<?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>Coding-Agents on Bits-Entangled</title><link>https://stondo.github.io/tags/coding-agents/</link><description>Recent content in Coding-Agents on Bits-Entangled</description><generator>Hugo -- 0.162.1</generator><language>en-us</language><lastBuildDate>Mon, 05 Oct 2026 15:30:00 +0200</lastBuildDate><atom:link href="https://stondo.github.io/tags/coding-agents/index.xml" rel="self" type="application/rss+xml"/><item><title>Attention Fails Before the Guard Does: Measuring Context Management at 700K Tokens</title><link>https://stondo.github.io/posts/clm-context-pressure-measured/</link><pubDate>Mon, 05 Oct 2026 15:30:00 +0200</pubDate><guid>https://stondo.github.io/posts/clm-context-pressure-measured/</guid><description>Everyone has opinions about long-context agents; nobody had measurements of what the machinery actually does under sustained pressure on a local fleet. I built a harness that drives a real coding-agent session past a 524K-token model window, unattended, and measured five things: where promotion fires, what compaction destroys, when verbatim recall collapses, why the truncation guard I was testing never fired, and how the whole story reorders my assumptions about agent memory hygiene.</description></item></channel></rss>