RAZE Debug Mentor

This post is my submission for DEV Education Track: Build Apps with Google AI Studio.

What I Built

I built RAZE — Debug Mentor, an AI-powered debugging mentor designed to help developers understand why their code is failing instead of simply giving them a corrected solution.

RAZE follows a structured debugging workflow:

Problem → Execution → Observation → Questions → Root Cause → Fix → Why It Works → Engineering Takeaway

Instead of immediately generating an answer, RAZE looks at the user's code, execution result, expected vs. actual behavior, and debugging context before guiding them toward the underlying issue.

One of the problems I specifically tested was a Python shallow-copy bug involving nested dictionaries. RAZE was able to identify that copying the outer list did not create independent copies of the nested objects, explain the shared-reference behavior, and recommend copy.deepcopy() as the appropriate fix.

The original RAZE project is built with Python + FastAPI. For this challenge, I imported the project into Google AI Studio, where AI Studio recreated the application as a modern TypeScript-based web app that I could run and iterate on directly in the AI Studio environment.

Demo

Live App / AI Studio Applet:
https://ai.studio/apps/9dd65d7a-41e2-460e-906c-380c386e5ba5

The demo shows the complete debugging flow: submitting a problem, running the code, examining the observed behavior, answering mentor questions, and receiving a root-cause explanation and fix.

My Experience

The most interesting part of this challenge was seeing how quickly Google AI Studio could turn an existing project into a runnable web application.

I started with an existing Python/FastAPI version of RAZE and imported the GitHub repository into Google AI Studio. AI Studio analyzed the project and recreated the application using TypeScript, allowing me to run and interact with it directly inside the AI Studio environment.

What I liked most was that I didn't have to start the application from scratch. I could bring an existing idea and codebase into the environment and then focus on getting the experience working and refining the product.

Building RAZE also taught me an important lesson about AI-assisted development:

AI can build quickly, but the developer still has to understand, test, verify, and improve what is being built.

I used AI heavily during development, but I also tested the actual debugging behavior with real programming problems, especially bugs involving mutable default arguments and shallow copies.

That process made RAZE more than just an AI chatbot that generates code. The goal is to make it feel like a calm senior developer sitting beside you, helping you understand the bug before fixing it.

Google AI Studio made experimenting with that idea significantly faster, especially when moving from an existing project into a runnable web experience.