Developer Technique: Using HTML Output Files to Make AI Coding Agent Results Readable

Software developer Eivind Kjøsbakken published an article on Towards Data Science on September 10, 2026, describing a productivity technique he uses roughly 100 times per day when working with AI coding agents like Claude Code. He identified two core problems with standard terminal output: important information is scattered throughout a conversation with no visual hierarchy, and plain text is cognitively harder to process than formatted content. His solution is to instruct the AI agent to compile key results into an HTML file and open it automatically in a browser, enabling headers, bold text, highlights, and diagrams that reduce reading effort. He implements this permanently by editing a user-level Markdown configuration file so the instruction applies to every session without retyping. For complex reporting tasks, he also uses specialized skill sets such as a test-report template and an architecture-diagram tool called Archify, which generates SVG-based diagrams inside HTML files.