The Nines

The Nines is a live uptime comparison of Claude and OpenAI, built with Vite and vanilla JavaScript.

A GitHub Action fetches status data every hour from public Statuspage APIs and commits the result. Vercel picks up the push and redeploys.

How It Works

The site pulls a 90-day rolling window of per-service status from status.claude.com and status.openai.com. Daily statuses are normalized into compact status strings and turned into:

• aggregate health bars
• per-category comparison cards
• weighted daily winners
• streaks, ties, and comeback moments

Historical cells are day-bucketed. Today's cell is refreshed from the current component summary so the live page and the latest bar stay aligned.

API and chat products are weighted more heavily than coding products.

Methodology

• Claude source: status.claude.com
• OpenAI source: status.openai.com
• Window: 90 days (rolling)
• Status classes: operational, degraded, partial outage, major outage, maintenance
• Daily scoring:
  • operational = 100%
  • degraded = 60%
  • partial outage = 30%
  • major outage = 0%
  • maintenance = 80%

Repo Structure

index.html              page structure and metadata
src/
  main.js               app logic, scoring, rendering
  styles.css            layout, theme, responsive styles
  data.js               seed/fallback data
public/
  data/status.json      live status (committed by CI)
  og.png, favicon.svg   static assets
scripts/
  fetch-status.js       fetches APIs, normalizes, writes status.json
  smoke-test.js         validates output against live APIs
.github/workflows/
  fetch-status.yml      hourly cron action

Running Locally

npm install
npm run fetch   # pull latest status data
npm run dev     # start Vite dev server

Testing

npm run fetch && npm test

The smoke test validates structure, uptime math, incident attribution, real-time status alignment, and component coverage.

Limitations

• Daily rollups compress incidents into a simpler comparison model.
• Some source systems expose richer data than others, so the comparison involves normalization.
• This is an independent interpretation of public status data, not an official benchmark.

License

MIT. See LICENSE.