AI Coding Episode 9: Creating UI Prototypes with Alibaba 秒悟 and Google Stitch

In the era of AI coding, the software development cycle—from idea to design, implementation, testing, and deployment—relies on clear, machine‑readable requirements. The article explains that requirements must be taught to AI agents using structured formats such as EARS (Easy Approach to Requirements Syntax) and SDD (Specification‑Driven Development) to avoid ambiguous or erroneous outputs.

Alibaba 秒悟 Platform

The author explores Alibaba’s newly released online coding platform 秒悟. After logging in, the interface resembles Baidu’s 秒哒, offering a visual canvas for building applications. Two skills are pre‑added: "Brainstorming" and "UI Design".

Example user prompt (translated):

I want to create a web application that starts with a beautiful landing page suitable for promotion.
The page should later help PMs and developers improve daily software development efficiency.
The first daily function for PMs is to input a requirement and receive an EARS‑rewritten description.
The second daily function is to input the raw requirement and get a Spec document driven by the SDD standard.

Using 秒悟’s "Refine" feature, the raw prompt is transformed into a structured description. The resulting UI mockup and the refined prompt are shown in the following screenshots.

After switching to Agent mode, 秒悟 automatically generates a functional UI based on the refined prompt, producing pages that match the intended EARS requirement‑optimization and SDD spec‑generation workflows.

Google Stitch

Google Stitch is introduced as an AI‑driven UI design canvas that creates complete interfaces from natural‑language descriptions. It is positioned not as a coding tool but as an "AI‑native design canvas" that supports vibe design, voice interaction, infinite canvas, MCP integration, and export to Figma.

When a simple natural‑language prompt is entered, Stitch quickly generates multiple web page designs. The author demonstrates creating a Chinese‑style page by adjusting the prompt, resulting in a polished UI prototype.

Stitch also offers an "Export to Interactive Prototype" feature, producing clickable prototypes that can be exported as Figma files, HTML, or React code. The exported React project is shown below.

The export includes a DESIGN.md file that documents the design standards of the generated UI. This markdown file can be read by AI agents to ensure consistent UI generation across projects.

Conclusion

The demonstration shows that AI coding dramatically shortens the gap between ideas and implementation. 秒悟 and Stitch illustrate a possible future where concepts become interactive prototypes within minutes, and requirements are precisely understood and executed by AI agents. This workflow empowers engineers, product managers, and creators to focus on creativity rather than repetitive manual tasks.