The Startup Ideas Podcast

My AI Design Workflow That Doesn't Ship Slop

107 snips
May 12, 2026
Meng To, designer, educator, and founder of Design+Code and Aura, breaks down design.md as portable design DNA for AI-made work. He walks through a live workflow for landing pages, motion, slides, and mobile mocks. They explore design drift, remixing with reusable skills, why taste sets work apart, and how a solo creator runs multiple products with agents.
Ask episode
AI Snips
Chapters
Transcript
Episode notes

Design MD Carries A Design Across Tools

  • Design.md acts like a portable blueprint for a design’s soul, carrying typography, color, spacing, and effects into any AI prompt.
  • Meng To compares HTML to the finished dish, design.md to the recipe, and skills to the ingredients.

Why One Shot Prompts Collapse On Page Two

  • One-shot prompts usually fail after the first screen because the design drifts into generic output on later pages.
  • Meng To says templates are too tied to Figma, Framer, or Webflow internals, while a foundational system gives agents flexibility without cloning a whole site.

Taste Is The Moat In Generic AI Design

  • Taste has become the real moat because baseline AI design quality is higher, but also far more generic.
  • Meng To treats design.md like design memory that moves between Lovable, Figma, Cursor, Codex, and local folders where agents can reuse workflow knowledge.
Get the Snipd Podcast app to discover more snips from this episode
Get the app