Living House: Adaptive UI and docs using in-browser inference (passo.uno)

An experimental application called Living House has been developed to explore the potential of adaptive user interfaces (UIs) powered by in-browser large language models (LLMs). Inspired by the dynamic environments in the video game Control, Living House reshapes its UI and documentation in real-time based on user interactions, such as clicks and hovering behaviors. The application uses a small LLM that operates locally through WebLLM, allowing for immediate modifications to the interface and documentation as user habits are analyzed, enhancing usability and relevance. This innovation is significant for the AI/ML community as it ushers in a new era of interactive design that adapts dynamically to user intent, striving to eliminate friction points in user experience. Living House’s underlying model logs various user signals to inform its adaptations, which are then validated before implementation to ensure stability. This experiment not only highlights the potential for AI to enhance user experience but also raises important questions about the implications of fluid UIs on terminology and documentation stability, suggesting a shared evolution of both interfaces and accompanying guides. As adaptive UIs become more prevalent, establishing trust and maintaining core terminology will be essential to navigate the challenges of intelligent systems.