AI compressed the 15% of our website rebuild that was typing, not the other 85%
The article describes how the Bitnoise team rebuilt their website using Claude Code and the Figma MCP tool. They found that AI-assisted coding only accelerated the "typing" portion (about 15% of the work), while the majority of the rebuild effort involved architecture, branding, design, and decision-making—areas where human judgment remained essential.
Background
Bitnoise is a Polish software development agency. This article describes their experience using Claude Code (Anthropic's AI coding assistant) together with a Figma MCP (Model Context Protocol) integration to rebuild their company website. The main point is that the AI handled the straightforward "typing" parts (generating code from designs) but did not meaningfully contribute to the hard parts: architecture decisions, content strategy, UX design, and refinement. The title's "15%" refers to the portion of the work that was pure coding output, while the other 85% — all the thinking, planning, and iterating — remained human-driven. This reflects a common pattern in AI-assisted development: the tool accelerates the mechanical aspects but still requires human expertise for the high-value work.