Cla64.ai

How a practice session works

Watch a whole session, start to finish - or read the same steps below. The example is a short bug-fixing exercise, Meeting Room Booking.

About 5 minutes, with captions. Prefer to read? Skip to the steps.

The steps

  1. 1

    Start with the instructions

    Every session opens on the instructions, full screen - no assistant yet, just the task. Your workspace starts in the background while you read, and the bar at the bottom shows how it's going. The clock only starts once the workspace is ready, so the wait doesn't cost you anything.

    The opening screen: the task's instructions, with 'Your workspace is starting' and a Start the exercise button along the bottom.
  2. 2

    Watch the video, then read the task

    Start with the short video at the top, then read the task below it. In this example there are two complaints - keep each one in mind, you'll need to explain every one. Once your workspace is ready, click Start the exercise.

    The instructions, with the second complaint outlined.
  3. 3

    The exercise and the assistant

    Your code is on the left, with the time left at the top. On the right is the assistant. It opens with a short welcome: how the session works, and where to find the instructions again - its Instructions tab.

    The workspace with booking.py open, and the assistant's welcome message outlined in the panel on the right.
  4. 4

    The assistant wears two hats

    It's your interviewer: it asks you questions, and answers yours about the task. And it's your AI assistant: it helps with practical things along the way. It won't write the solution for you - that part is yours.

    The Assistant panel, labelled Hat 1: Interviewer and Hat 2: AI assistant.
  5. 5

    Read the code before you change it

    Open the file the task is about and find the lines that decide the behavior people complain about - here, the three lines that decide what counts as a clash. Don't change anything yet.

    booking.py open in the editor, with the three lines of the overlap check highlighted.
  6. 6

    Run the tests first

    Click Run Tests. The terminal at the bottom is already open, and the results show up in the Unit Tests panel. A failing test tells you that something is wrong, not what: compare what it expected with what it got.

    The Terminal tab showing pytest's output for one failing test, with the line 'At index 1 diff: False != True' outlined.
  7. 7

    Ask about the rules (hat 1)

    Not sure how something should behave? Ask in the chat, the way you'd ask an interviewer - about the rules, not for the answer. Asking instead of guessing is one of the things you're scored on.

    The candidate asks whether back-to-back meetings clash, and the interviewer answers.
  8. 8

    Ask practical questions too (hat 2)

    Questions about tools - like how to run just one test file - get a direct answer. What it won't do is write the solution for you.

    The candidate asks how to run one test file and gets the pytest command back, plus the ▶ button for a single test.
  9. 9

    Fix one thing at a time

    Make the smallest change that explains one complaint - here, the <= signs that include the end time become <. Save (Ctrl+S) and run the tests again.

    The two less-than-or-equal signs outlined in booking.py.
  10. 10

    A passing test doesn't mean you're done

    The given test only covers the cases it checks. Think about the ones it doesn't - here, a new meeting that completely covers an existing one. Its start isn't inside the other meeting, and neither is its end.

    A timeline sketch: an existing 10-11 meeting, and a new 9-12 meeting covering it.
  11. 11

    Write your own test for it

    Put that case in a test of your own, in the tests folder, and run it. It fails - the system accepted the meeting. You've found the second problem.

    A new test file, test_overlaps.py, and its failing run in the terminal.
  12. 12

    Fix it, and test every fix

    Fix the rule, add a test for the first fix too, so each fix has a test that would have caught it, and run everything. All green - and this time you know why.

    Both tests in test_overlaps.py and a terminal showing 3 passed.
  13. 13

    The end-of-session conversation

    Click Finish session, then Start questions. The interviewer asks two or three questions about your solution, one at a time, and may follow up on what you said. A short answer, or "I'm not sure", is fine. You can still change your code while you answer. When it has no more questions, it says so, and the button changes to End session.

    The end-of-session conversation in the chat, with the button now reading End session.
  14. 14

    End the session

    Click End session and confirm. Only then does the session actually end, and your written feedback is prepared. If there's time before you finish, you can also tell the assistant you're ready for the optional next steps.

    The 'End this session?' confirmation dialog.

Now it's your turn

Read first, run the tests, ask about the rules, fix one thing at a time, and write your own tests for the cases nobody gave you. The interview guide covers these habits for any coding interview, not just ours.

Browse the exercises