Why Copying a Rival's Design Pixel-by-Pixel Made a Better Product

A Kibu engineer makes the case for deliberate imitation as a working method, framed around Virgil Abloh’s “3% approach”: when redesigning the Air Force 1, Abloh let himself change only 3% of the original. The constraint is the point. To find which 3% to alter, you have to study the other 97% closely enough to understand every decision behind it. The author and a coworker applied this by rebuilding their marketing site from scratch, recreating Mintlify’s 2025 site pixel-by-pixel rather than starting from a blank page.

The payoff was less about the clone and more about what the reconstruction taught them. Tracing someone else’s hover effects, color runs, and layout alignment surfaced the reasoning—and the tradeoffs—behind each choice, which in turn sharpened their own instincts about where to diverge: a more minimal nav, team photos on CTA buttons, video over screenshots. They shipped in under a month of weekend work on Framer, then migrated to a full codebase in March 2026, betting that AI-assisted “vibecoding” would outpace a drag-and-drop builder.

The broader argument: originality is overrated as a goal. People get rewarded for identifying and solving problems efficiently, and someone smarter has usually already done both. The author now treats prior art—blogs, videos, AI-surfaced precedents—as a first step in any project, with the real work being to absorb the full original and then add a small, intentional layer of your own.