Duct Tape AI

• 

  Frontend

Inspiration

I’ve been in situations where my code just refused to run, and I was too tired to debug properly. That’s when it hit me what if there was an AI that didn’t try to fix your code perfectly, but just patched it enough to compile or run? That’s where Duct Tape AI was born a “chaotic good” tool that patches broken code using random but working snippets.

What it does

Duct Tape AI takes broken code as input and tries to “fix” it but instead of making an ideal correction, it applies duct-tape fixes that make it run or compile somehow. It’s not elegant, but it works… most of the time. It’s a fun blend of humor and utility great for hackathons where creativity meets chaos.

How we built it

I used a React frontend where users can paste or upload broken code. The backend, built with FastAPI, handles the requests and communicates with an AI model (either via OpenAI API or HuggingFace). The model analyzes the input and returns a “patched” version of the code often inspired by random functional snippets stitched together. It’s then displayed back to the user as a quick “fix” suggestion that compiles (or at least tries to).

Challenges we ran into

One of the biggest challenges was making the AI return something that doesn’t just look fixed but actually compiles. Balancing chaos and functionality was tricky. I also had to deal with prompt-tuning issues to make the model generate creative yet plausible code patches.

Accomplishments that we're proud of

I’m proud that I managed to turn a funny idea into something that genuinely works. Seeing random broken code come back “fixed” is oddly satisfying. The reactions from other developers testing it were priceless.

What we learned

I learned how prompt engineering and model control play a major role in shaping AI output. I also deepened my understanding of backend–frontend communication and AI integration in real-world tools. Most importantly, I learned that creativity and humor can drive serious innovation.

What's next for Duct Tape AI

I plan to add more models and build a “Duct-Tape Ranking” system that scores how stable or cursed your patched code is. I also want to add a collaboration mode where multiple users can submit broken snippets and the AI merges them into a single chaotic but running program.

Built With

• and-deployed-using-vercel/render
• fastapi
• fastapi-(python)
• github
• openai/huggingface-apis
• react
• render
• sqlite
• tailwind-css
• vercel