Not in the code. In the silence right after the question lands, before a direction becomes obvious — the stretch where you are thinking and the interviewer is writing "struggled to structure the approach".
Interviewers cannot see you thinking. They can only hear whether you are talking.
An actual exchange. The middle column never leaves your display.
Interviewer
So, given a list of intervals, I want you to merge the overlapping ones. Take your time.
On your overlay · ~1s later
Sort by start, then sweep merging when current.start ≤ prev.end. O(n log n), dominated by the sort. Watch: empty list, single interval, intervals that touch exactly.
You, out loud
"Right — I'll sort by start time first, then do one pass merging forward. That puts us at n log n from the sort, and the sweep itself is linear."
Interviewer
Good. What happens if two intervals just touch — say [1,3] and [3,5]?
On your overlay · ~1s later
Ask them to define it: touching is usually merged, but say it is a business decision. Use ≤ to merge, < to keep separate.
You, out loud
"Depends what we want — I'd merge them with a ≤ check, but if adjacent intervals should stay separate that's a one-character change. Which do you want?"
You write the code. The overlay is a prompt to speak from, not a script to read.
Suggestions that arrive after you have already started talking are worse than none. Measured on real interviews over the last 30 days:
median time to first word on screen
75th percentile — three in four answers
answer turns measured in the sample
"Can you get that to O(n)?" and "what breaks on an empty input?" are what the interviewer writes their verdict from. AceRound tracks the whole conversation, so the second and third questions get the same treatment as the first — not just the opening problem statement.
It listens to the interview audio and reads your own screen, so it works on CoderPad, HackerRank, CodeSignal, Karat or a plain shared doc, over Zoom, Meet or Teams. The overlay sits outside what screen sharing and recording capture, and clicks pass through it — including on a full-screen share.
The same assistant picks up the question type from the conversation and changes what it puts on screen — trade-off framing for design rounds, STAR structure for behavioural ones.
Real-time support in English, Japanese, Korean, Spanish, Portuguese, Vietnamese and more — including hearing the question in one language and getting your points in another.
