Sustainable homework verification through crowd-sourced statistics
Stop wasting compute on duplicate AI queries — learn from how your peers approached the same problems.
Important
Prototype Status: This project is a somewhat working prototype and is currently still under development. No guarantees are made. This is far from production level. Collaboration is welcome!
Students are already using ChatGPT, Gemini, and other AI tools to verify their homework solutions. This is happening at scale—hundreds of students running identical queries for the same problems, generating massive amounts of redundant compute.
Each individual AI query consumes energy, uses GPU resources, and contributes to environmental impact. When 200 students ask an LLM "Is my answer to problem 5 correct?", we're wasting computational resources solving the same problem 200 times.
This app provides a sustainable alternative through crowd-sourced verification. Instead of everyone running separate AI queries, students can:
- ✓ Submit their solutions and see how they compare to peers
- ✓ Learn from community patterns and approaches
- ✓ Reduce computational waste by sharing verification data
- ✓ See statistics that aid learning, not just "correct answers"
This is harm reduction: accepting that AI verification is happening and making it more sustainable, transparent, and educational.
The application allows students to:
- Select or create tasks from their curriculum (organized by semester and course)
- Upload solutions to homework problems
- View statistics showing how their answers compare to other students
- Identify potential errors by seeing which questions have divergent answers
- Learn from patterns in how peers approached the same problems
Unlike ChatGPT where you can copy-paste blindly, this app requires genuine effort:
- Effort Required: You must submit your own answers before seeing any results. No freeloading.
- Quality Checks: Spam detection prevents gibberish submissions. Answers must be meaningful.
- Accuracy Gate: Requires ~10% similarity to reference solution to verify you actually tried.
- Duplicate Prevention: Fingerprinting prevents multiple submissions from the same person.
This tool promotes transparent peer learning rather than opaque AI dependency. When you see that 60% of your peers got a different answer, it prompts reflection and discussion—not just passive acceptance of "the correct answer."
Study groups and peer verification have always been part of education. This app digitizes that practice in a way that's:
- More sustainable than everyone running individual AI queries
- More educational than getting a black-box AI answer
- More transparent than secret ChatGPT usage
- Node.js 18+ and npm
- PostgreSQL (or use Docker)
-
Clone the repository
-
Install dependencies:
npm install
-
Set up your database:
cp .env.example .env # Edit .env with your database credentials -
Option A: Run PostgreSQL with Docker:
docker run -d --name task-verifier-db \ -e POSTGRES_PASSWORD=password \ -e POSTGRES_DB=task_verifier \ -p 5432:5432 \ postgres:16-alpine
-
Run database migrations:
npx prisma migrate dev
-
Seed the database with curriculum data:
npx tsx prisma/seed.ts
-
Start the development server:
npm run dev
-
Open http://localhost:3000 in your browser
- Framework: Next.js 16 (App Router)
- Database: PostgreSQL with Prisma ORM
- Styling: Tailwind CSS
- Language: TypeScript
This tool is designed for practice, self-study, and learning. Please use it ethically:
- → Check your understanding, don't blindly copy answers
- → Respect your institution's academic integrity policies
- → Use this to learn, not to shortcut learning
Note: This entire project was built using Antigravity, Google's AI coding environment based on VS Code. This was my first time trying out Antigravity, and this project is what came to mind as a test case. Kinda ironic to use AI to build a tool that reduces AI usage, but here we are 😄
Questions or concerns? This is an open conversation about AI in education.