In recent days, renowned security researcher Jane Manchun Wong has disclosed the latest research and development updates from design collaboration platform Figma, capturing industry attention. According to her revelations, Figma is developing an AI-powered app generator that will integrate various input methods, including text prompts, Figma design files, and image assets. The core functionality of this tool is powered by Anthropic's Claude Sonnet model.
The standout feature of this AI app generator lies in its cross-modal input integration capabilities. Users can describe app functionalities through natural language or directly import existing Figma design drafts and visual assets. The system will automatically generate interactive app prototypes based on the Claude Sonnet model. This design approach is expected to significantly lower the barriers to app development, enabling designers and product managers without technical backgrounds to complete the entire process from concept to implementation more efficiently.
Besides the app generator, Wong’s research also uncovered Figma's efforts in the website creation space. Her team is working on a new product called "Figma Sites." While specific features have not been fully revealed, it is speculated that this tool may follow the multi-modal input and AI generation logic seen in the app generator, offering a complete solution for visual design to code generation in web development.
The unveiling of these two products signifies Figma's transformation from a simple design collaboration tool to a full-stack digital product development platform. By integrating large-scale AI model capabilities, the company aims to create an end-to-end solution covering design, development, and deployment. This move reflects both a response to the current wave of generative AI technology and a strategic intention to push beyond the boundaries of traditional design tools. However, the actual performance of these products remains to be seen and will require further validation upon official release.
