1. Use recruiters and network: Wading through the sheer volume of applications was even nasty before COVID, I don't even want to imagine what it's like now. A good recruiter or a recommendation can save a lot of time.
2. Do either no take home test, or one that takes at most two hours. I do discuss the solution candidates came up with, so as long as they can demonstrate they know what they did there, I don't care too much how they did it. If I do this part, it's just to establish some base line competency.
3. Put the candidate at ease - nervous people don't interview well, another problem with non-trivial tasks in technical interviews. I rarely do any live coding, if I do, it's pairing and for management roles, to e.g. probe how they manage disagreement and such. But for developers, they mostly shine when not under pressure, I try to see that side of them.
4. Talk through past and current challenges, technical and otherwise. This is by far the most powerful part of the interview IMHO. Had a bad manager? Cool, what did you do about it? I'm not looking for them having resolved whatever issue we talk about, I'm trying to understand who they are and how they'd fit into the team.
I've been using this process for almost a decade now, and currently don't think I need to change anything about it with respect to LLMs.
I kinda wish it was more merit based, but I haven't found a way to do that well yet. Maybe it's me, or maybe it's just not feasible. The work I tend to be involved in seems way too multi faceted to have a single standard test that will seriously predict how well a candidate will do on the job. My workaround is to rely on intuition for the most part.
But hiring is nowhere near my full time job, thankfully :D
However, these days, people seem barely OK with paying for a recruiter. They think just because more people than usually are looking, they get to just lean back and great candidates show up.
IMHO, if anything, it got more difficult to hire. More noise to work through, people in existing roles are more reluctant to switch, candidates hustle more because they need a job etc. I absolutely think it's a useful service. But I don't know if it's easy to market.
And yet this interviewing problem seems to only affect tech companies.
When you're operating a huge furniture company, you want every dowel and every hole the same size. Then any fool can assemble it, and you don't have to waste time and space on a collection of slightly-too-large dowels.
To scale up often means focusing on consistency and precision, and less on expert judgement and care.
The literal only way to get rid of that is quantitative tests.
In the past Google made a lot of hires where they first give people a phone screen, then five in-person whiteboard interviews, then the interviewers send a dossier to a committee, then that committee decides whether to hire or not, then "team matching" lets hiring managers see the resumes.
So the interviews are basically conducted by random people, who have no idea what team the candidate will end up on.
Of course if you look at the number of employees Google has, and the average tenure, you can see why they make hires like Ikea makes chairs.
Usually what happens in a large org is (1) a department gets allocated a head count, and it eventually trickles down to team allocations; (2) a team needs a role that they can’t backfill from the rest of the company; (3) the team lead puts out an ad, receives resumes (directly or through HR), schedules interviews, and makes the hiring decision themselves. Exactly the same as a small company or startup in the last step.
The biggest problem with the take home tests are not people who don't show up due to not being able to finish the assignment, But that those people who do, now expect to get hired.
95% people don't finish the assignment. 5% do. Teams think submitting the assignment with 100% feature set, unit test cases, onsite code review and onsite additional feature implementation still shouldn't mean a hire(If not anything, there are just not enough positions to fill). From the candidate's perspective, its pointless to spend a week working and doing every thing the hiring team asked for and still receive a 'no'.
I think if you are not ready to pay 2x - 5x market comp you shouldn't use take home assignments to hire people. There is too much work to do, and receive a reject at the end. Upping the comp absolutely makes sense as working a week to get a chance at earning 3x more or so makes sense.
Be upfront that finishing the assignment doesn't guarantee a hire and very likely the very people you want to hire won't show up.
Please note that as much as you want good people to participate in your processes. Most good talent doesn't like to waste its time and effort. How would you feel if someone wasted your time and effort?
Uhhh yeah. That would really piss me off. Like reviewbombing glassdoor type of pissed.
In reality they need to look at this like onboarding a new team member.
What I added:
1. Instead of asking "do you have any questions for me?" at the very end, we started with that general discussion.
2. A few days ahead, I emailed the candidate the problem and said they are welcome to do it as a take home problem or we could work on it together. I let them know that if they did it ahead of time, we would do a code review and I would ask them about their design and coding choices. Or if they wanted to work on it together, they should consider it a pair programming session where I would be their colleague and advisor. Not some adversarial thing!
3. This the innovation I am proud of: a segment at the beginning of the interview called "teach me something". In my email I asked the candidate to think of something they would like to teach me about. I encouraged them to pick a topic unrelated to our work, or it could be programming related if they preferred that. Candidates taught me things like:
• How to choose colors of paint to mix that will get the shade you want.
• How someone who is bilingual thinks about different topics in their different languages.
• How the paper bill handler in an ATM works.
• How to cook pork belly in an air fryer without the skin flying off (the trick is punching holes in the skin with toothpicks).
I listed these in more recent emails as examples of fun topics to teach me about. And I mentioned that if I were asked this question, I might talk about how to tune a harmonica, and why a harmonica player would want to do that
This was fun for me and the candidate. It helped put them at ease by letting them shine as the expert in some area they had a special interest in.
I don't remember how I came up with the idea. Maybe I just like learning things.
One candidate even wrote after their interview, "that was fun!"
Have you ever had a candidate say that? This was the moment when I realized I might be on to something. :-)
Interviews are too often an adversarial thing: "Are you good enough for us?"
But the real question is would we enjoy working together and build great things!
People talk about "signal" in an interview. Someone who has an interest they are passionate and curious about and likes to share it with others? That's a pretty strong signal to me.
Even if it has nothing to do with "coding".
And I'm sure you're going to get different questions about different aspects of the things you're trying to teach from different interviewers. The wonderful thing about this is that it models trying to teach a concept to a fellow coworker. How you handle the questions during the teaching time says a lot about the candidate.
Or would it train people to choose topics very carefully, such that a little teaching skill goes a long way?
It's not like the interviewer getting a lesson in tuning a harmonica is going to bust out a harmonica and start putting his newfound knowledge to work, or revisit the subject in 6 months to see if he's retained the knowledge, or bring in a panel of harmonica tuning experts to check there weren't any major gaps or mistakes in the lesson.
If someone didn't want to do it, that's fine and wouldn't be held against them. In practice, I think one person out of 20 chose not to do it.
And if they weren't a great teacher, that was fine too!
The purpose of the segment is to give the candidate a chance, if they want, to shine at something they are interested in, and help put them at ease by letting them start out being the expert.
And then when they were teaching me, I made sure to pay attention, ask questions about anything I didn't understand - not to judge their teaching skills but to show I was interested and listening.
I can't fathom how making the whole process being all hostile makes any real sense!
We all want that great person to show up and really help, not just cope or put up a facade to survive.
Yet, that is exactly what happens. People that other people expect to bond with get put through the wringer....
Seems counterproductive to me.
Agreed. I have been through technical interviews where at the end, my feeling as a candidate was "I don't want to work with this asshole".
What if I decide to teach you something about the Quran and you don't hire me?
Perhaps this is just urban legend but from the stories I've heard hiring for FAANG-type companies there are people out there interviewing with their only goal being baiting you into a topic they can sue you over.
Worst instance I have heard of is when an interviewer asked about the books the candidate liked to read (since the candidate said they're an avid reader in their resume) and he just said that he liked to read the Bible. After not getting hired for failing the interviews he accused the company of religious discrimination.
I'm by no means an expert in US law and don't know the people this happened to directly so maybe it's just one big fantasy story but it doesn't seem that far fetched that
- If you are a rich corporation then people will look to bait you into a lawsuit
- If you give software engineers (or any non-lawyers) free rein on how to conduct interviews then some of them can be baited into a situation that the company could be sued over way more easily than a rigid leetcode interview
I think nothing came of the fellow who liked reading the Bible but I would imagine the legal department has a say in what kind of interview process a rich company can and can not use.
Simple, don't pick subjects, such as religion, sexuality, etc. It's not that deep.
a candidate that pulls the religious thing now is one that's going to drag their crazy shit into the office later. or complete misread the room / culture now, and in the future.
like it's a technical job interview, talk about something vague technical, even if it's just how the convection settings work in your oven.
Secondly, share that place by example. Radiating non fear can do wonders for the people around us, as can honest and frank conversation.
When all members of a team, or even small companies come to understand not all environments are about blame and shame, magic happens!
Don't have anything else to add, but that I agree with you.
There is literally no way to prevent a candidate from disclosing a protected characteristic during an interview. Some obvious examples: they might show up to the interview in a wheelchair, they might be wearing a religious garment, they might be visibly pregnant, and so on...
What legal doesn't want, is you asking questions directly intended to elicit that kind information when the candidate didn't volunteer it. Asking the candidate a direct question like "are you planning to have kids?" makes it sound like that information will be used in the hiring decision.
As you can see from the examples I gave, what they taught me was completely up to them.
It was just a fun exercise that I and every candidate enjoyed.
Forgive me, but I just don't see the problem here.
If you did decide to teach you something about the Quran, that would be great!
I don't know enough about these holy books, and I always welcome a chance to learn more.
There is no problem. But some people will create problems out of thin air, when there is none.
Completely unrelated, but I find that other people will create problems by making an off-colour statement, something mildly offensive, or disrespectful and then when they get a reaction they say "why are you making a problem where none exists".
DARVO/gaslighting behaviour, and I wish it was rarer than it is.
It's somewhat overblown - obviously anyone can try to submit a demand letter about anything. My experience with legal in the hiring process is they want to avoid obvious own goals, and document the process so that clear reasoning can be expressed. Then, unless you really do something obviously discriminatory, you can tell people who claim they've been discriminated against to go pound sand.
There are lots of good tactical reasons to settle claims like that, so lawyers may advise you to settle, but if you're of the "we don't negotiate" mindset, in my experience most lawyers are quite happy to gear up for a fight with the right groundwork in place.
You just need a small file, like a point file. Any gasheads remember those? And a single edge razor blade or the like to lift the reed.
To raise the pitch, you file the end of the reed, making it lighter so it vibrates faster.
To lower the pitch, you file near the attached end of the reed. I am not sure on the physics of this and would appreciate anyone's insight.
The specific tuning I've done many times is to convert a standard diatonic harp (the Richter tuning) to what is now called the Melody Maker tuning.
The Richter tuning was apparently designed for "campfire songs". You could just "blow, man, blow" and all the chords would sound OK.
Later, blues musicians discovered that you could emphasize the draw notes, the ones that are easy to bend to a flat note to get that bluesy sound. This is called "cross harp". For example, in a song in G you would use a C harp instead of one tuned in G.
The problem with cross harp is that the 7th is a minor 7th and you have no way to raise it up to a major 7th if that would fit your song. And the 2nd is completely missing! In fact you just have the tonic (G in this case) on both the draw and blow notes where you might hope to hit the 2nd (A). There is no A in this scale, only the G twice.
To imagine a song where this may be a problem, think of the first three notes of the Beatles song All My Loving. It starts with 3-2-1. Oops, I ain't got the 2. Just the 1 twice.
This is where the file comes in. You raise the blow 1st to a major 2nd. And you raise the minor 7th to a major 7th in both octaves.
Now you have a harp with that bluesy sound we all love, but in a major scale!
It's pretty clear why that's the opposite of filing off material close to the tip, so obviously the tone goes lower.
In my mental image, filing close to the base of the reed gives the reed a similar shape as putting extra material next to the tip (thinner at the base, thicker next to the tip), and that's why it behaves the same.
The solder on the tips reminds me of a doctor visit years ago. I thought I may have broken a finger, and when I got to the doctor's office I mentioned to the receptionist that I had a high-deductible (HSA-compatible) insurance plan.
The physician's assistant said, "We could send you over for an X-ray, but since you're paying out of pocket, we can start with a simple test." He pulled a contraption out of his desk drawer and asked, "Do you know what this is?"
I said, "Yeah, a tuning fork. And based on the size and those heavy weights on the tips, it must be tuned to a rather low frequency."
It looked like this one:
https://www.stethoscope.com/adc-tuning-fork-128hz-500128/
He said, "Yep. So we get it vibrating and then touch the base to your finger. If you have a fracture, it will hurt because of the broken bone ends jiggling against each other. Then we will go for the X-ray to get more details. If it doesn't hurt, you are good to go. Is that OK?"
"Sounds good to me!"
It didn't hurt at all, and I just had to pay for a simple office visit instead of an expensive X-ray.
Once we are there, it then is all about techniques. Working from first principles is going to get us there, but there are likely pitfalls, traps, all manner of gotchas laying in wait...
And there we now have a basis for further discussion.
https://chatgpt.com/share/67a2aade-5870-8012-b211-1bd749a2d7...
The opening question is a copy of what was done before (probably by someone who doenst work at IBM anymore) and all the new stuff is stolen from outsiders.
I am tempted to take some offense at your comment. But I have to assume that you mean well.
You are correct that I don't work for IBM any more. What does that have to do with anything?
Thanks for sharing.
My take is the negatives may just be people leaning hard on, "if it seems too good to be true...
I don't know that many programmer/developer jobs where you can just put your headphones on and code without ever talking to another person.
Being able to explain/teach your work is part of "doing the job well" for a developer (IMO).
... Classic interviewing techniques of "explain how X algo works" or "write code to solve Y" will unfairly bias against interviewees that don't test well under pressure, but would otherwise be a good coworker.
... "Teach me something interesting to you" or "tell me stories about past experiences" will unfairly bias against interviewees that are shy, soft spoken, or are mildly socially awkward, but would otherwise be a good coworker.
Given the above awareness of where bias can emerge, how should the interview be done in order to get a candidate that knows what they're doing, and works well with the rest of the team?
Other comments mention relying more on recruiters and referrals, but that isn't always an option.
It takes time, money and people to bring someone in, and hiring is actually quite a risk for many companies (unless they're huge and/or in a hiring frenzy).
If a candidate doesn't work out, that's a lot of time and money down the drain, and potentially lost work, and disruption to teams and timelines, etc..
Most people don't get this part, and I think that's why they don't understand why interview processes are structured the way they are.
You really want to do the best possible evaluation, on all fronts, at the start. The longer a bad candidate stays in your pipeline or company, the more expensive and disruptive it gets.
As a wishlist I like it, I just don't see how you're going to assess all that in an interview. You'll notice that the technique of the day ("teach me something") doesn't address any of the dot points and that holds for ... pretty much any technique. Interviews are a weak process for assessing anything.
His name was Richard Feynman.
https://www.youtube.com/watch?v=nYg6jzotiAc
(Highly recommended video.)
I should also note that "teach me something" was intended to give the candidate a chance to be the expert at the beginning the interview.
It wasn't something they would be judged on, it was an ice-breaker. And a fun one, based on the feedback I got from candidates.
It is a nerd test. You can get every nerd talking if you ask them to tell you about a special skill or project they really care about.
I remember quite a lot of these lunch conversations:
Me: Hey, what's up? Do you like the salad?
Introvert Nerd: Mmmm-Hmmm.
Me: Great weather outside!
Introvert Nerd: Mmmm-Hmmm.
Me: I was hiking last weekend in the mountains and got caught in a storm.
Introvert Nerd: Mmmm-Hmmm.
Me: Does your work project proceed well?
Introvert Nerd: Mmmm-Hmmm.
Me: I've heard that you regularly cook medieval dishes and you created a food medievality detector using a Raspberry Pi and a horseshoe?
Introvert Nerd: Oh, yes! You know, measuring the medievality of a dish is not as simple as it sounds! Obviously, there are no American foods like tomatoes or potatoes allowed, but did you ever think which spices were common in Europe in the High Middle Ages and why that changed in the Late Middle Ages... ...
I find interviews quite tense and I generally dislike having them (on both sides of the table), but throwing in a few disarming questions here and there throughout really takes the edge off. A conversational, or more cerebral, approach like that can suit some people far better than live code tests or quick-fire Q&A.
I wish more bosses were like you.
I am grateful that so many people in this thread found the ideas useful.
I work as a developer and as an interviewer (both freelance). Now I want to integrate your point 3. into my interviews, but not to choose better candidates, just to learn new stuff I never thought about before.
It is your fault that I see now this risk in my professional life, coming at me. I could get addicted to "teach me something". 'Hey candidate, we have 90 minutes. Just forget about that programming nonsense and teach me cool stuff'
I especially like that you're prepping them very clearly in advance, giving them every opportunity to succeed, and clearly setting a tone of collaboration.
In person design and coding challenges are a pressure cooker, and not real-world. However, giving people the choice, seems like a great way to achieve the balance.
Honestly, I'm really just commenting here so that this shows up in my history, so I can borrow heavily from this format next time I need to interview! :) Thanks again for sharing.
But I digress!
Asking them what they wanted or would teach me was very illuminating and frankly, fun all around! I was brought a variety of subjects and not a single one was dull!
Seems I had experiences similar to yours.
One of my questions was influenced by someone who I respected highly asking, "what books are on your shelf at home?"
This one almost always brought out something about the candidate I would have had no clue about otherwise. As time advanced, it became more about titles because fewer people maintain a physical book shelf!
IBM conducted a mass layoff a few months ago, and as our little team lost our only customer at the time (this is public information), many of us were let go. Ah well.
One door closes, another opens.
If anyone out there is curious about "who is this guy with the strange and interesting ideas about interviewing", you can find my LinkedIn and other contact info in my HN profile.
> // This line prevents X from happening
I've seen a number of those. The issue here is that you've already wasted a lot of time with a candidate.
So being firmly against take home tests or even leetcode, I think the only viable option is a face to face interview with a mixture of general CS questions(i.e. what is a hashmap, benefits and drawbacks, what is a readers-writer lock, etc) and some domain specific questions: "You have X scenario(insert details here), which causes a race condition, how do you solve it."
> I feel like take home tests are meaningless and I always have. Even more so now with LLMs
I am starting to think that we need two types of technical interview questions: Old school (no LLMs allowed) vs new school (LLMs strongly encouraged). Someone under 25 (30?) is probably already making great use of LLMs to teach themselves new things about programming. This reminds me of when young people (late 2000s/early 2010s) began to move away from "O'Reilly-class" (heh, like a naval destroyer class) 500 page printed technical books to reading technical blogs. At first, I was suspicious -- essentially, I was gatekeeping on the blog writers. Over time, I came to appreciate that technical learning was changing. I see the same with LLMs. And don't worry about the shitty programmers who try to skate by only using LLMs. Their true colours will show very quickly.
Can I ask a dumb question? What are some drawbacks of using a hash map? Honestly, I am nearly neck-bearded at this point, and I would be surprised by this question in an interview. Mostly, people ask how do they work (impl details, etc.) and what are some benefits over using linear (non-binary) search in an array.
The drawback is that elements in a hashmap can’t be sorted and accessing a specific element by key is slower then accessing something in an array by index.
Linear search is easier to implement.
These are all trivial questions you ask to determine if a person can develop code. The hard questions are whether the person is the cream of the crop. The amount of supply of developers is so high most people don’t ask trivial questions like that.
> What if I use an LLM but I understand the code?
For most companies, the better strategy would be to explicitly LET them use LLMs and see whether they can accomplish 10X what a coder 3 years ago could accomplish, in the same time. If they accomplish only 1X, that's a bad sign that they haven't learned anything in 3 years about how to work faster with new power tools.
A good analogy of 5 years ago would be forcing candidates to write in assembly instead of whatever higher level language you actually use in your work. Sure, interview for assembly if that's what you use, but 95% of companies don't need to touch assembly language.
Do you seriously expect a 10x improvement with the use of LLMs vs no LLMs? Have you seen this personally, are you 1 10th the developer without an LLM? Or is the coding interview questions you ask or get asked, how to implement quicksort, or something?
Let's make it concrete, do you feel like you could implement a correct concurrent http server in 1/10th the time with an LLM than what you could do it without? Because if you jut let the LLM do the work I could probably find some issue in that code or alternatively completely stump you with an architectural question unless you are already familiar with it, and you should not be having an LLM implement something you couldn't have written yourself.
Absolutely fucking yes.
Concidering this would probably at most take a days work to get at least a workable prototype done if not a full implementation, using an AI you should be able to do it in a lunch break.
> Put the candidate at ease - nervous people don't interview well
Another reason:
If you're only say one of four interviewers, and you're maybe not the last interviewer, you really want the candidate to come out of your interview feeling like they did well or at least ok enough, so that they don't get tilted for the next interview. Because even if they did really poorly in your interview, maybe it's a fluke and they won't fail the rest of the loop.
Which is then a really difficult skill as an interviewer - how do you make sure someone thinks they do well even if they do very poorly? Ain't easy if there's any technical guts in the interview.
I sure as shit didn't get any good at that until I'd conducted like 100+ interviews, but maybe I'm just a slow learner haha
To their credit, I think they would have hired "the old guy" if I'd aced their take-homes, but I was a bit rusty and not super thrilled about their problem areas anyway so we didn't get that far. And honestly it seems like a decent system for hiring well-paid cogs in your well-oiled machine for the short term.
Your method sounds like what we were trying to do ten years ago, and it worked pretty well until our pipeline dried up. I wish you, and your candidates, continued success with it: a little humanity goes a long way these days.
Wishing you all the best for your career!
Especially people with ADHD don’t remember details as long as others, even though ADHD does not make someone a bad hire in this industry (and many successful tech workers have it).
I do prefer take-home challenges to live coding interviews right now, or at least a not-rushed live coding interview with some approximate advance warning of what to expect. That gives me time to refresh my rust (no programming language pun intended) or ramp up on whichever technologies are necessary for the challenge, and then to complete the challenge well even if taking more time than someone who is currently fresh with the perfectly matched skills might need. I want the ability to show what I can do on the job after onboarding, not what I can do while struggling with long-term unemployment, immigration, financial, and family health concerns. (My fault for marrying a foreigner and trying to legally bring her into the US, apparently.)
And, no, my life circumstances do not make it easy for me to follow the common suggestion of ramping up in my “spare time” without the pressures of a job or a specific interview task. That’s completely different from when I can do on the job or for a specific interview’s technical challenge.
I would definitely not expect someone out of work for a while to have any meaningful side projects. I mean, if they do, that's cool, but I bet the sheer stress of not having a job kills a lot of creativity and productivity. Haven't been there, so I can only imagine.
For such a candidate, I'd probably want to offer them a time limited freelance gig rather quickly, so I can just see what they work like. For people who are already in a job, it's more important to ensure fit before they quit there, but your scenario opens the door to not putting that much pressure on the interview process.
Beyond immigration types of obstacles, somehow recruiters and hiring managers rarely consider how many bad managers, bad executives, and bad companies there are when evaluating gaps and short tenures in an employee's resume. But equally, discussing those matters during an interview risks being seen as unprofessional and unreasonably negative. "Why did you leave this company?" "Oh, the warnings I got about the leadership from a former employee before I joined turned out to be true, and they didn't want to hear professionally presented necessary feedback about emotional safety in an emergency incident response situation, so they fired me without a single meaningful 1:1 discussion other than trying to assign blame." / "Oh, the CEO was enough of a problem that the investors eventually replaced him in a subsequent funding round despite him being the majority shareholder, but that was long after I had resigned or been fired due to that CEO's particular problems." / "Oh, they communicated in a very idiosyncratic way which didn't work for me and which I haven't seen at any company before or since." / etc. Most of that doesn't fly in an interview, but equally, even if it did, saying too much of that sounds like making up excuses for oneself even when it's 100% true. Same thing for why a job search doesn't succeed quickly.
Ours is a messy and imperfect industry, and it sucks that interview candidates have to pretend otherwise to seem like they'll be reasonable employees. Meanwhile the companies and executives that act in those ways get to present whatever positive and successful image they want, and they get to praise each other for firing fast or cutting costs with attrition or layoffs.
One of the problems I see in big corporates is that feedback cycles for bad policy can be so long, whoever caused something is far away when the effects hit. AKA nobody really cares. But to be fair, I've always tried to care, and worked with many others who do.
If nothing else, it's a useful tool when doing reviews with your manager, or when you're trying to evaluate your personal growth.
It comes in really handy when you're interviewing or looking for a new job.
Julia Evans calls it a "brag doc": https://jvns.ca/blog/brag-documents/
It doesn't have to be super detailed either. I tell people to write 50% about what the work was, and 50% about what the purpose/value of the work was.. That tends to be a good mix of details and context.
Writing in it once a month, or once a sprint, is enough. Even if it's just a paragraph or two..
dude maintained a "me wall" of achievements, commendation medals, etc. plus a file full of other crap. he mentioned he got a lot of promotions because you have a section to fill out where you describe all the shit you did over the least year, and he always came with examples...
Honestly, putting those details in writing in a way that I retrain after leaving might leave me vulnerable to claims of violating my corporate NDA. But legalities aside, yes, it would help with these kinds of interview issues - prospectively only of course, not retrospectively.
Annoyingly, it's also exactly the kind of doc that's difficult for people with ADHD to create and maintain rigorously, even though people with ADHD are more likely than average to need the reminders of those details. A lot of ADHD problems are that kind of frustrating catch-22, especially when interacting with a world that is largely designed around more typical brain types.
If I can ask a follow-up question: Is it the act of writing it down, or the whole recall aspect of it that is challenging? Or something else?
I've personally starting using audio transcription with a voice recorder to capture ideas more easily when I'm not at my computer (I hate typing on my phone keyboard) and that has made building these kinds of diary/log/notes documents a lot easier because I can just speak my ideas into the recorder whenever it is convenient for me, instead of waiting until I'm at a computer.
If you're a company that does this, please dog food your problems and make sure the interview makes the effort feel valued. It also smells weird if you claim it's representative of a typical engineering discussion. We all know that consultancy is wrangling data, bad data and really bad data. If you're arguing over what optimiser we're choosing I'd say there's better ways to waste your customer's money.
On the other hand I like leetcode interviews. They're a nice equalizer and I do think getting good at them improves your coding skill. The point is to not ask ludicrously obscure hard problems that need tricks. I like the screen share + remote IDE. We used Code which was nice and they even had tests integrated so there wasn't the whiteboard pressure to get everything right in your head. You also know instantly if your solution works and it's a nice confidence if you get it first try, plus you can see how candidates would actually debug, etc.
You need to do this to establish some baseline competency... for junior hires with no track record?
Recruiters have been notoriously bad in my experience. Relying on network has the potential to create bias and avoid good candidates that simply don't have an "in".
So far, everyone that elected to use GPT did much worse. They did not know what to ask, how to ask, and did not "collaborate" with the AI. So far my opinion is if you have a good interview process, you can clearly see who are the good candidates with or without ai.
To me it's the lack of skill. If the LLM spits out junk you should be able to tell. ChatGPT-based interviews could work just as well to determine the ability to understand, review and fix code effectively.
Reading existing code and ensuring correctness is way harder than writing it yourself. How would someone who can't do it in the first place tell if it was incorrect?
"Oh, I don't remember how to do parameterized testing in junit, okay, I'll just copy-paste like crazy, or make a big for-loop in this single test case"
"Oh, I don't remember the API call for this one thing, okay, I'll just chat with the interviewer, maybe they remember - or I'll just say 'this function does this' and the interviewer and I will just agree that it does that".
Things more complicated than that that need exact answers shouldn't exist in an interview.
Agreed, testing for arcane knowledge is pointless in a world where information lookup is instant, and we now have AI librarians at our fingertips.
Critical thinking, capacity to ingest and process new information, fast logic processing, software fundamentals and ability to communicate are attributes I would test for.
An exception though is proving their claimed experience, you can usually tease that out with specifics about the tools.
They'll type in things like, "C# read JSON from file"
As opposed to something like:
> I'm working on a software system where ... are represented as arrays of JSON objects with the following properties:
> ...
> I have a file called ... that contains an array of these objects ...
> I have ... installed locally with a new project open and I want to ... How can I do this?
No current LLMs can solve any of the problems we give them so pasting in the raw prompt won't be helpful. But the set up deliberately encourages them to do some basic scaffolding, reading in a file, creating corresponding classes, etc. that an LLM can bang out in about 30 seconds but I've seen candidates spend 30 minutes+ writing it all out themselves instead of just getting the LLM to do it for them.
The difference compared to where we were just 1-2 years ago is staggering.
Edit: the above is with Claude-3.5-Sonnet
Low level - I'll write up the structure of what I want in the form of a set functions with defined inputs and outputs but without the implementation detail. If I care about any specifics with the functions I'll throw some comments in there. And sometimes I'll define the data structures in advance as well.
Once all this is set up it often spits out something that compiles and works first try. And all the context is established so iteration from that point becomes easier.
I honestly can't imagine this. If the AI says "However, a downside of approach B is that it takes O(n^2) time instead of the optimal O(nlog(n))", what do you think the odds are that it literally made up both of those facts? Because I'd be surprised if they were any lower than 30%. It's an extremely confident bullshitter, and you're going to use it to talk about engineering tradeoffs!?
> Once all this is set up it often spits out something that compiles and works first try
I'm sorry, but I'm extremely* doubtful that it actually works in any real sense. The fact that you even use "compiles and works first try" as some sort of metric that the code it's producing shows how easily it could slip in awful braindead bugs without you ever knowing. You run it and it appears to work!? The way to know whether something works -- not first try, but every try -- is to understand every character in the code. If that is your standard -- and it must be -- then isn't the AI just slowing you down?
"Please don't generate or rewrite code, I just want to discuss the general approach."
Bc I don't know any design patterns or idiomatic approach, being able to discuss is amazing.
Though quality and consistency of responses is another thing... :)
Being confidently incorrect is not a unique characteristic of AIs, plenty of humans do it too. Being able to spot the bullshit is a core part of the job. If you can't spot the bullshit from AI, I wouldn't trust you to spot the bullshit from a coworker.
This is honestly where I believe LLMs can really shine. I think we like to believe the problems we are solving are unique, but I strongly believe most of us are solving problems that have already been solved. What I've found is, if you provide the LLM with enough information, it will surface edge cases that you haven't thought of and implement logic in your code that you haven't thought of.
I'm currently using LLMs to build my file drag and drop component for my chat app. You can view the problem and solution statement at https://beta.gitsense.com/?chat=da8fcd73-6b99-43d6-8ec0-b1ce...
By chatting with the LLM, I created four user stories that I never thought of to improve user experience and security. I don't necessarily think it is about knowing the edge cases, but rather it is about knowing how to describe the problem and your solution. If you can do that, LLMs can really help you surface edge cases and help you write logic.
Obviously what I am working on, is really not novel, but I think a lot of the stuff we are doing isn't that novel, if we can properly break it down.
So for interviews that allow LLMs, I would honestly spend 5 minutes chatting with it to create a problem and solution statement. I'm sure if you can properly articulate the problem, the LLM can help you devise a solution and a plan.
I’m basically pair programming with a wizard all day who periodically does very stupid things.
At a previous job I made the mistake of letting it write some repository methods that leveraged SQLAlchemy. Even though I (along with my colleague via PR) reviewed the generated code we ended up with a preprod bug because the LLM used session.flush() instead of session.commit() in exactly one spot for no apparent reason.
LLMs are still not ready for prime-time. They churn out code like an overconfident 25-year-old that just downed three lunch beers with a wet burrito at the Mexican place down the street from the office on a rainy Wednesday.
That's...oddly specific.
The LLMs are much more eager to please and to write lots of code. When I was younger, I would get distracted and play computer games (or comment on HN..), rather than churn out mountains of mediocre code all the time.
My process right now when working LLMs is to do the following:
- Create problem and solution statement
- Create requirements and user stories
- Create architecture
- Create skeleton code
- Map the skeleton code
- Create the full code
At every step, where I don't need the full code, the LLM will start coding and I need to stop it and add "Do not generate code. The focus is still design".
This is why I built my chat app to let me manipulate LLM responses. If I feel it is not worth knowing, I'll just erase parts of it to ensure the conversation doesn't get side tracked. Or I will go back to the original user message and modify it to say
### IMPORTANT
- Do not generate more code than required.
The nice thing about LLM conversations are, every time you chat, the LLM treats it as a first time conversation, so this trick will work if the model is smart enough.
To go off on a tangent: yes, good developers can produce good code faster. And bad developers can produce bad code faster (and perhaps slightly better than before, because the models are mostly a bit smarter than copy-and-paste is).
Everyone potentially benefits, but it won't suddenly turn all bad programmers into great programmers.
I agree if devs iterated over the results it could be good, but that has not been what I have been seeing.
It is not a traditional tool because tools we had in the past had expected results.
Now in cursor I just write "extract this block of code into its own function and set up the unit tests" and it does it, with no configuration from my part. Before I'd have a snippet for the unit test boilerplate for that specific project, I'd have to figure out the mocks mysel, etc.
Yes, if you use AI to just generate new code blindly and check it in without no understanding, you end up with garbage. But those people were most likely copy pasting from SO before AI, AI just made them faster.
Each rewrite is a chance to add subtle bugs, so I take issue with the description of LLMs "working on existing code". They don't use text editors to manipulate code like we do (although it might be interesting if they did) and so will have different issues.
As far as I'm aware, tools like Aider don't do this, it does targeted changes.
The way I do it "copy/paste into chat interface", I just have the relevant context and just change the parts required.
I must be very different, as very little of my coding time is spent typing.
But typing throughput has never been my major bottleneck. Refactoring is basically never just straight code transforms, and most of my time is spent thinking, exploring or teaching these days
Some kinds of coding require very little thinking, though. Converting a design to frontend interface or implementing a CRUD backend are mostly typing.
Maybe it's because I'm used to working with constant interruptions, but until what I want is on the screen, I can't start thinking about the next thing. E.g. if I'm making a new class, I'm not thinking about the implementation of the inner functions until I've got the skeleton of the class in place. The faster I get each stage done, the faster I work.
It's why I devoted a lot of time getting efficient at vim, setting up snippets for my languages, etc. AI is the next stage of that in my mind.
Maybe you can keep thinking about next steps while stuff is "solved" in your head but not on the screen. It also depends on the type of work you're doing. I've spent many hours to delete a few lines and change one, obviously AI doesn't help there.
I have said it here before, because I would love to see some videos of HNers who complain AI gives them crap as we are getting amazing results on large and complex projects... We treat AI code the same as human code, we read it and recommend or implement fixes.
When I write code, the hard part is already done -- the mental model behind the program is already in my head and and I simply dump it to keyboard. (At least for me typing speed has never been relevant as a limiting factor)
But I read code I have to reassemble the mental model "behind" it in my head from the output artifact of the thought processes.
Of course one needs to read code of co-workers and libraries -- but it is more draining, at least for me. Skimming it is fast but reading it thoroughly enough to find bugs by reading requires making the full mental model of the code which takes more mental effort for me at least.
There is a huge difference in how I read code from trusted experienced coworkers and juniors though. AI falls in the latter category.
(AI is still saving me a lot of time. Just saying I agree a lot that writing is easier than reading still.)
That's a vital part of writing software though.
Reading is massively passive, and in fact much more mentally tiring if whole reading is in detective mode 'now where the f*ck are some hidden issues now'. Sure, if your codebase is 90% massive boilerplate then I can see quickly generated code saves a lot of time, but such scenarios were normally easy to tackle before LLMs came. Or at least those I've encountered in past few decades.
Do you like debugging by just tracing the code with your eyes, or actually working on it with data and test code? I've never seen effective use of such regardless of seniority. But I've seen in past months wild claims about magic of LLMs that were mostly un-reproduceable by others, and when folks were asked for details they went silent.
Much, much harder. Sure, you can skim large volumes of code very quickly. But the type of close reading that is required to spot logic bugs in the small is quite taxing - which is the reason that we generally don't expect code review processes to catch trivial errors, and instead invest in testing.
Examples from every day reality in my company; writing 1000s of lines of react frontend code is all LLM (in very little time) and reviews catch all the issues while the database implementation we are working on we spend sometimes one hour on a few lines and the LLM suggest things but they never help. Reviewing such a little bit of code has no use as it's the result of testing a LOT of scenarios to get the most performance out in the real world (across different environments/settings). However, almost everyone in the world is working on (similar issues like) the former, not the latter, so...
Maybe we just located the actual problem in this scenario.
Unfortunately I've made some career choices that mean I've very rarely been in that position - weird mobile hardware dependencies and/or massive clouds of micro services both render this technique pretty complicated to employ in practice.
It’s a symptom less of the career choice and more of poor coding practices.
Due to unfortunate constraints of operating in the real world, one can only run integration tests in-situ, as it were (on real hardware, with real dependencies).
Yes, I have to read what it writes, and towards the end it gets slow and starts making dumb mistakes (always; there's some magically bad length at which it always starts to fumble), but I feel like I got the advantages of pairing out of it without actually needing to sit next to another human? I'll finish the script off myself and review it.
I don't know if I've saved actual _time_ here, but I've definitely saved some mental effort on a menial script I didn't actually want to write, that I can use for some of the considerably more difficult problems I'll need to solve later today. I wouldn't let it near anything where I didn't understand what every single line of code it wrote was doing, because it does make odd choices, but I'll probably use it again to do something tomorrow. If it needs to be part of a bigger codebase, I'll give it the type-defs from elsewhere in the codebase to start with, or tell it it can assume a certain function exists
I feel like I'm going a little bit more insane whenever folks say this, because one of the primary roles of a software engineer is to develop tools and abstractions to reduce or eliminate boilerplate.
What kind of software are you writing, where generating boilerplate is the limiting factor (and why haven't you fixed that)?
Is it? Says who? Not only do I see entire clans of folk appearing who say ; DRY sucks, just copy/paste, it's easier to read and less prone to break multiple things with one fix vs abstractions that keep functionality restricted to one location, but also; most programmers are there to implement crap their bosses say, that crap almost never includes 'create tools & abstractions' to get there.
I agree with you actually, BUT this is really not what by far most people working in programming believe one of their (primary) roles entail.
I use both :)
Vim has a way to run shell programs with your selection as standard input as you'd know, and it will replace the selection with stdout.
So I type in my prompt e.g "mock js object array with 10 items each having name age and address" do `V!ollama run whatever` for example and it will fill it in there.
Now this is blocking and I have a hacky way to run it async and fill it in based on marks later in my vimrc. Neovim really since I use jobstart().
This also works with lots of other stuff, like quick code/mock generation e.g sometimes instead of asking an LLM I just write javascript/python inline and `vap!node`/python on it.
I was amazed at how great ChatGPT and DeepSeek and Claude etc are at quickly throwing together some small visualisations in Python or JavaScript. But they struggled a lot with Bird-style literate Haskell. (And just that specific style. Plain Haskell code was much less of a problem.)
Look at it this way - a powerful debugger gives you superpowers. That doesn't mean it turns bad devs into good devs, or that devs with a powerful debugger never write bad code! If somebody says a powerful debugger gives them superpowers they're not claiming those things; they're claiming that it makes good devs even better.
The reason is that I almost always have a complete understanding of everything that is happening in all code that I write. Not because I have a mega-brain; only because "understanding everything that is happening all the time" becomes rather easy if all of your code is as simple as you can possibly make it, using clear interfaces, heavily leveraging the type system, keeping things immutable, dependency inversion, and aggressively attacking any unclear parts until you're satisfied with them. So debuggers are generally not involved. It's probably a couple times per week that I enter debug mode at all.
It sounds a little like saying "imagine the driving superpowers you could have if your car could perfectly avoid obstacles for you!" Okay, sure, that'd be life-saving sometimes, but the vast majority of the time, I'm not just randomly dodging obstacles. Planning ahead and paying attention kinda makes that moot.
Also: I don't use debuggers much either. Is Superman's heat vision not a superpower because he rarely needs to melt things? :P
The conflux of events requiring me to debug someone else's code would be someone who wrote bad code without much oversight, then left the company, and there's no budget to fix their work properly. Not very common. Since I usually know what the code is supposed to be doing, such code can likely be rewritten properly in a small fraction of the original time.
In any other professional field that would be grounds for termination for incompetence. It's so weird that we seem to shrug off that kind of behavior so readily in tech.
I suspect people get away with saying "I don't know why that didn't work, I did what the computer told me to do" a lot more frequently than they get fired for it. "I did what the AI said" will be the natural extension of this
I'm not sure what you mean here? Are you saying that if you work with the same language for years that you're some how not proficient with using LLMs?
Don't take this the wrong way, but maybe you are.
For example, this weekend I was working on something where I needed to decode a Rails v3 session cookie in Python. I know, roughly, nothing about Rails. In less than 5 minutes ChatGPT gave me some code that took me around 10 minutes to get working.
Without ChatGPT I could have easily spent a couple hours putzing around with tracking down old Rails documentation, possibly involving reading old Rails code and grepping around to find where sessions were generated, hunting for helper libraries, some deadends while I tried to intuit a solution ("Ok, this looks like it's base64 encoded, but base64 decoding kind of works but produces an error. It looks like there's some garbage at the end. Oh, that's a signature, I wonder how it's signed...")
Instead, I asked for an overview of Rails session cookies, a fairly simple question about decoding a Rails session cookie, guided it to Rails v3 when I realized it was producing the wrong code (it was encrypting the cookie, but my cookies were not encrypted). It gave me 75 lines of code that took me ~15 minutes to get working.
This is a "spare time" project that I've wanted to do for over 5 years. Quite simply, if I had to spend hours fiddling around with it, I probably wouldn't have done it; it's not at the top of my todo list (hence, spare time project).
I don't understand how people don't see that AI can give them "superpowers" by leveraging a developers least productive time into providing their highest value.
You didn't blindly use it, you still used your expertise. You gave it a high quality prompt and you understood the reply enough to adjust it.
In other words, you used it correctly. It complimented your expertise, rather than replace it.
I'm unwilling to write code^1 that's not correct, or at least as correct as I'm able to make it. The single most frustrating thing in my daily life is dealing with the PoC other devs cast into the world that has more bugs than features, because they can't tell it's awful. I've seen code I've written at my least productive, and it's awful and I'm often ashamed of it. It's the same quality as AI code. If AI code allows me to generate more code that I'm ashamed of, when I'm otherwise too tired myself to write code, how is that actually a good thing?
I get standards for exclusively personal toy projects, and stuff you want others to be able to use are different. But it doesn't add value to the world if you ship code someone else needs to fix.
^1 I guess I should say commit/ship instead, I write plenty of trash before fixing it.
Last year I took a piece of code from Ansible that was extremely difficult to understand. It took me the better part of a day to do a simple bug fix in it. So I reimplemented it, using several passes of having it implement it, reviewing it, asking LLMs to make it simpler, more obvious code. For your review: https://github.com/linsomniac/symbolicmode/blob/main/src/sym...
It's indeed like having a super junior, drunk intern working for you if we are using the gps analogy.
Some work is done but you have to go over it and fix a bunch of things.
hey look, we found him!
The day when generative AI might hope to completely handle a coding task isn't here yet - it doesn't know your full requirements, it can't run your integration tests, etc. For now it's a tool, like a linter or a debugger - useful sometimes and not useful other times, but the responsibility to keep bugs out of prod still rests with the coder, not the tools.
Give examples and let it extrapolate.
New better models are coming out almost daily now and it's almost common knowledge that Copilot was and is one of the worst. Especially right now, it doesn't even come close to what better models have to offer.
Also the way to use them is to ask for small chunks of code or questions about code after you gave them tons of context (like in Claude projects for example).
"Not ready for prime time" is also just factually incorrect. It is already being used A LOT. To the point that there are rumors that Cursor is buying so much compute from Anthropic that they are making their product unstable, because nvidia can't supply them hardware fast enough.
Which is an argument for quality why? Bad coders are not going to produce better code that way. Just more with less effort.
Who's talking about quality anyway?
Code quality is not and was never the number one most important thing in business, popularity of a product/service or keeping a job. You may find that unfortunate (I agree), but it's just how it is based on my own 15yr+ experience.
The integration with the editor was neat but the quality of the suggestions were no different than what I'd had with Copilot much earlier, and the pathological cases where it just spun off into some useless corner of its behavior (recommending code that was already in the very same file, recommending code that didn't make any sense, etc.) seemed to happen more than with Copilot.
This was a ridiculously simple project for it to work on, to be clear, just a parser for a language I started working on, and the general structure was already there for it to work with when I started trying Cursor out. From prior experience I know the base is pretty easy to work with for people who aren't even familiar with it (or even parsing in general), so I think given the difficulties that Cursor had even putting together pretty basic things it might be that a user of Cursor would see minimal gains in velocity and end up having less understanding in the medium to long term, at least in this particular case.
Ive had ChatGPT do the same thing with code involving SQLAlchemy.
I think we're in a period similar to self-driving cars, where the LLMs are pretty good, but not perfect; it's those last few percent that break it.
That sounds like a lot of developers I've worked with.
I would love to go through mock interviews for myself with this approach just to have some interview-specific experience.
>> So far, everyone that elected to use GPT did much worse. They did not know what to ask, how to ask, and did not "collaborate" with the AI.
Thanks for sharing your experience! Makes sense actually.
As an interviewer, it's wild to me how many candidates think they can get away with it, when you can very obviously hear them typing, then watching their eyes move as they read an answer from another screen. And the majority of the time the answer is incorrect anyway. I'm happy that we won't have to waste our time on those candidates anymore.
Instead, we were looking to see if they followed instructions, and if they left anything out.
I never had a chance to test it out, since we hadn't hired anyone new in so long, but ChatGPT/etc would almost always fail this exam because of how bad it is at making sure everything was included.
And bad programmers also failed it. It always left us with a few candidates that paid attention, and from there we figure if they can do that, they can learn the rest. It seemed to work quite well.
I was recently laid off from that company, and now I'm realizing that I really want to see what current-day candidates would turn in. Oh well.
Companies seem to think that we program just for fun and ask to make a full blown app... also underestimating the time candidates actually spend making it.
Can you really blame them? If you're not a houshold name, why would you expect someone to spend hours researching your specific company?
On the other hand, it can come off as creepy if your a small company and suddenly someone nerds out about how your CEO said this one thing at a talk years ago and knows your lead has cancer based on his personal blog. I'd rather just treat it as a transaction of my skills and services for money. We are not a family (multiple layoffs have taught me so)
> it’s effective and worthwhile to spend more time curating higher quality samples than to scatter and hope.
Not in this market. Too many ghost jobs, too many people ghosting after multiple rounds. Too many hiring freezes when you spend a month talking with a company. If you want respect from candidates, don't disrespect them.
You sound like you’ve been burned. That sucks and I’m sorry, I sympathize. I’m hearing that the job market is very tough right now. A big part of that is because it’s extremely competitive. Taking it personally and assuming it’s disrespect isn’t going to help get the job though (even if there was disrespect… but that’s not the only explanation, so it’s a dangerous assumption).
Well, everyone has different experiences. I never felt like knowing about a company put me ahead in my early days. I guess I have a dump stat in Charisma (not surprised).
Like you said, the market is competitive. No one's going to take the nice guy over the one who blitz's an interview unless that nice guy has connections. Those few minutes of thousands of applications adds up to days of research. I just lack that time and energy these days.
>You sound like you’ve been burned. That sucks and I’m sorry, I sympathize.
several times, yes. It's honestly worse than my first job search out of college 10 years ago.
>Taking it personally and assuming it’s disrespect isn’t going to help get the job though
I only ask for basic decency. Keep a candidate in the loop, don't drag the process on for the sake of it, any take home should warrant a response (even if it's a template rejection letter). i.e. respect people's time.
I haven't been burned in a lot of my interviews, I'm not talking about bummers like the several times I was interviewing before a hiring freeze. I don't even treat non-responses as an interview process. But several of them just end with absolutely no communication nor closure after speaking for weeks with recruiters and hiring managers.
I don't know what to call that in a day and age where AI is supposedly increasing efficiency, other than disrespect. This has never happened before 2023, which makes the times all the more weirder.
The only time the bespoke approach works is if you have like 30 candidates only. But then there are still issues here because the candidate is still one in thirty so if he does a bespoke approach 30 times it takes an inordinate amount of time.
recently I spent a good 10 hours making a crossword solver. Hiring freeze a few days after I turned it in. I completely get GP's mentality.
The one who loses all power is the new junior straight out of school, which used to already be difficult to distinguish from many other candidates with similar resumes: Now they compete with thousands upon thousands of total fakes which claim more experience anyway.
Even something like citizenship is a differentiating factor: an undergrad who applies to, say, a national lab won't compete with foreign students by definition.
As a person looking for a job, I’m really not sure what to do. If people are lying on their resumes and cheating in interviews, it feels like there’s nothing I can do except do the same. Otherwise I’ll remain jobless.
But to this day I haven’t done either.
I haven't looked for development-related jobs this millennium, but it's unclear to me how effective a crutch AI is for interviews--at least for well-designed and run interviews. Maybe in some narrow domains for junior people.
As a few of us have written elsewhere, I consider not having in-person interviews past an initial screen sheer laziness and companies generally deserve whoever they end up with.
In my opinion, if a prospective employee is able to successfully use AI to trick me into hiring them, then that is a hell of a lot closer to the actual work they'll be hired to do (compared to leetcode).
I say, if you can cheat at an interview with AI, do it.
What you’re suggesting here isn’t any different than submitting a fraudulent resume because you disagree with the required qualifications.
If a candidate were up front with me and asked if they could use AI, or said they learned an answer from AI and then wanted to discuss it with me, I'd be happy with that. But attempting to hide it and pretend they aren't using it when our interview rules specifically ask you not to do it is just being dishonest, which isn't a characteristic of someone I want to hire.
What is being suggested here is not participating in the mind numbing process that is called ‘applying for a job’.
What? No, it isn't.
Regardless, if the job requirements state "X years of XYZ experience" and you have to have >X years of experience, then using AI to look up how to do a leetcode problem for some algorithm you haven't used since your university days is absolutely not "false pretenses" nor fraud.
well that's the neat part... they aren't going to. All this AI stuff just happened to coincide with a recession no one wants to admit, amplifying the issue.
So yea, even if I'm desperate I need to be mindful of my time. I can only do so many 4-5 stage interviews only to be ghosted, have the job close, or someone else who applied earlier get the position.
Because committing fraud to get hired is a pretty shitty way to live your life
What you're missing here is that this is an individual's answer to a systemic problem. You don't apply when it's _one_ obnoxious employer.
When it's standard practice across the entire industry, we have a problem.
> submitting a fraudulent resume because you disagree with the required qualifications.
This is already worryingly common practice because employers lie about the required qualifications.
Honesty gets your resume shredded before a human even looked at it. And employers refusing to address that situation is just making everything worse and worse.
Sadly, it’s not just the interview loops—the way candidates are screened for roles also sucks.
I’ve seen startups trying to innovate in this space for many years now, and it’s surprising that absolutely nothing has changed.
I don't want to be too crass, but I'm not surprised people who can startup a business are precisely the ones who hyper-fixate on efficiency when hiring and try to find the best coders. Instead of the best engineers. When you need to put your money where you mouth is, many will squirm back to "what works".
Does it? Mine is honest, fairly normal, and gets me through to interviews fine. What are common lies and why are they necessary?
Due to the prevalence of the practice this is tantamount to suggesting constructive unemployability.
People were up in arms about widespread doping during the Lance Armstrong era. But the only viable alternative to doping at the time was literally to not compete at all.
I work in security, and our questions are pretty basic stuff. "What is cross-site scripting, and how would you protect against it?", "You're tasked with parsing a log file to return the IP addresses that appear at least 10 times, how would you approach this?" Stuff like that. And then a follow-up or two customized to the candidate's response.
I really don't know how we could possibly make it easier for candidates to pass these interviews. We aren't trying to trick people, or weed people out. We're trying to find people that have the foundational experience required to do the job they're being hired for. Even when people do answer them incorrectly, we try to help them out and give them guidance, because it's really about trying to evaluate how a person thinks rather than making sure they get the right answer.
I mean hell, it's not like I'm spending hours interviewing people because I get my rocks off by asking people lame questions or rejecting people; I want to hire people! I will go out of my way to advocate for hiring someone that's honest and upfront about being incorrect or not knowing an answer, but wants to think through it with me.
But cheating? That's a show stopper. If you've been asked to not use ChatGPT, but you use it anyway, you're not getting the benefit of the doubt. You're getting rejected and blacklisted.
because it matches my experience. I work in games and interviews are more varied (math, engine/language questions, game design questions, software design patterns). I'd still say maybe 30% of them do leetcode interviews, and another 40% bring in leetcode questions at some point. I hate it because I need to study too many other types of questions to begin with, and leetcode is the least applicable.
Out of curiosity, did anyone just reply with `awk ... | sort | count ... | awk`? Its certainly what I would do rather than writing out an actual script.
I come from a math background rather than CS and code for fun / personal projects, so don't know the 'proper' names for some algorithms from memory. I could have done some leetcode prep / revision if I had any indication that it was coming up, though the interview was pretty much a waste of time. I told them that and made a stab at it, though they didn't seem interested in engaging at all and barely made eye contact during the whole interview.
On the other hand, if you have 1,000 candidates, and you only need 1, why not do it if the top candidate selected by this method can do well on the test and your work?
The companies that do this only do it because they can. They have to have hundreds of people applying. The companies that don’t do this basically don’t have many people applying.
Top athletes do it because they're essentially at the limits of human performance and those drugs are the only edge they can reasonably get.
Different people have a different threshold for cheating no matter the stakes. I imagine some people vheat even if they know the answer - just to be sure.
I don't/won't cheat, as I am a rather anxious person who can't really handle "covert ops". But at this point I totally understand those who do.
I think LLM performance on previously seen questions like interview questions is too good for it to be allowed. I wouldn't mind someone using an IDE or API docs, but you have to draw the line somewhere. It's like how you can't use a calculator that can do algebra on a calculus test. It just doesn't accomplish the goal of testing anything if you use too much tech, and using the current LLMs that all suck in general but can nail these questions that have a million examples online is bad. I would much rather see someone consult online references for information than to see them use an LLM.
Well, sounds like you know the solution. Or set your sights on a job that interviews a different way.
I think it's mostly leetcode "easy", anyway. Maybe some medium. Never seen a hard, except maybe from one smartass at Google (they were not expecting a perfect answer). Out of a dozen technical interviews, I don't think I've ever needed to know a data structure more exotic than a hash map or binary search tree.
The amount of deliberate practice required to stand out is probably not more than 10-20 hours, assuming you do actually have the programming and CS skills expected for a FAANG job. It's unlikely you need to do months of grinding.
If 20 hours of work was all that stood between me and half a million dollars a year, I'd consider myself pretty lucky.
If you’re totally unqualified, 20 hours of leetcoding won’t get you a job at Meta.
Why does it feel like that when you’re replying to someone who already points out that it doesn’t work? Cheating can prevent you from getting a job, and it can get you fired from the job too. It can also impede your ability to learn and level up your own skills. I’m glad you haven’t done it yet, just know that you can be a better candidate and increase your chances by not cheating.
Using an LLM isn’t cheating if the interviewer allows it. Whether they allow it or not, there’s still no substitute for putting in the work. Interviews are a skill that can (and should) be practiced. Candidates are rarely hired for technical skill alone. Attitude, communication, curiosity, and lots of other soft skills are severely underestimated by so many job seekers, especially those coming right out of school. A small amount of strengthening your non-code abilities can improve your odds much faster than leetcode ever will. And if you have time, why not do both?
But YMMV. I have 9 years and still can get interviews the old fashioned way.
Never buy into this mentality. Because once you do, it never goes away. After the interview, your coworkers might cheat, so you cheat too. Then your business competitors might cheat, so you cheat too. And on and on.
In most interview tasks you are not solving the task “with” ai.
Its AI who solves the task while you watch it do it.
The ultimate question that an interview is trying to answer is not "can this person solve this equation I gave them?", it's usually something along the lines of "does this person exhibit characteristics of a trustworthy and effective employee?". Using AI when you've been asked not to is an automatic failure of trust.
This isn't new or unique to AI, either. Before AI people would sometimes try to look up answers on Google. People will write research papers by looking up information on Wikipedia. And none of those things are wrong, as long as they're done honestly and up front.
Imagine you need to hire some people, and think about what you’d want. That’ll answer your question. Do you want people who don’t know but think AI will solve the problems, or do you want people who are capable of thinking through it and coming up with new solutions, or of knowing when and why the AI answer won’t work?
Keep in mind that the amount of time you spend in a real job solving clear and easy interview style problems that an LLM can answer is tiny to none. Jobs are most often about juggling priorities and working with other people and under changing conditions, stuff Claude and ChatGPT can’t really help you with. Your personality is way more important to your job success than your GPT skills, and that’s what interviewers want to see… your personality & behavior when you don’t know the right answer, not ChatGPT’s personality.
Eh, I wish more people felt that way, I have failed so many interviews because I haven't solved the coding problem in time.
The feedback has always been something along the lines of "great at communicating your thoughts, discussing trade-offs, having a good back and forth" but "yeah, ultimately really wanted to see if you could pass all the unit tests."
Even in interview panels I've personally been a part of, one of the things we evaluate (heavily) is whether the candidate solved the problem.
I admire this worldview, and wish for it to be true, but I can't help but see it in conflict with much of what floats around these parts.
There's a recent thread on Aider where the authors' proudly proclaim that ~80% of code is written by Aider itself.
I've no idea what to make of the general state of the programming profession at all at the moment, but I can't help but feel learning various programming trivia has a lower return on investment than ever.
I get learning the business and domain and etc, but it seems like we're in a fast race to the bottom where the focus is on making programmers' skills as redundant as possible as soon as possible.
Honest interviewers may not realize how dishonest other interviewers became in such recent times (2-3 years ago). Interviewing today compared to COVID times is night and day. Let alone the 10's Gold Rush.
The respect is long gone.
Remember that you are only catching the candidates who are bad at cheating.
Start with recording the session and blocking right-click, and you are halfway there. Its not hard.
The AI app has helped me surface top candidates. I don't even look at resumes anymore. There's no point. I interview the top 10 out of 200, and then do my references and select.
We actually allow using AI in our in-person technical interviews, but our questions are worded to fail safety checks. We'll talk about smuggling nuclear weapons, violent uprising, staging a coup, manufacturing fentanyl, etc. (within the context of system design) and that gives us really good mileage on weeding out those who are just transcribing what we say into AI and reading the response.
Just one time someone got mad and yelled at the interviewer about nothing specific, just stuff like I’m not who you are looking for, you will never find anybody to hire.
I love the idea of embedding sensitive topics that ChatGPT and other LLMs will steer clear of, within the context of a coding question.
Have you ever had any candidate laugh?
Any candidates find it offensive?
No one’s found it offensive, the prompt is mostly neutral just very “dangerous activity” coded.
I had just recently lead through several interview rounds for software engineering role and we have not had any issue with LLM use. What we do for the technical interview part is very simple - live whiteboarding design task where we try to identify what the candidate's focus is and might pivot at any time or dig deeper into particular topics. Sometimes, we will even go as detailed as talking about particular algorithms the candidate would use.
In general, I found that this type of interview is the most fun for both sides. The candidates don't feel pressure that they must do the only right thing as there is a lot of room for improvisation; the interviewers don't get bored with repetitive interviews over and over as new candidates come by with different perspectives. Also, there is no room for LLM use because the candidate has to be involved in drawing on the whiteboard and showing their technical presentation skills, which are very important for developers.
At that point it seems like trying too hard, but be aware there are theoretical approaches which are extremely hard to detect (the inevitable evolution of sticky notes on the desk, or wall behind the monitor).
I'm genuinely curious what questions you ask during the behavioral interview. Most companies ask questions like "recall a time when..." and I know people who struggle with these kinds of questions despite being good teammates, either because they find it difficult to explain the situation, or due to stress. And recruitment process is not a "basic conversation" — as a recruiter you're in far more comfortable position. I find it hard to believe anyone would use an LLM if you ask them question like "what were your responsibilities in your last role", and I do see how they might've primed the chat to help them communicate an answer to a question like "tell me about a situation when you had a conflict with your manager"
That normally prompts some follow ups about specific work, specific projects, if they know so-and-so moot at their old company. I call it behavioral because I don’t have another word but it’s not brainteasers and etc like consulting/finance interviews.
I wouldn’t be surprised if the effect of Google Docs and Gmail forcing full AI, is a generation of people who can’t even talk about themselves, and can’t articulate even a single email.
Is it necessary? Perhaps. Will it make the world boring? Yes.
Detection of AI in remote interviews, behavioral and technical, just has to be taught today if you are ever interviewing people that don't come from in-network recommendations. Completely fake candidates are way too common.
- share your screen
- download/open the coding challenge
- you can use any website, Stack Overflow, whatever, to answer my questions as long as it's on the screenshare
My goal is to determine if the candidate can be technically productive, so I allow any programming language, IDE, autocompleter, etc, that they want. I would have no problem with them using GPT/Copilot in addition to all that, as long as it's clear how they're solving it.
It proved to be awkward and clumsy very quickly. Some candidates resisted it since they clearly thought it would make them judged harsher. Some candidates were on the other extreme and basically tried asking ChatGPT the problem straight up, even though I clarified up front "You can even use ChatGPT as long as you're not just directly asking for the solution to the whole problem and just copy/pasting, obviously."
After just the initial batch of candidates it became clear it was muddying things too much, so I simply forbade using it for the rest of the candidates, and those interviews went much smoother.
Unfortunately in our industry it's pretty much all personal anecdotes on what works better and what doesn't.
The challenge should be in determining if ChatGPT is correct.
But yeah, the point is that once I applied it in practice it did quickly become confusing, so now I know from experience not to use it.
I think the other suggestions in this thread about how to use it are good ones, but they would present their own meta challenges for an interview too. Just about finding whatever balance works for you I guess.
But for me, it's just not how my brain works. If someone is watching me, I'll be so self-conscious the entire time you'll get a stream of absolute nonsense that makes me look like I learned programming from YouTube last night. So it's not worth the time.
You want some good programming done? I need headphones, loud music, a closed door and a supply of Diet Coke. I'll see you in a few hours.
Whereas my natural approach would be to take a long shower, workout etc and let my brain wander a bit before digging into it. But that wouldn’t fly during an interview..
That said, now we're just talking about take home challenges for interviews and you always hear complaints about those too. And shorter, async timed challenges (something like "Here's a few hours to solve this problem, I'll check back in later") are now going to be way more difficult to judge since AI is now ubiquitous.
So I really don't think there's any perfect methodology out there right now. The best I can come up with is to get the candidate in front of you and talk through problems with them. The best barometer I found so far was to set up a small collection of files making up a tiny app and then have candidates debug it with me.
This is a great one! I wish more companies tried that.
In an interview, the coding challenge is often to produce something new from scratch while being closely monitored by people you don't know, who control your financial future.
When working with a "junior," you'd already be fairly familiar with the code base, build system, and best practices. And with a junior, you're not likely to be solving things that require deep concentration, like never-before-seen problems or architectural work (or screwball interview-tests). And, unlike an interview, if something does require all my focus, it's very easy to defer. Take a break and think about it alone.
No, it's not "obvious" whatsoever. Actually it's obviously confusing: why you are allowing them to use ChatGPT but forbidding them from asking the questions directly? Do you want an employee who is productive at solving problems, or someone who guess your intentions better?
If AI is an issue for you then just ban it. Don't try to make the interview a game of who outsmart who.
> If AI is an issue for you then just ban it.
Yes, that was the conclusion I just said we rapidly came to.
- You get to see how they then review the generated code, do they spot potential edge cases which the AI missed? - When I ask them to make a change not in the original spec, a lot of them completely shut down because they either didn't understand the code generated well enough, or they themselves didn't really know how to code.
And you still get to see people who _do_ know how to use AI well, which at this point is a must for its overall productivity benefits.
if i see anything remotely challenging i dip out. interviewing is just a numbers game nowadays so i dont waste time on interviews if they seem like they're gonna burn me out for the rest of the day. granted i have 11 years experience
Too many people are the opposite that I would literally never tell you
And this works.
what can we do to help that?
I’ve had interviews where AI use was encouraged as well.
but so many casual tirades against it dont make me want to ever try being forthcoming. most organizations are realistically going to be 10 years behind the curve on this
Some of the questions in our interview loop have been posted in github... which means every AI has trained on them specifically. They are, therefore, useless if you have AI turned on. And if you interview enough people, someone will post on github, and therefore your question will have a pretty short shelf life before it's in training and instantly solved.
I do not want AI. The human is the value add.
I understand that people won't feel super comfortable with this, and I try not to roast the candidate with leetcode. It should be a conversation where I surface technical reality and understanding.
Once you get to the interview process, it's very clear if someone thinks they can use AI to help with the interview process. I'm not going to sit here while you type my question into OpenAI and try to BS a meaningful response to my question 30 seconds later.
AI-proof interviewing is easy if you know what you're talking about. Look at the candidates resume and ask them to describe some of their past projects. If they can have a meaningful conversation without delays, you can probably trust their resume. It's easy to spot BS whether AI is behind it or not.
If you do a rote interview that's easy to game with AI, it will certainly be harder to detect them cheating.
If you have an effective and well designed open ended interview that's more collaborative, you get a lot more signal to filter the wheat from the chaff.
I understood their point but my point is a direct opposition to theirs, that at some point with AI advances this will essentially become impossible. You can make it as open ended as you want but if AI continues to improve, the human interviewee can simply act as a ventriloquist dummy for the AI and get the job. Stated another way, what kind of "effective and well designed open ended interview" can you make that would not succumb to this problem?
People don't really call the police, nor sue over this. But they can, and have in the past.
If it gets bad, look for people starting to seek legal recourse.
People aren't developers with 5 years experience, if all they can do is copy and paste. Anyone fraudulently claiming so is a scam artist, a liar, and deserves jail time.
So you create an interview process that can only be passed by a skilled dev, including them signing a doc saying the code is entirely their work, only referencing a language manual/manpages.
And if they show up to work incapable of doing the same, it's time to call the cops.
That's probably the only way to deal with scam artists and scum, going forward.
It costs employers money to on board someone, not just in pay, but in other employees training that person. Obviously the case must be clear cut, but I've personally hired someone who clearly cheated during the remote phone interview, and literally couldn't even code a function in any language in person.
There are people with absolutely no background as a coder, applying to jobs with 5 years experience, then fraudulently misrepresenting the work of others at their own, to get the job.
That's fraud.
As I said, it's not being prosecuted as such now. But if this keeps up?
You can bet it will be.
Because it is fraud.
I won't name names, but there are a lot of Consulting companies that feed off Government contracts that are literally this.
"Experience" means a little or a lot, depending on your background. I've met plenty of people with "years of experience" that are objectively terrible programmers.
If the AI premise is true, then it's this or good programmers, and good companies will never meet.
In-person interviews, second round comes with a plane ticket. This used to be the norm.
Likewise with hiring; at a small company you're looking to fill an immediate need and are losing money every day the role isn't filled. You wouldn't bring in every viable candidate, you'd bring in the first viable candidate.
FAANG hiring practices assume a budget far past any exit point in your mind.
To put the whole concern in a nutshell: If AI was good enough to fool a seasoned engineer in an interview, that engineer would already be using the AI themselves for work and not need to hire an actual body.
To be fair, many humans do too, but many promising candidates even at the mid-level band of experience who thrive at organizations I've approved them into are able to eventually get to a good enough balance of many tradeoffs (technical and otherwise) with a pretty clean and compact amount of back and forth that demonstrates thoughtfulness, curiosity and efficacy.
If someone can get to that level of capability in a technical interviewing process using AI without it being noticeable, I'd be really excited about the world. I'm not holding my breath for that, though (and having done LOTS of interviews over the past few quarters, it would be a great problem to have).
My solution, if I were to have the luxury of having that problem, would be a pretty blunt instrument -- I'd instead change my process to actually have AI use of tools be part of the interviewing process -- I'd give them a problem to solve, a tuned in-house AI to use in solving the problem, and have their ability to prompt it well, integrate its results, and pressure check its assumptions (and correct its mistakes or artifacts) be part of the interview itself. I'd press to see how creatively they used the tool -- did they figure out a clever way to use it for leverage that I wouldn't have considered before? Extra points for that. Can they use it fluidly and in the heat of a back and forth of an architectural or prototyping session as an extension of how they problem solve? That will likely become a material precondition of being a senior engineer in the future.
I think we're still a few quarters (to a few years) away from that, but it will be an exciting place to get to. But ultimately, whether they're using a tool or not, it's an augment to how they solve problems and not a replacement. If it ever gets to be the latter, I wouldn't worry too much -- you probably won't need to do much hiring because then you'll truly be able to use agentic AI to pre-empt the need for it! But something tells me that day (which people keep telling me will come) will never actually come, and we will always need good engineers as thought partners, and instead it will just raise the bar and differentiation between truly excellent engineers and middle of the pack ones.
If you look solely at my GitHub you'd likely reject me right away.
I wish I had the time and energy for passion projects in programming. I so wish it was so. But commercial work has all but destroyed my passion for programming, though I know it can be rekindled if I can ever afford to take a properly long sabbatical (at least 2 years).
I'll more agree with your parent / sibling comments: take a look at the resume and look for bad signs like too vanilla / AI language, too grandiose claims (though when you are experienced you might come across as such so 50/50), or almost no details, general tone etc.
And the best indicator is a video call conversation, I found as a candidate. I am confident in what I can do (and have done), I am energetic and love to go for the throat of the problems on my first day (provided the onboarding process allows for it) and it shows -- people have told me that and liked it.
If we're talking passion, I am more passionate about taking a walk with my wife and discussing the current book we're reading, or getting to know new people, or going to the sauna, or wondering what's the next meetup we should be going to, stuff like that. But passion + work, I stand apart by being casual and not afraid of any tech problems, and by prioritizing being a good teammate first and foremost (several GitHub-centric items come to mind: meaningful PR comments and no minutiae, good commit messages, proper textual comment updates in the PR when f.ex. requirements change a bit, editing and re-editing a list of tasks in the PR description).
I already do too much programming. Don't hold it against me if I don't live on the computer and thus have no good GitHub open projects. Talk to me. You'll get much better info.
I would also add meticulous attention to documenting requirements and decisions taken along the development process, especially where compromises were made. All the "why's", so to speak.
But yes, commercial development, capital-A "Agile" absolutely kills the drive.
And yep I didn't want to make my comment too big. I make double sure to document any step-by-step processes on "how to make X or Y work", especially when I stumble upon a confusing bug in a feature branch. I go out of my way to devise a 100% reproducible process and document it.
Those, plus yours, and even others, are what makes a truly good programmer IMO.
And tbh, at the senior level they rarely care about personal projects. I must have had 60+ interviews and I feel a lack of a github cost me maybe 2 positions. When you job is getting a job, you rarely have the time for passion.I'm doing contract work in the meantime; prevents gaps from showing, more appealing than a personal project, and I can talk about that to the extent of my NDA (plenty of tech to talk about without revealing the project)
Same. I could afford not working throughout most of 2023 but I had to deal with ruined health + my leeway didn't last as long as I hoped so I was back on the treadmill just when I was starting to enjoy some freedom and a peace of mind.
> And tbh, at the senior level they rarely care about personal projects. I must have had 60+ interviews and I feel a lack of a github cost me maybe 2 positions.
I have no idea how much it costed me but I was told in no uncertain terms 10+ times that having a GitHub portfolio would have meant no take-home assignment, and skipping parts of the interview I already attended. So it definitely carries weight _and_ can help shorten hiring processes.
So I don't feel it was a deal-breaker for the people who interviewed me either but I think it would have netted me more points, so to speak.
Assuming you are graded and are the same person:
Without portfolio: 7/10
With portfolio: 8/10
...for example.
> I'm doing contract work in the meantime
Same x2, but it's mentally draining. No stability. That removes future bandwidth that would have been used for those passion projects.
TL;DR a lot of things conspire to rob you of your creative potential. :(
The long term goal should be that you should NOT be energetic, yet be able to pretend that you are. We'll see where you are at the 33 year mark :)
Also if you have a novel or disclosure sensitive passion project, GitHub may be avoided even as a very conservative bright line.
As stated above I think it can be good to find common points to enhance the interview process, but make sure to not use it as a filter.
That said, I'm waiting for an "interview assistant" product. It listens in to the conversation and silently provides concise extra information about the mentioned subjects that can be quickly glanced at without having to enter anything. Or does this already exist?
Such a product could be useful for coding to. Like watching me over the shoulder and seeing aha, you are working with so-and-so library, let me show you some key parts of the API in this window, or you are trying to do this-and-that, let me give you some hints. Not as intrusive as current assistants that try to write code for you, just some proactive lookup without having to actively seek out information. Anybody knows a product for that?
The only way I can see that working is if it spends hundreds of hours watching you to understand what you know and don't know, and even then it'll be a bit of a crap shoot.
This was 2-3 years ago in a remote interview. The candidate would hear the question, BS us a bit and then sometimes provide a good answer.
But then if we asked follow up questions they would blow those.
They also had odd 'AV issues' which were suspicious.
So you just subjectively say "this resume is too perfect, it must be bullshit"? How the fuck is any actual, qualified engineer supposed to get through your gauntlet of subjectivity?
I'm sure some small % of people get away with it by using LLaMA-FooQux-2552-Finetune-v3-Final-v1.5.6 or whatever, but realistically, the majority is going to be obvious to anyone that's been force-fed slop as part of their job.
I am imagining an AI saying my CV is AI-generated, when in reality, I do not even use Auto-correct or Auto-suggest when I (type)write! :-)
Generally, this is how to figure out if a candidate is full of crap or not. When they say they did a thing, ask them questions about that thing.
If they can describe their process, the challenges, how they solved the challenges, and all of it passes the sniff test: If they are bullshitting, they did crazy research and that's worth something too.
It's as if we were testing for replicants in Blade Runner: The AI response will rarely figure out you are aiming to look for something frustrating, that they are actually proud of, or figure out when you are looking for a hot take you can then disagree with.
AI doesn't just change the interviewing game by making it easy to cheat on these interviews, it should be changing your hiring strategy altogether. If you're still thinking in terms of optimizing for cogs, you're missing the boat—unless you're hiring for a very short term gig what you need now is someone with high creative potential and great teamwork skills.
And as far as I know there is no reliable template interview for recognizing someone who's good at thinking outside the box and who understands people. You just have to talk to them: talk about their past projects, their past teams, how they learn, how they collaborate. And then you have to get good at understanding what kinds of answers you need for the specific role you're trying to fill, which will likely be different from role to role.
The days of the interchangeable cog are over, and with them easy answers for interviewing.
"talk about their past projects, their past teams, how they learn, how they collaborate"
You have now excluded amazing engineers who suck at talking about themselves in interviews. They may be great collaborators and communicators, but freeze up selling themselves in an interview.
- “big” tech companies like Google, Amazon, Microsoft came up with these types of tech interviews. And there it seems pretty clear that for most of their positions they are looking for cogs
- The vast majority of tech companies have just copied what “big” tech is doing, including tech interviews. These companies may not be looking for cogs, but they are using an interview process that’s not suitable for them
- Very few companies have their own interview process suitable for them. These are usually small companies and therefore the number of engineers in such companies is negligible to be taken into account (most likely, less than 1% of the audience here work at such companies)
Bugs need to be fixed. Features need to be implemented. If it weren't for cogs, you'd have people just throwing new projects over the fence and dropped 6 months after release. Don't want to be another cog? Join a startup. Plenty of those hiring. The reality is that when you work at a large company, you're one of 50,000 people. By definition, only 1% are in the top 1%.
Someone has to wash the dishes and clear the tables. Let's stop looking down at jobs just because it's not hot and sexy. People who show up and provide value is great and should be appreciated.
The interview process being a circus of how many hoops you'll jump through. Which in this case is upwards of 3 months of trivia, beauracracy, and politics. And these days they don't even give you the grace of a response; they may just ghost you.
But being a cog itself is personally fine. Work to live, not live to work. But leading people on to drop them on the tip of a hat is disrespectful of everyone's time. At least a 1-2 stage interview for a dishwasher or table busser is only wasting a few hours per role applied. Time is the most valuable resource we have, of course people want to use it carefully.
Human cogs are going to be phased out. I'm not an AI doomer who thinks engineers are going to be replaced across the board, but the need for a human being who functions like a robot is going away fast. We need humans to do what humans do well, and humans don't do well as cogs in a machine—machines are better at that role.
The days of leetcode interviews are numbered not because they're too easy to cheat at, but because they were always optimizing for the wrong traits in most companies that cargo culted them, and even the companies that used them correctly (Big Tech) are going to rapidly need a different type of interview for the new types of hires they need.
The council itself is made of "busywork" worker bees. Slave hiring slaves - the vast majority of IT interviewers and candidates are idiot savants - they know very little outside of IT, or even realize that there is more to life than IT.
This was the norm until perhaps for about the last 10-15 years of Software Engineering.
I didn't say that. I said that this style of interview was designed to hire pluggable cogs. As others have noted, that was the correct move for Big Tech and was cargo culted into a bunch of other companies that didn't know why their interviews were shaped the way they were.
> there would be "out of the box thinkers" who would go against your ideas and cause dissent. At that point, guess what? You're going to figure out how to hire compliant people who will execute on your strategy.
In answer to your original question: yes, I'm actively involved in hiring at a 100+ person engineering org that hires this way. And no, we're not looking to figure out how to hire compliant people, we're hiring engineers who will push back and do what works well, not just act because an executive says so.
> You have now excluded amazing engineers who suck at talking about themselves in interviews. They may be great collaborators and communicators, but freeze up selling themselves in an interview.
Only if you suck at making people comfortable and at understanding different (potentially awkward) communication styles. You don't have to discriminate against people for being awkward, that's a choice you can make. You can instead give them enough space to find their train of thought and pursue it, and it does work—I recently sat in on an interview like that with someone who fits your description exactly, and we strongly recommended him.
This is the job of a good interviewer. I've run the gauntlet from terrible to great answers to the exact same questions depending on the interviewer. If you literally just ask that question out of the blue, you'll either get a bad or rehearsed response. If you establish some rapport, and ask it in a more natural way, you'll get a more natural answer.
It's not easy, but neither is being on the other side of the interviewer, and that's never been accepted as an excuse
That’s exactly what we always needed, long before LLMs arrived. That’s why all the interviews I’ve seen or give already were designed to have conversations.
I’m agreeing with you, but I’ve never seen these ‘interchangeable cog’ interviews you’re talking about.
I think every interviewer, hiring manager ought to know or be trained on these tools, your intuition about candidate's behaviour isn't enough. Otherwise, we will soon reach a tipping point where honest candidates will be at a severe disadvantage.
Im really happy it's finally broken though. Dumbest fad our industry ever had.
Every alternative I can think of is either worse, or sounds nice but impractical to implement in practice at scale.
I don’t know about you, but most interviewers out there don’t have the ability to judge the technical merit of a bullshitters’s contribution to a class or internship project in half an hour, specially if it’s in a domain interviewer has no familiarity with. And by the way, not all of them are completely dumb, they do know computer science, just perhaps not as well as an honest competitor.
They are juniors, I don't expect them to be experts, I expect eagerness and passion. They spent 4 or more years focusing on schooling, show me the results of your projects. Let them talk and see how well they understand what they did. Side projects are even better to stand out.
And you know... apparently people can still fail fizzbuzz in 2025. If you really question their ability to code, ask the basics, not if they can write a Sudoku verifier on the spot. If you aren't a sudoku game studio I don't see the application outside of "can they work with arrays?"
>I don’t know about you, but most interviewers out there don’t have the ability to judge the technical merit of a bullshitters’s contribution to a class or internship project in half an hour, specially if it’s in a domain interviewer has no familiarity with.
everyone has a different style. Personally I care a lot less about programming proficiency and a lot more about technical communication. If they only wrote 10 lines of code for a group project but can explain every aspect of the project as if they written it themselves, what am I really missing? The odds of that sort of technical reaspning being accompanied by poor coding is a lot rarer than the alternative of a Leetcode wizard who can't grasp architectural concept nor adjust to software tooling, in my experiences.
For students specifically, the strongest signal is if they've done research with a past collaborator of mine and my collaborator vouches for them. It's a great signal because it's very high barrier to entry and absolutely does not scale.
Realistically, being able to speak confidently about something they did/built during their education is a decent proxy. If they can handle open-ended follow-up questions like "what did you learn?" and "what trade-offs did you make?" and "how would you tweak it under X different requirements?" then that's a great signal too.
These aren't "gotcha" questions, but they insist on the candidate being reasonably competent and (most importantly) an actual human who can think on their feet.
100% of all job interviews are a proxy. It is not possible to perform an interview in ~4 hours such that someone sufficiently engages in what the job “actually entails”.
A leetcode interview either is or not a meaningful proxy. AI tools either do or not invalidate the validity of that proxy.
Personally I think leetcode interview are an imperfect but relatively decent proxy. And that AI tools render that proxy invalid.
Hopefully someday someone invents a better interview scheme that can reliably and consistently scale to thousands of interviews, hundreds of hires, and dozens of interviewers. That’d be great. Alas it’s a really hard and unsolved problem!
>> ... leetcode interview are an imperfect but relatively decent proxy.
I think all this is just the status quo that should be challenged instead of being justified.
When I conduct interviews (environment: a FAANG company), I focus on (a) fundamental understanding and (b) practical problems. None of the coding problems I pose are more than O[N] in complexity. Yet, my hiring decisions have never gone wrong.
How do you know? Every interview you conduct is for your team? And you’re a manager who participates in their annual review? No one bats 1000%.
> I focus on (a) fundamental understanding and (b) practical problems.
What do you ask? If you want to challenge the status quo you have to offer a replicable alternative. Not hand waive.
My apologies. I should have been more careful in making the claim. You are right in challenging me**.
>> What do you ask? If you want to challenge the status quo you have to offer a replicable alternative
I cannot share the specific problems publicly as then those become available for candidates to practice out, whereas the idea is to give them a fresh problem to think about. Further, the list of interviewers is often made available to the candidates ahead of the time, which can create a loophole as interviewers tend to have a relatively small set of problems they keep repeating for different candidates***.
In general however, I always use real-life problems. Similar examples from text books may perhaps be: (a) Find the day of the week for a given date, (b) Output a given number in textual form (e.g., 14630 -> Fourteen thousand, six hundred and thirty), (c) etc. Most of the problems I pose are motivated from actual use cases, while the remaining are taken from my own prior interviews as a candidate. As I noted before, each one is at best Big-O[N] in time complexity, just like the Fizz Buzz problem. I have brought in more complex ones only when testing for specialized skills for specialized positions.
I also focus a lot on fundamentals. There have been candidates hired for my own teams which could not fully solve the problem I posed, but who however were unambiguously going in the right direction and would not say anything wrong or stupid during their reasoning towards it. I chose to go ahead with them with some risks in my mind, however, these people proved to be stellar in performance. When fundamentals are understood well, gaps can be picked up on the job as well.
I still do stand by my original assertion that proxy problems can be avoided for interviewing. In general, the best method to solve a problem is to solve that problem itself and not another.
** Appendix: Here're the actual data points:
- For interviews within my own team where I have been an interviewer, my claim likely holds. However, this is a small number of hires (say around six-eight) so the error margins would be large. Also, I would not have information about false-rejects, hence '100% accuracy' cannot be claimed even in theory.
- I have been a interview 'bar raiser' at Amazon. Bar Raisers at Amazon are people highly trusted for interview outcome decisions and process control.
*** I have even had cases where (a) someone interviewed a candidate for a second time after a time gap and posed exactly the same problem, (b) a recruiter leaked out questions an interviewer asked frequently to the candidate.
"Cheating" on leetcode is a net positive for society. Unironically.
anyone I know who actually got a job through leetcode style in the last 2 years cheated. they would get their friends to mirror monitor and then type the answers in from chatgpt LOL
The coding challenge is supposed to be solved with AI. We can no longer afford not to use LLMs for engineering, as it's that much of a productivity boost when used right, so candidates should show how they use LLMs. They need to be able to explain the code of course, and answer questions about it, but for us it's a negative mark of a candidate proclaims that they don't use LLMs.
Do you state this upfront or is it some hidden requirement? Generally I'd expect an interview coding exercise to not be done with AI, but if it's a hidden requirement that the interviewer does not disclose, it is unfair to be penalized for not reading their minds.
I am personally of the view you should be able to use search engines, AI, anything you want, as the task should be representative of doing the task in person. The key focus has to be the programmer's knowledge and why they did what they did.
They also take this approach of "whatever tool works," but their coding test is "here's some symptoms of the SVG generator misbehaving, figure out what happened and fix it," which requires digging into the commit history, issues, actually looking at the SVG output, etc.
Once you've figured out how the system architecture works, and the most likely component to be causing the problem, you have to convert part of the code to use a newer, undocumented API exposed by a RPC server that speaks a serialization format that no LLM has ever seen before. Doing this is actually way faster and accurate using an AI, if you know how to centaur with it and make sure the output is tested to be correct.
This is a much more representative test of how someone's going to handle doing actual work knocking issues out.
It's not a hidden requirement per se to use LLM assistance, but the candidate should have a good answer ready why they didn't use an LLM to solve the challenge.
Also, what is a good answer for not using one? Will you provide access to one during the course of the interview? Or I am just expected to be paying for one?
We are providing an API key for LLM inference, as implementing the challenge requires this as well.
And I haven't heard a good answer yet for not using one, ideally the candidate knows how to mitigate the drawbacks of LLMs while benefiting from their utility regardless.
Again, what would be a good answer? Or are you just saying there isn’t one?
"I considered various approaches for solving this problem. Initially, I thought about using an LLM, as it's great for natural language processing and generating text-based solutions. However, for this particular challenge, I felt that a more algorithmic or structured approach was more appropriate, given the problem's nature (e.g., the need for performance optimization, a specific coding pattern, or better control over the output). While LLMs are powerful tools, they may not always provide the precision and control required for highly specific, performance-critical tasks, so I chose to solve the problem through a more traditional method. That said, if the problem had been more open-ended or involved unstructured data like text generation, I would definitely consider leveraging an LLM."
This answer reflects the candidate's ability to critically assess the problem and use the right tools for the job, showing maturity and sound judgment.
- GP, probably
"LLM-esque AI, especially in my industry, is under heavy scrutiny and I want to wait for the dust to settle before exploring options with such tools."
I was never asked as such, but I do have an answer to that.
Frankly, if an interviewer told me this, I would genuinely wonder why what they're building is such a simple toy product that an LLM can understand it well enough to be productive.
You ask a rote question and you'll get a rote answer while the interviewee is busy looking at a fixed point on the screen.
You then ask a pointed question about something they know or care about, and suddenly their face lights up, they're animated, and they are looking around.
It's a huge tell.
I just tried and it's: hard. It feels like being ask to keep one's breath like writing something.
I need to focus too much on keeping my eyes closed, I don't have enough bandwith left to thing about anything relevant.
This works especially well if I don't know the area they're strongest in, because then they get to explain it to me. If I don't understand it then it's a pretty clear signal that they either don't understand it well enough or are a poor communicator. Both are dealbreakers.
Otherwise, for me, the most important thing is gauging: Aptitude, Motivation and Trustworthiness. If you have these three attributes then I could not possibly give a shit that you don't know how kubernetes operators work, or if you can't invert a binary tree.
You'll learn when you need it; it's not like the knowledge is somehow esoteric or hidden.
Then they emailed me a month later after my "invitation" expired. It looked like it was written by a human: "Hey, we're really interested in your profile, here's a new invite link, please complete this automated pre-screen thingie".
So I swallowed my pride and went through with that humiliating exercise. Ended up spending two hours doing algorithmic leetcode problems. This was for a product security position. Maybe we could have talked about vulnerabilities that I have found instead.
I was too slow to solve them and received some canned response.
The price of people bulk applying with no thought is I have to bulk filter.
[1]: Including but not limited to: having to manually fill a web form because the system couldn't correctly parse a CV; take-home coding challenges; studying for LeetCode interviews; sending a perfectly worded, boot-licking cover letter.
I want to see how the candidate reasons about code. So I try to ask practical questions and treat them like pairing sessions.
- Given a broke piece of code, can you find the bug and get it working?
- Implement a basic password generator, similar to 1Password (with optional characters and symbols)
If you can reason about code without an LLM, then you’ll do even better with an LLM. At least, that’s my theory.
I never ask trick questions. I never pull from Leetcode. I hardly care about time complexity. Just show me you can reason about code. And if you make some mistakes, I won’t judge you.
I’m trying to be as fair as possible.
I do understand that LLMs are part of our lives now. So I’m trying to explore ways to integrate them into the interview. But I need more time to ponder.
- Spin up a Digital Ocean droplet
- Add the candidate’s SSH key
- Have them implement a basic API. It must be publicly accessible.
- Connect the API to a database. Add more features.
- Set up a basic deployment pipeline. Could be as simple as script that copies the code from your local machine to the server.
Anything would be fair game. The goal would be to see how the candidate converses with the LLM, how they handle unexpected changes, and how they make decisions.
Just a thought.
i would just look at stack overflow for your second point lol...
Is everyone writing a "Projects" section by rewording what they wrote in "Experience"?! For me, "Projects" should strictly be personal projects. If not, maybe that's what I'm missing.
They don't have to be public to the whole world, you can have links that are only in your resume.
But if they're on GitHub, they have to be public, since there aren't unlisted repositories.
An unlisted repository would be one that is public to those who know the URL.
I’ve also been asked for the first time in ages to come to the companies office to do interviews.
Last time I did this they told me it is but that they are at late stages of interviewing so I shouldn't bother applying for that one, but they got down my details and had other jobs that matched what I was looking for. Recruiters are sales people and you just reversed cold called them making their job easier. The majority of applications are AI bots and people who don't live in the country the job is listed in. By making a phone call you are up the top of the list of "most likely to be a legitimate applicant".
This even applies to businesses in some cases. You trying to walk in and talk to someone is a security threat compared to the times where you could do that and walk out with a job offer. US companies absolutely hate unsolicited calls from non-businesses.
I'm too busy doing actual paid work for companies for that.
that'd be a huge issue for most candidates (and basically all top candidates) because "exactly what you want to hire you for" is probably not open source code to begin with.
>If you are one of the 1 in 4000 applications who gets an interview then you're already 70% likely to get an offer and the interview is mostly a formality.
That has not been my experience at all in 2023/2024.
The candidate’s first response? “Memory updated”. That led to some laughs internally and then a clear rejection email.
This is because my brain couldn't fathom what is likely the reality here -- that someone was just pumping your email thru AI and pumping the response back unedited and unsanitized, and so the first thing you got back was just the first "part" of the AI response.
...Christ.
Unfortunately what we end up doing is have to make some assumptions. If something seems remotely fishy, like that “Memory updated” or typeface change (ChatGPT doesn’t follow your text formatting when pasting into your email compose window), it raises a lot of eyebrows and very quickly leads to a rejection. There’s other cases where your written English is flawless but your phone interview indicates you don’t understand the English language compared to when we correspond over email/Indeed/etc.
Mind you, this is all before we even get to the technical knowledge part of any interview.
On a related hire, I am also in the unfortunate position where we may have to let a new CS grad go because it seemed like every code change and task we gave him was fully copy/pasted through ChatGPT. When presented with a simple code performance and optimization bug, he was completely lost on general debugging practices which led our team to question his previous work while onboarding. Using AI isn’t against company policy (see: small team with limited resources), but personally I see over reliance on ChatGPT as much, much worse than blindly following Stack Overflow.
Long live plain text email.
() ascii ribbon campaign - against HTML e-mail
/\ www.asciiribbon.org - against proprietary attachments1. The manual was mostly a bunch of phrases that were grammatically correct, but didn't really convey much meaning
2. The second half of the manual talked about a different machine than the first half
3. It was full of exceptionally bad mistranslations, and to this day "trained signaturee of the employee" is our inside joke
Imagine asking ChatGPT to write a manual except ChatGPT has down syndrome and a heart attack so it gives you five pages of complete bullshit. That was real manual that got shipped a 100 000€ or so machine. And nobody bothered to proofread it even once.
But yes, I read it the same way. Pretty funny way to respond to a recruiter after they say "no AI please".
To then select a good developer I'd test communication skills. Have them communicate what the pros/cons of several presented solutions are. And have them critique their own solution. To ensure they don't have canned answers, I might just swap the problem/solutions for the in-person bit. The problem they actually solved and how they did it isn't very important. It's whether they could read and understand the problem, formulate multiple solutions, describe why one would be chosen over another. Being presented with a novel problem and being asked on the spot to analyze it is a good exercise for developers (Assuming software development is the job we're discussing here).
Just take the time to talk to people. The job is about reading and writing human language more than computer programming. Especially with the arrival of AI when every junior developer is now micro managing even more junior AI colleagues.
tightened market is one thing, the absolute insanity of the recruitment process in last couple years with now AI thrown into the mix is really something to behold, test these waters at your own peril
someone who actually wants to hire will want you and wants to do whatever they can to get a good candidate.
For some random SMB in shipping or something to be bashing people over the head with 10 step-leetcode-full-panel-10-hour-systems-design interviews they just don't get it. For starters they probably don't even have the talent to properly evaluate the prospect. So who are they helping?
- Much less resposnes
- For your sanity you want to make sure there aren't obvious signs of something being a ghost job, an H1B hire, an internal hire, or a farce because "we're always growing" (lying).
- expect longer processes. Hasn't happened to me but 7+ stage interview processes is not uncommon these days, even outside of tech
- accept that some processes will be frozen under your nose. especially because of longer processes crossing quarters
- expect less respect in the process. They feel like they do not want you. You are expendable
- don't bother negotiating in this market. You get a number and take it unless you already have a job. Even then they may simply pass you for someone more desperate. BTW wages are being suppressed; you're probably not getting pre 2023 salaries right now.
Yeah... if you're not being abused at work, I'd just weather it out.
"logically, necessarily" as Marc Andreessen says? https://futurism.com/the-byte/ai-investor-goal-crash-human-w...
Now if they could also crash factors like material, energy and profit... consumers could pay pennies using their centinaire incomes.
The biggest thing I've noticed is take home challenges have lost all value. Since GPT can plausibly solve almost anything you throw at it, and it doesn't give you any indication of how the candidate thinks.
And to be fair, I want a candidate that uses GPT / Cursor / whatever tools get the job done. But reading the same AI solution to a coding challenge doesn't tell me anything about how they think or approach problems.
Sometimes you have to. In my previous analyst stint a writing sample was pretty non-negotiable unless they could oint to publicly-published material--which was much preferred. ChatGPT isn't much use there except to save some time. It's very formulaic and wouldn't pass though, honestly, some people are worse on their own.
Not being able to code is by far the easiest failure mode to deal with and I can deal with it more quickly by looking at resumes and firing people for outright lying about their abilities very quickly.
What is much harder to detect is the person who gives up at the first sign of trouble. Or someone who likes to over abstract everything. Or someone who likes to spend all day nitpicking PRs.
The absolute most damaging employee is the technical tornado midlevel who has prolific output and is good at figuring out how to get PRs through the code review process.
I can only bring to begin imagine the kind of damage a person like that could do with an LLM, an inattentive manager, and a buddy willing to rubber stamp PRs.
I assume your from the US? The issue with firing people quickly after hiring them is:
* You need to pay whatever headhunter or other personnel that is involved. Some offer a cashback or reduced rate to find a new person, but your still out of the pocket. Do it a few too many times and the headhunter dump you.
* You just wasted a ton of time of other employees, setting up this persons accounts, social benefits, contract negotiation, training time, and more. Again, this is not free.
* Now you are back to looking for other people. You know those people that you rejected, well, the change is reduced that they will come back to your company, maybe they got hired somewhere else, maybe your rejection left a bad impression as you hired that lying person over a actually well qualified person that just did not ace the interview (happens to so many people, often because of nerves).
The trick is, to have a that person (and others in the interview process) work a few days in your company, but this is not covered by a official work contract but a interim contract (of course, you pay those people, some companies are real bastard and do not pay).
That can weed out the people that are good at lying about their skills. But to be honest, if you can not spot them during the interview stage, ...
We had people coming in for Senior positions, and their CVs looked good, they knew how to talk the talk. But when asked for some incredible basic SQL question, they fumbled so much, because they used ORMs their entire life. But loved to place SQL skills on their CV. Seen multiple of those... And so often they then switch to going after the interviewer, thinking that by attacking the interviewer before the boss, that will help them. Yea, "Senior material"...
With AI its going to be the same. People get comfortable using AI, but when you ask code questions without that AI to back them up, that skill gap will exposé themselves very fast during the actual interviews.
I love using AI because it can remove so much boilerplate / repeat / great for finding doc info but you need to know your language. If you grew up with only AI, and do not grow beyond AI to solve every issue, your nothing more but a copy/past dev (AI is just a extension of that dev type).
> Or someone who likes to over abstract everything.
No no, the most annoying ones are those that come in and then want to implement everything into a style they are familiar with! Code rewrites to fit them, framework changes because they are familiar with them... Nothing more fun as hiring somebody to write Angular for 13k/month (talking almost 10 years ago), and then this person pushing Vue, to even rewriting part of the codebase into Vue. Like, we hired you with a insane salary because we are desperate for a Angular developer, if we wanted to rewrite everything to Vue, we will have hired somebody cheaper.
And I had made it clear that they should use their own words.
It happened twice, that the candidate on the other side was clearly typing during the interview, taking a pause for a second or two, and then reading from the screen. That's very obvious as of today, but I can see how it will become a problem one day with the future AI development in terms of the speed of responses, and better voice recognition techniques (so, no typing needed).
The interview is a chance to see how a candidate performs in a work like environment. Let them use the tools they will use on the job and see how well they can perform.
Even for verbal interviews, if they are using ChatGPT on the side and can manage the conversation satisfactorily then more power to them.
There's nothing wrong with a candidate going "Normally I'd prompt ChatGPT and get a skeleton project going" or saying "Look, I don't run around with the entire standard library in my head. I look that stuff up and sometimes that's with an LLM". The problem is when they can't go through the steps of solving a program, without the AI. I don't care about the details, or if you ask Copilot to do the API query code, because you don't want to write the error handling, that actually fairly reasonable, but if you can't prompt it to add the logic for a HTTP 403 then what's the point? In that case I'd rather hire someone who takes longer, but who knows that the 403 should probably redirect an unauthenticated user to the login page.
And generally, the more junior people are just completely lost without it. They've become so dependent on it, they can't even google anything anymore. Their search queries are very weirdly conversational questions and the idea of reading the docs for whatever language or library they're using is totally foreign to them.
I think it's really hampering the growth of junior devs - their reasoning and thought processes are just totally tuned to this conversational form of copy and paste programming, and the code is really bad. I think the bottom half of programmers may just LLM themselves out of any sort of job because they lose the ability to think and problem solve... Kinda sad imo.
Ask even the shallowest question and they are lost and just start regurgitating what feels like very bad prompt based responses.
At that point it's just about closing down the interview without being unprofessional.
This is borne out by results downstream with clients. No client we've sent more than a couple of people has ever had concerns about quality, so we're fairly confident that we are in fact detecting the cheating that is happening with reasonable consistency.
I actually just looked at our data a few days ago to see how candidates who listed LLMs or related terms on their resume did on our interview. On average, they did much worse (about half the pass rate, and double the hard-fail rate). I suspect this is a general "corporate BS factor" and not anything about LLMs specifically, but it's certainly relevant.
- adapted Java's Regex engine to work on streams of characters
- wrote a basic parametric geometry engine
- wrote a debugger for an async framework
- did innovative work with respect to whole-codebase transformation using macros
Among other things.
As for ChatGPT in the context of an interview, I'd only use it if I were asked to do changes on a codebase I don't know in limited time.
It's frustrating though when you've done a lot of work, as you've listed. I think in a good interview maybe going over that code and getting the chance to explain things you did, why you did, or issues you had, could also go a long way.
Interviewing is tough, more so at scale.
My portfolio site is just one of the sites on my personal website that also hosts many of my projects. It wasn't much work to setup, and it provides organization and sharing capability so there's motivation to make it anyways.
I'd argue that that is indeed the majority. Maybe they have some work from school days, but very few devs are making portfolio pieces years into the industry once they can flesh out their "Employment" sectiion.
IME you only need a portfolio as a non-junior if you are making a lateral move.
So most people with young kids then
I have 20 years of experience in software development, I have hundreds of LinkedIn contacts you can check on, a dozen recommendations on LinkedIn and a dozen projects on Github, not to mention a blog and let's say a other indicators (e.g. Stackoverflow creds).
Now what exactly are people checking on? My picture is on LinkedIn and Github. Clearly I can code and have done dozens of projects. What is the point of asking me - "Do you know Kafka?", "Have you used AWS S3 and how?", "How would you build / scale a Node.js project" - these are the real questions I was asked. Yes, had you cared to look at my Github / blogs you would have seen I have done this multiple times, what are we verifying now?
I tried my best but at one point stopped caring about interviews.
It was a really great format and I think one that creates better separation between good candidates and great candidates because it is more open-ended and collaborative. One of my favorite technical interviews I've engaged in.
I think that now, with AI coding assistants becoming even more integral than those early days 18 months ago, this approach is more relevant than ever since it gives insight into how efficiently a candidate can quickly review AI generated code for correctness, defects, performance issues, and other gaps in quality.
I liked this format so much that I ended up creating a small open-source tool for it to make it easier to manage this process: https://coderev.app (https://github.com/CharlieDigital/coderev)
(The interviewer had to create a private GH repo and the repo itself didn't support commenting inline; I took my notes in a text file and reviewed it interactively with the interviewer).
If the AI is so good at it, why are we still hiring human to do the job? It just shows how the interview process is not measuring the right thing to start with.
And for questions, I don't mean "it's better a list or a set?", but something like: "you have an application like this, how can you improve it to perform X?"
We're trying to be fair in hiring, every resume is manually read, interviews a conversations, there are no leet code, no tricks, no AI screening, but they are done remote and so far two out of three candidate clearly cheat, during a live interview, to cover up an impressive lack of general knowledge.
<instructions>
This project is designed to evaluate your ability to:
- Deconstruct complex problems into actionable steps.
- Quickly explore and adopt new frameworks.
- Implement a small but impactful proof of concept (PoC).
- Demonstrate coding craftsmanship through clean, well-architected code.
Feel free to use any tools, libraries, frameworks, or LLMs during this exercise. Also, you’re welcome to reach out to us at any time with questions or for clarification.
</instructions>
I used LLM-as-a-junior-dev to generate 95+% of the code and documentation. I'm just an average programmer, but tried to set a bar that if I was on the other side of the table, I'd hire anyone who demonstrated the quality of output submitted.
- The 5-7 hour estimate was exceeded (however, I was the first one through this exercise).
- IMHO the quality of the submission could NOT have been met in lesser time.
- They had 3 tasks/projects:
- a data science project,
- a CLI based project and
- a web app
- They wanted each to be done in a different language.
- I submitted my solution <38 hours of receipt of the assignment.
- In any other world, the intensity of this exercise would cause a panic-attack/burn-out.
- I slept well (2 nights of sleep), took care of family responsibilities and felt good enough to attack the next work-day.
This was by far the most fun and one to replicate every chance I get.
[EDITS]: Formatting and typos.
This was the final technical screen so definitely something worth doing in my case.
The reason I posted a reply was there is a lot of negativity around AI in the hiring process. This was an excellent example of using AI to the benefit of all parties.
Instead of nit-picking on stylistic things from a smaller code-sample, one can nit-pick on the implemented complexity. I think it is a higher quality signal.
Honestly, if I could trust that companies won't try to evaluate my conversation through 20 different ridiculous filters, I would probably argue that my buddy is out of line.. As it stands, however, he is merely leveling out the playing field. But, just life with WFH, management class does not like that imposition one bit.
The job you are therefore hiring for is now trivial. If it weren't, no amount of AI could pass your interview process.
An interview has to be hard enough to filter those that are unqualified but also easy enough that the right person can pass with some minor preparation. If an interviewer asked me for the equivalent of production-ready code to add support for custom hardware in the Linux kernel I'd either reply with my freelance hourly rate or I'd end the interview.
Yeah, ideally. Meanwhile the market evolved to asking Leetcode hards on the spot.
As usual, companies take a good concept and milk it to the point where you need to cheat or make studying leetcode your full time job to ace it. I'm not to sympathetic.
(PS yes, I've been asked to do what was essentially spec work a few times. Woudl have been smarter to reply with an invoice, but I was too tired and just said no).
this is implying interview process actually showcases someone would be good at a job when it isnt
when has leetcode ever translated to web dev... fuckin never lol.
Why do you let your interview process for web devs use leetcode?
If I was incompetent, I could've shoved the problem into o1 on ChatGPT and probably solved the problems, but I wouldn't have been able to provide insight into why I made the design choices I made and how I optimized my solutions, and that would've ultimately gotten me thrown out of the candidate pool.
I had a few coding challenges, all were preinterview and submitted online or shared in a private repo. One company had an online quiz that was actually really interesting to take, the questions were all multiple choice but done really well to tease out someone's experience in a few key areas.
For what its worth I don't use LLMs and the interview loop went about as I'd expect in a tough job market.
What I usually do is case study that I also do not know at the start of the interview. The case study does not imagine spherical cows and are not usually leetcode style. Its a case of role playing a problem. We brainstorm it together and I determine if I can work with them.
I would say that if it wasn't a pattern. So let's not pretend they're not cheaters. Call them out.
Nontech roles are just a sea of prompt-generated answers, they don't tie into the applicants' experience and are usually about 200 words of waffle. If you're applying for something, type out a few sentences then give it to a prompt to refine. Don't just paste the question in.
Tech roles we focus on the combination of soft skills and technical skills and so we've gone back to a 'pair' programming exercise and a whiteboard architecture exercise. In reality, it's just a spectator sport that we nudge them forward if needed.
In Java looking for them to cover the basics of debugging a problem, writing a test, and mocking out some services. I would say 50% are unable to or unwilling to write a test to prove the error or don't know how to mock out a service.
Whiteboard exercise looking for them to explain a system they've worked on in the past. We question some of their decisions and see how they handle themselves. Not many get defensive (though some do).
What I liked about that process is that it relied less on their ability to suss out a solution to some problem they'll never have to solve on the job and focused more on average activities. Sometimes I'd get a candidate who would go "wait, is this a trap?" and start asking a lot of questions - good! Now I got to see them refine requirements.
Having them review a PR is a good exercise too, you can see how they are at feedback.
I've seen a few things change.
The interviews themselves are for the most part unchanged. Occasionally I see someone who seems to be using AI during the interview. It's sort of obvious. They fist give a vague answer. It's as if they repeat the question. Then they answer while not maintaining eye contact. When asked a followup or a "why" question, they fall apart. They do poorly.
I find more people pass the written screen nowadays and then bomb the interview. I'm guessing they use AI. C'est la vie.
It seems that a lot more people are looking for work now than 3+ years ago. It seems that for roles that require top-notch coding ability, there are relatively few people on the market. Im guessing that great people are staying at their jobs longer now. Makes sense.
After a candidate gives an answer, probe. Ask why they think that. Ask what alternatives they might consider. Watch their body language and how quickly they answer.
For context, I hire very experienced developers in Poland for very demanding remote roles in the US.
The story is completely different for airgapped dark room jobs, but if you know you know.
That would be a bit awkward with my 32:9 primary monitor.
Not like I'd do anything. I use one screen to work, one for video, and one for work visuals. Seeing one screen would show if I had any hidden windows anyway.
When I was part of interviews on the other side for my former employer, I encountered multiple candidates who appeared to be using AI assistance without notifying the interviewers ahead of time or at all.
Only the trivial problems. We don't use AI during interviews but many try and it's always obvious. Delay after any question problem; textbook perfect initial answer; absolutely nothing when asked to go deeper on a specific dimension.
It's nice because interviews that are scheduled for an hour are only lasting ~20 minutes in these situations, and we can cut them short.
Everyone hired after that is more suspect, and if they screw up too much or don’t perform well we just fire them quickly during the probation period, whereas previously it was rare for people to get fired during the probation period.
In the recent couple of years I have seen a lot more people ace the test and not do very well during the actual interview. Take-home exams feel like they would always be ineffective now.
I think it ultimately comes back to impact (like always) which has remained largely unchanged.
tbh it's very hard, in general, to assess someone skills in ~1 hour, with the diverse set of problems we face everyday and people/companies often focus too much on "previous relevant experience" (do you know this?) instead of thought process and depth of understanding (how much have you understood what you have done) which, to me, gives more insights on someone personality and attitude and ability to learn new things and becomes harder to cheat on (you can memorize "cracking the code interview" and be a hero on leetcode and still have no idea about how to write good software, take decisions or work in a team :)
If they can talk through the technology and code fluently, honestly, I don't care how they do the work. Honestly I feel like the ability to communicate is a far more important skill than the precise technology.
This is of course presumes you have a clue about the technology you're hiring for.
I feel that smaller things like syntax etc make perfect sense. But for larger things that involve a slightly higher complexity it becomes a bit grey. I likenit personally to writing. When I write things down as I'm trying to work things out or even trying to learn something I find I retain the data so much better and have a better picture in my mind of what's going on. That might just be personal preference for learning but if I copy straight from claud I know 100% I'm not going to remember anything about it the next day.
You can see things in the emails like:
"I provided a concise, polite response from a candidate to a job rejection, expressing gratitude, a desire for feedback, and interest in future opportunities."
The main takeaway is that if you make design your interview questions to match the actual skill you're looking for, AI won't be an issue because it doesn't have those skills yet. In short: ask questions that are straightforward in surface but deep beneath with trade-offs that must be weighted by asking questions to in interviewer.
[1]: https://softskills.audio/2025/02/03/episode-446-wading-throu...
It’s better to know when to use a Linked list than how to make one (because I’d just use the one in the library).
So the candidate can prompt well good. But how much of the knowledge can they apply to a problem or are they just masters of hacker rank (sic).
But more often than not most interviewers are lazy and just use canned hacker rank style questions or if it’s not laziness it’s being too overworked to craft a really good interview.
Even remotely, normally a coding interview isn't a candidate typing things for 45 min on a screen. There are interactions, follow-up questions, discussions about trade-offs and so on... I suppose it's possible for a good candidate to cheat and get some extra points, but the interview isn't broken yet.
You could also let the candidate use AI, and still gather all the relevant signals.
I had this one interview where I had to remove nodes from a linkedlist. Pretty trivial, right? It is, but as I was writing out the solution, probably for the 200th time, I thought "Ive never used a linkedlist, ever"
And I thought, I just cant do it anymore. I cant reverse one single more binary tree. I cant listen to some CTO tech bro in his early 30s explain to me why I need to be in an office he'll never set foot in for the 'company culture'.
Ive had so many interviews just like this. Ive done 7 hour on sites where you break down algorithms, system design, app development. Its brutal out there, no one seems to have any idea what their looking for so they just put you through the ringer.
So now im not interviewing, Im happy and content in my bank programming job. Im probably not a very good programmer, thats okay, Im a wonderful partner, family member, musician. Ill get by, hopefully, for another decade until everything collapses.
The kind of core conclusion is that companies need to do in person interviews now. There's no other way to prevent cheating.
Even if you're using ChatGPT heavily it's your job to ensure it's right. And you need to know what to ask it. So you still need all the same skills and conceptual understanding as before.
I mean, I didn't observe interviews change after powerful IDE's replaced basic text editors.
I've done a lot of interviews over Zoom, and whenever someone cheats, by passing someone else's work off as their own (the weirdest thing I've ever encountered was someone having a friend on trying to feed them answers, which he admitted to later) it is so painfully obvious if you grill them a bit and throw a few curveballs.
It's the people who have no idea what the hell they're doing and who try to pass of a solution they can't explain and by definition those people you will catch if you know how to interview.
My company at the same had been using the same coding exercise for years, and many candidates inevitably published their code to github, so copilot was well trained on the exercise.
I had a candidate that used copilot, still flubbed the interview, while ignoring perfectly good suggestions from the LLM.
Screen share to avoid cheating via AI, same as we were doing before AI when people could get friends or Google to help them cheat.
In my unfortunate experience, candidates who covertly rely on LLMs also tend to have embellished resumes, so I try to root them out before asking technical questions.
Last time I interviewed I spent about half of it standing at a whiteboard.
Which is the case for >90% of the companies I interviewed with (big and small)
That seemed to thwart AI use, at least thwart one-shotting, and require understanding and experience with working in an organization
I liked that
It is really important to watch people code.
Anyone can fake an abstract overview.
Two main things I've seen:
1. Recommendations are a heck of a lot more important
2. Internal applicants are suddenly at a big advantage
The gap between interview and actual on job duties is very wide at many — delusional - companies.
Its a sad world out there these days.
As a result, several of my friends who assist in hiring at their companies have already returned to "on-site" interviews. The funny thing about this is that these are 100% remote jobs - but the interviews are being conducted at shared workspaces. This is what happens when the level of trust goes down to zero due to bad actors.
As someone else noted, this used to be utterly standard. And frankly I’d probably just pass on someone who balked. Plenty of fish in the sea.
How good one is in understanding a problem (the scope and source), and how good are they at designing a solution that is simple, yet will tackle other "upcoming" problems (configurable vs hard coded, etc).
This is what one should care about.
* StackOverflow early days raised the same question, and so were Google. trust me. I am old enough to remember this.
Online, you don't know if the person you're interviewing is an AI or not.
They could have an AI resume, AI persona, AI profile, and then someone shows up who looks like that and it could be a deepfake, then they do the coding challenges, and then you hire the person remotely, and they continue to do great work with the AI, but actually this person doesn't exist, they just created 100 such people and they're just AI agents.
It sucks if you want to hire humans. And it sucks for the humans. But the work gets done for the same price and more reliably. So dunno
How LLMs will evaluate a skill they are making obsolete is a question I am not sure I understand.
(You hardly needed a throwaway for this comment, though)
I dont know but Ill always remember the funniest thing I noticed once during my career in England..
A company called tripadvisor based in a very very small town where I was at the time a senior dev, working on my own things, had never reached out. yet I saw their ads and finally an actual article in a newspaper, basically where they were bragging about their latest hire. Let call him Pablo, who had apparently aced every single technical interview question, and so they had hired him, after interviewing tens of people. They were so happy with their hire, the article had been based on him and the "curiosity" that they were the first company to have hired him after he had failed I think it was something like 50 interviews.
Obviously they couldn't believe how lucky they were to have finally found someone who could have completed all the technical tests perfectly.
Now I have nothing against Pablo, and rooted for when when I read the article. But I found it hilarious, and still do almost a decade later, that this top tier company, based in a university town with perhaps the most famous university, had not realised they had simply over fitted for someone who could answer their leetcode selection perfectly. Not only not realised this but then commissioned the article.
Eventually they reached out for an interview with me where the recruiter promised there was no test for me, then I was "surprised by one" in a room with a manager who hadnt had time to read my expection ( which is fine ), but when he walked in and saw I hadn't done it I was "walked out". The whole interview having taken less than about 10 minutes, when I was the most qualified senior developer for hundreds of miles who was available at the time. No Im honestly not tryign to brag Im just saying the town was so small there just couldnt have been more during that short time period I was available.
I know this reads bitter,( my life is great now ) , I just remember it because, my point was at the time I was at my peak, and would have accepted the job if offered, but I was walked out within 10 mins.
Honestly just sharing this insight, the moral of the story I think for me is companies never were great hiring and well if anything the advent of LLMs might actually improve as LLMs start to assess people based on their profiles and work? One can hope, I dont, I want an edge in this market with my company.
Just asked to add a small feature or make a small change after the present the initial code (without AI help). That makes it really easy to see who doesn’t understand the code they are presenting.
I don’t care how much AI you use if you understand the code that it writes and can build on top of it.
In-person dialogue with multiple team members and sufficiently complex take-home assignment remains a pretty good method. LLM are excellent API docs and should not be avoided.
Who is everyone?