<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Cognitive Modes on Matt Graham</title><link>/tags/cognitive-modes/</link><description>Recent content in Cognitive Modes on Matt Graham</description><generator>Hugo</generator><language>en-au</language><lastBuildDate>Wed, 18 Mar 2026 00:00:00 +0000</lastBuildDate><atom:link href="/tags/cognitive-modes/index.xml" rel="self" type="application/rss+xml"/><item><title>Your AI Output Is Fine. That's the Problem.</title><link>/blog/testing-the-theory/</link><pubDate>Wed, 18 Mar 2026 00:00:00 +0000</pubDate><guid>/blog/testing-the-theory/</guid><description>The convergent/divergent framework from my AI pipeline project was a theory built on one domain. I tested it against Anthropic&amp;rsquo;s knowledge-work-plugins across six domains, then against a professional reasoning benchmark. Pipeline won in five of the original six, then lost on the benchmark — and the failure sharpened the thesis into something more honest: pipelines earn their cost only when the correct output can&amp;rsquo;t be produced without seeing the specific input data.</description></item><item><title>Context Carries Cognitive Mode</title><link>/blog/context-carries-cognitive-mode/</link><pubDate>Thu, 05 Mar 2026 00:00:00 +0000</pubDate><guid>/blog/context-carries-cognitive-mode/</guid><description>I removed a quality-checking section from an AI prompt. The prompt got smaller. The output got deeper. Same model, same data, 70% of the context window unused. This wasn&amp;rsquo;t a capacity problem.</description></item><item><title>Context Carries Cognitive Mode — Evidence &amp; Methods</title><link>/blog/the-evidence-for-thinking-in-modes-evidence/</link><pubDate>Thu, 05 Mar 2026 00:00:00 +0000</pubDate><guid>/blog/the-evidence-for-thinking-in-modes-evidence/</guid><description>The companion evidence post to &amp;lsquo;Context Carries Cognitive Mode&amp;rsquo; — containing the full experiment designs, version comparisons, and claims tables. Each section is self-contained. Every claim is mapped to its evidence and its limitations.</description></item></channel></rss>