<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Alex Sun on Alex's Blog</title><link>https://alexsun98.github.io/</link><description>Recent content in Alex Sun on Alex's Blog</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 23 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://alexsun98.github.io/index.xml" rel="self" type="application/rss+xml"/><item><title>The new software tar pit and how AI made it stickier</title><link>https://alexsun98.github.io/series/mythical-man-month/01-the-tar-pit/</link><pubDate>Fri, 04 Sep 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/series/mythical-man-month/01-the-tar-pit/</guid><description>&lt;h2 id="is-the-tar-pit-better-or-worse-now"&gt;Is the tar pit better or worse now?&lt;/h2&gt;
&lt;p&gt;I keep coming back to Brooks&amp;rsquo;s opening picture of big software projects as prehistoric tar pits, where teams got stuck and sank deeper the more they struggled. He wrote that in 1975. Rereading it in 2026, I wanted to know whether the pit had got better or worse. It is the same pit at the same depth. What changed is the water on top of it. It is clearer than it has ever been, and a pit that looks shallow is a more dangerous pit.&lt;/p&gt;</description></item><item><title>Ten subagents in an hour is 1975 asking for ten developers in a month</title><link>https://alexsun98.github.io/series/mythical-man-month/02-the-mythical-man-month/</link><pubDate>Wed, 23 Sep 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/series/mythical-man-month/02-the-mythical-man-month/</guid><description>&lt;h2 id="the-first-thing-i-reach-for"&gt;The first thing I reach for&lt;/h2&gt;
&lt;p&gt;The last chapter was about mistaking speed for progress. This one is about the first thing I reach for when speed is not enough. The work is massive and behind, so I split it up. Brooks wrote a whole chapter on why that does not work, but clearly, the version of me that opens three parallel agent sessions at once has not read it carefully enough.&lt;/p&gt;</description></item><item><title>Preface</title><link>https://alexsun98.github.io/series/notes-from-the-stack/00-preface/</link><pubDate>Mon, 14 Sep 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/series/notes-from-the-stack/00-preface/</guid><description>&lt;p&gt;Before a humanoid robot leaves the line, it has to pass 77 inspections and 41 tests that simulate real operating conditions. 24 assembly steps to build it. 118 chances to be told no.&lt;/p&gt;
&lt;p&gt;That is reporting from March 2026 on the line in Robot Valley of Shenzhen. It puts the line at one robot every 30 minutes.&lt;/p&gt;
&lt;p&gt;What I saw was in March, a day with my family in the newly built Shenzhen Science and Technology Museum, not far from the Robot Valley, where working lines and static displays and video walls sit next to each other. The thing I could not put down was a production line that builds robots. A whole day was not enough, and nobody got bored, which I had not expected. The line in that reporting opened a week after that visit.&lt;/p&gt;</description></item><item><title>The Mythical Man-Month, is this outdated in the age of AI?</title><link>https://alexsun98.github.io/series/mythical-man-month/00-why-i-went-back/</link><pubDate>Mon, 24 Aug 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/series/mythical-man-month/00-why-i-went-back/</guid><description>&lt;h2 id="why-i-went-back-to-it"&gt;Why I went back to it&lt;/h2&gt;
&lt;p&gt;I still remember someone proposing to add people to a project that was already late. It is a reasonable instinct, the work is behind, so find more hands. Someone else stepped in and mentioned Brooks&amp;rsquo;s &lt;em&gt;The Mythical Man-Month&lt;/em&gt;. Nobody was added. The book won the argument. The project took a few more months anyway, and I cannot remember what we said it would take at the start.&lt;/p&gt;</description></item><item><title>Now</title><link>https://alexsun98.github.io/now/</link><pubDate>Fri, 24 Jul 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/now/</guid><description>&lt;p&gt;The cards above are the whole story, updated whenever the focus shifts.&lt;/p&gt;
&lt;p&gt;This is a &lt;a href="https://nownownow.com/about"&gt;now page&lt;/a&gt; that I referred to. If you have your own, I
would love to see it.&lt;/p&gt;</description></item><item><title>One-command blogging with Claude and MCP: from idea to deployed post</title><link>https://alexsun98.github.io/posts/claude-mcp-blog-automation/</link><pubDate>Fri, 10 Jul 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/posts/claude-mcp-blog-automation/</guid><description>&lt;blockquote&gt;
&lt;p&gt;Just a sample post for test, here so the post layout has something real to render.
Replace it with your own writing.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;Publishing a post on a static site is a chain of small chores: draft the
markdown, fix the front matter, commit, push, wait for the build, then click
around to make sure nothing broke. None of it is hard. All of it is friction.&lt;/p&gt;
&lt;p&gt;With Claude and a few MCP servers, the whole chain collapses into one prompt.
This post walks through the setup.&lt;/p&gt;</description></item><item><title>Cold starts, warm feelings: .NET APIs on AWS Lambda</title><link>https://alexsun98.github.io/posts/serverless-dotnet-lambda/</link><pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/posts/serverless-dotnet-lambda/</guid><description>&lt;blockquote&gt;
&lt;p&gt;Sample post, here to fill out the list page and tag filters. Replace it
with your own writing.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;The standard objection to .NET on Lambda is cold start time. It is a real
cost, but it is paid less often than people assume, and most of it can be
engineered away before you spend money on provisioned concurrency.&lt;/p&gt;
&lt;h2 id="where-cold-starts-actually-hurt"&gt;Where cold starts actually hurt&lt;/h2&gt;
&lt;p&gt;A cold start happens on the first request to a new execution environment.
For a steady-traffic API, that is a tiny fraction of requests. The places it
genuinely hurts:&lt;/p&gt;</description></item><item><title>About</title><link>https://alexsun98.github.io/about/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://alexsun98.github.io/about/</guid><description>&lt;h2 id="how-i-work"&gt;How I work&lt;/h2&gt;
&lt;p&gt;Boring and dependable beats clever. Simplicity is always the deliverable.&lt;/p&gt;
&lt;p&gt;I also believe self-evolving agents, the ones that improve their own tools and checks, will end up doing real work in real businesses, not as a demo. They are the things that let a small team deliver far more than its size suggests. I think that’s one of the things that pushed me to build Mirana as the very first version. The story is honestly just beginning.&lt;/p&gt;</description></item></channel></rss>