TweetLike - Who Said That?

Inspiration

We wanted to mix pop culture and AI in a way that’s both fun and meaningful. Tweets are short, chaotic, and often iconic — so we turned that into a guessing game.

What it does

TweetLike is a web app that shows you funny or chaotic tweets, and you guess which celeb posted them. After the game, you find out some were AI-generated. You can also write your own tweet and see which celebrity you sound like using our ML model.

How we built it

We used Streamlit for the frontend, pandas for data handling, and scikit-learn to train a text classification model. Tweets were curated from the web and some were AI-generated. Game logic and styling were built to mimic Twitter’s vibe.

Challenges we ran into

We had to balance fun UI with functional ML. Handling avatars in HTML, tuning the model with limited data, and managing Streamlit re-renders were tricky.

Accomplishments that we're proud of

We built a working game with clean UI, a tweet classifier that gets ~60% accuracy, and an AI deception twist that got real reactions. Everything runs live, no errors.

What we learned

We learned how to train and use a model in production, integrate ML into a web app, and debug weird frontend edge cases under time pressure.

What's next for TweetLike - Who Said That?

We’d love to add multiplayer, track streaks, improve the model with more tweets, and explore GPT-based tweet generation for deeper fake vs. real challenges.

Built With

• eli5
• joblib
• machine-learning
• pandas
• python
• sklearn
• streamlit