<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Michael Czeiszperger</title>
    <link>https://czei.org/</link>
    <description>Recent content on Michael Czeiszperger</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Wed, 22 Apr 2026 10:00:00 -0400</lastBuildDate>
    <atom:link href="https://czei.org/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Multi-LLM Spec-Driven Software Development</title>
      <link>https://czei.org/blog/multi-llm-spec-driven-development/</link>
      <pubDate>Wed, 22 Apr 2026 10:00:00 -0400</pubDate>
      <guid>https://czei.org/blog/multi-llm-spec-driven-development/</guid>
      <description>&lt;p&gt;Over the past year I have shipped around twenty major revisions of a 25-year-old, 550,000-line codebase using a workflow that seems obvious to me (and a growing number of professional software engineers) and foreign to many working programmers I talk to. Either they are still in the Cursor-style inner loop: small prompts, small edits, small completions, or trying to one-shot complex instructions. They are not running an agentic system at full capability, and they are paying for it in slow, flaky, one-model output. It&amp;rsquo;s impossible to quantify, but my unscientific impression is I&amp;rsquo;ve made more changes to this product in the past 12 months than previous teams of four could have achieved over several years.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Modernizing a 550K-Line Java Codebase with AI-Assisted Development</title>
      <link>https://czei.org/blog/modernizing-load-tester/</link>
      <pubDate>Tue, 31 Mar 2026 10:00:00 -0400</pubDate>
      <guid>https://czei.org/blog/modernizing-load-tester/</guid>
      <description>&lt;p&gt;&lt;a href=&#34;https://www.webperformance.com&#34;&gt;Web Performance Load Tester&lt;/a&gt; has been in continuous development for over 25 years. I designed the original architecture and wrote the first version. Over the years, a dozen developers have contributed to what is now 550,000 lines of hand-written Java. Longer than the entire Lord of the Rings trilogy. Today I&amp;rsquo;m the sole developer again, and for version 7.0, I modernized the platform and started moving the product to the cloud.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Building Agentic AI Systems for Web Performance Load Tester 7.0</title>
      <link>https://czei.org/blog/agentic-systems-load-tester/</link>
      <pubDate>Thu, 26 Mar 2026 10:00:00 -0400</pubDate>
      <guid>https://czei.org/blog/agentic-systems-load-tester/</guid>
      <description>&lt;p&gt;Configuring a web load test by hand is one of those tasks that&amp;rsquo;s just interesting enough to require expertise and just tedious enough to make you resent having it. After 25 years of tediously pouring through HTTP headers, I decided to teach an AI to do the work instead. Then I built a second system to handle the part nobody enjoys: staring at load test metrics for hours and condensing them into a client report that will almost certainly be read no further than the executive summary.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Building an AI Financial Analysis System for Copilot Money</title>
      <link>https://czei.org/blog/ai-financial-analysis-copilotmoney/</link>
      <pubDate>Thu, 09 Jan 2025 10:00:00 -0800</pubDate>
      <guid>https://czei.org/blog/ai-financial-analysis-copilotmoney/</guid>
      <description>&lt;p&gt;Is it cheaper for two people to go to the movies at &lt;a href=&#34;https://www.redcinemas.com&#34;&gt;Red Cinemas&lt;/a&gt; or &lt;a href=&#34;https://www.gtcinemas.com/locations/greensboro-ale-house-and-elite&#34;&gt;Golden Ticket&lt;/a&gt;?&lt;/p&gt;&#xA;&lt;p&gt;&lt;a href=&#34;https://www.copilot.money&#34;&gt;Copilot Money&lt;/a&gt; (which I strongly recommend for personal financial management) can tell you exactly what you spent at each theater, down to the penny. What it can&amp;rsquo;t do is group those transactions by visit, include the Fandango ticket purchases and parking fees that belong to the same outing, and compare the all-in cost per person across venues. Its budget-vs.-actuals only works for the current month. Custom analysis means exporting to a spreadsheet.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
