As ChatGPT gets better at solving technical interviews I don’t understand why companies still insist on giving Leetcode style problems in their online assessments.<p>I get that they need to cut a large amount of people before live interviews but hackerrank/code signal have the ability to test “real world” scenarios like debugging,sys design,using APIs,etc…<p>In live interviews I feel like DSA interviews make even less sense. When is a candidate ever going to have to explain/implement a niche algorithm without access to their co workers, ChatGPT, or any other help<p>But curious to see what everyone thinks!
> I get that they need to cut a large amount of people before live interviews<p>You've just answered your own question. Leetcode is meant to be a filter, nothing more.<p>> but hackerrank/code signal have the ability to test<p>I'm not familiar with those things, but I don't think companies using things like leetcode actually care how good they are at skills/knowledge testing. I think they just want a coarse filter that doesn't depend on a human making initial assessments. Leetcode does that.<p>And it works! For instance, I can't imagine being willing to engage in a leetcode test to get a job, so I am automatically filtered out of some positions without them doing a thing.
You're missing the point. Live DS/A interviews are not about whether or not you can solve that particular problem. They are about evaluating how you approach and work through a novel task. How do you handle ambiguity? Are you able to talk through and explain the decisions you make? How are you able to adapt if the parameters of the problem are changed slightly? Do you have a notion of how your code will consume resources (CPU, Memory, Disk, Network, Time) as the amount of input grows?<p>The problem itself is simply the medium by which the interviewer tries to understand these things about you.<p>The fact that most FAANG companies don't even have you run your code during the interview should clue you into this.