How it works
Three steps to
objective proof.
Pick → Build → Score. Each step is designed to capture what matters and ignore what doesn't.
THE SHIPYARD
Pick a problem.
Browse the Shipyard and find a real problem that interests you — building features, fixing production bugs, extending existing systems. These aren't toy exercises. They're drawn from the kind of work companies actually pay for.
Ship a feature from scratch against a spec and a deadline.
Diagnose a broken codebase and land a working patch.
ONE COMMAND
Clone and build.
One command clones the challenge repo, connects GoShip MCP, and starts Claude Code. The MCP server runs silently alongside your editor, observing how you work — prompting, iterating, debugging — without ever touching your source code.
No separate install step. No config files. Just run the command and start building.
$ claude mcp add goship -- npx -y @goship/mcp-server
● GoShip MCP server added
$ claude
● Claude Code ready. Start building.
THE SCORING
Six dimensions. One score.
GoShip MCP observes your workflow (not your code) and scores across six dimensions: AI Fluency, Output Quality, Judgment, Speed, Prompt Precision, and Recovery Speed. Each one measures something different about how you work with AI — not just the output, but the process.
- AI Fluency
- 92
- Output Quality
- 87
- Judgment
- 78
- Speed
- 95
- Prompt Precision
- 84
- Recovery Speed
- 81
Judgment 78 — three AI suggestions accepted without review.
PRIVACY
Your code stays yours.
We measure how you work, never what you build. No source code leaves your machine. Period.
$ goship --ship-it
Ship it. Prove it. Get paid.
Your score goes on your profile. Companies see it. Bounty winners get paid.