A Tale of Vibe Coding by using Google´s Antigravity and Gemini 3.0 Pro

Google released Antigravity, an AI IDE Core. It has options to select among some models: Gemini 3 Pro, Claude Sonnet 4.5,  and GPT-OSS 120B.

For testing purposes, using the default Gemini 3.0 Pro, it was prompted:

<prompt> Generate an app to help teaching how to program quantum computers to undergraduate students by using IBM's Qiskit. </prompt> 

After a few minutes, Antigravity generated the Qiskit Learning App, a Next.js application designed to help students learn quantum computing.

Features Implemented

1. Challenge System

• Challenge List: A home page displaying available challenges.

• Challenge Data: A flexible data structure (lib/challenges.ts) that supports multiple challenges with starter code, solutions, and hints.

• Mock Execution: A simulated Qiskit execution engine (lib/mockExecutor.ts) that validates code logic and returns mock quantum circuit drawings.

2. Interactive Workspace

• Code Editor: Integrated Monaco Editor for a full-featured coding experience.

• Split View: Instructions and Assistant on the left, Code and Output on the right.

• Output Console: Displays execution status, success/failure messages, and ASCII-art circuit diagrams.

3. AI Assistant

• Chat Interface: A built-in chat component (AssistantChat) where students can ask for help.

• Mock Responses: The assistant currently provides predefined hints to guide students.

Verification Results

Build Verification

I ran npm run build to verify the application builds correctly for production.

• Result: Success (Exit code: 0)

• Output:

• ✓ Compiled successfully

• ✓ Linting and checking validity of types    

• ✓ Collecting page data    

• ✓ Generating static pages (4/4)       

• ✓ Finalizing page optimization

Code Quality

• Linting: Resolved all ESLint errors, including unescaped quotes and syntax issues in the mock executor.

• Type Safety: TypeScript checks passed during the build process.

Next Steps

• Real Execution: Replace the mock executor with a backend service (e.g., Python/Flask) running actual Qiskit code.

• More Content: Add more advanced challenges (Teleportation, Grover's Algorithm).

• User Progress: Implement local storage or a database to track completed challenges.