That changes how you prepare. Practice built for one of these rounds is wasted on another. You could rehearse Amazon's format, walk into Stripe's, and be asked why you didn't use the AI more.
Here is where this comes from. Candidates post about these rounds on a large interview forum. We counted 38 threads on Amazon's round, 17 on Stripe's, 16 on DoorDash's, 11 on Instacart's and 3 on Coinbase's. Everything below about those five companies comes from those threads. For Google and Meta we used the companies' own statements and press reports, plus what candidates describe. For Shopify we used candidate posts and what its head of engineering has said in public.
What each company runs
| Company | The round | The assistant | What stands out |
|---|---|---|---|
| Amazon | SDE online assessment: a 40 minute algorithm question, then 60 minutes on a buggy repository | Built into the assessment | It helps you find the file and explains the code, but won't write the fix |
| Stripe | One coding round, since spring 2026 | Inside HackerRank; Gemini Flash and Claude Opus reported | Interviewers push both ways on how much you use it |
| DoorDash | AI Code Craft, onsite | Your own IDE and AI tools, screen shared | Interviewers push for speed |
| Instacart | Online assessment, 90 minutes on CodeSignal | Claude in the editor, switched off for the bug fix | Passing the visible tests isn't the whole score |
| Coinbase | Onsite: a bug, a design question, a timeout to diagnose from logs, a code review, all on an existing service | Inside the IDE, sees the task and your open files | Screen share is mandatory |
| Code comprehension round, piloted for junior and mid level roles on select US teams | Gemini, the approved assistant for the pilot | Read, debug and optimize an existing codebase | |
| Meta | Piloted since October 2025; candidates report it replaces one of the two onsite coding rounds | Built into CoderPad | Candidates report about 60 minutes |
| Shopify | AI assisted pair programming, screen shared | Whatever AI tools you use, in your own IDE | Whether you can read the code the AI writes |
Each company name opens its own page: the task, the assistant, and what trips candidates up. We also have pages for Pinterest, LinkedIn, Rippling and Canva. They are not in the table yet.
Three things that are actually different
Who writes the code
At Amazon, you do. In the firsthand reports the assistant helps you find the right file and explains the code, and stops there. One candidate says it refused as soon as they got close to the answer. The forum's own summary of the round says some versions will edit code, so read the instructions on the day. Plan on typing the fix yourself.
At Stripe, DoorDash and Shopify, the AI writes and you answer for it. Stripe interviewers ask why you trust code you didn't write. Shopify's round tests whether you can read the code the AI produced. A DoorDash interviewer kept urging one candidate to accept a change. The candidate accepted it, had no time to read it, and expected a big deduction.
Our take: this is the real split. Before you practice anything else, find out which side of it your company is on.
Where it runs
Four of the eight put you in their sandbox: Amazon's assessment, Stripe's HackerRank, Instacart's CodeSignal, Meta's CoderPad. Two put you in your own editor with your own tools, screen shared: DoorDash and Shopify. Coinbase gives you an assistant inside the IDE that sees the task and your open files, and screen share is mandatory.
This matters more than it sounds. In a sandbox you get whatever model they picked. One Stripe candidate's advice was to pick Gemini Flash because the other options were slow. In your own editor, your setup is part of what is being judged.
What the score is
Passing the tests is not the score. One Amazon candidate passed four of six tests and still heard from the recruiter. Another didn't pass them all and went straight to the onsite. An Instacart candidate passed every visible test and found out afterwards that there were other tests. Another Instacart candidate scored 650 out of 1000 and was rejected.
What else is scored depends on the company. DoorDash told a candidate that velocity is a key criterion. Stripe asked a candidate who wrote a skeleton by hand why they weren't using AI, and asked the same candidate, when the AI wrote more, why they were so confident in its code. Nobody on the Stripe side is grading the code alone. They are grading the judgment around it.
One thing to ignore
Several prep sites describe Google's round as 60 minutes in CoderPad with three phases. Google has not said that. The press report that broke the story in May did not say it either. It does match how Meta's round is usually described. What Google did confirm is narrower: a code comprehension round, read, debug and optimize an existing codebase, Gemini as the approved assistant, piloted for junior to mid level roles on select US teams. Don't rehearse a script for a round nobody has described.
What to do about it
- Find your company's row before you practice. The assistant, the environment and the grading are different in each one. Generic AI interview prep is the new "grind 500 problems": it feels productive and it misses the point.
- Practice under the round's constraints, not with a better AI. No candidate we read blamed the model. They lost on reading code they didn't write, on tests they didn't run, and on questions they couldn't answer. Practice with an assistant that won't do the thinking for you, on a codebase, with someone asking why afterwards.
- Read the code before you accept it, even when pushed. The DoorDash candidate who accepted a change too large to read expected to lose points for it. Speed counts, and so does knowing what you shipped.
- Run tests on what you accept. The Stripe workflow that was graded well had the assistant read the README and propose a plan, then the candidate reviewed its work, wrote tests and ran them. The Instacart candidate who passed all the visible tests still didn't pass.
- Be ready to say why. Why you trust this code. Why you didn't use the assistant here. Stripe asks both in the same round.
What we don't know
Most of this comes from candidate threads, not from the companies. Coinbase has three threads, so treat that row as early. Google's round is a pilot and may change. Meta's duration and the round it replaces come from candidates, not from Meta. If you sat one of these rounds recently and it went differently, tell us and we will update the page.
What every version of the round grades, and how to prepare for the parts they share: the AI coding interview guide.
Practice on Nyrion
A real codebase, an assistant that won't do the thinking for you, and an interviewer that asks you about the code afterwards. Problems are tagged by company. No login required to look.