About the Project 📑
Introducing RobinHood - the ultimate hack for all your GPT-generated text woes. This bad boy uses cutting-edge GPT model to automatically detect and rephrase any GPT-generated text on Notion. But wait, there's more - you also have the power to tweak the intent, context, and styling of the write-up, making it virtually impossible for anyone to catch you using GPT on your assignments. Say goodbye to the fear of getting caught and hello to effortless, top-notch writing with RobinHood.
Inspiration 💡
The inspiration for this project comes from the recent explosion of GPT-generated text and the growing concern about the potential for it to be used to cheat on assignments. With the development of detection algorithms like GPT-Zero, there has been an increased call for regulation in the use of GPT-generated text. Our team believes that the best approach is to empower humans to collaborate with AI in a responsible and ethical way as simply banning GPT-generated text is not a long-term solution. That's why we created RobinHood, a tool that allows users to take advantage of the power of GPT while still maintaining their own unique voice and style. By providing a way to detect and rephrase GPT-generated text, we hope to promote a more authentic and transparent use of AI in writing.
What it does ⚙️
RobinHood is a web-extension that automatically detects GPT-generated text on Notion and rephrases it using the state-of-the-art GPT-3 model. It gives users the ability to change the intent, context, and styling of the write-up, making it completely unique and original.
How we built it 🏗️
- TypeScript
- React 18
- Chrome Extension Manifest Version 3
- Vite
- Tailwindcss
- SASS
- Notion
- FastAPI
- NLTK
- Huggingface 🤗
- GPT-2
- OpenAI API
Challenges we ran into 🧱
- Getting fast and accurate result for GPT generated text
- Difficulties of reading and showing data from notion web dom element
- Difficulty with hot reloading in development
Accomplishments that we're proud of 🦚
- Beautiful UI
- Functioning Prototype
- Empowering students
What we learned 🏫
- Interact with OpenAI API for completion task using prompt engineering techniques
- Optimise model for production usage using ONNX runtime
- Creating a web extension that parses the content from Notion before calling our backend for inference
- Creating/injecting a modal that appears when users hover over a text to perform text revisions
What's next for Robinhood: Catch Me If You Can ⌛
- Extension support for other WYSIWYG editors (e.g. Google Docs, Microsoft Word)
- Stay up-to-date for SOTA GPT text detection algorithm
- Enhance GPT completion features (e.g. Text Summarisation, Text Completion, Text Classification)
- SaaS support for general user with PAYG subscription model
- Dashboard for past history and history exporting
Built With
- chrome
- gpt
- huggingface
- nltk
- notion
- openai
- prompt-engineering
- react
- sass
- taiwindcss
- typescript
- vite

Log in or sign up for Devpost to join the conversation.