←back to thread

642 points vkoskiv | 4 comments | | HN request time: 0.611s | source
1. rvz ◴[] No.45530672[source]
This is the easiest way to hire engineers with high quality open source contributions with a public track record.

All it takes is just to check that the commit shows up in upstream projects such as Linux and anyone can see the code, the reviews and the authors email in the AUTHORS file which verify that this contribution / patch is indeed from the author who committed that change.

This is a very old form of social proof which saves lots time and makes Leetcode redundant. (Which can now be completely cheated with LLMs.)

replies(2): >>45532702 #>>45539638 #
2. akdor1154 ◴[] No.45532702[source]
Be careful.. any measure becomes a target, and in doing so voids its usefulness. Getting a kernel patch in is probably already somewhat afflicted by this, but imagine the lkml if every Bay Area wannabe company started soft-requiring this as a screen!
replies(1): >>45536435 #
3. rvz ◴[] No.45536435[source]
> Be careful.. any measure becomes a target, and in doing so voids its usefulness.

Except, this tests for a wide range of tasks related to a typical SWE role: programming proficiency, reasoning through rigorous code-review, clean code, open-source and testing.

Given that Leetcode and Hackerrank tests can be cheated / beaten with LLMs, it not only has been gamed to death, but it tests for nothing that is related to the role.

> but imagine the lkml if every Bay Area wannabe company started soft-requiring this as a screen!

Then this saves time and the proof is on the interviewee to show high quality functional open source changes to high profile projects like the Linux kernel for example and the bar is set high on purpose.

Imagine if hundreds of candidates have "interview cheating" software installed, just to pass the coding interview. Both the interviewer and the interviewee lose.

4. fifticon ◴[] No.45539638[source]
as others have warned.. a scary counterexample is the npm ecosystem, where people are gaming it with questionable npm package spam, to get the 'astroturf footprint' of being the author of widely installed packages. (which get their footprint by piggybacking on actual useful packages, akin to stapled-together law packages in the us congress).