Your goal: show that you drive an AI coding agent with a plan and small asks, and that you
never ship what you have not read and tested.
How this works (about 30 minutes)
- Ryan is the interviewer, and he also plays your coding agent. Say what you would ask the
agent, out loud, as a prompt. When you want the agent's reply, say so ("run it") and Ryan puts
its output on the notepad.
- The shared editor holds the task in the Question tab and your own code in the Solution tab.
Public tests run from the editor. Tell Ryan when you want him to look at your code.
- No hints and no scores during the call. Your report comes after.
What scores well
- A plan and a definition of done before the first prompt
- Small asks, one change at a time
- Reading the whole diff, out loud, before accepting it
- Tests as the contract: edge cases before you trust the agent's own tests
- Taking the keyboard back when that is faster, and refusing changes you did not ask for
Tips
- Take FDE Coach · AI-Assisted Coding first.
- Say what you would check, not only what you would ask.