https://blog.chughes.co/tags/tooling/Recent content in Tooling on Chris Hughes | Development JournalHugo -- 0.164.0en-usFri, 22 May 2026 00:00:00 +0000https://blog.chughes.co/posts/2026-05-22-memory-files-prime-llm-style/Fri, 22 May 2026 00:00:00 +0000https://blog.chughes.co/posts/2026-05-22-memory-files-prime-llm-style/<p><em>Written with Claude.</em></p> <p><img alt="Two printing plates and the matching prints: a dense substrate produces a dense response; a clean substrate produces a clean response. Same press, different plate." loading="lazy" src="https://blog.chughes.co/images/2026-05-22-memory-files-prime-llm-style/hero.svg"></p> <h2 id="every-session-loads-the-gotchas-file-and-the-gotchas-file-has-a-style">Every session loads the gotchas file, and the gotchas file has a style</h2> <p>I kept needing to invoke <code>/minto</code> or ELI7 to break down Claude&rsquo;s responses into plain language. Every session. Reliably. The root cause, I eventually understood, wasn&rsquo;t the model: CLAUDE.md and the memory-bank files don&rsquo;t just carry constraints, they prime the style the model writes back in. Dense comma-stacked gotcha bullets in, dense comma-stacked responses out.</p>