Player 1 · Free for developers
Code with AI. Get scored on how.
Take a real coding challenge in Claude Code or Cursor. A recorder logs your prompts, tool calls and test runs, and the session comes back with six scores. Publish them or keep them private.
- per challenge
- 15-60 min
- per challenge
- dimensions scored
- 6
- dimensions scored
- Claude Code or Cursor
- Your tools
- Claude Code or Cursor
- AI Fluency
- 92
- Output Quality
- 87
- Judgment
- 78
- Speed
- 95
- Prompt Precision
- 84
- Recovery Speed
- 81
00:44 -4 judgmentAccepted a diff unreviewed, so Judgment lost points.
How a score is builtWorld 1 · How it works
We score the session as well as the code.
Pick a challenge and build it in your own editor. Each score you get back comes with the session events behind it.
- 1
Pick a challenge
Build something from scratch, fix a broken app, or add a feature to one. Pick your level. Challenges run 15-60 min.
StartBuild a Task ManagerBuild · Mid · 60 min - 2
Build with your AI
Work in Claude Code or Cursor as you normally would. The GoShip recorder logs your prompts, tool calls and test runs, so how you got to the code counts.
- 00:11Rejected a suggestion
- 00:26Rewrote it by hand
- 00:31Tests pass
- 00:44Accepted a diff unreviewed
- 3
Get your score
Published rules score six dimensions and a model reviews four of them. Each score shows its evidence. You decide whether it goes public.
Who it is for
“Shippers”
Shippers take an idea to a live product with AI. It doesn't matter who typed the code.
That includes vibe coders who describe an app and let the AI write it, designers shipping their own ideas in Cursor, and PMs who build the prototype instead of writing the spec. GoShip scores how each of them works with the AI.
- 00:04Asked for a plan before any code+ fluency
- 00:11Rejected a suggestion+ judgment
- 00:26Rewrote it by hand+ precision
- 00:31Tests pass+ quality
- 00:44Accepted a diff unreviewed-4 judgment
- 00:52Build fixed after one failure+ recovery
The recorder logs events, never file contents.
How scoring worksYour GoShip score
Six scores, one total.
Each one measures a part of how you work with AI. The weights and the rules that set them are on the methodology page.
The weighted total. You see all six parts beside it.
- AI Fluency20%
- 92
- Context you give, tools you call, models you pick
- Output Quality25%
- 87
- Whether tests and builds pass, plus structure
- Judgment20%
- 78
- When you accept, edit or override the AI
- Speed15%
- 95
- Time to a working result
- Prompt Precision10%
- 84
- Specific prompts, few retries
- Recovery Speed10%
- 81
- Failing test to the next passing one
For teams
Hire people who ship. Or post a bounty.
Free to start. Invite candidates to any ready-made Shipyard challenge at no cost. They work with their own AI tools, and you get six scores plus the session log that shows how each person got to the result.
Sample data · sorted by score
Candidates 1 and 3 shipped similar code. Their sessions scored 91 and 62.
The Shipyard
Open challenges
Immigrant Resource Navigator
Play music on 3D printer motors
A toy that plays grandma's stories
Try one challenge. See your score.
Challenges run 15-60 min. You decide whether the score goes public.
goship.tech/profile/you
2 scored challenges, public by your choice