Claude Style Patch

A developer named Andrew Roxby published a free, open source configuration patch on GitHub designed to trim away annoying writing tics from Anthropic's Claude AI models. The project provides a text file that users can add to their settings, forcing the assistant to drop stiff phrasing, unnecessary colons, and overly compressed sentence structures.

AI models frequently adopt repetitive stylistic habits, like announcing a point before making it or framing sentences around awkward labels. Roxby discovered that asking the AI to write simply is rarely effective, but setting up strict rules that ban specific habits while offering direct rewrites produces much cleaner text. For writers and coders using Claude, the patch offers a quick fix for generating text that reads more like natural human prose.

While the instructions significantly boost text quality in daily use, the developer noted that the AI still tends to forget the rules during long, complex conversations. It remains an open question whether AI companies will refine their default training to make models sound less robotic out of the box, or if users will always need custom patches to maintain natural language.