AI Eyes

Product Introduction

1. Definition: AI Eyes is a local-first, permission-first AI co-presence prototype. Technically, it is a browser-based application built with React that implements a granular, user-controlled media capture and session management layer for AI companions.
2. Core Value Proposition: It exists to establish a new paradigm for human-AI interaction where AI agents can share a user's digital context—screen, audio, and microphone—explicitly and transparently. Its primary value is in shifting from hidden monitoring to explicit, visible permission and user sovereignty over sensory data shared with AI.

Main Features

1. Granular Source Selection & Control: Users can select the exact screen or application window to share, independently toggle system audio capture and microphone access, and see a live preview. This is implemented using the WebRTC getDisplayMedia and getUserMedia APIs, providing a technical foundation for selective context sharing.
2. Named AI Companion Presence: The interface allows users to invite specific, named AI agents (like "Codi" and "Lyra" in the prototype) into a session. This feature moves beyond generic AI assistance towards personalized AI co-presence, establishing a clear mental model of who is "in the room."
3. Explicit Session Lifecycle Management: The prototype provides clear controls for pausing/resuming all sensing, switching interaction modes (Quiet, React, Explain), and completely clearing the session. This implements a crucial privacy-by-design principle, giving users continuous, visible agency over the AI's access.

Problems Solved

1. Pain Point: It addresses the opaque and all-or-nothing data access common in many AI assistant integrations. Users often cannot control what specific digital context an AI sees or hears, leading to privacy concerns and mistrust.
2. Target Audience: The primary personas are privacy-conscious power users, developers collaborating with AI pair programmers, and researchers prototyping ethical human-AI interaction models. It's also highly relevant for product managers defining the future of AI companion applications.
3. Use Cases: Essential for scenarios like: an engineer wanting an AI to see only a specific code editor window during pair programming; a designer sharing only a Figma canvas for feedback; or a user wanting to temporarily let an AI "listen in" on a meeting for note-taking, with the ability to instantly pause.

Unique Advantages

1. Differentiation: Unlike screen-sharing tools built for humans (e.g., Zoom) or AI systems that silently access system data, AI Eyes is architected from the ground up for consensual AI co-presence. It puts granular control and visibility front-and-center, unlike most AI APIs which abstract away the capture mechanism.
2. Key Innovation: Its core innovation is the conceptual and technical framework of a "permission-first capture layer." It separates the problem of secure, user-approved context gathering from the problem of AI inference, proving that local-first capture and session control are viable and necessary precursors to multimodal AI understanding.

Frequently Asked Questions (FAQ)

1. What is AI Eyes and how does it work? AI Eyes is a prototype application that lets you explicitly share your screen, system audio, and microphone with named AI companions. It works by using your browser's media APIs to capture only the sources you select, creating a controlled session that does not yet send data to an AI model, establishing the foundational permission layer.
2. Is AI Eyes spying on me or sending my data to the cloud? No. The current v0.1.0 release is a local-first prototype that handles capture and session control entirely within your browser. It explicitly does not connect to any external AI provider or cloud service, focusing solely on the permission and control interface.
3. What AI models does AI Eyes use with GPT-5.6? AI Eyes v0.1.0 does not perform live AI analysis. It was built with Codex (powered by GPT-5.6) as an engineering co-pilot to write the React and JavaScript code. The "live semantic understanding" using an AI model is stated as the next milestone, not a current feature.
4. Who should use the AI Eyes prototype? The prototype is currently most valuable for developers and technologists interested in the future of human-AI interaction, UI/UX researchers studying consent models for AI, and product teams looking to implement ethical, user-trusted AI companion features in their own applications.
5. How is AI Eyes different from a screen recorder? While both capture screen content, AI Eyes is structured around a session with invited AI agents, includes independent audio/mic controls, and features like pause/resume and mode switching specifically designed for interactive AI co-presence, not just passive recording.