FORGE
Services About Blog
Google Stitch and the disappearing first draft
Strategy 20 March 2026 · 3 min read

Google Stitch and the disappearing first draft

Google Labs shipped Stitch. Prompt in, UI out. The crowd is calling it a designer killer. The initiated recognise it as something more specific — and more interesting.

RO
Robert Okoroafor
FORGE

Google Labs shipped Stitch this week. You describe a UI, optionally drop in a screenshot, and it returns HTML and CSS. Iterate in plain English. No Figma. No handoff. No sprint.

The crowd's reaction was immediate: this kills designers. That read is wrong — but not for the reason designers usually claim.

What Stitch actually is

Stitch is a first-draft machine. It takes the lowest-value, highest-friction part of UI work — the blank canvas — and removes it. You get something to react to. Something to correct. Something to throw away.

That is genuinely useful. The first draft has always been the worst use of an experienced designer's time.

What Stitch cannot do: it has no knowledge of your system. It does not know your token structure, your spacing rhythm, your colour contract, or why the primary button is exactly 52px tall. It produces something that looks like a UI. It does not produce something that is your UI.

The constraint was never the pixels

Every serious design system has the same problem: the knowledge that makes the output good lives in people's heads. The senior designer who knows why the grid breaks at 1024px. The developer who knows which component is deprecated. The PM who knows the conversion data behind that CTA.

Stitch externalises the execution. It does not externalise the knowledge. That gap is where the work actually is.

What this means

Tools like Stitch compress the gap between idea and first draft to near zero. That is a forcing function. If your value is in producing first drafts quickly, that value is gone. If your value is in the knowledge system that makes every draft better than the last — that value compounds.

Forge is a knowledge system that runs on AI. Every canon rule, every spacing decision, every gradient recipe is explicit and transferable. When a tool like Stitch appears, we point it at a brief and a canon and the output is already closer to right than anything it produces cold.

The question is not whether AI will replace designers. The question is: what is your canon?

What could actually change this

If Stitch learns to consume a design system — real tokens, real constraints, real component inventory — it becomes substantially more powerful. That is the product Google has not yet shipped. When it does, the first draft will not just be fast. It will be consistent.

Until then: the moat is the knowledge, not the pixels.

New project

Start a
project.

Discovery call
Loading calendar…