デザイナー4人で半年。「思想が染み込んだデザインシステム」がAI時代の揺るがない開発基盤に
haru · 20 October 2026
Read the originalFour designers on a product team at Goodpatch spent six months writing the product's design rules by hand, and AI-generated screens now match that product without manual fixes.
Goodpatch, a Japanese design company, published an account by haru, a user interface designer, of a product team whose AI-generated screens looked almost right but not like their product. Four designers spent six months building a design system by hand with the engineers, instead of adopting a ready-made component library. They then wrote the rules into DESIGN.md files and Agent Skills for AI tools to read.
- Each of about 50 components carries its Figma part, a profile of what it is, and rules for correct and incorrect use.
- The team wrote down what the product never does, and departed from common practice where the product needed it, such as a large palette of greys and greens.
- Figma and code match exactly, and about ten engineers replaced the old components during a refactoring period agreed with the product manager.
- Engineers report that AI-written screens are usable as generated, and design reviews no longer flag wrong padding or corner radii.
Design teams can treat written usage rules and refusals, not only components, as what AI tools need to produce screens that match the product. The article is in Japanese, and Goodpatch also sells design system services.
Give every component a profile and rules for correct and incorrect use, and write down what the product never does.
Keep components identical in Figma and code, then write their usage rules into DESIGN.md files and Agent Skills so that AI tools build screens from the same rules.
23 October 2026: teams get dependable work from AI agents by writing the rules first and keeping the checks with people and ordinary code that the agents cannot override.