How ViberBlitz Works

The timer starts the moment you click Start on a challenge. It stops the moment you click Finish. This is wall-clock elapsed time — not build time, not AI response time, not the time your app takes to boot.

Your start time is recorded server-side when the run is created. Your finish time is recorded server-side when you submit. The elapsed time shown on the leaderboard is the difference between those two server-side timestamps, not a client-side measurement.

Your GitHub repository must be public. The verifier clones it over HTTPS — private repos will fail.

The repo must include a package.json (or equivalent) with a start or dev script that launches the app on a local port. The verifier runs npm install (or pnpm install if a pnpm-lock.yaml is present) followed by npm run dev or npm start.

There is no restriction on framework, language, or architecture. React, Vue, Svelte, plain HTML — all accepted, as long as the app starts and the tests pass.

When you submit, our verifier clones your repo to a temporary directory, installs dependencies, and starts the app on a randomly assigned free port. It then runs a Playwright test suite specific to the challenge you attempted.

The test suite is deterministic and public — the exact tests are listed on each challenge card. There is no hidden criteria. If your app passes all listed tests, the run is logged. If any test fails, the run is rejected and you see the full test output explaining what failed.

The verifier has a 120-second timeout per run. If the app fails to start or the tests do not complete within that window, the run is rejected.

Tool stack disclosure is mandatory. On submission, you must select at least one tool from our canonical list (Cursor, Bolt.new, Lovable, v0, Windsurf, Claude Code, ChatGPT, GitHub Copilot, Replit, Gemini, or Other).

The tool stack you report is taken on trust — we do not verify it programmatically. The purpose of disclosure is to build a public dataset of which tools produce the fastest verified runs across different challenge types. Misreporting your stack undermines the data that makes ViberBlitz useful.

The most obvious form of cheating is pre-building an app before clicking Start. We cannot fully prevent this in the current beta. We rely on the community's interest in accurate data — a leaderboard full of pre-built submissions is worthless to everyone.

We record the GitHub repository creation date and push timestamps on every submission. Runs where the repo was created significantly before the run start time may be flagged for review.

Top runs may be subject to manual review. Replay links are strongly encouraged for any run in the top 10 of a challenge. Runs found to be fraudulent will be removed.