<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
    <title>Aunttwister&#39;s Blog</title>
    <description>A homelab journal by Pavle Curcic — AI agents that run my Proxmox homelab and self-healing Slack bots. Benchmarks and self-hosting guides, with the numbers.</description>
    <link>https://blog.curci.cc/</link>
    <atom:link href="https://blog.curci.cc/rss" rel="self" type="application/rss+xml" />
    <pubDate>Wed, 04 Nov 2026 14:02:27 GMT</pubDate>
    <lastBuildDate>Wed, 04 Nov 2026 14:02:27 GMT</lastBuildDate>
    
    <item>
        <title>What Paperclip gets right, and why I still run simple-lcp</title>
        
        <description><![CDATA[Paperclip has 86k stars and I have a 304 MB container. I measured both, kept mine, and here is the honest case for a small control plane.]]></description>
        
        <link>https://blog.curci.cc/blog/what-paperclip-gets-right-simple-lcp</link>
        <guid>https://blog.curci.cc/blog/what-paperclip-gets-right-simple-lcp</guid>
        <pubDate>Sat, 26 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>I&#39;m betting local AI hardware appreciates faster than gold</title>
        
        <description><![CDATA[DRAM is up 200-400%, token prices fell 84%, and three memory makers control the shortage. Why I&#39;m betting local AI hardware beats gold over five years.]]></description>
        
        <link>https://blog.curci.cc/blog/local-ai-hardware-vs-gold</link>
        <guid>https://blog.curci.cc/blog/local-ai-hardware-vs-gold</guid>
        <pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>I ran an AI pentest on my homelab — and misread my own results</title>
        
        <description><![CDATA[Running an AI pentest with a local model against 38 homelab targets — 25 came back with findings, including a HIGH in my gateway that&#39;s now fixed.]]></description>
        
        <link>https://blog.curci.cc/blog/ai-pentest-homelab-local-model</link>
        <guid>https://blog.curci.cc/blog/ai-pentest-homelab-local-model</guid>
        <pubDate>Wed, 09 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>Qwen3.8-Flash-Next performance on my DGX Spark: every attempt</title>
        
        <description><![CDATA[Qwen3.8-Flash-Next on a DGX Spark: mmap fix, hybrid fp8, a 45 t/s int4 fork that wasn&#39;t deterministic, and the abliterated stack now in prod.]]></description>
        
        <link>https://blog.curci.cc/blog/qwen38-flash-zgx-optimization-saga</link>
        <guid>https://blog.curci.cc/blog/qwen38-flash-zgx-optimization-saga</guid>
        <pubDate>Mon, 07 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>Qwen3.8-Flash-Next on my DGX Spark: tuning decode speed</title>
        
        <description><![CDATA[Tuning Qwen3.8-Flash-Next on a single DGX Spark: PLE mmap fixes, hybrid fp8 MoE, batch scaling limits. Real measured numbers from my homelab.]]></description>
        
        <link>https://blog.curci.cc/blog/dgx-spark-qwen38-flash-tuning</link>
        <guid>https://blog.curci.cc/blog/dgx-spark-qwen38-flash-tuning</guid>
        <pubDate>Sat, 05 Sep 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>Intel iGPU Passthrough to LXC: Encode Benchmark</title>
        
        <description><![CDATA[Intel iGPU passthrough to LXC: a hardware video encoding benchmark with the config that worked and the real encode numbers.]]></description>
        
        <link>https://blog.curci.cc/blog/intel-igpu-passthrough-lxc-encode-benchmark</link>
        <guid>https://blog.curci.cc/blog/intel-igpu-passthrough-lxc-encode-benchmark</guid>
        <pubDate>Thu, 27 Aug 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>I Gave My Homelab Agent Long-Term Memory — Here&#39;s What Happened</title>
        
        <description><![CDATA[My AI agent couldn&#39;t remember anything across sessions. I plugged in Hindsight with embedded PostgreSQL, then spent weeks putting out fires it started.]]></description>
        
        <link>https://blog.curci.cc/blog/homelab-agent-long-term-memory-pgvector</link>
        <guid>https://blog.curci.cc/blog/homelab-agent-long-term-memory-pgvector</guid>
        <pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>I Reverse-Engineered My AC Because the App’s Scheduler Was Broken</title>
        
        <description><![CDATA[The Midea app&#39;s scheduler was broken for months. I reverse-engineered it, built a Flask API, and gave my homelab agent direct temperature control.]]></description>
        
        <link>https://blog.curci.cc/blog/reverse-engineering-midea-ac-scheduler</link>
        <guid>https://blog.curci.cc/blog/reverse-engineering-midea-ac-scheduler</guid>
        <pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>From Prototype to Product</title>
        
        <description><![CDATA[Building a homelab dashboard taught me the gap between a prototype that works and a product people can actually use. The hard parts were never technical.]]></description>
        
        <link>https://blog.curci.cc/blog/from-prototype-to-product</link>
        <guid>https://blog.curci.cc/blog/from-prototype-to-product</guid>
        <pubDate>Sat, 14 Mar 2026 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>How to Host Multiple Apps on a Single IP</title>
        
        <description><![CDATA[How I host multiple apps on a single IP: DNS split-brain, two reverse proxies, and a failed wildcard cert taught me the layout that works.]]></description>
        
        <link>https://blog.curci.cc/blog/how-to-host-multiple-apps-on-a-single-ip</link>
        <guid>https://blog.curci.cc/blog/how-to-host-multiple-apps-on-a-single-ip</guid>
        <pubDate>Sat, 25 Oct 2025 00:00:00 GMT</pubDate>
    </item>
    
    <item>
        <title>Nvidia P40 Total Guide and Lessons Learned</title>
        
        <description><![CDATA[Learn how to run an NVIDIA Tesla P40 on Linux and Proxmox with GPU passthrough, power fixes, and a 3D-printed cooling solution.]]></description>
        
        <link>https://blog.curci.cc/blog/nvidia-p40-total-guide-and-lessons-learned</link>
        <guid>https://blog.curci.cc/blog/nvidia-p40-total-guide-and-lessons-learned</guid>
        <pubDate>Wed, 22 Oct 2025 00:00:00 GMT</pubDate>
    </item>
    
</channel>
</rss>