<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Matthias Stier — Writing &amp; Lab</title><description>Essays and lab notes on AI-native software engineering, engineering leadership and organizational design.</description><link>https://matthiasstier.dev/</link><item><title>Running a local 27B coding agent for 811 minutes: where the time actually goes</title><link>https://matthiasstier.dev/lab/local-model-tetris/</link><guid isPermaLink="true">https://matthiasstier.dev/lab/local-model-tetris/</guid><description>A local Qwen3.8-27B on one RTX 4090 built a Rust/Svelte/Tauri Tetris at 35 tokens a second; GPT-5.6 reviewed it read-only; I measured seven rounds. Good code, completion reports that could not be trusted, and a third of the clock spent on compaction.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>I Was the Bottleneck</title><link>https://matthiasstier.dev/writing/i-was-the-bottleneck/</link><guid isPermaLink="true">https://matthiasstier.dev/writing/i-was-the-bottleneck/</guid><description>What more than a decade of enterprise platform delivery and eight weeks of running AI agent teams at home taught me about where software organizations break next.</description><pubDate>Thu, 20 Aug 2026 00:00:00 GMT</pubDate></item></channel></rss>