Traceback: A Code Reasoning Tutor
Learn to reason through code, not just fix it.
Debugging is the vehicle, not the destination. Traceback uses bugs as a concrete anchor for developing deeper understanding of how code actually works — the kind of understanding that transfers to new problems, not just fixes the current one. The tutor won't give away answers, instead it asks questions that guide you to reason through the problem yourself.
Grounded in learning science
Guided Discovery
Your lead your own acquisition of understanding. Rather than being told what's wrong, you're nudged toward the right direction through questions, arriving at the concept yourself.
Cognitive Task Analysis
The three-phase flow (Submit → Diagnose → Fix) models expert debugging behavior: observe, hypothesize, verify. Making those implicit expert steps explicit provides scaffolding for learners.
Metacognitive Scaffolding
The hypothesis revision stack and end-of-session retrospective make your evolving thinking visible as a concrete artifact, turning the reasoning process itself into something the you can reflect on and learn from.
How a session works
Paste your code and describe what you observe — what's the symptom?
Form a hypothesis about the cause and refine it through Socratic dialogue.
Apply your fix once you understand the root cause, with the tutor available for questions.
A note on AI: Traceback is powered by Claude, an AI made by Anthropic. AI can make mistakes — this tutor is designed to support your thinking, not replace it.