An AI interviewer runs a 45-minute system design round on any of our 20 questions: it asks, you answer, it probes what you said, and it ends with a scored debrief. It has the full reference design in front of it — the requirements, the numbers, the diagram and every deep dive — and you do not. That is the whole point.
No sign-in to hear the interviewer open the round. Pro from $6.58 a month, billed annually.
A conversation, not a form. One question at a time, and no answer visible until you have given yours.
The AI interviewer opens the way a real one does: the problem in two sentences, then "start with the requirements". Nothing is revealed in advance — no diagram, no numbers, no list of alternatives.
Short paragraphs, the way you would speak. Ask clarifying questions and you get a number and a commitment back, as in the real thing.
Name a component and it asks what backs it. Give a number and it asks where it came from. Pick an option and it asks what that costs you. It reacts to your answer, not to a script.
In a timed round it spends the 45 minutes deliberately: it cuts a rabbit hole, moves you on when requirements run long, and forces the close. Untimed, it follows you as deep as you want to go.
A score out of 100, what was strong, what was missing, and one concrete thing to practise next — against the reference material for that question, which it had all along and you did not.
Each one teaches patterns the others do not, and each can be run as an AI interview, read in study mode, or worked through step by step in practice. Design a URL Shortener is free in every mode; the rest are Pro.
bit.ly at scale: turn long URLs into short codes, redirect in milliseconds from anywhere, count the clicks.
Real-time team messaging: channels, DMs, presence, history and search for millions of concurrent users.
Accept card payments for an e-commerce platform, move money correctly exactly once, and be able to prove it.
The home timeline of Twitter, Instagram or Facebook: posts from people you follow, ranked, fresh, and fast for hundreds of millions of users.
Upload, transcode, and stream video to a billion viewers: a processing pipeline on one side, a CDN problem on the other.
Map tiles, place search, turn-by-turn routing with live traffic, and ETAs for a billion users.
Group online game players into fair, low-latency matches within seconds, at millions of concurrent players.
Match riders with nearby drivers in seconds, track trips live, and price them, for millions of concurrent users worldwide.
Deliver push, email, SMS and in-app notifications to hundreds of millions of users: the right channel, once, on time, and never at 3 a.m.
Real-time collaborative editing: many cursors in one document, every keystroke converging to the same text, offline edits merged on reconnect.
Serve a large language model behind an API: queue the requests, batch them onto scarce GPUs, stream tokens back, and keep every accelerator busy without blowing the latency budget.
A Redis-like cache as a service: hundreds of nodes, terabytes of memory, sub-millisecond gets, and a cluster that survives losing a machine mid-request.
Fetch billions of pages a month politely, discover new links, skip duplicates, and keep the copy fresh without getting the crawler banned.
A distributed service that decides, in under a millisecond, whether a request from a client is allowed under its quota.
Keep a folder identical on every device: upload huge files in resumable chunks, push only what changed, and reconcile two machines that both edited the same file offline.
Sell 60 000 seats in ten minutes to a million people at once: hold a seat while someone pays, never sell it twice, and keep the site up when demand is a hundred times supply.
Suggest the top ten completions after every keystroke, in under fifty milliseconds, from what the world searched for in the last hour — without suggesting anything that gets you in the newspaper.
Count a million ad clicks a second, show advertisers a dashboard that is seconds behind, stop a campaign the moment its budget runs out, and be exactly right when the invoice is cut.
Run millions of cron jobs and one-off tasks on time across a fleet, exactly once each, and keep going when a worker dies halfway through one.
Match buy and sell orders in microseconds, in a strict and provable order, never lose a fill, and stay correct when the matching engine dies mid-trade.
Every question is also a full written answer, in the five steps an interview runs in. It is what the AI interviewer has open while it asks you, and it is there for you too — before the round, to learn it, and after, to see what you missed.
Functional requirements with the constraint that makes each one interesting, an explicit out-of-scope list, and non-functional targets with real numbers.
Back-of-envelope estimates with the arithmetic shown. In practice mode you type each number and are graded against the derivation.
An interactive high-level design. Click any component for what it does, what backs it, and the non-obvious decision inside it.
Five user flows per question: the happy path, the read path, the scale-breaking case, a failure or retry, and a background job. Each step lights up the components and calls, with an API sketch.
Deep dives compare three or four realistic alternatives, pick one with concrete keys, thresholds and TTLs, then ask the follow-ups an interviewer asks next.
The main event. An AI interviewer asks one question at a time, probes what you say, moves you on when a section runs long, and ends with a debrief — a score, what was strong, what was missing.
Everything visible. Flows can autoplay across the diagram. Read a question end to end in about 40 minutes, the night before.
Every answer hidden until you ask. Type the estimates, draw or click the components a step touches, press Check, and see what you missed. A session timer counts only active time.
The theory behind the questions: one page per building block, written to be said out loud in an interview. Free to read.
One product, three ways to pay. Every plan unlocks the AI interviewer on every question, including the ones added after you buy.
For one focused interview loop: unlimited AI interviews while it lasts.
A year for the price of three months. For prep that runs over months.
The interviewer and every question added later, for good.
Payments by Stripe. Cancel subscriptions any time from the billing portal; 7-day refund on a first purchase. Full details.
A live conversation with an AI interviewer that runs a 45-minute system design round on one of our questions. It states the problem, asks you one thing at a time, probes what you said — what backs that component, where did that number come from, what does that choice cost you — and ends with a debrief: a score out of 100, what was strong, what was missing, and one thing to practise next. It has the full reference design for that question in front of it, which you do not, and it never reads it out to you.
Your choice at the start. In a timed round the AI interviewer owns a 45-minute clock and spends it the way a real one does: it cuts a rabbit hole, moves you on when requirements run long, and forces the close in the last few minutes. Untimed there is no clock at all and it follows you as deep as you want to go on whatever you want to go deep on.
Yes. On every question the interviewer will open the round and ask you its first question with no account and no plan. Answering it back is where Pro starts, because from that point every turn is a live call to the model. Design a URL Shortener is free in full, including the whole interview.
The same AI interviewer, drawing one of the questions reported at that company and running the round in the style that company’s round is reported to take — what it weighs, how it probes, where candidates go wrong. The house style comes from public interview guides and candidate reports, not from the company. It is practice, not a prediction.
Reading tells you the answer. Being asked makes you produce it under pressure, out loud, in your own order, with someone following up on the weak part. That is the gap this closes: the AI interview for rehearsal, and practice mode underneath it for the mechanics — typing each estimate and being graded, clicking the components a step touches and being corrected, answering the follow-ups before the model answer opens.
The AI interview is the main event: a conversation where nothing is revealed and you are scored at the end. Study shows every answer, to read end to end. Practice hides them and makes you produce your own, step by step, and grades what you produced. Study to learn it, practice to check you can reproduce it, AI interview to rehearse saying it under pressure. You can switch between the three from the header of any question.
Five parts, in the order an interview runs: functional requirements with explicit scoping, non-functional requirements with numeric targets, back-of-envelope estimates with the arithmetic shown, a high-level design diagram with five user flows that light up the components and calls of each step (with API sketches), and five or six deep dives that compare realistic alternatives, pick one, and then ask the follow-up questions an interviewer asks next. It is all of that material the AI interviewer is grading you against.
It grades only the transcript, against our reference material for that question, and it is told to be honest: 50 is borderline, 70 is a solid senior hire signal, 85 and above is rare. A round that barely started scores low, and the debrief says so rather than inventing an assessment. Treat the score as a signal about coverage and depth, not as a verdict from the company you are applying to.
Yes, to your account, so you can look back at a round and see how it went. What you type is sent to the model that plays the interviewer. See the Privacy Policy for how long it is kept and how to have it deleted.
Design a URL Shortener is open to everyone in all three modes, with no sign-in. On every other question the AI interviewer’s opening question is free too, so you can hear how a round starts before deciding.
The AI interviewer on every question, the company mock interviews, every question on the site and every question added later, practice mode with graded answers, and your session history synced across devices. All three plans include the same thing; they differ only in how you pay.
Monthly and annual plans renew until you cancel. Cancel from the billing portal at any time and you keep access until the end of the period you paid for. Lifetime is a single payment with no renewal.
Yes. Open the billing portal from the pricing page and switch plans; Stripe prorates the unused part of the current period.
If Pro is not what you expected, email within 7 days of your first purchase and it is refunded. Renewals are not refundable, so cancel before the renewal date if you are done.
Not to read the free question, and not to hear the AI interviewer open a round. Signing in with Google is required to answer it, to buy Pro, and to sync history across devices. Without an account, your progress lives in your browser.
Public interview guides and candidate reports. They indicate where a question has been reported, not a verified dataset, and a company name never implies endorsement.
Senior and staff loops at large tech companies, where the interview is 45 to 60 minutes, expects numbers, and spends most of its time on deep dives. Mid-level candidates use it too; study mode reads as a complete walkthrough.