Fork Yeah!

RepoRadar (R², “R-squared”) is a generative UI for discovering, exploring, and instantly prototyping from the top 10,000 GitHub repositories on Earth. Users navigate via whichever modality fits their brain in the moment - dragging sliders, pulling vertices on a Street Fighter-style radar, typing, or talking out loud - and watch the recommended repo list reshape on the fly. Then they say “make this repo look like Stripe’s homepage” and watch a working prototype spin up on Cloudflare Workers or Containers seconds later.

Full PRD: https://docs.google.com/document/d/1EVPnZXZvcWzxg9c1WvQ2xKnpyr1vmAriEaTs_tCuO_A/

Public Repo: https://github.com/RepoRadar/reporadar/

Why R²
The name has two halves working together. Repo Radar for the discovery half - we are scanning the open-source universe and surfacing what matters to you. R² because the core of the system is two layers of recursion: the AI ranks the repos, then the AI generates the UI that displays the repos, then the AI generates the prototype that uses the repos. Generative UI on top of generative UI on top of generative ranking. R squared.

The problem
Developers and engineering leaders constantly need to discover open-source projects that fit a specific taste - “something hot right now I can hack on tonight,” “stable production-grade libraries with strong community,” “rising stars I should know about before everyone else does.” Current discovery is binary search by hand across GitHub Trending, Hacker News, awesome-lists, and X/Twitter. None of these let a user express their taste directly, and once a developer finds a candidate repo, they still have to spend an hour cloning, configuring, and seeing if it can do what they actually want.

R² collapses both halves of the workflow. Discovery via expressive multi-modal input. Evaluation via instant prototyping in a sandbox. Both ends generate UI rather than retrieving fixed views.