Taking a Ducker Assessment
You have been invited to a Ducker assessment. Here is exactly what happens, so none of it is a surprise.
Before you start
Use a laptop or desktop with a recent Chrome or Edge, 8 GB of RAM or more, and a stable connection. Phones and tablets will not work. See What You Need.
Close other tabs and heavy applications. The workspace runs on your machine.
The sequence
- Open your invitation link
- A device check runs — a minute or two
- The workspace downloads and caches
- You see Ready to Start
- You click Start when you are ready
- The task loads, and then the timer begins
Nothing is timed until you press Start. Take a moment.
What is in the workspace
An editor, a file explorer, a terminal, a test runner, the task instructions, a timer, an application preview and a database viewer where the task uses them — and an AI coding agent.
How to do well
Read before you write. Run the tests. Look at the existing code. Understand the shape of the problem.
Use the agent — you are supposed to. That is the point of the format. But direct it. Precise instructions get better results than "implement this feature".
Review what it gives you. The agent implements what you ask, plainly. It will not proactively add validation, error handling, tests, or security work. If those matter here, you have to notice and ask.
Think about what the brief did not say. Real requirements are incomplete. What happens on a duplicate request? Two people at once? Bad input? An empty result?
Leave time to review. The last ten minutes are worth more than the first ten. Read your own submission before you send it.
Submit when you would stand behind it. Not when the tests pass.
What is being measured
Not whether you can code without AI. Whether you can use AI and still think like a strong engineer — architecture, judgement, edge cases, and whether you take responsibility for what you submit.