2026 Stripe AI Coding Interview, Experienced SWE Onsite and Phone Screen: 6 Experiences
- Company
- Stripe
- Reports
- 6 interview reports
- Level
- Full-time
- Stages
- Phone screen → Onsite
Five of these six experienced candidates got the same Stripe AI coding task, a transaction rule parser that starts with simple equality checks and grows into AND/OR logic and then parentheses. None of the six reported an offer.
Coverage: Experienced full-time software engineers (an SDE-2, a mid-level engineer, a backend engineer, a senior candidate, a staff-level loop and one general SWE loop). Five took the AI-assisted coding round in the virtual onsite and one in the phone screen. Interviews ran from April to October 2026. No locations were given.
What came up in Stripe's AI coding round
The task hardly varies. Every candidate who named it described the rule parser. The one who forgot the prompt described a session where the AI wrote all the code and the candidate reviewed it. Here is how often each topic appeared across the 6 reports.
| Topic | Reports | Most recent report |
|---|---|---|
| Transaction rule parser / evaluator | 5 of 6 | October 21, 2026 (report date) |
| Tests and edge cases inside the round | 4 of 6 | October 21, 2026 (report date) |
| Talking through the approach and the AI's code with the interviewer | 4 of 6 | October 21, 2026 (report date) |
| AI assistant built into the interview platform | 4 of 6 | October 21, 2026 (report date) |
| Task given in stages (parts) | 3 of 6 | October 21, 2026 (report date) |
| Picking the model | 2 of 6 | October 21, 2026 (report date) |
| Not knowing what the round is graded on | 2 of 6 | October 21, 2026 (report date) |
One candidate said the assistant itself made several mistakes. Another mentioned that whether to write code by hand or with AI was agreed with the interviewer.
A candidate can show up in several rows, so the rows don't add up to 6.
Questions we found
AI Coding
| Question | Frequency | All last seen dates |
|---|---|---|
| Transaction rule parser: equality conditions, then AND/OR, then parentheses (practice versions: Evaluate Ordered Action Rules, Transaction Rule Expression Evaluator, Parse Boolean Rule Expressions) | 5 reports | Reports: October 21, 2026 (report date); September 29, 2026 (report date); September 23, 2026 (report date); September 14, 2026 (report date); April 30, 2026 (interview). Additional FastPrep history: November 5, 2026; October 2, 2026; August 30, 2026; August 29, 2026; August 24, 2026; July 28, 2026 |
The staff-level candidate in entry 05 couldn't remember their prompt, so they aren't counted here.
6 detailed candidate timelines, loops, and screens
01: Two parts done, a no-hire on the AI round only
Report date: October 21, 2026 · Interview period: October 2026 · Role: SDE-2 · Language: Python · Result: No hire on the AI round, rejected overall
Earlier steps
- An OA, then a recruiter call, then a technical screen.
- An integration round and a debugging round. Both came back as Hire.
AI coding round (fourth interview)
- HackerRank with a built-in assistant. The candidate never switched away from the default model, and they later counted that as their mistake. They said the assistant got several things wrong.
- The task was a transaction rule engine. Rules were plain key/value conditions at first, then conditions joined with AND/OR. In a follow-up they warned that the problem hides a lot of edge cases.
- They finished two parts, talked through their approach the whole time and wrote tests they felt good about.
The verdict for this round was no hire. The candidate said they couldn't tell why so many people fail it.
02: Three stages, from string matching to an AST
Report date: September 29, 2026 (same-author follow-up October 10) · Interview period: July–September 2026 · Role: Software engineer, experienced hire · Language: Python · Result: Rejected; they rated the loop the hardest level
AI Programming Exercise (onsite)
- The platform had its own AI built in.
- The task was ordered transaction rules. Input: one transaction as a dictionary of fields, plus an ordered list of rules. Output: accept or block.
- Part 1: string equality conditions. Rules are checked in order and the first one that matches decides. If none match, the answer is accept. The field name and the constant can appear on either side of
=. Inputs are always well formed. - Part 2: keep Part 1 working, then add Boolean fields and AND/OR. The interviewer asked what happens when a merchant name itself contains the word "and" or "or".
- Part 3: parentheses, mixed with equality and AND/OR. They discussed building an abstract syntax tree (AST, a tree of the expression's pieces) and evaluating it recursively. Most of Part 3 was talk, not finished code.
Looking back, the candidate said the round needs two things at once: using the AI to move fast, and reading what it produced closely enough to catch where it drifts from the spec. Their example: "the first matching rule decides" is easy to turn into "any block rule wins", which is wrong. They didn't think the prompts had to be fancy.
The same loop had separate integration and debugging rounds.
03: The interviewer let Opus finish
Report date: September 23, 2026 · Interview period: July–September 2026 · Role: Software engineer, mid level · Result: Rejected
AI coding (onsite)
- The task was the rule parser earlier candidates had described.
- Candidates could pick the model. This one used Opus. Whether to write with the AI or by hand was something they settled with the interviewer.
- They kept a running conversation with the interviewer, but admitted they couldn't fully follow all of the AI-written code in the moment. They had the impression the interviewer was only half following it too.
- Near the end the interviewer said he trusted Opus to finish the remaining requirements, so there was no need to keep going.
Rest of the onsite
- Integration, debugging and a behavioral round. The debugging round got stuck when the local debugger wouldn't run.
- A week later the recruiter said the overall signal was positive but short of the bar, mainly because of debugging and integration. The AI round wasn't named as a problem. They were invited to apply again after the cooldown.
They said their mid-level loop had no system design.
04: A Codex agent and test cases you write yourself
Report date: September 14, 2026 · Interview period: July–September 2026 · Role: Senior software engineer · Platform: hosted editor (the candidate wasn't sure if it was CoderPad or CodeSignal), no local setup · Result: Rejected
Round 1: debugging
- A bug-fixing round in a local IDE. They lost about half an hour installing the IDE and plugins because they had skipped the setup notes in the invite email.
Round 2: AI coding
- The round was framed as using a Codex agent to solve a coding problem.
- Two inputs, as the candidate remembered them: a list of rules written as strings, such as "card name is not X" or "merchant name is not Y", and an array of strings, one per transaction.
- The goal was to get the agent to write code that passes the test cases. The candidate had to write test cases too.
Rounds 3 and 4
- An integration round, with no AI allowed.
- A manager behavioral round on past projects and leadership.
The loop was for a senior role but had no system design and no classic algorithm round. They heard back in two or three days. They blamed the debugging round and called the others average.
05: A phone screen where the AI writes everything
Report date: August 18, 2026 · Interview period: July–September 2026 · Stage: Phone screen · Level: Staff-level loop · Result: Rejected after the onsite
Phone screens (two rounds)
- One round was coding with AI. The candidate's read: you mostly show how you work with the assistant. Let it write all the code. Your part is to review it: ask what algorithm it chose and how it set up the functions and classes, then judge the code it wrote. They didn't remember the task.
- The other phone round was system design.
Onsite (four rounds)
- Two behavioral rounds, a project deep dive and an integration round. AI coding wasn't part of it.
- Feedback: the technical side was fine, but the behavioral answers sounded execution-level and didn't show staff scope.
06: The earliest report, and an interviewer who said the format was new
Interview date: April 30, 2026 · Role: Backend software engineer · Platform: HackerRank · Result: Failed this round; the recruiter suggested the other rounds went well
AI coding (onsite)
- The candidate noticed that Stripe's coding round had become an AI round. In a reply they said the interview's title included AI.
- HackerRank had an AI tool built in. They found it good and guessed it was Claude underneath.
- The interviewer pasted the problem: a rule parser that decides whether a condition is true or false. You solve it with the AI.
- Three parts. They said the AI made all three doable. Tests were required, though the AI wrote them.
This was their second AI interview in a short time. With classic coding they knew what was graded: complexity, clean code, corner cases. Here they couldn't tell, because pasting the prompt into Claude more or less produces the answer. In a reply, they said the interviewer told them the AI format had started only in the past few weeks. Months later they added that this was the round they failed.
Patterns across the reports
-
The same staged rule task. Five candidates (01, 02, 03, 04, 06) described a rule parser for transactions. Entry 02 gives the full ladder: equality, then AND/OR, then parentheses. Entry 01 matches its first two stages, and 06 also had three parts. Entry 04 remembered "not equal" rules over a list of transaction strings, which may be a variant. How far people got varied: two parts (01), three parts with the last one mostly discussed (02), all three (06).
-
The AI comes with the platform. Four candidates used an assistant built into the interview tool: HackerRank in 01 and 06, the platform's own AI in 02, and a hosted editor with a Codex agent in 04. Model choice came up twice. Candidate 01 left the default on, and 03 picked Opus.
-
Tests are part of the job. Candidates 01, 04 and 06 all wrote or were asked for tests. In 06 the AI wrote them. In 02 the interviewer pushed on one edge case: a merchant name that contains "and" or "or".
-
Explaining the AI's code. Candidates 01, 02, 03 and 05 all described talking through either their plan or the generated code. Their experiences differed. Candidate 05 saw reviewing the AI's code as the whole job. Candidate 03 couldn't keep up with what Opus wrote, and the interviewer told them to stop before the end.
-
Failing the AI round alone, without knowing why. Candidates 01 and 06 both failed this round while their other rounds went well, and both said they couldn't tell what it was grading.
Keep practicing
- Transaction Rule Expression Evaluator: the rule task from entries 01, 04 and 06, with AND/OR and parentheses evaluated against transaction rows.
- Evaluate Ordered Action Rules: the version from entry 02, with ordered accept/block rules where the first match wins.
- Parse Boolean Rule Expressions: the version from entry 03, which focuses on parsing a rule into a tree.
- Stripe AI coding question bank: Stripe's project-coding problems in one list (sign-in needed).
- Stripe interview guide: Stripe's interview process and question list.
- Stripe hiring insights: hiring data for Stripe.
- Interview experiences: the full Experience feed.
- Interviewing for an internship? Intern AI rounds are covered in 2026 Stripe Intern AI Coding: A Prep Guide from 10 Candidate Experiences and in one intern's full loop, Stripe SWE Internship Reconciliation, AI Coding, and Bikemap Loop.
AI coding rounds at other companies
- Meta: Meta AI coding interview: two onsite rounds where you finish a hard algorithm inside a mostly written codebase with an AI model, then face stress-test suites.
- Google: Google AI coding interview: an online assessment with HackerRank's built-in assistant, and a live round on a multi-file project where using AI was optional.
- Rippling: Rippling AI coding interview: HackerRank's built-in AI on Rippling's familiar problems, plus an HR codebase at the onsite.
- DoorDash: DoorDash AI coding interview: AI CodeCraft: build and run several small services live with your own AI tool.
- LinkedIn: LinkedIn AI coding interview: one long, multi-part task in an IDE whose agent edits files directly.
- Shopify: Shopify AI coding interview: pair programming with your own AI on Shopify's well-known question set.
- Datadog: Datadog AI coding interview: bring any AI tool and share your screen; the tasks are easy, and the follow-ups decide the round.
- Capital One: Capital One AI coding interview: new grad Power Day agentic round: fix a bug, then add a feature to a large full-stack app with Claude Code.
- Bloomberg: Bloomberg AI coding interview: intern rounds that start in a small file and move to a large codebase, debugging with AI help.
- xAI: xAI AI coding interview: new grad take-homes and onsites where you build with AI.
- Adobe: Adobe AI coding interview: AI showed up as one step in a director round and in a take-home; the rest stayed classic.
- Snowflake: Snowflake AI coding interview: design and build a rate limiter or a retrying task queue while the AI writes the code; one AI Engineer onsite was in person with your own laptop.
- Snap: Snap AI coding interview: a backend-only chat service from an empty Python project with Claude or Codex, or a feature in an existing codebase.
- Ramp: Ramp AI coding interview: build inside a repo Ramp provides; one repo logged every prompt for the interviewer to read.
- Anthropic: Anthropic AI coding interview: review four pull requests, then improve one with Claude Code.
- Apple: Apple AI coding interview: CoderPad with a sidebar agent, alongside Apple's other recent interviews.
More companies with AI rounds on FastPrep:
- Amazon: Amazon interview report: an AI-assisted debugging OA (1 experience).
- Atlassian: Atlassian interview report: project coding and AI tool use (11 experiences, two parts).
- Circle: Circle interview report: three AI rounds: building, debugging and PR review (1 experience).
- Volta: Volta interview report: a deliberately vague data task with a token budget (1 experience).
Across companies: 2026 AI Coding OAs at 15 Companies: 32 Reported Problems and 2025–2026 AI Coding Phone Screens and Onsites at 43 Companies: 76 Reported Problems.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.