<?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>Kv-Cache on Pavel Nasovich's Blog</title><link>https://forcewake.me/tags/kv-cache/</link><description>Recent content in Kv-Cache on Pavel Nasovich's Blog</description><generator>Hugo -- 0.157.0</generator><language>en-us</language><copyright>Copyright 2026</copyright><lastBuildDate>Sun, 05 Apr 2026 23:39:32 +0200</lastBuildDate><atom:link href="https://forcewake.me/tags/kv-cache/index.xml" rel="self" type="application/rss+xml"/><item><title>TurboQuant Under the Hood: Google's 3-Bit Attack on the LLM Memory Wall</title><link>https://forcewake.me/turboquant-kv-cache-compression/</link><pubDate>Thu, 26 Mar 2026 10:53:00 +0000</pubDate><guid>https://forcewake.me/turboquant-kv-cache-compression/</guid><description>A deep technical walkthrough of TurboQuant, PolarQuant, and QJL: how Google turns random rotations, optimal scalar quantization, and a 1-bit residual sketch into a practical attack on the KV-cache memory wall without pretending every model problem is solved.</description></item></channel></rss>