LearnLock

LearnLock — An AI that debugs your understanding

Inspiration

Modern learning is broken.

We consume endless tutorials — YouTube, GitHub, PDFs — but most of what we gain is familiarity, not understanding. Traditional quizzes only check surface recall. They can’t detect why you believe something.

We wanted to build a system that doesn’t test memory — it tests mental models.

That’s how LearnLock was born.

What LearnLock Does

LearnLock is an adversarial AI tutor that:

Ingests real learning sources (YouTube, GitHub, PDFs, articles)
Extracts ground-truth conceptual claims
Models what the user believes
Actively interrogates that belief until it collapses or becomes correct

Instead of asking “What is X?”, it asks:

“If you believe X works this way, what happens when Y?”

This forces the learner to expose hidden misunderstandings.

How We Built It

LearnLock is built around a Duel Engine, a multi-stage AI system:

Claim Extraction
LLM converts source material into falsifiable conceptual claims.
Belief Modeling
Each user response is turned into a one-sentence mental model.
Contradiction Detection
The belief is checked against claims to find missing mechanisms, wrong assumptions, or boundary errors.
Adversarial Interrogation
The AI generates a surgical follow-up question that targets the most critical flaw.
Spaced Repetition (SM-2)
Errors determine future review scheduling.