Mocks and debriefs

Mocks are rehearsal. Mock opens the Mock Interview hub, where you choose how to practise: a coding problem, an interview quiz, flash cards, a system design review, or a live AI interviewer that asks one question at a time and ends with a debrief that turns into plan changes. The post-interview debrief is the record of the real thing, written by you.

The Mock Interview hub

The hub shows five modes, your next interview, your recent runs and how many AI runs your credit balance covers.

Every AI button tells you the cost and your balance before you press it. If a run fails, or the AI returns something unusable, the credits are returned.

Quizzes and flash cards

Each mode offers three starts:

• An AI set tuned to your next interview, using its company, role, level and round, plus the skills your evidence says are weakest.
• A custom concept you type, with a difficulty, a length and a focus.
• Random from the library, which is free. When the sprint has readiness evidence, the pick leans toward your weakest skills.

A finished quiz shows your score, a bar per skill and, for each question, why interviewers ask it. A finished quiz counts as quiz evidence for readiness, and a finished deck counts as flash card evidence.

Coding arena

Choose a language (Python, Java, TypeScript, JavaScript, C++ or Go), a difficulty, a practice mode (Learn / Interview / Timed / Blind) and a topic, or let the topic follow your interview. The problem opens beside the editor with examples, constraints, visible tests and hints you reveal according to the mode. Your draft is saved in the browser as you type.

Run executes sample or custom tests in a sandbox for free (no AI review, no Studio). Final Submission freezes your code, runs the full suite in the sandbox when available, then spends credits for an AI scorecard and unlocks the Visual Learning Studio at /app/sprint/mock/coding/[runId]/studio.

Before Final Submission, you predict your own score. The report then shows the scorecard verdict, the score next to your prediction, scorecard dimensions, executed (or AI-narrated) tests, time and space complexity against the Optimal Reference, strengths and weaknesses tied to your lines, bugs with fixes, missed edge cases, a strategy compare, an improvement plan and likely follow-up questions.

Each run gets one Final Submission review. Rematch starts a new problem, which is a new AI run. Guided repair and retry modes can seed a fresh attempt from the Studio or the report.

How the coding review stays honest

• Sandbox execution is the source of truth. Run and Final Submission execute your code in an isolated sandbox for Python and JavaScript / TypeScript. Every test row that ran in the sandbox is labelled Executed. When the sandbox cannot run (or for languages without a local tracer), the reviewer may still narrate rows and labels them AI narration (sandbox did not run). Runtime always wins over narration.
• AI narrates; it does not invent execution. The Visual Learning Studio plays Trace IR recorded from the sandbox. Line-by-line cinema comes from that IR, not from the model guessing what the program did.
• Feedback cites your lines. Every strength, weakness and bug must quote a line of your code. Before the report reaches you, each quote is checked against that line, and anything that does not match is removed. Selecting a line reference jumps to that line in the editor.
• No verdict without evidence. If too little of the review survives that check, the report says there is not enough evidence for a verdict instead of guessing, and shows no overall score, calibration or scorecard chart either. A failed executed or narrated test caps the verdict at Lean no, and a score under 45 is always No hire.
• Unobserved runs. If most of the submitted code was pasted rather than typed, or the editor's record of your typing does not match the code you submitted, the run is marked unobserved. It still gets a review, but it earns no XP, does not count as readiness evidence, and keeps the Reference Solution locked. The Reference Solution also stays locked when the submission is little more than the starter code.
• Your prediction is saved first. The predicted score is stored with your code before the review starts, so it cannot be changed once the score is known.

An observed coding review counts as mock evidence, one of the strongest kinds for readiness.

System design

The system design mode opens Evaluator AI inside the sprint, on the open design task that trains your weakest skill when there is one. You can pick another question from the canvas.

Starting a live mock interview

Open Mock, choose Live AI interviewer, then pick:

1. The interview you are rehearsing for, which sets the company, round and level the interviewer plays.
2. A format:

3. Optionally a problem from the design tasks on your current plan. Otherwise the interviewer picks one suited to the company.

During the mock

Each interviewer turn carries a tag saying what it is doing: Deeper dive, Validation, Challenge, Exploration or Pivot. You answer in the text box; Cmd or Ctrl with Enter sends. The counter shows which turn you are on, and a timer sized to the format keeps pace without ending the session.

When the interviewer wraps up, or you reach the answer limit for the format, Finish and debrief becomes the main action. You need to answer at least one question before finishing.

The mock debrief

The debrief shows:

• An overall score for the rehearsal, with its band.
• What you did well and where the interview weakened.
• The moment that mattered: your own words at the turning point, next to a stronger version of the answer.
• New preparation tasks: short repairs with their minutes.

Update my plan proposes a replan that schedules those repairs, and shows what they would replace before anything changes. A completed mock is recorded as mock evidence, one of the strongest kinds for readiness.

Credits

Each interviewer reply, each mock debrief and each post-interview summary uses 50 AI credits from the sprint's monthly 25,000, as does every AI run in the Mock Interview hub. Library quizzes and decks are free. If credits run out, the page says so where it happened and links to a top-up.

After the real interview

Open Debrief for the interview and record:

• the questions asked and the follow-ups;
• what surprised you and where you struggled;
• how prepared you felt, from 1 to 5;
• the outcome, if you know it, and the next round date.

Your notes are saved before the AI summary runs, so nothing you write is lost if the summary fails. The summary lists what matched your preparation, what did not, any new weak areas and changes for the next round. Rebuild my plan proposes repairs for those areas.

A debrief with a known outcome and a preparedness rating is recorded as real-interview evidence for that interview.

When the outcome is an offer, the page shows what the sprint added up to: preparation days, tasks completed, designs, mocks, and readiness from first to latest. From there you can archive the sprint.

What is shared, and only with consent

A debrief is private by default. Nothing from it leaves your account unless you tick Contribute to Interview Intel on the form.

If you do, the company, role, level, region, round, date, questions and follow-ups from that debrief can be added to Interview Intel. Your name and email are never shared. Your notes on surprises and struggles, your preparedness rating and your outcome are not part of what is shared. You can change your choice by editing the debrief.