As a tech lead, from last year, all my new hire interview is fundamentally changed, no concept, no algo, no design.
Just a real world problem, even not clearly defined yet, allow candidate use any AI tool they like, ask me questions or do research for problem clarification, and work it out.
I'm watching all this process in 1-1.5 hours to see if he is a problem solver. 99% will be solved by AI with your proactivaly and smart prompt or questions in current work, so the thinking and prompting process is key.
> 99% will be solved by AI with your proactivaly and smart prompt or questions in current work, so the thinking and prompting process is key.
I am still a junior but this seems like you are interviewing the AI rather than the candidate. Also why bother with a technical interview if you expect AI to do their job?
Actually not a real technical interview for this case, it's a real world problem solving, including business analysis(for the uncleared problem), coding, and testing to deliver to me. What I'm looking for a individual builder(or a one-person tech team) instead of an expert on a specific tech stack.
My employer's process is basically exactly the same: leetcode, system design, behavioural; we just tell candidates not to use AI. Hiring is one of the scarier decisions a manager can make so I think they will stay pretty conservative. Personally my philosophy is that the interview is an information-gathering session, not a workday simulation. So it makes sense to test your fundamentals even if in practice you may be delegating them most of the time.
I’m not preparing much because I have quite a few years of experience under my belt.
Basically I read the JD, find some stories from my work that I can tell, brush up the CV for a bit and then that’s it. I don’t prepare for LC interviews and if I get one I just decline.
Maybe AI-assisted coding? I just interviewed with Amazon and they are quite looking on how you use AI to finish a task with a wide scope. Leetcode is not the main part now though.
If you're hiring software developers and you care about IQ, you don't need to test it implicitly; you can safely test for it explicitly, and there are several large, deep-pocketed plaintiffs lawyer targets who routinely do so. The idea that general cognitive testing is verboten in US employment is almost entirely an Internet myth.
People use Leetcode because they believe it tests for programming aptitude.
Nope, you'll get a lawsuit for discrimination if you explicitly test for IQ. That's why they do it implicitly. The scheme is very simple: hire people with the highest IQ, and since they have high IQ, they will figure at least something out.
It tests for multiple things, at its best: A basic work ethic to understand fundamental CS concepts. Sure maybe plenty of people can't write a binary search in two minutes unless they practice live coding a bit, but plenty of people do study, so it self selects for that type.
There are also people who, no matter what, could not live code simple tree traversals or bin search or something, and it filters on that.
Finally, there's a pattern matching aspect to it. Some of the best interview questions I got involved very simple algorithms, but it was obfuscated by the problem. So the 'trick' was to just think through the problem and ask questions. Not to have memorized something obscure.
be ready for zingers, like " what benefit do i get if i hire you for thousands of dollars a month, instead of paying a couple hundred for a few AI sessions?"
no, but i would make plans for an answer somewhat better than " i dont know, thanks for your time."
such as " hiring me will ensure that your AI sessions are few and limited to a couple hundred dollars expense, bare minimum, a human must prompt an AI or it does nothing. as a [professional] i have insight regarding structuring prompts, as well as fast response to code based remediation for incidents involving off the rails output, and abberant alignment adoption."
also: interviews can be more about,how you respond to being knocked off your footing, rather than gathering rote factoids about "you".
What about answering something like "I don't know but if the work can be done for hundreds on AI instead of thousands on me then I refuse to let you waste your money like that"?
Make this thing that would be impossible without AI. The test is to see if you actually architect it properly and understand principles of how things connect together.
Make this thing that would be impossible without AI. Now make these modifications without any AI.
Make this thing. You may use low quality AI like Composer 3 or none at all, but if you use none, we'll probably think of you as some kind of boomer.
Here's a bunch of technical problems that we don't know the answer to. If you give answers or insights we haven't considered, then you're bringing value to the team (e.g. git/PR policy, microservices, feature flagging, localization, security)
As a tech lead, from last year, all my new hire interview is fundamentally changed, no concept, no algo, no design. Just a real world problem, even not clearly defined yet, allow candidate use any AI tool they like, ask me questions or do research for problem clarification, and work it out. I'm watching all this process in 1-1.5 hours to see if he is a problem solver. 99% will be solved by AI with your proactivaly and smart prompt or questions in current work, so the thinking and prompting process is key.
> 99% will be solved by AI with your proactivaly and smart prompt or questions in current work, so the thinking and prompting process is key.
I am still a junior but this seems like you are interviewing the AI rather than the candidate. Also why bother with a technical interview if you expect AI to do their job?
Actually not a real technical interview for this case, it's a real world problem solving, including business analysis(for the uncleared problem), coding, and testing to deliver to me. What I'm looking for a individual builder(or a one-person tech team) instead of an expert on a specific tech stack.
It's interviewing the capacity to use the tools in a useful manor rather testing the tools themselves.
How has that worked out? What are you looking at when you compare that to what you were doing before last year?
This is so fucking dumb you're going to hire a candidate who got A/B tested into the 'smarter' SoTA model or something.
My employer's process is basically exactly the same: leetcode, system design, behavioural; we just tell candidates not to use AI. Hiring is one of the scarier decisions a manager can make so I think they will stay pretty conservative. Personally my philosophy is that the interview is an information-gathering session, not a workday simulation. So it makes sense to test your fundamentals even if in practice you may be delegating them most of the time.
I’m not preparing much because I have quite a few years of experience under my belt.
Basically I read the JD, find some stories from my work that I can tell, brush up the CV for a bit and then that’s it. I don’t prepare for LC interviews and if I get one I just decline.
Maybe AI-assisted coding? I just interviewed with Amazon and they are quite looking on how you use AI to finish a task with a wide scope. Leetcode is not the main part now though.
I had just interviewed with Amazon and it was purely leetcode, with the exception of their leadership principle question. What role did you apply to?
SWE internship, but the team is focusing on machine learning
Leetcode is implicit IQ testing. That is why they will likely keep it despite AI.
If you're hiring software developers and you care about IQ, you don't need to test it implicitly; you can safely test for it explicitly, and there are several large, deep-pocketed plaintiffs lawyer targets who routinely do so. The idea that general cognitive testing is verboten in US employment is almost entirely an Internet myth.
People use Leetcode because they believe it tests for programming aptitude.
Nope, you'll get a lawsuit for discrimination if you explicitly test for IQ. That's why they do it implicitly. The scheme is very simple: hire people with the highest IQ, and since they have high IQ, they will figure at least something out.
It tests for multiple things, at its best: A basic work ethic to understand fundamental CS concepts. Sure maybe plenty of people can't write a binary search in two minutes unless they practice live coding a bit, but plenty of people do study, so it self selects for that type.
There are also people who, no matter what, could not live code simple tree traversals or bin search or something, and it filters on that.
Finally, there's a pattern matching aspect to it. Some of the best interview questions I got involved very simple algorithms, but it was obfuscated by the problem. So the 'trick' was to just think through the problem and ask questions. Not to have memorized something obscure.
be ready for zingers, like " what benefit do i get if i hire you for thousands of dollars a month, instead of paying a couple hundred for a few AI sessions?"
Just tell them they will feel better about the money they spent on the ad.
did someone actually ask you that?
no, but i would make plans for an answer somewhat better than " i dont know, thanks for your time."
such as " hiring me will ensure that your AI sessions are few and limited to a couple hundred dollars expense, bare minimum, a human must prompt an AI or it does nothing. as a [professional] i have insight regarding structuring prompts, as well as fast response to code based remediation for incidents involving off the rails output, and abberant alignment adoption."
also: interviews can be more about,how you respond to being knocked off your footing, rather than gathering rote factoids about "you".
What about answering something like "I don't know but if the work can be done for hundreds on AI instead of thousands on me then I refuse to let you waste your money like that"?
Typical interview types we see:
Make this thing that would be impossible without AI. The test is to see if you actually architect it properly and understand principles of how things connect together.
Make this thing that would be impossible without AI. Now make these modifications without any AI.
Make this thing. You may use low quality AI like Composer 3 or none at all, but if you use none, we'll probably think of you as some kind of boomer.
Here's a bunch of technical problems that we don't know the answer to. If you give answers or insights we haven't considered, then you're bringing value to the team (e.g. git/PR policy, microservices, feature flagging, localization, security)