psst

Note: This analysis is based solely on the self-reported and unverified account provided by the authors. No independent verification or historical data exists for this project.

• The description states that psst is an "interruptible, spatially aware AI tutor."
• It presents math problems as narrated notebook lessons that unfold step-by-step at a human pace.
• Students can pause the lesson by tapping or pressing Space, annotate expressions, and ask questions.
• The system maps these annotations back to specific mathematical steps and provides concise answers in the margin with visual connectors.
• Voice interaction is supported via WebRTC and OpenAI Realtime API for low-latency speech transcription and response.
• A privacy-safe teacher dashboard aggregates interruption events without exposing student identities or raw data.

Inference: The product appears to be a full-stack web application built using Next.js, React, TypeScript, and integrates with OpenAI APIs. It includes custom logic for spatial interaction (e.g., pointer-to-expression mapping), real-time audio processing, and structured model outputs.

• The authors claim that psst aims to recreate the experience of sitting beside a brilliant friend while they work through a problem in a notebook.
• They position it as an alternative to traditional chatbot-style AI tutors, which they say fail to capture where understanding breaks down.
• The core innovation is treating interruptions not as failures but as signals for adaptive learning.
• The system is described as turning “confusion” into structured data at the level of individual reasoning steps.

Claim: Psst introduces a new interaction primitive for AI learning: point, interrupt, and continue from shared context.

Inference: This reflects an evolution from generic AI tutoring tools toward more granular feedback mechanisms.

• The description does not explicitly name target customers or define an ideal customer profile (ICP).
• It implies use cases in educational settings where students learn math, physics, chemistry, or other domains with sequential reasoning.
• The focus is on learners who benefit from visual and interactive explanations, particularly those who struggle with abstract concepts.

Not evidenced: No explicit mention of specific user segments, age groups, or institutional buyers.

• There is no evidence in the description of a business model or pricing strategy.
• The project is presented as a hackathon submission and lacks any indication of monetization plans, subscription tiers, or sales channels.

Not evidenced: No information on how psst would generate revenue or be sold.

• Built with Next.js, React, TypeScript, and integrates OpenAI APIs (Responses API, Realtime API).
• Uses SVG-based spatial scene rendering; KaTeX for math notation.
• Implements custom playback state machine, collision-aware placement engine, and schema validation via Zod.
• Voice interaction uses WebRTC and OpenAI speech models.
• Data stored in SQLite with idempotent write paths.
• Includes Vitest test suite covering playback, spatial targeting, session state, and API behavior.

Inference: The technical stack suggests a modern, full-stack web application designed for real-time interaction and rich UI experiences. Custom-built components imply strong engineering focus on user experience.

• The project was submitted to the OpenAI 2026 hackathon.
• No evidence of revenue, customers, or product adoption beyond the authors’ own account.
• The team size is listed as four members.
• There is no mention of prior versions, user testing, or iterative development outside the hackathon.

Not evidenced: No data on traction, usage metrics, or market validation.

• The description does not reference existing competitors or platforms in the AI tutoring space.
• It positions itself as a novel approach compared to traditional chatbots and static video lessons.
• The focus on spatial awareness, interruption handling, and structured feedback sets it apart from typical AI learning tools.

Not evidenced: No competitive landscape analysis or comparison with other products.

• The project is described as a hackathon submission with no evidence of commercial viability or product-market fit.
• Lack of revenue, customers, or traction raises questions about scalability and long-term sustainability.
• Heavy reliance on OpenAI APIs may pose risks related to cost, availability, and lock-in.
• Privacy features are emphasized, but there is no indication of compliance with educational data regulations (e.g., FERPA).

Inference: Without real-world usage or feedback, the risk of misalignment between intended functionality and actual user needs is high.

1. What specific educational outcomes or learning gains have you observed from early testing?
2. How do you plan to scale beyond a hackathon prototype?
3. Are there any existing partnerships with schools, edtech platforms, or educators?
4. What is your roadmap for monetization and go-to-market strategy?
5. Can you describe the process of collecting and analyzing confusion signals in more detail?
6. Have you considered how to handle edge cases like overlapping annotations or ambiguous inputs?