It should be said that there were others - like https://interviewing.io/https://karat.com/ et al - that figured out the importance of moving beyond "grind LeetCode" years hackerrank did, but it's good to see them acknoweldge this, too.
As someone who is top x% on Leetcode (I think I also held the longest daily streak), I would say Leetcode is a test of perseverance. I became a worse developer because I spent time on Leetcode instead of my open source projects, but it taught me marathon long grind. I truly believe its unintended positives are somewhat worthwhile.
Well, I am happy that people are noticing that LC interviews were not the best idea, but I am sad to see that they are noticing it for the wrong reasons (LLMs will replace the need for algorithms).
> Tasks: Algorithmic questions → real work in code repositories
> Evaluation: Code correctness → AI fluency and judgment
I am never trading algorithmic knowledge for "real work", unless I am hiring for something that does not really require any kind of intellect, which would make me question why I am hiring in the first place. Same goes for code correctness. I wouldn't really like working with anyone who can't read and assess the code they are trying to ship (whether they wrote it or generated it).
That said, I really hate(d?) LeetCode. I recently passed two long interview loops at two FAANG companies, and the number of LC questions was frustrating. Some of them were problems I practically had to memorize because their solutions relied on very specific tricks that rarely come up in real engineering work (take, for example, the in-place index manipulation in First Missing Positive, the binary search trick in Median of Two Sorted Arrays, or the pivot trick in Next Permutation). There were so many of them, and yes, some came up. I did well and got offers, but part of me felt sad for those who might not have had the same opportunity simply because they weren't lucky enough to have practiced and memorized the same ad hoc problems I did.
Replacing this process with engineers prompting a CLI is not evaluating much at all. In fact, it might be even less reliable because agent behavior is not deterministic. It's possible that one candidate will simply get luckier with their agent than another.
What I have always proposed is to keep algorithmic knowledge as part of the process. At least IMO, it is a crucial part of engineering intellect, but it must be combined with other assessments: communication, problem clarification, code correctness, etc. Otherwise, HackerRank is just inventing another cult, using tokens rather than problems.
As much as I detested the idea of Leetcoding the reason they became the de-facto for hiring was two fold:
- Software development for the most part did not require any particular specialization but it paid really well because of the explosive growth of the internet so there was an oversupply of talent.
- Most companies had no idea what they wanted from hires and how to test them so started copycat-ing puzzle based and algorithm and data science questions from the days when programming was mostly numerical computing
It devolved into a race to the bottom as more and more people lured by the 300k packages lined up and companies well aware of "No Silver Bullet" threw these people at increasingly menial tasks.
Companies at Leetcode and Hackerrank have no relevance now that you can ask an AI agent to do a hyper specialised candidate selection based on the role, with most roles now being redundant every six months.
I wouldn't go that far. My last job had someone apply that blatantly lied on their resume. Had we just hired on vibes or a trial basis, it would have been awful for what was then a tiny business with a very narrow runway.
This is Hackerrank quickly realizing that their platform was becoming obsolete due to developers using AI agents and had to quickly adapt to that.
They only waited for AI to get this good to then post this fake letter with new assessments which tests for outcomes written by a machine but not raw understanding.
Sorry we increased the AI/word rate to 20% in order to summon agentic investment and sales harnesses as if reciting a satanic Super Intelligent invocation.
Technically all of us are top x%, for some x!
> Tasks: Algorithmic questions → real work in code repositories
> Evaluation: Code correctness → AI fluency and judgment
I am never trading algorithmic knowledge for "real work", unless I am hiring for something that does not really require any kind of intellect, which would make me question why I am hiring in the first place. Same goes for code correctness. I wouldn't really like working with anyone who can't read and assess the code they are trying to ship (whether they wrote it or generated it).
That said, I really hate(d?) LeetCode. I recently passed two long interview loops at two FAANG companies, and the number of LC questions was frustrating. Some of them were problems I practically had to memorize because their solutions relied on very specific tricks that rarely come up in real engineering work (take, for example, the in-place index manipulation in First Missing Positive, the binary search trick in Median of Two Sorted Arrays, or the pivot trick in Next Permutation). There were so many of them, and yes, some came up. I did well and got offers, but part of me felt sad for those who might not have had the same opportunity simply because they weren't lucky enough to have practiced and memorized the same ad hoc problems I did.
Replacing this process with engineers prompting a CLI is not evaluating much at all. In fact, it might be even less reliable because agent behavior is not deterministic. It's possible that one candidate will simply get luckier with their agent than another.
What I have always proposed is to keep algorithmic knowledge as part of the process. At least IMO, it is a crucial part of engineering intellect, but it must be combined with other assessments: communication, problem clarification, code correctness, etc. Otherwise, HackerRank is just inventing another cult, using tokens rather than problems.
- Software development for the most part did not require any particular specialization but it paid really well because of the explosive growth of the internet so there was an oversupply of talent.
- Most companies had no idea what they wanted from hires and how to test them so started copycat-ing puzzle based and algorithm and data science questions from the days when programming was mostly numerical computing
It devolved into a race to the bottom as more and more people lured by the 300k packages lined up and companies well aware of "No Silver Bullet" threw these people at increasingly menial tasks.
Companies at Leetcode and Hackerrank have no relevance now that you can ask an AI agent to do a hyper specialised candidate selection based on the role, with most roles now being redundant every six months.
This is Hackerrank quickly realizing that their platform was becoming obsolete due to developers using AI agents and had to quickly adapt to that.
They only waited for AI to get this good to then post this fake letter with new assessments which tests for outcomes written by a machine but not raw understanding.
The first question doesn’t make sense given the last, you should ask the ai the first one.
Algorithmic interviews have ZERO value in a world where Opus 5.5 is better than any coder on earth at writing algorithms.