finishing-expensive/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/ai-made-starting-free-finishing-expensive/</guid>
      <pubDate>Sat, 11 Jul 2026 00:00:00 GMT</pubDate>
      <description>AI-assisted teams start more work than ever and ship about the same. Little&apos;s Law applied to agents, and the case for one-thread-one-task and WIP limits.</description>
    </item>
    <item>
      <title>Vibecoded MVP stopped shipping? Manage your AI like a junior</title>
      <link>https://sg4.tech/blog/vibecoded-mvp-stopped-shipping/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/vibecoded-mvp-stopped-shipping/</guid>
      <pubDate>Thu, 11 Jun 2026 00:00:00 GMT</pubDate>
      <description>A vibecoded MVP stalls the way junior-built codebases always have. Manage AI coding agents with junior-developer practices — a real rescue case with numbers.</description>
    </item>
    <item>
      <title>Average lead time is lying to you — how to forecast delivery with percentiles</title>
      <link>https://sg4.tech/blog/forecast-delivery-with-percentiles/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/forecast-delivery-with-percentiles/</guid>
      <pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate>
      <description>Average lead time hides what your delivery system really does. Forecast with median, spread, and p85/p95 percentiles to commit to dates you can actually hit.</description>
    </item>
    <item>
      <title>How to diagnose where engineering delivery is actually broken</title>
      <link>https://sg4.tech/blog/diagnose-broken-engineering-delivery/</link>
      <guid isPermaLink="true">https://sg4.tech/blog/diagnose-broken-engineering-delivery/</guid>
      <pubDate>Fri, 22 May 2026 00:00:00 GMT</pubDate>
      <description>An engineering leader&apos;s metrics playbook: split T2M into Lead Time and Cycle Time, find the WIP bottleneck with Little&apos;s Law, and set WIP limits.</description>
    </item>
  </channel>
</rss>
