About

The advent of powerful AI tools means that anyone can submit code faster than ever, even without really knowing what it does. Recently, there's been a significant influx in low-effort contributions to popular open-source projects. This creates a heavy burden on the maintainers, who don't have the time to review all the spam.

At Hack_NCState 2026, we built a solution in just 24 hours. When a contributor submits a pull request to a repository managed by Inquizitive, the app asks Gemini to generate a quiz using the pull request's diff and description. It posts a link to it in the pull request comments, and when the submitter completes it, a label is added to the pull request that indicates whether they passed.

Rather than just asking what was changed, the content of the quizzes is focused on the context surrounding the changes and their possible implications on security and functionality. This ensures that contributors understand the nature and impact of their changes before asking maintainers to review them. The quizzes aren't meant to be foolproof; rather, they serve as a way for maintainers to easily filter out low-effort contributions to save time reviewing pull requests.

Inquizitive is a full-stack web application written in TypeScript. On the frontend, we're using React and TailwindCSS, and on the backend, we're using Hono and Prisma. We're using a GitHub App to subscribe to new pull requests on users' repositories and the GitHub API for user authentication and interacting with pull request comments as a bot. Finally, we're using SQLite for the database and the Gemini API for AI quiz generation. Our project won MLH's Best Use of Gemini API category for our unique use of Gemini's structured output feature to solve a legitimate problem.