ARIA

Wearable display technology - such as Meta’s Ray-Ban smart glasses, is rapidly emerging as the next evolution of personal computing. Unlike phones, which require deliberate interaction, wearables offer an always-on, hands-free experience that integrates seamlessly into daily life. These devices perceive the world exactly as the user does, capturing both visual and auditory context continuously throughout the day.

However, this new level of immersion introduces a unique challenge: an AI assistant must now adapt to an ever-changing environment and be able to provide value in a wide range of unpredictable, context-specific tasks.

ASI:1 provides the ideal solution. Its agentic marketplace enables specialized autonomous agents to collaborate dynamically, handling diverse tasks each with expertise. Together, they form a flexible, intelligent ecosystem perfectly suited for the demands of wearable computing, where real-world context and constant adaptation are key.

As wearable display devices are still emerging and relatively difficult to access in Australia, this project was simulated on a mobile device. Throughout this section, the term “wearable device” refers to this simulated mobile implementation.

The project was developed using React Native with the Expo framework for the front-end interface, supported by a TypeScript backend. Information is gathered from the device’s camera and microphone, which capture visual and auditory inputs respectively. These inputs are then processed through OpenAI’s LLM APIs, which transcribe the audio and generate detailed descriptions of the visual context. Once this contextual data is available, the user’s commands are executed by ASI:1 agents.

The system first employs an ASI:1-fast agent to determine whether visual data or agentic reasoning is required. Since both processes are computationally intensive, this decision layer ensures that simple queries are handled quickly while avoiding unnecessary processing. For more complex tasks, the operation is delegated to ASI:1-fast-agentic, which autonomously coordinates within the ASI agent network to identify and engage the most suitable expert agent for execution. This backend was implemented in TypeScript.