<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Inference on Local First AI</title><link>https://localfirstai.eu/tags/inference/</link><description>Recent content in Inference on Local First AI</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Tue, 28 Apr 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://localfirstai.eu/tags/inference/index.xml" rel="self" type="application/rss+xml"/><item><title>The Memory Bandwidth Cliff: Lessons from an AI Runaway</title><link>https://localfirstai.eu/posts/2026-04-28-incident_003_alpha_post/</link><pubDate>Tue, 28 Apr 2026 00:00:00 +0000</pubDate><guid>https://localfirstai.eu/posts/2026-04-28-incident_003_alpha_post/</guid><description>An investigation into the super-quadratic prefill latency and memory bandwidth bottleneck observed on the Gemma 4 26B stack.</description></item></channel></rss>