The AI Pair Programming Round

The Shopify AI pair programming interview is a live coding session of about 75 minutes. You work on your own machine, in your own IDE, sharing your screen, with an AI coding assistant running. Shopify's job postings ask you to "come prepared with your own IDE," and recruiters now tell candidates that using AI is encouraged.

Most candidates find the problem easy. A small program, often a robot moving on a grid, grows as the interviewer adds requirements. What decides the round is how you run the assistant: what you ask it for, what you keep, what you throw out and how clearly you explain each call.

What Shopify says about AI in the interview

Less than you might expect. Engineering postings describe the technical round in one line: "You will be expected to complete a live pair programming session, come prepared with your own IDE." The careers page names the technical stages a craft assessment and craft interviews, and one way it describes the interviews is taking the lead on a problem "with your interviewer as your co-pilot." Neither page sets a rule for AI inside an interview.

The culture around the round is loud, though. The careers page says "AI is the baseline" and that everyone gets unlimited AI access. In April 2025 CEO Tobi Lütke posted his internal memo making reflexive AI use a baseline expectation for employees. That memo covers how staff work and doesn't set interview rules. The interviews point the same way, though, and one candidate was already using Copilot in the round before the memo went out.

Why the round allows AI

The clearest answer comes from Farhan Thawar, Shopify's Head of Engineering, on The Pragmatic Engineer podcast in July 2025. Asked whether Shopify uses AI in its interviews, he said yes and added, "We're embracing it." His reasoning was blunt: "If they don't use a copilot, they usually get creamed by someone who does."

He also said what he watches for. The AI "will sometimes generate pure garbage," and he likes that, because it gives him something to discuss. He'll stop on generated code and ask the candidate, "Is this good code? Is this not good code? Are there problems?" He'd like candidates to use AI for 90 to 95 percent of the work and do the rest themselves. That figure is his own preference, and Shopify hasn't published it as a rule. It still tells you what the interviewer checks, which is whether you can spot the bad line yourself.

The Shopify angle

Fix the one character yourself. Thawar called out a habit that annoys him: a candidate spots a one-character bug, then writes a prompt asking the AI to fix it. Let the assistant do the bulk of the typing, and when you already know the fix, make it by hand. Small moments like that show the interviewer you're the one steering.

Where the round sits in the Shopify loop

Shopify's careers page lists three stages: the Life Story, then a craft assessment that starts with an online assessment, then craft interviews. Candidates describe the online assessment as a set of short logic and reasoning questions plus a batch of values questions. The Life Story interview is a separate conversation about your career, and the Shopify interview process lesson walks the full loop.

Many loops include two AI pairing rounds. The first works as the technical screen, and the second comes at the onsite, often next to a project deep dive. Order and count vary by team, and a screen can run in CoderPad instead, so ask your recruiter which version you'll get. Shopify gives coding interviews to director-level hires too, because in Thawar's words, "people should be as close to the details as possible."

The setup: your machine, your IDE, your AI

You join a video call, share your screen and work in your own editor with your own AI tool. Candidates report using Cursor, Claude Code (including inside a JetBrains IDE), Cline in VS Code and GitHub Copilot, and none report a banned tool. Pick the setup you use every day, because the round moves at the speed of your normal workflow.

• Length: About 75 minutes. Usually 75 minutes, sometimes 60 or 90. Shopify's own 2022 guide to the round gave 75 to 90.
• Environment: Your machine, screen-shared. Your IDE, your terminal and your AI tool. Have the language, test runner and assistant installed and signed in before the call.
• Repo: Often a git repo. Reports describe an empty repo or one with only a README, and one candidate got a repo holding several small questions. Some describe pushing the work to a branch at the end.
• Language: Your choice. Any language unless you hear otherwise. Shopify's 2022 guide had front-end candidates build in HTML, CSS and JavaScript, with TypeScript and React fine.
• AI policy: Allowed, usually encouraged. Recruiters tell candidates AI is encouraged. An interviewer can still block it, so ask before the day.

One candidate reported losing early minutes setting up git on the call. The day before, clone a throwaway repo, create a branch and push it once, so the tooling is warm when the clock starts.

What the interviewer checks

Shopify doesn't publish a rubric. From what candidates describe and what Thawar has said, these matter most:

• You direct the AI. One interviewer reportedly told a candidate, "We don't want the AI to work for you. We want you to tell it what to do." Decide the structure yourself, then hand the assistant well-scoped pieces.
• You review what it writes. The interviewer may point at generated code and ask whether it's good, or what you'd improve before shipping it. One candidate said they were marked down for accepting AI output without reviewing it.
• You talk the whole time. "They expected me to explain my approach, not just produce code," one candidate wrote. Going quiet while the AI types is a common way to fail.
• You test it. Write tests as you go. Interviewers have pushed back on AI-written tests that only covered the happy path.
• Your code survives the next requirement. Each new requirement should slot in without a rewrite. Class and file structure, naming and extensibility all get looked at, and some candidates get a late question about concurrency.

Prompt wording seems to matter little. A candidate who got an offer said the interviewers didn't care about the prompts as long as the candidate showed they were competent. Put your attention on the code you keep.

Why finishing the problem isn't enough

Because the problems are easy, finishing proves little. Candidates have finished the build, even heard praise in the room, and still been turned down. One finished a rover problem with Claude Code. Another heard "This design is really good" about an auto-refund system and was turned down the next day.

One interviewer put it this way: "how many you get through doesn't matter, what matters is how you do it." Shopify's 2022 guide to the round said much the same: a well-approached, considered solution that's incomplete beats a complete one that's sloppy.

Problems candidates report

The problems are easy algorithmically. Candidates call them "on the very easy side" or "not hard," and the challenge sits in the design as requirements pile up. These shapes come up most. Your problem may differ, so use them to practice the format.

• A rover or robot on a grid. The most reported shape. Something like: parse turn and move commands, then report position and heading. Follow-ups add obstacles, several rovers or a 3D map, and one version ends with a review of the AI-written code.
• Secret Santa from a CSV. Read names and emails, give each person someone to buy for, and make sure nobody draws themselves. A follow-up can add pairs that aren't allowed.
• A classic data structure dressed up as a small service. Once the basic version works, the follow-ups change its limits, its storage or its rules.
• A Wordle-style word game. Load a dictionary, take guesses, give letter feedback that handles repeated letters and track stats, with unit tests.
• An automatic refund system built from a list of requirements, part design and part code.

The rover is worked through in Simulation & State Machines. The data-structure builds are in Hash Maps & Design and Trees & Graphs.

How to prepare for the Shopify AI pair programming interview

Practice the exact setup you'll use on the day, with the assistant on and a timer running. Each session should feel like the real round: a short brief, a working first version, then two or three new requirements.

Rehearse the talking too. Record one session and listen for long silences while the AI types. Fill them by saying what you asked for and what you'll check when the answer comes back.

How it differs from the classic pairing round

Expect a higher bar for speed. With an assistant doing the typing, the people you're compared with get further through the requirements than anyone typing by hand did in 2022. That's Thawar's "get creamed" point, and one commenter on Blind put it more simply: if AI is allowed, "it's going to be a very high bar."

Common questions