LearnGraph

The description states that LearnGraph:

• Ingests Markdown notes and transforms them into a personal concept graph.
• Builds deterministic and GPT-5.6-enriched prerequisite relationships between concepts.
• Uses adaptive flashcards to assess mastery, with grades generated by GPT-5.6.
• Updates mastery scores using a weighted formula based on previous mastery and attempt signal (correctness, response time, error type).
• Exposes learner context through read-only MCP tools to AI agents.

The system includes four key runtime capabilities powered by GPT-5.6:

1. Structured concept extraction
2. Prerequisite-edge proposals
3. Adaptive flashcard generation
4. Evidence-grounded answer grading

It also exposes four read-only MCP tools:

1. get_learner_context
2. get_mastery_summary
3. find_gaps
4. get_concept

The backend is built with FastAPI, Pydantic, SQLAlchemy, and Supabase PostgreSQL; the frontend uses Next.js, React, TypeScript, Tailwind CSS, and react-force-graph-2d.

Confidence Low — this is a self-reported technical architecture and functionality, not independently verified.

The author states that LearnGraph seeks to support students who increasingly learn inside general-purpose AI tools rather than dedicated platforms. It positions itself as a “learning-memory layer” that enhances existing workflows.

It claims to distinguish between:

• "Claimed knowledge" (what the student has seen)
• "Demonstrated mastery" (what they actually understand)

This distinction is central to its positioning — it aims to improve AI tutoring by grounding explanations in what students have proven they know, not just what they’ve encountered.

Inference The project appears to be a response to current limitations in AI tutoring systems that lack robust personalization and feedback mechanisms.

The description does not clearly define the target customer or ideal customer profile (ICP). It implies usage by students learning within general-purpose AI tools, but no explicit segmentation or persona details are provided.

Confidence Not evidenced — no stated user types, demographics, or use cases beyond general student learning.

There is no evidence of a business model or pricing strategy in the description. The project appears to be a hackathon submission with no indication of monetization plans, customer acquisition strategies, or revenue streams.

Confidence Not evidenced — no mention of how the product would generate value or income.

The team built:

• A backend using FastAPI, Pydantic, SQLAlchemy, Supabase PostgreSQL
• A frontend with Next.js, React, TypeScript, Tailwind CSS, and react-force-graph-2d
• An ingestion pipeline combining deterministic Markdown signals (headings, wiki-links, tags) with optional GPT-5.6 enrichment
• A mastery engine using a weighted update formula:

New mastery = 0.65 × previous mastery + 0.35 × attempt signal

They used Codex for implementation and verification, and recorded substantial tasks against commits to maintain an auditable development history.

Inference The team has technical depth and appears to have implemented a functional prototype, though no production data or performance metrics are provided.

There is no evidence of traction, adoption, or maturity beyond the hackathon submission. No customers, users, revenue, ARR, headcount growth, or product usage data are mentioned.

Confidence Not evidenced — this is a prototype-level effort with no demonstrated market impact.

No competitive landscape or positioning relative to existing tools is described. The author does not reference competitors, nor do they explain how LearnGraph differs from other AI tutoring platforms or concept graphing systems.

Confidence Not evidenced — no competitive analysis or differentiation strategy provided.

• Unproven learning loop: The described mastery update mechanism and adaptive flashcard logic are untested in real-world conditions.
• Dependency on GPT-5.6: Heavy reliance on a single model raises risk if that model becomes unavailable or changes behavior.
• No user feedback or validation: No evidence of actual student interaction, testing, or iterative improvement.
• Limited team size (4 members): A small team may struggle to scale the product beyond prototype stage without additional resources.
• Uncertainty handling in prerequisites: While uncertainty metadata is included, it's unclear how this impacts real-time tutoring decisions.

Inference The project lacks validation and scalability assumptions. It remains a proof-of-concept rather than a viable commercial offering.

1. What specific data or metrics would indicate that the mastery model is working correctly?
2. How do you plan to validate that the system improves learning outcomes over time?
3. Are there any known edge cases where the concept graph fails to reflect true understanding?
4. Has the team tested the system with real students, and if so, what were the results?
5. What are your plans for scaling beyond the hackathon prototype?
6. How do you intend to monetize or commercialize this product?