Design systems for AI agent usability
This audit gives you a concrete checklist for what agents actually need from a design system, which is a different bar than what humans need from documentation.
Reddit Proto how Reddit's teams built 700 production prototypes with AI and design systems
Reddit's pattern of hiding Git behind a design-system MCP and using real data instead of mocks is a reusable blueprint for getting a prototyping tool adopted across an entire org.
Design systems need performance audits like accessibility reviews
Treating performance as a release gate for shared components, the same way accessibility already is, would catch problems before they propagate to every team downstream.
Code reviewbecame a quality check; it should verify outcomes
Reframing code review as outcome verification rather than style policing changes what you ask reviewers to check once a chunk of the code is AI-generated.
Going nowhere fast when AI polish hides weak input
A sharp reminder that AI makes weak input data look confident, so the discipline of feeding it real interview notes and tickets instead of guesses matters more, not less.
White & Case rebrand swaps navy restraint for kinetic systems
A law firm rebrand built on kinetic type and a photography-to-data tool is a useful case for how far motion and strategy can push a famously conservative category.
Figma's agent automates motion and custom shaders
Worth watching to see whether agent-generated shaders and motion are actually usable starting points or just impressive demos before you rely on them in a real project.
Soviet poster archive shows constraint as a design principle
Nearly 5,000 archived Soviet and Eastern European posters are a fast way to study how extreme typographic constraint still carries a full message.
Design leaders who make are back because smaller teams demand it
The shift toward player-coach design leaders validates the structural bet of running a small practice that ships rather than scaling headcount.
Design systems for AI agent usability
This audit gives you a concrete checklist for what agents actually need from a design system, which is a different bar than what humans need from documentation.
Pick one component in the system you maintain and ask what an agent reading only its code and tokens (no docs, no Slack history) could infer about how to use it correctly.