I have written about the difference between generated software and created software, as well as how to interview a senior software engineer, in 57 architecturalist papers. Additionally, I have explored on how to interview a senior engineer.
So when Samuel Bashears wrote that a human barely built a coding agent at a LEET coding competition, I was thrilled.
This was the best news I had heard in ages.
It was about a tweet by Sam Altman:
When I first joined SGI in 1997, I took a class on how to interview. The presenter concluded that the best predictor of future performance is past performance.
The tech industry took a different direction, focusing on puzzles.
And so the LEET Coding Test began. In 2013, while searching for a job, I failed to secure a position because I was unable to complete the LeetCode coding questions.
Somehow, that didn’t stop me from becoming an architect of the most highly penetrated management software company and then turning that product around, so that the company was purchased for five times what it was worth when I joined.
And even while I was doing that, a co-worker was so annoyed at my perceived lack of coding skills that they anonymously trashed me on my blog and TheLayoff.com.
The notion that these questions are valid predictors of anything useful is an absurdity.
Worse, they cultivate a belief system that being able to do them is the essence of great software engineering.
It’s not.
Excellent software engineering involves understanding customer requirements, the limits of the software system, and how to engineer a solution that fits within the budget while gaining leverage for the next set of features.
Does that mean coding questions are off the table? No. However, there is a vast difference between building an optimal hash table and the kind of work that involves learning a large code base, figuring out customer requirements, and thinking through the possible places to improve the code to address those requirements.
Please find a way to incorporate that part of the interview process, but make it relevant to your work.
And if building optimal hash tables is what you do, then by all means, ask that question.
May this mania on Leet Coding go the way of the dodo bird.
Leave a Reply