DEV Community

Cover image for I won my first hackathon by building something I barely understood
Ashwinhegde19
Ashwinhegde19

Posted on • Originally published at ashwinhegde19.github.io

I won my first hackathon by building something I barely understood

A few weeks ago I was using OpenClaw to set reminders and run basic cron jobs.

That's it. Nothing impressive.

Then I walked into the HSRFC x OpenClaw Builders event in Bengaluru. 4 hours, real constraints, ship something or go home.

I had never built a complete agent before. No plan. Just enough curiosity from weeks of tinkering to figure things out as I went.

What I built
A GitHub Issue Resolver β€” an autonomous agent that discovers, ranks, and analyzes open GitHub issues. Fully local, no cloud, no servers. Everything runs on your own machine.

The idea was simple: point it at a repo, and it would:

Pull open issues
Rank them by complexity and impact
Suggest fixes with context from the actual codebase
Nothing groundbreaking on paper. But getting an agent to do all of that reliably, in 4 hours, with tools I'd barely used before β€” that was the real challenge.

What actually happened
Did it break multiple times? Yes.

Was I debugging in the last 20 minutes? Also yes.

There's a specific kind of stress that comes from knowing your demo is in 15 minutes and your agent just threw an error it hasn't thrown before. You don't think clearly. You just try things faster.

The thing that saved me wasn't some clever architecture decision. It was that I'd spent enough time tinkering with OpenClaw beforehand that I had intuition for where things go wrong. Not knowledge β€” intuition. There's a difference.

The lesson
Still took first place.

But the lesson wasn't about the win. It was this β€” you don't need to know everything before you start. You just need enough curiosity to keep going when it breaks.

I think a lot of developers (myself included) wait too long to build things. We want to understand the full picture first. Read the docs. Watch the tutorials. Make sure we won't look stupid.

But the people who build interesting things are the ones who start before they're ready and figure it out along the way.

What's next
The project still needs work. The ranking algorithm is basic, the error handling is rough, and there are edge cases I haven't touched. But it's real, it runs, and it's on ClawHub.

I'll probably keep hacking on it. Maybe write about the technical side in a follow-up post.

Built with OpenClaw 🦞

Top comments (0)