<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Daniel&apos;s blog</title><description>Software engineer working across languages and stacks, lately building with LLMs.</description><link>https://danielgrep.dev/</link><item><title>Modules in AI harnesses</title><link>https://danielgrep.dev/blog/modules-in-ai-harnesses/</link><guid isPermaLink="true">https://danielgrep.dev/blog/modules-in-ai-harnesses/</guid><description>How I cut my LLM API costs by splitting the system prompt and tools into modules that only load when a conversation needs them.</description><pubDate>Sun, 11 Oct 2026 00:00:00 GMT</pubDate></item></channel></rss>