Design Systems Are Now Infrastructure for AI
In today's newsletter, I've got a lot of great content for you, including:
- Meta Just Open-Sourced Its Own Design System
- Custom Figma skills, built from scratch
- Thorsten Ball defines what ownership means at work
- And 7 more
Most Interesting
Your Design System is What AI Learns From
Design systems stopped being a consistency nicety the moment AI started generating UI from your codebase. A weak or drifting system doesn't just cause inconsistency: AI will compound the problems across every future output.
Design + Development
Product Design is Dead; Long Live Product Design
The era of painting pretty pictures in Figma and lobbing them over a wall is ending, and the power is shifting to designers who work in the real medium (code) so what they design is exactly what ships.
CSS Easing Explained: Ease-in, Ease-out, Ease-in-out & Cubic-bezier
Carmen Ansio explains CSS easing by connecting familiar pen tool handles to cubic-bezier curves. The article clarifies ease-in, ease-out, and ease-in-out, shows how keyword easings map to exact handle positions, and offers practical rules of thumb for choosing the right easing for common UI motion patterns.
Meta open-sourced Astryx: a design system with 150+ accessible React components, theming tools, templates, and a CLI for building and customizing UI.
Tech + Innovation
An interactive tool for orienting yourself around AI tools and concepts. Content loads dynamically, so details are thin, but it looks built to help you find your footing in the AI landscape.
Getting Started with Custom Skills in Figma
Figma's Amy Lima and TJ Pitre walk through building custom skills from scratch.
A Field Guide to Fable: Finding Your Unknowns
Thariq from Anthropic argues that working effectively with Claude Fable means systematically discovering and clarifying your unknowns: before, during, and after implementation, so the "map" of your prompts matches the real "territory" of your code and constraints.
Work + Mindset
Design Engineering is Eating the Middle
Marcelo Chaman argues design engineers no longer just translate between teams. They own the whole path from intent to shipped product. As software gets easier to build, taste plus execution ability is what actually separates good work from fast work.
Tom Tunguz argues teams should design AI systems around routing before picking models. He breaks routing into three layers, classifier, router, and model selector, and explains how queueing non-urgent tasks lets 70-80% of traffic run on cheap local or async models.
Thorsten Ball sent his engineering team a message on what real ownership means: defining the actual problem, thinking through edge cases and failures, testing well, shipping to production, communicating changes, and following up after release.
Resources & Market Signals
Edition #120
Design Systems Meet AI, Process Evolves
Edition #144
