The AI coding speedrun. Start the clock, build with Claude Code, deploy. How fast can you go?

---

Everyone says AI makes you 10x faster. Nobody's actually timing it. zero-to-prod sets up the challenge, starts the clock, watches your files as you build, and records your best time. Empty directory to production — the only rule is you have to actually ship.

$ npx github:NickCirv/zero-to-prod start express-api

  Challenge: express-api
  Build a REST API with JWT auth, 5+ routes, tests, and deploy config.

  Working directory: /tmp/zero-to-prod-express-api-1709123456/
  CLAUDE.md written. Open Claude Code in that directory to begin.

  Timer started. GO.
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Watching for: tests ✗  deploy config ✗  routes ✗
  Elapsed: 00:01:47

  [Tests detected] ✓
  [Deploy config detected] ✓
  [5+ routes detected] ✓

  Confirm deployment? (y/n): y

  Run complete: 2m 11s
  Personal best for express-api! Saved.

Install

No npm account needed — runs straight from GitHub:

npx github:NickCirv/zero-to-prod start express-api

Requires Node.js 18+ and Claude Code (npm install -g @anthropic-ai/claude-code). No API key for the timer itself.

Commands

npx github:NickCirv/zero-to-prod start <challenge>   # Start a timed speedrun
npx github:NickCirv/zero-to-prod challenges           # List all challenges with full specs
npx github:NickCirv/zero-to-prod leaderboard          # View your personal best times
npx github:NickCirv/zero-to-prod submit               # Manually log a completed run

Command	Description
start <challenge>	Creates a fresh working directory, writes a CLAUDE.md spec, starts the timer, and watches for completion criteria
challenges	Lists all built-in challenges with their stacks and pass/fail criteria
leaderboard	Shows your personal best times per challenge, stored locally
submit	Interactive prompt to manually record a run (challenge, time, files, tests)

Challenges

Challenge	Build This	Completion Criteria
express-api	REST API with auth	5+ routes, JWT, tests, deploy config
react-app	Dashboard with routing	3+ pages, responsive, deploy config
next-app	Blog with SSG	5+ pages, markdown, SEO, Dockerfile
flask-api	Python API with auth	5+ endpoints, SQLAlchemy, tests
cli-tool	Published npm CLI	3+ commands, tests, --help

How it works

1. Run zero-to-prod start <challenge>
2. A clean directory is created with a CLAUDE.md challenge spec inside
3. Open Claude Code in that directory — it reads the spec and starts building
4. The live file watcher detects tests, deploy config, and routes as you build
5. When all criteria are met, confirm deployment — the timer stops
6. Your time is saved to the local leaderboard

Rules

1. Start from an empty directory — no templates, no copying existing code
2. Use Claude Code or any AI coding assistant
3. The app must pass tests and include deploy configuration
4. Time stops when you confirm deployment

What it is NOT

• Not a framework or scaffold generator. It sets the challenge and watches — Claude builds the actual app.
• Not a cloud leaderboard. Times are stored locally; this is about beating your own personal best.
• Not opinionated about your AI tool. Optimised for Claude Code but works with Cursor, Copilot, or any assistant.

---

_{Node 18+ · MIT · by NickCirv}