<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Blog on Portfolio</title><link>https://luca-pelzer.github.io/blogs/</link><description>Recent content in Blog on Portfolio</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Tue, 16 Dec 2025 00:00:00 +0000</lastBuildDate><atom:link href="https://luca-pelzer.github.io/blogs/index.xml" rel="self" type="application/rss+xml"/><item><title>Building LearnTogether: Because Studying Alone is Boring</title><link>https://luca-pelzer.github.io/blogs/building-learn-together/</link><pubDate>Tue, 16 Dec 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/building-learn-together/</guid><description>&lt;h1 id="building-learntogether-because-studying-alone-is-boring">Building LearnTogether: Because Studying Alone is Boring&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I&amp;rsquo;m preparing for the German IT apprenticeship exam (Fachinformatiker IHK), got bored studying from PDFs, and built a multiplayer learning app instead. It has flashcards, quizzes, duels, races, and a realistic exam simulator. Productive procrastination at its finest. Try it at &lt;a href="https://learn.engels.wtf">learn.engels.wtf&lt;/a>.&lt;/p>
&lt;hr>
&lt;h2 id="the-problem-death-by-pdf">The Problem: Death by PDF&lt;/h2>
&lt;p>You know what&amp;rsquo;s incredibly boring? Studying for an exam by reading the same PDF for the 47th time.&lt;/p></description></item><item><title>My AI Team: How Multi-Agent Workflows Made Me Feel Like I Have Employees</title><link>https://luca-pelzer.github.io/blogs/multiagent-ai-workflow/</link><pubDate>Tue, 16 Dec 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/multiagent-ai-workflow/</guid><description>&lt;h1 id="my-ai-team-how-multi-agent-workflows-made-me-feel-like-i-have-employees">My AI Team: How Multi-Agent Workflows Made Me Feel Like I Have Employees&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I built a multi-agent AI setup using OpenCode with 11 specialized agents. Each one is an expert in their domain (code, storage, security, docs, etc.). Combined with MCPs (Model Context Protocol) and LSPs (Language Server Protocol), they can actually interact with my systems. It genuinely feels like delegating to a team instead of doing everything myself.&lt;/p></description></item><item><title>Building My Own Email Server with Stalwart: A Journey to 10/10 (Except Microsoft)</title><link>https://luca-pelzer.github.io/blogs/stalwart-email-server-guide/</link><pubDate>Thu, 11 Dec 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/stalwart-email-server-guide/</guid><description>&lt;h1 id="building-my-own-email-server-with-stalwart-a-journey-to-1010-except-microsoft">Building My Own Email Server with Stalwart: A Journey to 10/10 (Except Microsoft)&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I spent mass time setting up my own mail server using Stalwart in an LXC container on Proxmox. I got a perfect 10/10 score on mail-tester.com, all major providers accept my emails&amp;hellip; except Microsoft, who decided my Hetzner IP is guilty by association. Here&amp;rsquo;s the complete guide, including all the DNS records, firewall configs, and the ongoing Microsoft saga.&lt;/p></description></item><item><title>When Your Server Lags Only When You're Not Watching: Fixing IRQ Imbalance</title><link>https://luca-pelzer.github.io/blogs/fixing-network-lag-irq-imbalance/</link><pubDate>Thu, 11 Dec 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/fixing-network-lag-irq-imbalance/</guid><description>&lt;h1 id="when-your-server-lags-only-when-youre-not-watching-fixing-irq-imbalance">When Your Server Lags Only When You&amp;rsquo;re Not Watching: Fixing IRQ Imbalance&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: My Proxmox server had mysterious network lag that disappeared the moment I opened &lt;code>htop&lt;/code> to investigate. Turned out all network interrupts were being handled by a single CPU core. After enabling &lt;code>irqbalance&lt;/code>, tuning RPS, and adjusting network buffers, the lag vanished for good. Here&amp;rsquo;s how to diagnose and fix IRQ imbalance on your own server.&lt;/p>
&lt;hr>
&lt;h2 id="the-observer-effect-bug">The Observer Effect Bug&lt;/h2>
&lt;p>You know that feeling when your car makes a weird noise for weeks, but the moment you drive it to the mechanic, it runs perfectly? That was my Proxmox server last month.&lt;/p></description></item><item><title>Why I Built My Personal Site: More Than Just Another Tech Blog</title><link>https://luca-pelzer.github.io/blogs/why-i-built-my-personal-site/</link><pubDate>Wed, 10 Dec 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/why-i-built-my-personal-site/</guid><description>&lt;h1 id="why-i-built-my-personal-site-more-than-just-another-tech-blog">Why I Built My Personal Site: More Than Just Another Tech Blog&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I built my own blog using Hugo, running in an LXC container on Proxmox, with Flatnotes as my writing companion. It&amp;rsquo;s not just about sharing knowledge—it&amp;rsquo;s about building a personal knowledge base that actually sticks in my brain.&lt;/p>
&lt;hr>
&lt;h2 id="the-problem-wait-how-did-i-fix-that-again">The Problem: &amp;ldquo;Wait, How Did I Fix That Again?&amp;rdquo;&lt;/h2>
&lt;p>You know that feeling when you solve a tricky problem at 2 AM, feel like a genius, then three months later encounter the exact same issue and have absolutely no idea what you did?&lt;/p></description></item><item><title>From CLI to SDK: Making Sleepless-OpenCode's Internal Agent Truly Invisible</title><link>https://luca-pelzer.github.io/blogs/sleepless-sdk-migration/</link><pubDate>Sun, 12 Jan 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/sleepless-sdk-migration/</guid><description>&lt;p>I was building &lt;a href="https://github.com/Luca-Pelzer/sleepless-opencode">sleepless-opencode&lt;/a>, a 24/7 AI agent daemon that processes coding tasks in the background. Think of it as a task queue where you submit work via Discord, and an AI agent executes it while you sleep.&lt;/p>
&lt;p>The daemon uses a specialized agent called &lt;code>sleepless-executor&lt;/code> to run tasks. This agent is purely internal—it&amp;rsquo;s not meant for direct user interaction. Users should never select it manually from the agent picker.&lt;/p>
&lt;p>But there was a problem: &lt;strong>to work with the OpenCode CLI, the agent had to be configured as &lt;code>mode: primary&lt;/code>, which made it visible in the UI.&lt;/strong>&lt;/p></description></item><item><title>Agent Design Patterns: What Makes a Good AI Agent</title><link>https://luca-pelzer.github.io/blogs/agent-design-patterns/</link><pubDate>Sat, 11 Jan 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/agent-design-patterns/</guid><description>&lt;h1 id="agent-design-patterns-what-makes-a-good-ai-agent">Agent Design Patterns: What Makes a Good AI Agent&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I&amp;rsquo;ve built 28 agents across infrastructure, development, content, and research domains. Along the way, I discovered that the difference between a mediocre agent and a great one comes down to a few key patterns. This post breaks down what actually works.&lt;/p>
&lt;hr>
&lt;h2 id="why-this-matters">Why This Matters&lt;/h2>
&lt;p>You can give an AI a vague system prompt like &amp;ldquo;You are a helpful coding assistant&amp;rdquo; and it&amp;rsquo;ll work&amp;hellip; kind of. It&amp;rsquo;ll be inconsistent, sometimes overstep, sometimes under-deliver, and you&amp;rsquo;ll spend more time correcting it than it saves you.&lt;/p></description></item><item><title>The Inevitable Pivot: Why All My Projects Now Lead to AI</title><link>https://luca-pelzer.github.io/blogs/focusing-on-ai/</link><pubDate>Sat, 11 Jan 2025 00:00:00 +0000</pubDate><guid>https://luca-pelzer.github.io/blogs/focusing-on-ai/</guid><description>&lt;h1 id="the-inevitable-pivot-why-all-my-projects-now-lead-to-ai">The Inevitable Pivot: Why All My Projects Now Lead to AI&lt;/h1>
&lt;p>&lt;strong>TL;DR&lt;/strong>: I started this blog to document my homelab adventures. Somewhere along the way, I built 110 prompts and 28 specialized agents across infrastructure, development, content, and research. Every project I touch now seems to involve AI workflows. This isn&amp;rsquo;t a pivot—it&amp;rsquo;s just where everything naturally led.&lt;/p>
&lt;hr>
&lt;h2 id="the-realization">The Realization&lt;/h2>
&lt;p>You know that moment when you look up from your keyboard and realize you&amp;rsquo;ve been working on something completely different from what you intended?&lt;/p></description></item></channel></rss>