Guides · Coding assessment platforms

HackerRank CodePair: what the live pad looks like from your side

2026 · ~6 min read · OfferPilot AI

CodePair is a live pair-programming session in a shared browser editor, not a timed test you hand in. You and the interviewer type in the same file and see each other's cursor, and there is no submit button, just an editor that autosaves while somebody watches. You can normally run code and read the output in a console, video is either a panel inside CodePair or a separate Zoom call, and screen sharing is a distinct request you approve. Most of the outcome is set in the first five minutes, before you type a line.

What you see when you open the link

The invite is a URL in the calendar entry. It opens in the browser and normally asks for a name rather than a HackerRank login, though some companies wire it to an account, so click it early.

  • One large editor in the middle. The shared file. It may be empty, hold a signature the interviewer pasted, or sit beside a question pane if they pulled the problem from HackerRank's library.
  • A language dropdown near the top. Whatever is selected is what the run button compiles, and it defaults to whatever the interviewer set, not always the language you told the recruiter. Change it in the first thirty seconds, before you have code to lose.
  • A console or output pane, often collapsed. Run results land here, and there is generally a box for custom input fed to stdin.
  • An editor, not your IDE. Syntax highlighting and bracket matching, yes. Past that, assume less than you are used to: no project-wide rename, no auto-import because you referenced a library, and completion that may not flag the misspelled variable you will meet at run time instead. Type the imports first, before you need them.
  • A diagram or whiteboard mode, for when the round drifts into schema or system design.
  • Things you cannot see. The interviewer has a private notes and rating panel, so when they go quiet and type, they are writing there rather than in your file. The code history can also be replayed later, so the record of the round is how the file evolved, not a final snapshot.

The interviewer is typing in your file, not a copy of it

CodePair is collaborative in the shared-document sense: their cursor appears in your buffer with their name on it, and either of you can edit any line at any moment. They use it to paste an example input under your function, add three lines of harness, fix the typo you have stared at for forty seconds, or write the follow-up as a comment below your solution while you finish the first one. None of that is a trap.

Two consequences. There is no private scratch space, so think in a comment block at the bottom and say that is what you are doing. And your typing is the performance: they see the wrong variable name, the backspacing, the moment you rename i to left. Nobody grades that. What reads badly is thirty seconds of a motionless cursor followed by twenty perfect lines arriving at once, because they watched both halves.

Can you run code, and will there be test cases?

Running code is standard. Pick the language, press run, output appears in the console. Compile errors come back the same way, in front of the person interviewing you. Test cases are the part candidates assume wrongly: whether you get any depends on where the question came from.

How the question got into the padWhat you can run againstWhat that means for you
Pulled from the HackerRank question libraryQuestion text in its own pane, usually with sample casesRead the sample input format closely, it dictates your signature
Interviewer pasted a prompt into a blank padNothing. The pad is a file and a run buttonYou write the driver and the assertions yourself
Interviewer dictates it out loudNothing, and the wording is written down nowhereType your restatement into the pad so you both agree on it

So assume you are writing your own checks. Add three cases before you run and name them aloud: the example, an empty or single-element input, and the edge case you flagged earlier. Print statements are fine. If the run button misbehaves, say so rather than debugging the platform in silence, since the interviewer can switch languages or hand you a harness.

The video panel and the screen-share request are separate things

CodePair has its own audio and video calling, so a round can run in one browser tab. Plenty of interviewers ignore it and send a Zoom, Teams or Meet invite anyway, leaving you on a call in one window and the pad in another, so check the calendar for a second link. With the built-in panel, expect a browser prompt for microphone and camera before anything transmits, and a video tile you can usually collapse.

Nothing shares your screen automatically. Screen sharing is a separate, explicit request, and it comes up less often than candidates fear: usually to see something outside the pad, a local IDE, a take-home repository, or why your run keeps failing. Your browser, not HackerRank, then shows the picker: entire screen sends every monitor surface including notifications, window sends one application, and tab sends one browser tab and nothing else, which is the right answer when they only want to watch the editor.

The first five minutes, said out loud

The real difference from a solo assessment is that thinking has to be audible. Interviewers form a view before the first line of real code exists, and the fastest way to lose the round is to start typing and reach minute seven of a solution nobody agreed to.

Restate the problem in one sentence

"Given an unsorted array and a target, I return the indices of the two values that sum to it, and there is exactly one answer." Your words, not theirs. If you have misread it, you find out ninety seconds in, not nine minutes in.

Ask exactly two constraint questions

Two, not six. Six sounds like stalling. The first is scale, because it decides which complexity is acceptable: "roughly how large is n, thousands or millions?" The second is the ambiguity that forks your code: duplicates, already sorted, empty input, value or index.

Name the brute force and why you are leaving it

"The obvious version is every pair, O(n squared) time, O(1) space. Fine at a thousand elements, not at a million." Ten seconds, and it gives them the chance to say "actually, code that one".

State the approach and its Big-O before you touch the keyboard

"I would rather do one pass with a hash map from value to index, checking for the complement before I insert. O(n) time, O(n) space, and it handles duplicates because I check first." Then the sentence that buys you the rest of the hour: "does that sound reasonable to start with?"

Type the skeleton, then fill it in

Signature, two or three comments mirroring the approach, then the body. If you stall, they show the interviewer where you are, and they keep you narrating.

Where an assistant is the wrong tool for this round

CodePair is the format where live help helps least. The interviewer is inside your editor watching characters appear, so a finished solution arriving all at once after a pause is conspicuous in a way it never is in an asynchronous test. The round is conversational too, and an approach you did not reason through yourself collapses at the first follow-up.

  • Capture exclusion does not address the risk here. The overlay in Code Copilot is excluded from screen capture at the operating system level, so it is absent from Zoom, Teams, Meet, OBS and screenshots. That covers screen sharing and recording only. It does nothing about a webcam, a lockdown browser, eye tracking or a person watching you, and in CodePair the thing being observed is the pad itself, live.
  • Windows 10 and 11 only, with no macOS, Linux, mobile or extension build.
  • The rules belong to the company. Some employers allow documentation and assistance in a live round, many do not, so read the invite and ask if it is ambiguous.

Preparation tools fit before the round. Practise until restate, two questions, brute force, approach and complexity is a habit: narrate the opening to an empty room, then check your reasoning against a worked solution with a dry run and a complexity line. The question banks and the company Q&A generator in OfferPilot AI cover the rounds either side of the pad, and the free tier is 5 minutes a day with no card.

Related guides

Practise the first five minutes until it is boring.

Work through real prompts with intuition, a dry run and Big-O, so narrating in a live pad is a habit rather than a performance.