Please read this before you begin. It explains how you will be assessed and what is recorded. You are being told, not asked to waive anything.
A machine scores your work. A person decides.
Your code is run against an automated test suite and scored against a rubric written in advance. Software calculates that score.
It does not decide anything about your application. The score and your code are shown to a named person at the company you applied to, and that person makes the decision. You are not subject to a decision based solely on automated processing.
Where AI is used, and where it is not
- Choosing your task. A model reads your CV to pick an exercise suited to your background, and may adjust the wording of the task so it refers to an industry you know. The difficulty and what is being measured do not change.
- Scoring. No model scores your work. Scoring is a fixed test suite and a rubric that was calibrated before you saw the task, and it produces the same result every time it runs.
- Interview preparation. After your work is scored, a model reads the code you wrote and suggests questions for your interviewer to ask you about it. The questions are anchored to five moments of your session that the record itself picks — your first test run, your biggest change, your longest pause, your largest insertion and your last fix — and your interviewer may walk through those with you, showing your own code as it stood at each one. The model suggests questions only. It does not score you, it does not rank you, and it does not recommend whether you should be hired.
- Pair, if you use it. A navigator in the workspace answers the questions you type. It explains and points the way; it does not write your solution, it sees none of your files or the brief, and it never scores you. Your prompts and its replies are part of your record: your interviewer reads how you prompted, looking for engineering direction.
What is recorded while you work
- The code and files you write.
- The commands you run in the terminal, and whether they succeeded.
- Automatic snapshots of your work, so nothing is lost if you disconnect. The history of those snapshots is kept as the record of how you built your solution, and it is what your interview can be about.
- When this browser tab loses and regains focus.
- Your prompts to Pair and its replies, if you use it.
- The size of anything you paste — its length and line count, never the text itself — and each time the workspace's clipboard limit stops a copy, cut or paste. The workspace limits how much text can move through the clipboard at once.
This is shown to the reviewer as context for your work — for example, that you ran the tests, saw a failure, and fixed it. It is not used to score you and a sparse record counts against nobody.
What is not recorded
No screen recording. No camera or microphone. No keystroke logging. Not the text you paste — only how much. Nothing about other applications or other browser tabs.
If you need an adjustment
If a disability or health condition means the time limit or the format would disadvantage you, contact the company that invited you before you start. They can arrange an adjustment. You do not have to explain your condition to us.
Your data
Your code and CV are kept for a period the company sets, then deleted. The detailed record of your session — the snapshot history, the raw record of commands and saves and, where you used it, the in-workspace assistant transcript — goes sooner: 7 days after a decision closes your application or the position closes, and in any case within 30 days of your assessment ending. Your score, your final code and the decision stay for the period the company sets. If the company leaves the platform, everything identifying you is deleted with it; only anonymised exercise statistics (score, duration — never your name or your code) are kept, to keep our exercises calibrated. You can ask the company for a copy of your data or ask them to delete it.