Govt Inisght

The Story Behind Government Insight

Inspiration: A Journey with AI

I didn’t start Government Insight with a big idea or mission. Honestly, I just wanted to see what would happen if I let AI take the lead.

So I asked it: “Give me a project worth building.” Out came Government Insight (originally GovTracker)—a tool that filters through the chaos and gives you real-time government updates, clean and simple.

I didn’t fight the idea. I ran with it. This wasn’t just an experiment in coding—it was a test to see how far AI could go. Could it help not just write code, but come up with the idea, shape the logic, and build something real? Turned out, yes. And then some.

What I Learned: AI’s Not Just a Code Machine

The biggest surprise? AI wasn’t just completing lines of code—it was thinking with me. Tools like GitHub Copilot and Google Gemini didn’t feel like passive helpers. They were more like an extra brain in the room.

Bolt helped me zip through frontend work, fill in gaps, and polish things I might’ve skipped. Gemini? That one was the backend whisperer. It helped debug strange Supabase behaviors, clean up messy API logic, and guide the overall architecture.

More than anything, I learned this: AI isn't magic, but it’s fast. When paired with real intent, it becomes a powerful co-builder that helps you ship faster and think bigger.

How It Was Built: Bolt, Copilot, Gemini

Tech stack was simple but powerful:

The backend parses news sources and creates a summary of the news report and picks up images from the article or just creates a new one with AI.

Bolt handled the full-stack boilerplate — React frontend, Tailwind CSS, and Supabase for auth and storage.
GitHub Copilot sat beside me for the day-to-day coding. From writing logic to fixing syntax, it filled in blanks faster than I could.
Google Gemini took on the backend heavy lifting — writing routes, transforming data, debugging edge cases, and improving Supabase interactions.

These tools didn’t just do their jobs — they freed me up to focus on core logic and user experience. No setup hell. No yak shaving.

The Hard Part: Convincing AI to Do Routing Properly

The trickiest part? Routing.

Not building it—but getting AI to actually generate the right code for it. Dynamic and nested routes in React are already finicky. But trying to get Copilot or Gemini to output something that works — without scroll bugs or broken back buttons — was a real adventure.

Turns out, AI doesn’t always guess your intent. I had to learn how to prompt really specifically — like explaining a task to a junior dev. Eventually, I figured out how to ask the right questions, break things down, and guide it step by step. That became its own skill: prompting not just for code, but for correct, production-grade code.

Looking Back

Government Insight started as a curiosity. Now it’s a project I’m proud of — not just because it works, but because it showed me what building with AI actually looks like in practice.

It’s not perfect. It still has rough edges. But it’s real. And it was built with the help of machines that are learning to build with us.

Even the demo video was AI-generated — I recorded the walkthrough, wrote the script with Gemini, and used ElevenLabs to generate the voiceover.

It’s not perfect — but it’s real. And it shows what’s possible when you build with intent and let AI handle the rest. Theres still work to be done but the ride so far it has been amazing.