The rubric
Each assessment has a rubric: a list of criteria, each with points. The hiring team sets the rubric, and can change it later. You do not see the rubric before or during the assessment.The tests
Some assessments have test cases. On those assessments:- While you work, Run uses only the visible sample cases.
- After you submit, all tests run once on your submitted code, unless the code runner fails or code running is switched off. This includes the hidden tests.
- You never receive the input or the expected result of a hidden test.
- The page you see after you submit shows no test results.
The score
An AI reviewer scores your submission against the rubric. By default the AI reviewer runs three times. For each criterion, it keeps the middle of the three results.- If the rubric has a correctness criterion and the hidden tests ran on your submitted code, the points for correctness come from the tests.
- If the hidden tests do not run because the code runner fails, the AI reviewer estimates correctness itself.
- The other criteria are always the AI reviewer’s judgment.
What the product does not do
- It has no facial recognition.
- It does not try to read emotions.
- It does not use your face or an identity document to check who you are.
Your device, camera and microphone
The product keeps a record of your device and network. It is used to flag when two candidates share a device or network. Your camera and microphone are used only if the hiring team switched them on for your assessment. The start page tells you if your assessment uses them. This assessment needs a camera and a microphone. Without them, Start stays locked.- If they are on, your browser keeps a live camera and microphone stream open. This is a signal that you are present.
- The product’s code does not send or store video or audio.
- The page records an event when you turn your camera or microphone off or on.
If you use the AI assistant
Use only the built-in AI assistant. Other AI tools are not allowed.- Your conversation with the assistant is recorded with your submission.
- The hiring account that owns the assessment can read all of it.
- Its active teammates can see the first 140 characters of each question you asked the AI assistant.
- For each suggested edit you review, whether you accepted or rejected it is also recorded (best effort).
- On the start page, in the “What to expect” list, when AI is allowed: “Use the available AI assistant deliberately, then verify its output”.
- On the start page, in the same list: “Show your reasoning; leave the work in a reviewable state”.
- Under the task description, when AI is allowed: “Direct the assistant deliberately and verify its output. Those choices become reviewable context alongside the work.”
