Communication and time management run through every task in the round. They decide how much of your work the interviewer can actually see and credit.
Communication#
These habits keep your reasoning visible from the first minute to the last.
- Narrate why, not what. "I'm typing a for loop" tells the interviewer nothing. "Iterating in sorted order so the first match is the smallest" shows your reasoning.
- Summarize at each checkpoint. At the end of every task, take about 15 seconds: what you did, what's verified, and what's next. For example: "Bug fixed, all tests pass, and I added a regression test. Moving to the feature." The interviewer is writing notes, and this gives them clean evidence to write down.
- Answer their questions yourself first. When the interviewer asks "what do you think?", give your view before you turn to the AI. Going straight to the tool looks like you don't have a view.
- Use the time while the AI generates. Talk through the next piece, or what you'll check in the output.
- Take hints well. Acknowledge the hint and apply it. If you disagree, explain why in one sentence and stay open to being wrong. A hint is the interviewer helping you, so don't argue with it.
- When you're stuck, think out loud. Say what you know, what you've ruled out, and what you're trying next. Silence is the worst thing you can give them.
Time budget (60 minutes)#
If your round has all three tasks, the hour splits into orientation, the three tasks, and a short wrap-up.
| Time | Task |
|---|---|
| 0–5 | Orientation |
| 5–15 | Fix the bug |
| 15–40 | Build the feature |
| 40–55 | Optimize |
| 55–60 | Wrap-up |
Check the clock every time you move to the next task. Two rules:
- Bug fix past ~20 minutes: ask for a hint. Ten silent minutes cost you more than a hint does.
- Feature not working by ~40: cut the scope. Get a correct basic version running, and say what's missing.
Recovery#
Each common problem has a standard move.
| Situation | Move |
|---|---|
| The AI is wrong repeatedly | Write that part yourself, and say why |
| You're stuck | Narrow your hypothesis, ask the AI a focused question, or ask the interviewer for a nudge |
| You broke something | Go back to your last working version, then rerun the tests. Save a copy or commit whenever the tests go green |
| Time is almost up | Stop coding and explain what's left, how you'd do it, and its complexity |
Wrap-up#
Use the last few minutes to cover what you'd improve with more time, and what production would need: input validation, error handling, more tests, and how it would scale. That shows judgment beyond the hour.