Stitch introduces design-system-first UI prototyping with prompt-to-code in early access

• Google DeepMind's early-access Stitch demo shows Stitch turning a plain-language prompt into a mobile or web UI, then iterating that concept into a clickable prototype and code.
• In the same design system walkthrough, Stitch appears to start by generating a design system first, locking in fonts, colors, and style before the screen design expands.
• The demo also shows a chat-style workflow where changing the "vibe" updates the design system across the project, while an existing system can be pulled in instead system controls.

The clearest takeaway from the early-access demo is that Stitch is being framed less as a mockup generator and more as an end-to-end UI drafting tool. VentureTwins' Stitch demo walks from a simple prompt for "Finn's Fudge" to a designed interface, then on to a clickable prototype and code, with the attached demo video showing the workflow in motion.

That matters for creative teams because the output is not just a single polished screen. The pitch in the thread is iterative: prompt a web or mobile UI, refine it conversationally, and keep the result usable as a prototype rather than a static concept image.

The most concrete product detail is that Stitch appears to create a design system at the start of each project. In VentureTwins' design system walkthrough, that system sets consistent fonts, colors, and styles first, which suggests Google is pushing coherence before speed.

The same post says creators can chat with the agent to change the overall vibe and have those changes propagate automatically, or import an existing design system instead system controls. That makes the early demo look aimed at fast exploration without fully abandoning brand rules or component consistency.