May 21, 2026
Pull Request or Pull a Fast One?
The Interview That Ships to Production: replacing whiteboards with pull requests
Job interview or free work? Commenters clash over candidates fixing the company’s app
TLDR: AngelList says its interview is better because candidates improve a real company app instead of doing fake puzzles. Commenters are split: some call it a smarter test of real work, while others say it looks uncomfortably close to getting free labor from job seekers.
A hiring blog from AngelList tried to sell a bold new dream: ditch the whiteboard, ship a real fix instead. Candidates first do a take-home coding task based on how venture funds actually pay out money, then the finalists are invited into the company’s real interview platform and asked to submit a pull request — basically, a proposed code change — to improve it. The company says this reveals what actually matters: can someone read messy real-world code, spot problems, use AI tools wisely, and ship something useful?
But in the comments, the vibe was less “wow, the future of hiring” and more “wait… are we watching unpaid labor get rebranded as innovation?” One skeptical reader flatly asked whether AngelList was just taking candidate work straight into production. Another went full scorched-earth, mocking the whole thing as “free labour from employment candidates” and joking they may as well start charging an entrance fee too. That set off the biggest drama: is this a smarter, more realistic interview, or a slick way to get applicants to improve your product for free?
Not everyone hated it. One supporter argued that reviewing a real pull request is a far better test than the usual sweaty puzzle-solving circus, because it shows how people think, write, test, and communicate. Even so, the internet clearly smelled blood: jokes about rage-bait, concerns over ownership rights, and side-eye at candidates literally fixing bugs in a company tool during the hiring process turned this from a hiring-process blog into a full-on comments-section cage match.
Key Points
- •AngelList says it replaced whiteboard-style technical interviews with a take-home coding exercise and a follow-up pull request in its assessment platform.
- •The take-home task asks candidates to implement a three-tier venture fund distribution waterfall in JavaScript, covering real financial mechanics such as return of capital, preferred return, and carried interest.
- •Candidates are allowed to use a Claude AI assistant, and AngelList records AI chat and keystrokes with millisecond timestamps to review how they work and validate results.
- •Candidates who pass the first stage are asked to work in the actual Next.js codebase of the testing platform and ship a real improvement via pull request.
- •The article cites examples of candidate contributions including a tabbed AI/test-case sidebar, a fix for a submission-related data-loss bug with tests and retry logic, and improved Claude calibration by Nathan Amin.