the Virtue Machine

About Page for the virtue machine

Inspiration

The Virtue Machine was inspired by the growing amount of hostile and unproductive language online. Instead of blocking or censoring speech, the project explores how AI can transform negative language into something constructive and human.

What it does

The app takes user-submitted text and rewrites it into a more positive, respectful version using an instruction-tuned language model, while preserving the original intent.

How we built it

We built a full-stack app using Vite on the frontend, Node.js and Express on the backend, PostgreSQL with Prisma for data modeling, and an AI layer powered by carefully designed instruction prompts.

Challenges we ran into

Designing prompts that changed tone without losing meaning was the biggest challenge. We also had to manage UI state complexity and handle real-world issues like file uploads and API coordination.

Accomplishments that we're proud of

We delivered an end-to-end AI transformation pipeline with clean architecture, extensible APIs, and a focused product vision beyond a simple demo.

What we learned

Small changes in prompt wording can drastically affect AI behavior. We also learned the importance of scope control and designing systems that are modular and easy to evolve.

What's next for the Virtue Machine

Next steps include user accounts, persistent posts and comments, multiple transformation styles, and better ways to evaluate tone improvement over time.