System design interviews, with the diagram already drawn
The hard part of a design round is not knowing the components. It is producing a coherent structure out loud, in order, under time pressure. Whisperer answers in that structure — requirements, estimate, components, trade-offs — with the diagram alongside.
A dedicated answer mode · diagrams, not prose
The answers below are an example. Your own base connects in the app.
…
…
Let's start with Q3 — how did conversion move?
Conversion is up 14% — let me pull up the funnel.
And the cost per lead?
Cost per lead is €12.40 — 9% under plan. Most of the gain came from organic.
Not a video — this is Whisperer itself, and the panel really answers. The microphone starts only when you press it.
Why structure
Design rounds are graded on order, not on cleverness
Most people who fail a design interview know the material. What goes wrong is the shape: they start naming technologies before establishing scope, skip the estimate, and arrive at trade-offs with no time left. The interviewer sees an unstructured answer and reads it as unstructured thinking.
Having the skeleton in front of you frees the part that is actually being assessed — the reasoning inside each step.
1 Requirements what is in scope, what is explicitly not 2 Estimate orders of magnitude, not precise numbers 3 Components the boxes and what each one owns 4 Data flow the path of one request, end to end 5 Trade-offs what this choice costs, and the alternative
- Heard The prompt arrives The question is taken from the running transcript rather than retyped, so nothing is lost in the paraphrase.
- Built Structure with a diagram The design mode returns components and data flows as a diagram next to the reasoning, instead of a wall of text.
- Yours You take it out of order Interviewers interrupt. Use it as a map you can jump around, not a script to read from the top.
Preparation
The rehearsal is worth more than the live use
Run three practice rounds against it before you run one real one. Not to memorise answers, but to find where your own structure collapses — usually at the estimate, which is the step people skip when nervous.
The line
What this does not make you
A design round assesses whether you can hold a system in your head and defend the choices in it. A structure on screen does not do that for you, and the follow-up questions are where the difference shows.
The interviewer will push on one box and ask why not the alternative. That answer has to come from you, and its absence is more obvious than a hesitation would have been.
Whisperer has to be running during the round to help you in it. A recording can be uploaded afterwards, but only for the transcript, and without system audio it will not hear the interviewer at all.
Questions
What people ask first
Does it draw the diagram for me?
The design mode returns component diagrams alongside the reasoning. What you do with them — redraw on the shared whiteboard, or narrate — is yours; interviewers usually want to watch you draw.
Is this cheating?
In a rehearsal, plainly not. In a live assessment the question is whether the tool replaces the skill being examined, and structure is part of that skill. The interviews page works through where the line falls.
Can it see the whiteboard the interviewer is sharing?
If you take a screenshot, yes — a vision mode reads what is on screen, including diagrams and code.
Which model handles design questions?
The one set for that role. Frontier models are part of the Max plan and are the noticeable choice here, since design answers reward depth over speed.
Will the interviewer see the panel if I share my screen?
No — the window is excluded from screen capture by the operating system. That covers software capture only, not a camera pointed at your monitor.
Do three rehearsals first
Pick a design you know well and answer it out loud against the structure. The gap between what you said and what the structure expected is the thing worth fixing.