Technical Guide: Implementing Knowable for Real-Time Academic Coaching

Practical Summary and Core Functionality

Knowable is a visually-grounded AI tutoring platform designed to assist students with SAT, ACT, and general high school coursework. Unlike traditional chatbots that rely solely on text input, Knowable utilizes the camera on a macOS device to observe a student's physical notebook or textbook in real-time. The system, powered by an AI tutor named Milo, provides personalized, Socratic-style guidance rather than direct answers. By analyzing handwritten work, diagrams, and equations, the platform offers hints that nudge students toward solving problems independently, ensuring that the learning process remains grounded in the student's own practice. The platform is designed to be non-intrusive, meaning it only responds when the student actively requests assistance, allowing for independent study sessions that are supported by AI when needed.

System Prerequisites and Hardware Requirements

To utilize Knowable, users must have a Mac running macOS Ventura or later. The platform leverages the built-in webcam or an iPhone connected via Continuity Camera to capture the workspace. No specialized external hardware is required. The system is designed to be accessible via the Mac App Store or directly through Safari. Because the platform relies on frontier vision models to interpret handwriting and printed prose, the quality of the camera feed is essential for accurate recognition of mathematical equations, scientific diagrams, and textual responses. Users should ensure their workspace is well-lit and that the camera is positioned to clearly capture the textbook or notebook page being studied.

Setup and Workflow Integration

The setup process is designed to be completed in minutes. Users place their notebook or textbook on their desk and position the camera to capture the page. Once the application is active, the system enters a "Desk View" mode. Interaction is controlled by the user through a push-to-talk mechanism. To initiate a query, the user presses and holds the opt+M key combination to speak or type a question. This action captures the current camera frame, allowing Milo to analyze the specific problem the student is currently working on. The system then provides a response via text, rendered LaTeX math, and text-to-speech (TTS) audio. This workflow ensures that the AI is always contextually aware of the student's current progress.

Documented Interaction and Feedback Loop

Knowable is explicitly engineered to avoid providing direct solutions. When a student asks for help, Milo analyzes the visual input and generates a guiding question. This approach is based on the research principle that students retain information more effectively when they discover the solution themselves. For example, if a student is dividing polynomials, the AI might ask about the properties of a remainder of zero rather than performing the division for the student. This interaction loop continues until the student reaches the correct conclusion, with the system tracking progress and competence across various topics. The platform provides detailed progress tracking, including study habits and competence metrics, which can be shared with parents or teachers if desired.

Data Privacy and Security Architecture

According to the official documentation at https://useknowable.ai, Knowable maintains strict data handling protocols. Camera frames are processed in memory during a session and are not written to disk or stored in a database. Chat transcripts and session metadata are encrypted in transit and at rest using AWS services. The platform contractually guarantees that user inputs and outputs are not used to train the underlying foundation models, nor are they shared with third-party model providers. This architecture ensures that student work remains private and is not utilized for advertising or external model training. The system is designed to be safe for educational environments, with clear policies regarding data retention and privacy.

Limitations and Usage Considerations

While Knowable supports a wide range of subjects including math, science, English, history, and languages, its effectiveness is contingent upon the clarity of the student's handwriting and the quality of the camera feed. The system is strictly limited to macOS environments. Furthermore, the platform is designed to be non-intrusive; it does not volunteer information or interrupt the student's workflow. If a student prefers to work independently, they simply do not trigger the opt+M command. Users should be aware that while the platform offers a free tier with monthly credits, advanced features and higher usage limits are gated behind the Knowable Plus subscription model. One-time credit packs are also available for users who require additional capacity beyond their monthly allotment.