Practice AI-Coding Interviews

Q: What is an AI coding interview?

An AI coding interview is a technical interview where you solve a realistic engineering problem with an AI assistant at your side, and you're assessed on how well you direct it, verify what it gives you and make the decisions yourself. On Cla64.ai you practice exactly that: a real project in a cloud workspace in your browser, an AI you can work with, and an interviewer that asks about your reasoning.

Q: How is Cla64.ai different from LeetCode?

LeetCode is for algorithm puzzles: a short function, a known trick, an accepted or rejected verdict. Cla64.ai tests different skills, with AI in the room: understanding a codebase, debugging, designing and building something new, testing, and explaining your decisions - and you get a written review of how you worked.

Q: Does the AI write the code for me?

Not on Cla64.ai. Some companies now let candidates use AI to generate code in their interviews - Meta's AI-enabled round, for example - but you still have to understand and explain every line. Cla64.ai practices that part: the AI answers language, library and tooling questions directly and gives one hint at a time when you're stuck, while the logic and design stay yours. Your review describes how you used the AI.

Q: How long does one practice interview take?

60 minutes for most exercises; a few run 30 or 45, and a few longer ones 75 or 90. Free interviews stop at 40 minutes. The clock starts when you start reading the task, and you can finish earlier.

Q: How is my work assessed?

Exercises that come with starting code are checked with hidden tests. Blank-project exercises have no fixed tests: you demo your solution in the terminal, and an AI reviews your code, the demo and the conversation. Either way you get a written review - what went well, what to improve and how a strong solution would approach it - with no score and no ranking.

Q: Is Cla64.ai free?

3 exercises are free, no account needed. Sign up free to see your feedback. There are also premium sessions, from $10 (in your local currency where we price it).

Work in a cloud-hosted workspace with an AI assistant and interviewer.

See catalog

No installation · Real code execution · Detailed feedback · Start for free

Watch a session

InstructionsExercise58:41 left

Run⌄Run Tests

FILES

⌄ src

PY__init__.py

PYrate_limiter.py

⌄ tests

PYtest_traffic_pattern.py

.gitignore

PYconftest.py

PYmain.py

README.md

PYrate_limiter.py×

12class RateLimiter:
13    def __init__(self, max_requests: int, window_seconds: float, initial_history: dict = {}):
14        self.max_requests = max_requests
15        self.window_seconds = window_seconds
16        self.history = initial_history
17
18    def is_allowed(self, user_id: str, timestamp: float) -> bool:
19        """ Returns True if this request should be allowed ... """⋯
24        bucket = self.history.get(user_id, [])
25
26        recent = [t for t in bucket if timestamp - t <= self.window_seconds]
27        allowed = len(recent) < self.max_requests
28
29        bucket.append(timestamp)
30        self.history[user_id] = bucket
31
32        return allowed

Unit TestsTerminal+

candidate:~/project$

.. 2 passed in 0.04s

AI interviewer

Ask or reply to the interviewer

Not a coding puzzle

Real engineering. Not coding puzzles.

Unfamiliar code, bug reports, tradeoffs and follow-ups. And AI, the way you use it at work.

How it works

One interview. Four steps.

1
Choose an interview
Pick a scenario and a language: Python, TypeScript, JavaScript, Java or C++.
2
Open your workspace
Watch a short video brief while your workspace gets ready. Nothing to install.
3
Work with the AI
Ask it for syntax, APIs or a hint. It follows your code, asks why, and adds requirements as you go.
4
Get your feedback
Hidden tests run against your code, and you get a written review of how you worked.

The scenarios

Practice real engineering scenarios

Each interview drops you into a small codebase with a task from a real job.

Debug
Rate Limiter
Users are blocked under quota in production. Find the cause and prove the fix with a test.
Build
Bus Control Center
Listen to a live stream of bus positions and snapshot their progress every 3 seconds.
Design
Push Notification Router
Route alerts across email, SMS and push from a blank project, then demo it live.
Incident
Load Balancer
Start from an incident report, restore correct routing, and explain what broke.

How the AI works with you

AI that helps and challenges you

Helps when you ask: Syntax, APIs, a library call, a hint when you're stuck. Ask it the way you'd ask AI at work.
Challenges your decisions: It reads your code as you write it and asks why.
Leaves the engineering to you: The design calls are yours to make and defend. Your report shows how you used AI along the way.

Quick one: what's the Python call for dropping items off the front of a deque?
popleft() on a collections.deque, O(1). Why move off the list, though? Your version passes the tests.
It's rebuilt on every call, so each request scans the whole window. With a deque, each expired timestamp is dropped exactly once.
Good. Partners are about to send 10,000 requests a minute. How would you prove memory stays bounded?

Your interview report

Your interview. Your feedback.

No score and no ranking. What the tests found, how you reasoned and used AI, and what to do differently next time.

Session review

Rate limiter in production

Language: Python
Time used: 52 of 60 min

Test results

3 of 4 passed, including hidden tests

✓test_traffic_pattern
✓test_denied_not_countedhidden
✓test_instance_isolationhidden
✕test_boundaryhidden

What went well

Went from the support ticket to the exact line in under ten minutes, and said out loud why raising window_seconds wouldn't help.

What to improve

A request exactly window_seconds old still counts against the user. Ask which side of the boundary is inclusive before choosing <=.

How a strong solution would approach it

Reproduce each ticket as a failing test first, then fix the three causes separately: let a request expire exactly at the window edge, stop recording denied requests, and give each limiter its own dictionary.

Rate limiter in production

Python, 52 of 60 min

✓
✓
✓
✕

test_boundary failed

What went well

Went from the support ticket to the exact line in under ten minutes, and said out loud why raising window_seconds wouldn't help.

What to improve

A request exactly window_seconds old still counts against the user. Ask which side of the boundary is inclusive before choosing <=.

Pricing

Practice the free exercises with no sign-up. Premium exercises come in packs.

Free

$0

No card. No account to start.

3 free exercises, up to 40 minutes.
Do them as often as you like.
Sign up free to get your written feedback.

Session packs

The premium exercises, at full length, with written feedback. Use within one month after purchase; nothing renews.

1 interview
$10 each
$10
Buy
3 interviews
$6.67 each, save 33%
$20
Buy
5 interviews
$5 each, save 50%
$25
Buy

Prices are in USD, tax included. No extra charges at checkout. Free exercises never use a pack - they stay free, with feedback, once you sign up.

Q&A

What is an AI coding interview?

An AI coding interview is a technical interview where you solve a realistic engineering problem with an AI assistant at your side, and you're assessed on how well you direct it, verify what it gives you and make the decisions yourself. On Cla64.ai you practice exactly that: a real project in a cloud workspace in your browser, an AI you can work with, and an interviewer that asks about your reasoning.

How is Cla64.ai different from LeetCode?

LeetCode is for algorithm puzzles: a short function, a known trick, an accepted or rejected verdict. Cla64.ai tests different skills, with AI in the room: understanding a codebase, debugging, designing and building something new, testing, and explaining your decisions - and you get a written review of how you worked.

Does the AI write the code for me?

Not on Cla64.ai. Some companies now let candidates use AI to generate code in their interviews - Meta's AI-enabled round, for example - but you still have to understand and explain every line. Cla64.ai practices that part: the AI answers language, library and tooling questions directly and gives one hint at a time when you're stuck, while the logic and design stay yours. Your review describes how you used the AI.

How long does one practice interview take?

60 minutes for most exercises; a few run 30 or 45, and a few longer ones 75 or 90. Free interviews stop at 40 minutes. The clock starts when you start reading the task, and you can finish earlier.

How is my work assessed?

Exercises that come with starting code are checked with hidden tests. Blank-project exercises have no fixed tests: you demo your solution in the terminal, and an AI reviews your code, the demo and the conversation. Either way you get a written review - what went well, what to improve and how a strong solution would approach it - with no score and no ranking.

Is Cla64.ai free?

3 exercises are free, no account needed. Sign up free to see your feedback. There are also premium sessions, from $10 (in your local currency where we price it).

Get started

Be ready when the real interview starts.

Practice the interview. Use the tools. Get challenged. Learn where you need to improve. Start free, no sign-up.

Practice AI-Coding Interviews

Concert Tickets

Real engineering. Not coding puzzles.

One interview. Four steps.

Choose an interview

Open your workspace

Work with the AI

Get your feedback

Practice real engineering scenarios

Debug

Build

Design

Incident

AI that helps and challenges you

Your interview. Your feedback.

Rate limiter in production

Test results

What went well

What to improve

How a strong solution would approach it

Rate limiter in production

What went well

What to improve

Pricing

Free

Concert Tickets

Session packs

Q&A

What is an AI coding interview?

How is Cla64.ai different from LeetCode?

Does the AI write the code for me?

How long does one practice interview take?

How is my work assessed?

Is Cla64.ai free?

Be ready when the real interview starts.

Concert Tickets