A coding interview, out loud, with an interviewer who reads your code
Solving problems in silence is not what a coding interview asks of you. The real thing is solving one out loud while someone watches the editor, asks why, and changes the problem the moment you get it working. That is what this rehearses.
A voice interview with an AI interviewer that states a real problem, reads your code as you write it, runs it against hidden tests, and pushes on your solution — complexity first, then a constraint that breaks your approach. The report grades the code as well as the conversation.
First session free (7 min, scored report) · no card details · no subscription
What happens in the round
The whole problem, up front
The interviewer states it in the first turn — the task, what goes in and what comes out, the constraints, and a worked example — not drip-fed across the round.
A real editor, 16 languages
Python, JavaScript, TypeScript, Java, C++, C, C#, Go, Rust, Ruby, PHP, Swift, Kotlin, Scala, R and SQL. Code runs in all of them except SQL.
It reads your code as you write it
Every time it speaks, the interviewer has your current code in front of it, and the output of your last run. It asks about your approach, your edge cases and your correctness from what is actually on the screen.
Run it, test it, debug it
Run your code and see what it prints. In Python and JavaScript, run it against hidden test cases — ordinary ones first, edge cases after — and see which ones fail. Then debug it out loud: the interviewer sees the same output, and asks what you think broke.
The follow-ups real interviewers ask
A working solution is where the round starts, not where it ends. Time and space complexity, justified. Then a constraint that breaks your approach: the input no longer fits in memory, it arrives as a stream, the values repeat, it has to run concurrently.
Silence means you are working
A candidate who goes quiet is usually typing. It will not offer to skip the question because you stopped talking; if it wants to hear from you, it asks you to narrate what you are trying.
Hints, not answers
Stuck, you get a nudge. Ask it to just tell you the answer and it asks how you would approach it, then offers a smaller version of the problem — the way a good interviewer keeps you moving without solving it for you.
Long rounds run two problems
Under 35 minutes, one problem, pushed as deep as it goes. At 35 minutes or more, a second problem arrives at the halfway mark, the way a longer onsite coding round often runs.
A report that reads the code
The scored report grades your solution alongside what you said, and keeps your final code in its own tab, next to the full transcript.
The problems
49 problems, 34 medium and 15 hard, each with hidden test cases and follow-up questions of its own. The level you choose sets the mix: an intern never meets a hard problem, a staff engineer meets one more often than not.
Medium (34)
Group anagrams · Merge overlapping intervals · Top k frequent numbers · Longest substring without repeating characters · Product of array except self · Count subarrays summing to k · Course schedule · Minimum meeting rooms · Word break · K closest points to the origin · Number of islands · Three numbers summing to zero · Decode a nested encoded string · Days until a warmer temperature · Search a rotated sorted array · Longest consecutive sequence · Spiral order of a matrix · Rotate a square matrix · Container with most water · Jump game · House robber · Unique paths with obstacles · Sort an array of three values · Longest palindromic substring · Remove the fewest parentheses to make a string valid · Kth largest element · Set matrix rows and columns to zero · Task scheduler with a cooldown · Generate valid parentheses · Combination sum · Word search in a grid · Find the duplicate number · Letter combinations of a phone number · Fewest coins to make an amount
Hard (15)
Rate-limited grouping · Edit distance · Largest rectangle in a histogram · Merge k sorted lists · Word ladder · Median of two sorted arrays · Alien dictionary order · Longest increasing path in a matrix · Basic calculator · Count n-queens arrangements · Largest square of ones · Regular expression matching · Trapping rain water · Minimum window substring · Sliding window maximum
The other technical rounds
System design
One open-ended prompt — a URL shortener, the backend for a ride-sharing app — driven through requirements, architecture, data model, APIs, scaling and trade-offs. Sketch schemas and API signatures in the editor; the interviewer asks what breaks first at ten times the load.
Start a system design round →Technical deep-dive
The interviewer picks projects and technologies from your resume and drills down: why each choice was made, how it worked underneath, what the hard problems were — until it finds the edge of what you really know.
Start a technical deep-dive →Mock interview practice · Interview preparation by role · The interviewers
Common questions
Can I practise a live coding interview with AI?
Yes. Choose Software Engineer, or any engineering role, and the Live Coding round. The interviewer states a problem, you solve it in the editor while talking it through, and it asks its follow-up questions about the code you actually wrote.
Does the AI interviewer actually see my code?
Yes. Your current code, and the output of your last run, go to the interviewer every time it speaks. It reacts to what is in the editor — the unhandled empty input, the loop that runs one step too far — rather than to what you said you wrote.
Which programming languages can I use?
Python, JavaScript, TypeScript, Java, C++, C, C#, Go, Rust, Ruby, PHP, Swift, Kotlin, Scala, R and SQL. Code runs in all of them except SQL. The hidden test cases run in Python and JavaScript.
Can I run my code against test cases?
In Python and JavaScript, yes: every problem has hidden test cases, ordinary ones first and edge cases after, and you see how many passed and which ones failed. In 13 more languages you can run your code and see its output.
What follow-up questions does it ask?
Once your solution works: its time and space complexity, and why — it does not accept a number without the reasoning. Then it changes the problem the way real interviewers do: the input no longer fits in memory, it arrives as a stream, the values can repeat, it has to run concurrently. Each problem also carries follow-ups of its own.
What kind of problems are they?
49 problems — 34 medium and 15 hard — across arrays, strings, matrix, dynamic programming, two pointers, backtracking, hash map and heap. The level you choose sets the mix: an intern never gets a hard problem, a staff engineer gets one more often than not.
Is it free?
Your first session is free and needs no payment details. It runs for 7 minutes and ends with a scored report. After that it is a one-time pass, not a subscription — nothing renews.
Is there system design practice as well?
Yes. The System Design round gives you an open-ended prompt, and the same editor for sketching schemas, APIs and pseudocode, with the interviewer asking what breaks first at ten times the load and why you chose each component.
Does this help during a real interview?
No, and deliberately not. Nothing here runs while you are being interviewed. It is built to be used before, so that you do not need help during.
The first round is free
No card, no trial that quietly turns into a subscription. One round, scored, with your code and the transcript — enough to know whether rehearsing with an interviewer that reads your code is useful to you.