Evaluate how candidates code with AI.
Candidates solve a coding task in the browser with an AI assistant, and you get a report on how they worked.
Get an access code
- 1 / 4The AI assistantThe candidate picks from 290+ AI models and prompts the assistant while coding.
- 2 / 4The AI budgetEvery interview has a set AI budget. The candidate decides how to spend it.
- 3 / 4The terminal and testsA real terminal to run the tests. Our own tests grade the work after the candidate submits.
- 4 / 4The editorEvery prompt and edit is recorded, so you see how the candidate got there.
- AI assistant · 290+ models
- A set AI budget
- Their terminal and the tests
- Every prompt and edit recorded
A set AI budget and 290+ AI models to choose from. How the candidate spends it on planning, prompting, coding and debugging is up to them, and we mark the whole workflow.
How a score is made
A binary marking scheme.
Every check in each category passes or fails. Drag or type to set how much each category weighs.
Sample candidate72
- Correctness
- Code quality
- AI process
- Efficiency
The prompt review
How they Prompted
After the session, we review the candidate’s prompts and write short observations, each quoting a prompt or a recorded step. Advisory, never a score.
Prompting reviewSample
- StrengthT1 Asked for the change and its constraint together; the proposal kept the public signature.
Add a token bucket per API key in limiter.py. Keep the public signature.
- ConcernT4 Asked to fix the tests without the failing output; the next two proposals missed.
fix the failing tests
Hire for how engineers code with AI.
We’re onboarding our first customers. Companies join with an access code from the founders.
Demos are arranged directly with the founders.