Meta: Design standards as code

In 2026, AI coding assistants are letting more people across disciplines generate UI and copy, including people who’ve never been close to design. Designers were catching the same violations in pull requests: incorrect use of title case, jargon in help text, inconsistent terminology and error messages that didn’t match the rest of the product. Instead of insisting the design team own every review of the user interface, experience and copy, the team needed a scalable approach — encode design standards directly into the code base so that no matter who was vibe coding, the AI assistant supporting them would start acting more like a designer.

Objective

  • Stop fragmentation of user experience and inconsistent terminology use in vibe coded product solutions

Challenges

  • Designers were catching user experience and content design errors in pull requests and knew reactive review wouldn’t scale

  • No blueprint to follow — designers were building AI skills, but I wanted to take things a step further by building a rule that wouldn’t require a user to download or be aware of a skill

  • The rule had to be precise enough for an AI assistant to enforce without unintended effects in the codebase

My role

  • Formed hypothesis, project plan and testing plan for body of work — defined its importance and gained buy-in and collaborative support from senior and director-level product designers and engineers

  • Shipped responsible changes to codebases without breaking critical functionality

  • Wrote and shipped a rule that the AI coding assistant loads automatically whenever someone edits files in a target codebase

  • Collaborated on a unified standards approach with product design partners — they owned the UI patterns and components and I owned the content patterns and terminology

  • Created a team of agents to test the rule and make sure AI coding assistants would actually catch violations

Impact

  • 100% pass rate across evaluation cases

  • The rule shifted content design from reactive review to proactive enforcement at the moment of authorship, and it scaled to every person who had an AI assistant touch the codebase

Next

Meta: Community Notes waitlist onboarding