Benefit Bridge – Project Story

Why we built it

People often give up on government aid because the websites are hard to use. We met seniors and students who didn’t know which programs they could get. We wanted a page where you type one sentence and see the right programs right away.

What it does

User picks their citizenship status.
User types a short sentence (example: “I lost my job and need help with rent”).
The app shows matching benefits with direct “Apply” links.

How we built it

Frontend: HTML + TailwindCSS + HTMX (for quick page updates).
Backend: Go (net/http, html/template).
AI: OpenAI GPT-3.5 pulls keywords from the sentence.
Data: A JSON file with benefit programs, their keywords, and links.

What went well

HTMX let us keep the site fast without a heavy JS framework.
The keyword whitelist stopped the AI from inventing fake programs.

Challenges

Getting dark-mode CSS to load correctly with the Tailwind CDN.
Expanding our program list quickly and keeping keywords accurate.

What we learned

Small tools (HTMX, Go templates) can build a usable product in 24 hours. Clear prompts and whitelists keep AI results safe. A single JSON file is enough to test an idea before moving to a real database.

Built at the UWB Hackathon 2025.