Google's design.md: a design system in a markdown file
Podcast by Greg Isenberg × Meng To (designer, founder of Design+Code, creator of the products Aura / New Form / Dream Cut) on design.md — Google's open-source convention, equivalent to agents.md / skills.md / soul.md but for the design system (typography, colors, spacing, WebGL/Three.js animations, reveal rules).
By Greg Isenberg// Source youtube.com ↗/Reading 2 min/.md// Auto-verified translation
#design.md#Google#design system#design team in a file#Meng To#Greg Isenberg#Design+Code#Aura
Greg Isenberg hosts designer Meng To (founder of Design+Code, creator of Aura, New Form, Dream Cut) to break down design.md — Google's open-source convention that carries "the soul of the design system" in a markdown file, equivalent to agents.md / skills.md / soul.md but dedicated to the visual layer (typography, colors, spacing, WebGL / Three.js, reveal animation rules). The episode is a pedagogical manifesto on how to create jaw-dropping designs without being a designer.
Central diagnosis: design drift. One-shot tools (V0, Lovable, Framer) generate an impressive first screen then drift into generic pages on subsequent screens — no shared design memory between prompts. The design.md solution acts as a transferable blueprint: it carries typography, colors, spacing, WebGL, and animation rules across Lovable, Cursor, Codex, Claude Code, Google Stitch, Aura, V0, Figma. Meng formalizes a triad: design.md = recipe, HTML = finished dish, skills = ingredients. The skills (63 in New Form: lasers, skeuomorphic, 3D, copywriting, batch design) are copyable prompts that boost a design from "0 to 50 or 50 to 80" in one click.
Meng's complete 9-step workflow: Reference → design.md → Generate → Inspect → Systemize → Iterate (1000+ prompts) → Remix → Expand → Export. He claims up to 10,000 iterations per product, ~$500,000 spent on tokens, 4 products in parallel solo. Structuring distinction: iteration (incremental, 90% of the time) vs remix (category change, 10%).
On-screen demo chain: Aura (design.md gallery + remix + prompt), Google Stitch 3.1 (infinite canvas, "unfair they give it for free"), OpenClaude/Codex (multi-file MD reading across nested folders), Variant.com (creative remix phase), Replit slides / Hyperframes / Remotion (expansion into slides + motion).
Meta-thesis: taste is the only remaining moat. "If something looks like another thing, it reduces the value by 10× to 100×." The purple gradient — a wow 5-10 years ago — has become a flight signal. The baseline is high but generic: survival hinges on recency of adoption of models/tools/conventions and on judgment per minute (human micro-decisions on what the agent produces).
Greg closes on the idea of second brain design: capturing cross inspiration from the real and digital worlds to recall it when creating. Meng agrees: "don't let AI do everything. If you do, make sure it's within a workflow you already master."
Guest identity. "Mang" / "Ming" / "Mangto" in the transcript = Meng To, founder of Design+Code (designcode.io), designer recognized for 10+ years, self-taught (dropped out at 18, first company at 22). French-speaking background — Greg notes "you come from the same town as me. You speak French". Multi-startup founder: Aura, New Form, Dream Cut (upcoming Mac app), Design+Code rebuild from scratch.
Which design.md?. a Google convention, open-source, recently released (announced in a tweet with a 10-minute video). Same logic as agents.md / skills.md / soul.md but for design.
Skills. = "ingredients" (typography, lasers, skeuomorphic, 3D skills…)
The agent receives all three and produces a result consistent across all media (web, mobile, slides, motion).
Named problem: design drift. — "You start something, you're kind of satisfied, but then you get into the other stuff and it becomes completely different." Symptom: a brilliant one-shot followed by generic pages. Root cause: no shared design memory between prompts.
Solution: transferable design memory. — design.md travels between Lovable / Cursor / Codex / Claude Code / Stitch / Aura / V0 / Figma. Allows you to "commit it to memory" by telling the agent "remember this".
Meng To's complete workflow. 1. Reference (always start by copying — "every designer who starts drawing always copies first") 2. Load design.md + HTML as prompt attachments 3. Generate (first shot) 4. Inspect 5. Systemize (lock the system) 6. Iterate — "this is where I get to a thousand iterations" (Meng has gone up to 10,000 iterations on some products) 7. Remix when changing category 8. Expand to other media: slide, promo video, Instagram, intros, Hyperframes, Remotion 9. Final Export
Iteration vs Remix. iteration = satisfied, small increment (90% of Meng's time); remix = wanting a new product category (10%).
Free resources. Aura → "design.md" section with downloadable design systems (MD + HTML). Variant.com (variant.com) for the creative remix phase. Figma / Framer / Webflow communities provide templates but Meng criticizes: "closed behind techniques and rectangles and pixels of Figma, but how do you tell that to an agent?" — hence the value of design.md as an exportable blueprint.
Skills (New Form, 63 skills). copywriting ("improve your copy 100%"), lasers ("everyone clicks on it, I don't know why"), skeuomorphic (Steve Jobs' "lickable" buttons), 3D (Three.js), batch design, font smoothing, system font, body font, secondary color/button. Each skill is a copyable prompt, free, monetized only through tokens.
WebGL / Three.js. what takes a design "from 0 to 50 or 50 to 80 real fast". Lasers, animations, 3D objects placed on typography. Major differentiator vs generic purple-gradient sites.
Critique of the purple gradient. 5-10 years ago = wow; today = signal "you just run". Broader metaphor: "baseline is so high but also generic" — the challenge is to exceed the baseline, not reach it.
Redefined competitive mode."as people starting new ideas, the only moat we have is we can click a button and say we found this new model that literally cooked all the designers". Survival through recency of adoption of novelties (models, tools, conventions like design.md).
Claimed indie economics. Meng To spends ~$500,000 on tokens for his products, runs 1,000–10,000 iterations per product, manages 4 products in parallel solo (Aura, New Form, Dream Cut, Design+Code rebuild), keeps Design+Code's original team out of human loyalty more than operational necessity.
Steve Jobs quote (acknowledged as apocryphal)."if you want to go fast, go alone; if you want to go far, go in a team." Meng uses it to justify his dual solo-indie / team-loyalty posture.
"Judgment per minute". a new KPI for the agentic designer — less pixel-pushing, more micro-decisions per minute on what the agent produces. "AI is not intelligent enough to make smaller decisions yet, but eventually decisions will become bigger."
Skills-maxi (indirect quote). Meng refers to a previous Greg guest who defines themselves as "skills-maxi" — Meng confirms: "I don't put anything in my agents.md because I have my design.md per project and my skills per workflow. If you only change agents.md, it costs more tokens AND it's too general."Specialization of MD files by dimension (project / workflow / design / agent).
Greg Isenberg (CTA). free workshop "this Thursday at 12pm Eastern" (= Thursday May 7, 2026, the day after publication) — RSVP via show notes, on ideabrowser.com — covers startup categories in the AI age + relevant AI tools for business + live ideabrowser.com demo.
Connection to the corpus watch.
Strong convergence with the agents.md / skills.md fiches (Vincent Superpowers, Forrestchang Karpathy Skills, Vercel Gao agents.md, Lattice atoms/molecules/refiners). design.md is the domain-specific extension of this family to the design system.
Direct link with Cognitive Surrender / Comprehension Debt (Osmani 2026-05-05): Meng says "the first screen is not the hard part… you have no leverage to change it" — exactly Osmani's surrender transposed to design.
Link with Karpathy Software 3.0: design.md is a perfect example of "prompt is programming, context is the lever". Design becomes programmable via file.
Frizzo / Cherny / Curran. "team of one, solo holding company" — Meng embodies the "10× engineer = 10× designer" thesis championed by Cherny (100% generated code, 150 PR/day) applied to design.