You know, I DO know how code works since I 've been writing it since 1977. I have worked at software companies both big and small for decades, ran my own for a decade, and I have done plenty of hiring and sat in numerous code reviews. My last job was on a microservices cloud architecture very similar to what Twitter does.
There are no surefire ways to discover how valuable someone is in an interview. Sometimes we give them coding tests. Sometimes just a lot of questions. What we would LOVE to do is have a look at the code they had written in the past, and quiz them about it. But when candidates come from another company their old code is off-limits. Musk is in a nearly unique position in that regard.
Sometimes prospective employees will bring open source project links so we can see their code (this ismbecoming more common), and this is a big advantage for them if they are good. I told my own son to participate in open source stuff to build experience and create resume artifacts for this reason.
Musk has to decide whether to keep or let go dozens of developers. As a way of doing that, his method is not necessarily bad, and could be quite good given the limitations of what he can look at in a short period of time.
It all comes down to the actual implementation at the meeting. How were they questioned, how were their answers evaluated, etc.
For example, there is a block diagram of Twitter’s architecture in one of the screen shots. How much do you want to bet Musk said something like, “Let’s start at the top - who wants to draw a high level diagram of Twitter’s architecture for us to refer to?” That’s what I would do. Then I’d have noted who volunteered for it, and I’d ask them some pointed questions along the way to see if they really understand what they are drawing.
Musk could ask lots of questions that would be useful without him having to understand the architecture at all. Things like, “Where would you say the key performance bottlenecks are with this architecture?”, or “How do you handle transactions between all these services? What if one of many dependent calls fails?” The answere you get to such questions (and more detailed ones if the candidate nails the easy ones) tell you a lot.
Believe me, there are a lot of programmers out there who claim to write code but who would fail Musk’s demands. I have interviewed professional programmers who couldn’t describe the difference between a stack and a queue. I’ve seen programmers leave bugs in code because they didn’t understand memory management. A common failure is when you ask a candidate to write a method that is best solved with recursion and see if they realize it, or if they give you spaghetti code.
We had one engineer (MIT graduate engineer no less) who was so incapable of writing decent code he eventually became a QA, then an IT guy who managed the build servers. But he was still listed as a software engineer in the company. His checkin history and a few questions would have made that instantly clear.
The idea that Musk’s technique is ridiculous on its face is, well, ridiculous on its face. Without sitting in on his meeting you have no odea what was asked and how the information was used. Maybe all he was looking to do was get rid of the bottom 10% of engineers who are useless, and pass all the rest. We do 't know.
)