Inspiration

As vibe coding gets popular, more people can ship faster than ever. The new bottleneck isn’t building—it’s being seen and evaluated fairly. We wanted a simple benchmark so good work doesn’t get lost in a sea of repos.

What it does

TopVibeCoder turns a repo link into:

  • a leaderboard for 3000+ apps (impact, techincial, creativity, ui, prompt engineering, etc)
  • auto eval any codespace (aggregated evals to reduce randomness)
  • auto demo generator (click to try, no setup)

How we built it

We built a pipeline that: 1) eval the code with 5 dimensions 2) uses Gemini 3 to produce structured evaluations, 3) uses Gemini 3 to generate and host demo (details here: https://www.kaggle.com/competitions/gemini-3/discussion/671193)

Challenges we ran into

  • Data collection
  • Demo Generation: story, smoothness

Accomplishments that we're proud of

  • Repo → rank + feedback + demo in ~1 minute
  • Evaluated thousands of projects and made them instantly comparable
  • Get lots of positive feedback from communities

What we learned

  • In the vibe-coding era, distribution and evaluation matter as much as shipping
  • Structured rubrics + aggregation create signal people can trust
  • Demos turn “code” into something others can actually experience

What's next for TopVibeCoder

Become a benchmark layer for vibe-coded apps:

  • better coverage across stacks
  • stronger anti-gaming and calibration
  • richer leaderboards and “learn-from-the-top” breakdowns

Built With

Share this project:

Updates