Skip to content

IsraelAfangideh/ruwt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

710 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ruwt

Monorepo for the Ruwt platform.

Projects

Directory What Stack URL
/dev AI-Efficiency Assessment Platform React + Vite, Cloudflare Pages/D1, Supabase Auth ruwt.dev
/social Ruwt Social Network (API + mobile) Hono + Bun on Fly.io, React Native (Expo), Supabase Postgres ruwt.fly.dev

ruwt.dev — How It Works

Ruwt measures how efficiently you use AI to solve coding challenges. The leaderboard doesn't just reward correctness — it rewards spending the least AI resources to get there.

User Flow

  1. Land on ruwt.dev — the landing page explains the three dimensions being measured: Model Selection, Prompt Efficiency, and Iterative Debugging.

  2. Register / Login — GitHub OAuth or email/password. New users get 5,000 free credits.

  3. Browse Challenges (/challenges) — filter by category (Model Selection, Prompt Efficiency, Debugging). Each card shows difficulty, category, and an efficiency cost goal.

  4. Pick a challenge — read the full description and constraints (time limit, max tokens, max cost), then choose:

    • Start Timed — countdown begins, pressure's on
    • Start Untimed — no timer, marked separately on the leaderboard
  5. The Arena IDE — the core experience:

    • Left side: Monaco code editor + Output panel (test results)
    • Right side: Tabbed panel — Description tab (problem + examples + constraints) and AI Chat tab (Llama 8B assistant, costs credits/tokens)
    • Status bar: real-time cost, token count, time remaining, credits
  6. Iterate — read the problem, write code, ask the AI for hints, run tests, see which fail, fix, repeat. Every AI chat message costs you — the leaderboard rewards efficiency, not just correctness.

  7. Run Tests — per-test pass/fail with input, expected output, and actual output. Failed tests auto-expand for quick debugging.

  8. Submit — finalizes your attempt as passed or failed.

  9. Leaderboard (/leaderboard) — ranked by challenges solved, then by average cost. Cheapest solvers win.

The Core Loop

"How good are you at using AI?"

Can you pick the right prompts, avoid unnecessary back-and-forth, and solve problems with minimal AI cost? That's the skill being ranked.


Development

Git Workflow

  • Main branch: main (deploys trigger here)
  • Development branch: develop
  • PR flow: develop -> main
  • Commit style: type(scope): description (e.g., fix(dev):, feat(social):)

Deploy

  • /dev: Pushes to main with changes in dev/** trigger GitHub Actions (deploy-dev.yml) -> Cloudflare Pages
  • /social: GitHub Actions (deploy.yml) -> Fly.io via flyctl

See CLAUDE.md for full architecture details, infrastructure config, and API reference.

About

Full-stack platform measuring how efficiently engineers use LLMs to solve coding challenges.

Resources

Contributing

Stars

Watchers

Forks

Contributors