We’re introducing new ways to design in real time with Stitch.

Source: Google Date: 2026-05-19 URL: https://blog.google/innovation-and-ai/models-and-research/google-labs/stitch-updates/

Summary

Google Labs updated Stitch, its AI design agent, with real-time streaming output to canvas and mid-generation steering — users can redirect the agent before it finishes rather than waiting for a completed artifact. Inputs span text, voice, existing code, and design files. Exports integrate with Google Antigravity for backend wiring or publish directly via Netlify. The framing is “design as ongoing dialogue” rather than prompt-then-receive.

Implications

• Agentic engineering: Stitch’s mid-generation steering is a concrete instance of the broader pattern — agent outputs become steerable streams rather than atomic completions. The same UX problem (how do you interrupt an agent mid-task?) applies to code generation, document drafting, and any long-horizon task. Google is solving it at the design layer first because visual feedback makes the interaction legible; expect the pattern to migrate.
• AI ecosystem/power dynamics: Antigravity integration as an export target pulls Stitch into Google’s emerging agentic stack (Antigravity CLI + SDK announced at I/O 2026). Design-to-deployment as a single Google-owned pipeline is a competitive pressure on Figma/Vercel’s own AI-first tooling.