hiring-without-whiteboards

poteto's list of "companies that don't have a broken hiring process" — 51.8k★ / 3,926 forks, MIT, created 2017-03-15, 2,175 commits from 421 contributors and 1,322 merged pull requests (counts read 2026-08-28). Shared via Telegram, routed here by the hub.

762 company entries, alphabetised into ten sections, each one line: a link to the careers page, the office locations, and a sentence describing how that company interviews. 757 carry the process sentence; five give only a name and a city. It is the first source in this spoke written from the hiring side of the table rather than the studying side — the thing interview-preparation said the corpus did not have.

What it is actually a list of

The title names the whiteboard, and the README immediately says the whiteboard is not the point:

"Whiteboards" are used as a metaphor, and are a symbol for the kinds of CS trivia questions that are associated with bad interview practices. Whiteboards are not bad – CS trivia questions are. Using sites like HackerRank/LeetCode probably fall into a similar category.

The stated test is two lines long: discussing a real-world problem, with or without a whiteboard, is fine; CS trivia, puzzles, riddles and brainteasers, with or without a whiteboard, are not. So the variable is what the question is about, and the room's furniture is incidental — eight entries say so directly by listing a whiteboard as one of the tools their good process uses, one of them offering the candidate a choice of laptop or whiteboard.

That distinction matters to this spoke because it is the same axis the curricula sit on. coding-interview-university and llm-interview-questions-hoang both teach recall of prepared material; what this list rewards is work resembling the job.

What the 762 processes are made of

Counting keyword mentions across the entry sentences (crude, but the entries are one sentence each):

The take-home is the substitute. Six of every ten listed companies replace the interview problem with work done at home, and pairing on real code is the runner-up. Nothing else is common.

Set that number against the repository's own advice. work-sample-assessment collects the argument; the short version is that RECOMMENDATIONS.md — thirteen lines, the only prescriptive document here — asks companies to pay the candidate, time-box under four hours, restrict to the standard library, and match the exercise to the actual work. Of the 762 companies on the list, 27 mention paying and 4 mention a time box. The list does not check its own recommendations before adding a company, and does not claim to.

Where the evidence comes from, and what that limits

Entries arrive by pull request, from the company or from someone who interviewed there. Nothing is verified. The only correction channel is the "Duds" rule: if your recent interview contradicts a description, open a PR to remove that company. So the list is a collection of claims about processes, self-reported and candidate-corrected, and it is T3 for exactly that reason — no audit, no methodology, no denominator.

And the correction channel has gone quiet. Commits by year:

Two thirds of the file was written before 2021. A hiring process is not a stable object — teams reorganise, a company adds a screening round, an engineering VP leaves — and the mechanism meant to catch that is candidate pull requests, which fell to eight in all of 2025. The list therefore reads as a 2017–2022 snapshot with a maintained veneer, and an entry's presence today is weak evidence about how that company interviews today. Last push 2026-08-24, so it is not abandoned; it is barely fed.

What it does not contain

No outcome data of any kind — not whether these processes hire better engineers, not whether candidates prefer them, not retention, not offer-acceptance. It is an advocacy list, and it argues from the resemblance between the exercise and the job rather than from evidence that resemblance predicts performance. Growth edge 4 asked for "T1/T2 on technical hiring practice or interview validity research"; this is the practice half at T3 and none of the validity half. See synthesis.

Then there is selection. Companies opt in, so the list describes the population willing to say publicly that they interview this way. Nothing here supports a claim about the share of the industry that does.

Related

work-sample-assessment · interview-preparation · lauren-tan · coding-interview-university · llm-interview-questions-hoang · cirr-standards · capture-the-flag