&lt;?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>Incompetence Notwithstanding</title><link>https://incompetencenotwithstanding.com/</link><description>Notes on language models, the things built with them, and what breaks.</description><generator>Hugo</generator><language>en-us</language><managingEditor>Lucas Hadfield</managingEditor><webMaster>Lucas Hadfield</webMaster><copyright>© 2026–2026 Lucas Hadfield</copyright><lastBuildDate>Mon, 31 Aug 2026 21:00:00 +1000</lastBuildDate><atom:link href="https://incompetencenotwithstanding.com/" rel="self" type="application/rss+xml"/><item><title>The Unreasonable Effectiveness of Reading Your System Prompt</title><link>https://incompetencenotwithstanding.com/posts/unreasonable-effectiveness-of-reading-your-system-prompt/</link><guid isPermaLink="true">https://incompetencenotwithstanding.com/posts/unreasonable-effectiveness-of-reading-your-system-prompt/</guid><pubDate>Mon, 31 Aug 2026 21:00:00 +1000</pubDate><author>Lucas Hadfield</author><category>llms</category><category>agents</category><description>I bet you haven't read it.</description><content:encoded>&lt;![CDATA[&lt;p&gt;I bet you haven&amp;rsquo;t read your system prompt.&lt;br&gt;
Not skimmed it.&lt;br&gt;
Not had it summarised for you.&lt;br&gt;
Actually read it.&lt;br&gt;
The whole thing.&lt;br&gt;
Every token.&lt;br&gt;
As the LLM sees it.&lt;br&gt;
With jinja templating rendered, with AGENTS.md appended, with few-shot examples, with tool schemas. Everything.&lt;/p&gt;
&lt;p&gt;Try it. You&amp;rsquo;ll doubtless find dozens of little issues. Poorly explained concepts. Overly verbose language. Redundant information. Inconsistencies between prompt and tool specs.&lt;/p&gt;
&lt;p&gt;It&amp;rsquo;s wild when you consider how important the system prompt is, and how little effort it takes to read it, and yet no one &lt;em&gt;really&lt;/em&gt; knows what&amp;rsquo;s in it.&lt;/p&gt;
&lt;p&gt;It really isn&amp;rsquo;t even that hard. Back of the envelope: a pretty big system prompt, with few-shot examples, and a bunch of tools might be 10k tokens. 0.75 words per token means it&amp;rsquo;s about 7.5k words. Reading speed is about 250 words per minute. So half an hour. Half an hour for you to read every single word. That&amp;rsquo;s nothing in the grand scheme of things.&lt;/p&gt;
&lt;p&gt;Every time I have done this, I have come away with a much better, cleaner prompt, lower costs and a quality bump.&lt;/p&gt;
&lt;p&gt;Try it out. Get your agent to render your system prompt as it would appear at runtime. And then sit down and read it. All of it.&lt;/p&gt;
]]></content:encoded></item></channel></rss>