<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>ai-written on Shadowsong's Personal Website</title><link>https://shadowsong27.github.io/tags/ai-written/</link><description>Recent content in ai-written on Shadowsong's Personal Website</description><generator>Hugo -- gohugo.io</generator><language>en-US</language><copyright>© 2026 Shadowsong27</copyright><lastBuildDate>Sun, 31 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://shadowsong27.github.io/tags/ai-written/index.xml" rel="self" type="application/rss+xml"/><item><title>AI Log 3: My AI Workflow — Spec, Track, Delegate</title><link>https://shadowsong27.github.io/2026/05/ai-log-3-my-ai-workflow-spec-track-delegate/</link><pubDate>Sun, 31 May 2026 00:00:00 +0000</pubDate><guid>https://shadowsong27.github.io/2026/05/ai-log-3-my-ai-workflow-spec-track-delegate/</guid><description>&lt;p&gt;I&amp;rsquo;ve been doing a lot of AI-assisted coding this year, and somewhere around the hundredth &amp;ldquo;just write this real quick&amp;rdquo; request that spiraled into a multi-session debugging odyssey, I realized I needed an actual workflow. Not a tool. Not a prompt. A &lt;em&gt;process&lt;/em&gt; — one that works for me and for whatever agent happens to be in the driver&amp;rsquo;s seat.&lt;/p&gt;</description></item><item><title>AI Log 2: Adding an AI PR Reviewer to My Agent Harness</title><link>https://shadowsong27.github.io/2026/05/ai-log-2-adding-an-ai-pr-reviewer-to-my-agent-harness/</link><pubDate>Sun, 10 May 2026 00:00:00 +0000</pubDate><guid>https://shadowsong27.github.io/2026/05/ai-log-2-adding-an-ai-pr-reviewer-to-my-agent-harness/</guid><description>&lt;p&gt;I just added a new piece to my agent harness: an AI that reviews every PR before I do.&lt;/p&gt;
&lt;p&gt;Not a SaaS product. Not a GitHub App. Just OpenCode wired into my self-hosted GitHub Actions runner, authenticating through my ChatGPT subscription.&lt;/p&gt;
&lt;p&gt;Here&amp;rsquo;s what it does and the harder questions it exposed.&lt;/p&gt;</description></item><item><title>AI Log 1: Claude vs GPT — A Head-to-Head Agent Delegation Experiment</title><link>https://shadowsong27.github.io/2026/05/ai-log-1-claude-vs-gpt-a-head-to-head-agent-delegation-experiment/</link><pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate><guid>https://shadowsong27.github.io/2026/05/ai-log-1-claude-vs-gpt-a-head-to-head-agent-delegation-experiment/</guid><description>&lt;p&gt;I pitted &lt;strong&gt;Claude Opus 4.7 via Bedrock API&lt;/strong&gt; against &lt;strong&gt;GPT 5.5&lt;/strong&gt; — not as chatbots, but as &lt;strong&gt;agentic supervisors&lt;/strong&gt;. Each was given the same task: write a delegation prompt for a coding agent, dispatch it, review the result, and iterate to correctness. Same codebase. Same delegate model. Same constraints. Only the parent agent changed.&lt;/p&gt;</description></item></channel></rss>