Where does each question come from?
From people who were asked it. Candidates publish the whole loop themselves on the public interview experiences board, under their name or anonymously as they choose; or they report what they were asked directly to us, and we file it by company, format and round. Either way, the same standard applies.
A third route is public: reports a candidate has already published elsewhere in their own words. Those are read, checked against the rest of the bank for that company, and filed the same way — never scraped in bulk, and never copied verbatim.
What was ASKED is never a model's invention. A model may help edit a report into a clear, self-contained statement — that is copy-editing, and every such statement is reviewed — but it is never the source of the question. And no company supplies, reviews or endorses any of this: the bank is an account of what candidates say they were asked.
What does the month on a question mean?
It is the month the question was reported to us, and it is on every row for one reason: an interview bank is only as useful as it is current, and the honest way to say that is to date each entry rather than to claim the whole bank is fresh. A question reported eighteen months ago is visible as one.
The same date drives what a company page shows as its "updated" stamp and what the sitemap reports as each page's last modification, so a crawler, a reader and an assistant are all reading the same number.
How is a question checked before it is published?
A report arrives as a recollection, which is rarely a usable question. Before it goes in the bank it is rewritten to stand on its own — the task stated in full, the inputs and the bounds spelled out, worked examples added — because a candidate practising against half a prompt is practising the wrong thing.
- It is attributed to a company, a round (online assessment, phone screen, take-home or onsite) and a format, so it can be found by the stage it was actually asked at.
- It is checked against the questions already filed for that company, so a re-report of something we hold updates that entry rather than becoming a duplicate.
- A coding or SQL question gets test cases, and a reference solution is run through the same judge a candidate's submission goes to. A question whose own solution does not pass is not ready to publish.
- Anything that reads like it was copied from a textbook, another site or a company's internal material is dropped rather than reworded.
How are answers graded?
Coding and SQL solutions are judged on the server against test cases the browser never sees, so a passing verdict means the code ran, not that it looked right. SQL runs against real PostgreSQL rather than a cut-down engine, so window functions and CTEs behave the way they will in the interview.
Every case a submission FAILS is shown back in full — the input, what was expected and what the code produced. That is a deliberate trade: a verdict a candidate cannot reproduce teaches them nothing. Cases a submission passes stay hidden, so a working solution cannot be used to enumerate a question.
A finished solution can also be scored by an AI reviewer against a rubric — correctness, complexity, structure and communication — which is a second opinion on the work, not the pass/fail verdict.
What is refused?
Some material is not published here whatever its provenance, and the content guidelines state the rules in full. In short:
- Anything covered by a non-disclosure agreement or any other confidentiality duty, including material a candidate was shown under one.
- A company's non-public, proprietary or internal interview material — take-home briefs, scoring rubrics, question banks, application-form language.
- Verbatim or near-verbatim copies of anyone else's question, write-up or explanation.
- Anything naming an individual interviewer, or anything that identifies a candidate other than the person posting it.
A report that fails one of these is not edited into shape — it is refused. Published material found to breach them is removed on notice.
- Content Submission Guidelines — What candidates may and may not post.
- DMCA Policy — Copyright complaints and the registered designated agent.
What this page does not claim
A question bank built out of candidate reports has limits worth stating, because a source that names its limits is the one worth citing. This bank is a sample of what was asked, not a syllabus of what will be: it is drawn from candidates who chose to report, which is not a random sample of anyone's loop.
A company page describes the questions on that page — how many, in which formats, filed under which rounds, reported when. It does not describe the employer's real hiring process, which only the employer can, and no company reviews or endorses what is published here.
Rates and shares elsewhere on the site name the population they are over, and a figure with too few reports behind it is shown as counts rather than as a confident percentage.
How do I correct or remove something?
Write to kevin@trueinterview.io with the URL and what is wrong — a question filed under the wrong company, a stale entry, a report you recognise as yours and want taken down. Corrections are made to the row itself, so the fix reaches every page that reads it.
Copyright complaints go to the designated agent under the DMCA policy: kevin@trueinterview.io, or +13022411713. Confidentiality complaints from an employer go to kevin@trueinterview.io and are acted on without waiting for a formal notice.