The directory records practical interviews, not zero whiteboards
Hiring Without Whiteboards makes a narrower argument than its name suggests. The README does not condemn a marker and a board. It rejects computer-science trivia, riddles, brainteasers, and unrelated algorithm puzzles as proxies for engineering ability. Discussing a real design problem on a board is acceptable. So are pair programming, reviewing existing code, and a job-related take-home exercise. That distinction makes the catalogue more useful than a blanket list of companies that promise no live coding.
The main README is an alphabetical directory. Each row can include a careers link, locations, remote status, and a short description of the interview. Some entries describe a single conversation. Others spell out several stages, including paid trial work, code review, system design, or pairing. The format is quick to search and easy to scan, but it is not normalized data. A location separator or phrase such as "take home" may vary between contributors.
A company entry is a lead, not a current promise
The project tells candidates what to do when reality disagrees with the directory: open a pull request to remove the company as a dud. That is an honest correction mechanism, and also the clearest warning on the page. Interview loops change by team, seniority, country, hiring manager, and quarter. A row written after one candidate's experience cannot guarantee what another candidate will receive.
Use the list near the start of a job search. Find employers whose stated process suits you, open the linked careers page, and ask the recruiter what the current technical stages are for the exact role. Ask whether an exercise is timed, paid, and reviewed with an engineer. A directory answer can save discovery time. Only the employer can confirm the process attached to today's vacancy.
GitHub reported 10 open issues and pull requests combined when we checked. The default branch was pushed on 2026-08-24, and recent merged pull requests updated Betterment and added Plivo. That combination matters more than the lack of a release tag. This catalogue changes through small content edits, so a tagged software release would add little. Current entry work shows that maintainers are still processing corrections.
What happened when we ran it
Our run at commit b96912c installed 243 Yarn packages in 11 seconds, consuming 25 MB on disk. The available test command succeeded in 6 seconds. There was no build script or target, so we skipped that step instead of inventing one. The fresh Debian container had 3 CPUs, 8 GB of RAM, no secrets, and Node 22.
The checkout itself was tiny: 15 files, about 1 line identified as source by the lab classifier, and 0.2 MB on disk. Most of the value lives in Markdown content, which explains why a source-line count says almost nothing about the directory's size. Our scan found 1 CI workflow file, no Dockerfile, and no tests directory. Those signals match a maintained document with automated checks, not an application.
Passing tests do not validate every employer claim. They show that the repository's available checks completed on the measured commit. Our sandbox did not call company recruiters, apply for jobs, or test the links as a candidate. Treat the 6-second result as evidence about repository hygiene only. Freshness still depends on contributors reporting changed processes and maintainers merging the correction.
The hiring advice sets a better bar than many entries prove
The separate recommendations document is the strongest part for hiring teams. It says screening questions should resemble the role. Take-home work should be paid, limited to under 4 hours, restricted to the standard library where appropriate, and based on the sort of task the candidate would perform. It also recommends an alternative for people who cannot spare take-home time, such as a shorter pairing exercise on their own laptop.
That advice is more demanding than the directory's inclusion threshold. A company may appear because it uses a take-home project, while its row says nothing about payment or a time limit. Some entries describe long interview sequences, and a non-trivia process can still consume too much unpaid time. Candidates should compare each row with the project's own recommendations instead of reading inclusion as an endorsement of every detail.
The README also allows candidate choice. Someone may prefer a whiteboard for discussing a problem, and the recommendation says to let them use it. This is a practical position: the problem is an assessment detached from the work, not one particular surface or interview setting. Hiring managers can borrow that principle without copying any listed company's full process.
Search works, but structured filtering does not
A browser find is enough if you know a company name or city. It is less helpful for questions such as which remote employers pay for exercises, which processes avoid take-homes, or which entries were verified within 6 months. The Markdown rows do not carry consistent dates, evidence levels, or fields. There is no API, vacancy feed, or guarantee that every careers URL leads to an open engineering role.
Turning the repository into a database would create another maintenance problem. The simple pull-request model lets a candidate correct one line without learning a schema or running a service. For personal research, copy a shortlist into your own spreadsheet and add the date you confirmed each process. The upstream list remains useful precisely because contribution is cheap, while your private notes can carry the stricter evidence a job decision needs.
Active edits make the list useful, not authoritative
Hiring Without Whiteboards is best used as a filter for where to investigate, not proof that an employer runs a humane process. The 2026-08-24 push shows it is still maintained, though it cannot prove older rows remain correct. Its definition of a good technical interview is specific enough to challenge recruiters with useful questions. The directory gets you names. Verification, compensation for exercises, and a role-specific explanation still have to come from the company.