What the heck does a good experienced dev interview even look like in 2026?
Posted by Undercoverwd@reddit | ExperiencedDevs | View on Reddit | 163 comments
I'm in charge of designing our interview process after a year of hiring freezes and I have no idea where to even start. I've interviewed hundreds of mid to senior level devs but I feel like none of the things we used as proxy's for experience are relevant anymore.
We're vasillating between making the interview so freaking hard not even the bots would do well and going crazy trying to avoid "cheating." It seems like we can't even have a normal "tell me about all your projects" conversation without it turning into a damn turing test.
I tried to ask my coworkers for their ideas and they just ran my question though their favorite clanker and got the stupidest ideas I've ever heard.
Has anyone figured this out?
LoveSpiritual@reddit
Why does “tell me about your projects” no longer work? Are people reading off LLM answers to that?
Cute_Activity7527@reddit
I really like this question as I can dive extremely deeply in case of technical topics. Can grill someone and instantly see if they did important stuff or only participated as part of the team.
Next question I ask - is about „How did you behave when you had conflict in the team or with management?” - this shows you how well someone can work with other ppl and how well they can handle politics.
Technical skill is only as good as politics you can wrap it with.
If this is a product role its also good to grill someone about customers handling. To spot good product engineers vs shitty ones.
AintNoNeedForYa@reddit
Do them in person
netanator@reddit
This would solve a lot of problems.
ninetofivedev@reddit
This sucks for everyone involved. So it's probably going to become the standard.
Maleficent_Tank2199@reddit
It was so much better in person, you would weed out shitty companies so early as a candidate.
Examples I personally experienced where: - Waiting 30 min in the Lobby because reception repeatedly told you to, interviewer getting cranky because you are "late". - Prospective new colleagues asking "how the market is, for personal reasons" - Someone getting an electric shock during the visit because they were updating a HW-in-the-loop setup but ignored all safety rules. - Hearing a PM I the Hallway say about the developers "The peasants are uppity again"
Maktube@reddit
I'll add one to your list: FAANG company that didn't know how doors work. Showed up for the first interview of the day, interviewer could not figure out how to open the door. So we called the front desk person, he couldn't figure it out either. We snagged a passing security guard, who also couldn't figure it out, but he knew where the building maintenance people were, so he went and fetched one of them and HE couldn't figure it out either. And I want to be clear here, I don't mean that they couldn't figure out which key to use, I mean that no one could figure out how you were intended to open the door. It was just a door handle that would not turn. There was no keyhole. There was no obvious place for a keycard or a badge, and none of the key cards or badges anyone had seem to do anything when waved at various parts of the door. Absolutely wild experience.
After about 10 minutes I suggested maybe we could just do the interview outside on the picnic table? No no, we had to do it in the interview room on account of The Equipment. We did finally get in, 40 minutes into the 1 hour interview time (it took a facilities maintenance guy from across the campus, who knew about the new door system they were apparently installing everywhere without telling anyone???). Shockingly, we did not really finish anything useful in that interview. The rest of the interviews went very well, though, I thought, but I heard back from the recruiter a couple weeks later who said that hiring committee was really on the fence, but they eventually decided to say no, not because they got bad signal, but because they didn't get enough signal from that first interview. And I was very tempted to say, "But... we did get enough signal from that interview, didn't we? Just... maybe not about me."
Izkata@reddit
Just going to leave us hanging about the door? What was the trick to opening it, lift/push/pull instead of turn?
Maktube@reddit
I was afraid someone would ask that XD
It was... almost a decade ago, I think, and I don't remember the specifics beyond that it turned out that you were SUPPOSED to be able to just walk up to the door and it would be unlocked. The idea was that no one would ever have to think about keys/locks again, if you've booked the meeting room then you will never experience it as anything other than an unlocked door. Only they'd installed the system before giving everyone the requisite hardware, which I THINK was a keycardish thing with some kind of RFID... dealio.
ninetofivedev@reddit
Yeah, nothing like taking PTO to deal with all of that. Sounds better
Maleficent_Tank2199@reddit
Yeah I understand that this is lot of an investment in time but you get information you are not getting otherwise. I have been burned now twice by remote only applications and then arriving at a clusterfuck of an organization that would have been apparent from lingering 10 minutes at the watercolor during an onsite-interview.
DandyPandy@reddit
It wasn’t that long ago that that was the norm.
slashedback@reddit
People are reading off LLM answers to everything. People running behavioral rounds of interviews have broadly reported candidates using LLMs during “tell me about a time…” style questions.
I have no idea what would make someone want to ask an LLM to tell them a story to then tell a human in an interview, but here we are in 2026.
Four_Dim_Samosa@reddit
because we have a nature of trying to get the same outcome but with least effort possible. I guess the system incentivizes it
ObjectiveConsistent2@reddit
I watched a youtube video about how someone used a job board specifically for women and non binary candidates to land the interview, and then chatGPT'd up 2 or 3 ficticious behavioral round stories.
It was pretty demoralizing
Four_Dim_Samosa@reddit
you can Prompt the LLM to make up a project you supposedly did
fued@reddit
because people tell you about the projects, then conveniently dont mention that they only worked on 5% of that project and barely interacted with the rest
LoveSpiritual@reddit
Thus the follow up questions? You ask them to go in detail, to explain how they balanced priorities (complexity vs speed, scale vs simplicity, etc). This has always been important. My experience is that it’s pretty hard to fake it.
Four_Dim_Samosa@reddit
also you can probe on impact metrics. why was certain metric chosen? how did you calculate the impact?
fued@reddit
I dunno, i have found people can fake that part super easily. Its better to just ask them how they would approach certain situations and how they would resolve it I have found
Groove-Theory@reddit
How is that faking? If they know everything about that system design, then they know it. Who cares if they did 5% of it.
I've never built a computer processor from scratch but I got my degree in computer engineering and I know exactly how to design a basic replica of a MIPS processor. I didnt invent it, I didnt build it, but I know it.
fued@reddit
The example I have is the guy talked up all the ci/cd pipelines on the project and all the tech, then when asked to do a simple pipeline had no idea and took weeks to do a 4h story
Groove-Theory@reddit
So.... you're telling me.....
You asked this person an incredible amount of ci/cd pipeline knowledge, and dove deep into it, asked follow up questions about it, asked him all about each detail of his project (whether he uses canary deployments or handling microservice pipelines or whatever), and you all heard his answers from his mouth as him knowing not only the project like the back of his hand, but saying the correct reasoning about CI/CD pipelines derived from it and the scenarios that come from it....
....
... and then he just didn't know?
I dont buy that. Either the person was not vetted as well as you let on or the project was not as simple as you say.
tronj@reddit
I have seen this happen once or twice with candidates. It’s pretty crazy. Guys had all the knowledge but for some reason just couldn’t deliver work. They didn’t last long.
poemmys@reddit
I was this guy. It’s a combination of severe untreated ADHD and some other shit sprinkled in. We get good at programming very quickly because we were obsessed with it at first, and we genuinely have the knowledge, but once that obsession/passion leaves… good luck getting any work done.
Complete 180 once I got on adderall. You gotta stay on top of guys like this and ask for itemized breakdowns of what they did each day, hour-by-hour, and call out anything fishy immediately.
They’re usually brilliant engineers if they can get properly medicated and “trained” to provide daily breakdowns so-to-speak to make sure they’re actually, you know, working. They just don’t make good code monkeys, you need the neurotypical type-A’s for that.
teknoise@reddit
Yep, as an adhd guy myself, it’s this, but sometimes also the opposite. Interviews suck, the prep sucks, the kissing babies and shaking hands networking sucks, we just want to code dagnabbit! So we suck ass during the interview process cuz the dopamine reward isn’t there, and miss out on the good jobs because we’re good at our jobs but not at the completely separate skill of interviewing and getting said jobs.
gfivksiausuwjtjtnv@reddit
I usually find interviews easy, talking about work easy, solving problems… its the actual doing work bit that gets me ;_;
Void-kun@reddit
Just for others keep in mind not all people with ADHD are the same.
If you did anything like this with me the opposite happens, this causes me to stop working and get frustrated. I have quit 2 jobs for doing exactly this, it just kept causing me to burn out.
CactusOnFire@reddit
As another ADHD Engineer, I came to realize that I work best when I am given a defined task and am otherwise given the autonomy to solve it without being over-managed. I've only delivered 1 task late in the last 5 years and it was the direct result of working for a micro-manager (I quit directly after that project ended).
webbed_feets@reddit
Also, people are slow AF when they first start. It happens at every job in every field.
ResoluteBird@reddit
I have skepticism about anyone with a top 1% poster badge on reddit. When do they have the time to post so much, and why do they feel the need to be posting so much
F0tNMC@reddit
This is the way. And in addition to design and implementation questions, asking concrete questions like, “how would you find the X for Y?” and “How would you change that value? What tools would you use?” can give you a lot of insight into how much hands on experience they have with the technologies.
Void-kun@reddit
"what type of problems did you encounter during development of this project?"
"Oh really, and how did you resolve that? Was it the best solution? What did you learn from it?"
bstaruk@reddit
Sounds like you are not a very skilled interviewer.
I find that people cannot fake that part at all.
jrodbtllr138@reddit
How you would approach questions in my experience have rarely been helpful in determining skill or ability to do the job.
It does give some insight into what they prioritize
AchillesDev@reddit
Yeah people are getting upset that they actually have to be good at interviewing.
curious_corn@reddit
And how is that wrong or diminishing? Aren’t we supposed to be a team working together on ADR’s and working as a “whole is more than a sum of its parts”? Having someone who’s lived through something play out, the decision process, the costs, the trade offs… it’s gold really.
Unless you’re hiring for a new Apple CEO position? Or setting yourself up for failure, hiring the next socially impaired centralizer who’s going to poison whatever collective output your team has, reducing it to whatever they can manage. Until they burn out
7HawksAnd@reddit
Or the inverse and you try not to go in the weeds to make sure you don’t go into an area they don’t care about, but then there’s no follow up Qs and they reject you for not having the right experience
codescapes@reddit
Right and that should be immediately apparent if you probe a little. Unless what, they have proprietary work information on their LLM about a project they don't understand and they're socially skilled enough to learn and speak about it ad hoc?
I just don't believe that. Even reading from a real teleprompter like a news presenter is very hard, let alone the cognitive overload of trying to technically understand what you're saying.
If someone is a poor communicator or seems stilted when asked what should be basic questions - multi second delays with weird "reading from a script" vibes - you thank them for their time and move on. In 2026 communication has never been more important, I will take a genuine person who is honest about their knowledge in a heartbeat compared to a suspicious 'expert'.
If someone is using an AI interview tool to get bullet pointed answers that they flesh out competently then fuck it, they deserve the job if they can pull it off because that still requires underlying talent.
Ok_Chemistry_6387@reddit
If you cant tell someone was not involved by talking to them. Maybe get off the interviews.
jellatin@reddit
Also so much more engaging. I love watching the eyes of candidates that do read off LLM prompts as their body language shift and their brain fully engages.
Often their eyes will start looking up as they think about questions like, “if I gave you two weeks at your most recent role to prioritize anything you wanted, what would it be / why / what’s been stopping you?” Or “what’s a technical decision you wish you could do-over / what choice would you make instead / why?”
Oftentimes they’ll actually start to smile and fall into the conversion rather than worry about giving the perfect answer.
ResoluteBird@reddit
Sometimes i look up to think, and worry that even though i have no other monitor or computer nearby they assume i am doing that.
do candidates need to be surrounded by mirrors? Ridiculous
Izkata@reddit
They're saying the opposite, they can tell the person has stopped looking at the LLM when they look up to think.
ResoluteBird@reddit
Ah, thanks, i misread the post!
SolidDeveloper@reddit
People looking up while thinking about something is very common in real life. I don’t see why you’re making this to be a problem related to LLMs all of a sudden.
jellatin@reddit
That’s my point. When you ask engaging questions and not rote memorization questions the candidates will stop reading the LLM output and fall back to their actual thinking habits, like looking up.
It’s just more noticeable nowadays when people switch from intently staring at bottom of their screen while answering to looking up, I take it as a good thing.
Wide-Pop6050@reddit
My bigger problem was with AI generated resumes that did not reflect the candidate at all but were very tailored to the position, or these weird scams people were running
Izacus@reddit
Because it never ever worked and there's studies about that.
whyamisogloopy@reddit
Link to the studies, please?
Undercoverwd@reddit (OP)
We've heard they are - is that fear over blown you think?
LoveSpiritual@reddit
It seems you’d have to have someone who is lying, reading off an LLM about things they aren’t very good at yet still able to respond quickly with follow ups and details and still sound natural. I don’t think there are very many people like that at all, and I’m not actually sure I wouldn’t want to hire someone who is. If they can pull that off, they’ve gotta be pretty smart.
Numerous-Draw-9540@reddit
just wait until the llms start negotiating salaries too
stedmangraham@reddit
They can lie about literally anything. It has always been possible to cheat.
You could put literally whatever you want on your resume and the only thing that a check can actually find is the start and end date of that person’s employment at a previous company.
You have to actually talk to them and ask about their work. Does it seem like they know what they are doing?
Before LLMs you could have just had someone in another room answering for you or something. The only real way to completely curb cheating is an in person interview
Groove-Theory@reddit
I dont understand the concept of cheating in an interview if
I mean theres lots of people who skip ahead of the line because they "know someone" at the company. Is that not getting a leg up? We treat that as culturally fine
As well, companies will lay us off regardless of past interview performance or work performance, as we've seen all too often.
"Anti-cheating" really only protects companies as well as their shitty, ineffective interview gatekeeping modalities. It doesn't help engineers much, if at all.
threepairs@reddit
Hear, hear!
SLW_STDY_SQZ@reddit
It depends on whether you or your organization has the ability to ask follow on questions and really drill down the open ended question with more detailed questions. It should be relatively easy to gauge their knowledge like that if you have the depth yourself.
ThingAboutTown@reddit
For real. You’ve gotta talk about the decision-making - the whys: why did you solve the problem that way? Why this architecture? Why this design pattern? We may not need to write the code, but we do need to understand why it’s written and structured the way it is.
You should be able to dig into the details of a complex/consequential problem and pull apart the considerations, the trade-offs, and the eventual decision.
Party-Split-3644@reddit
had a similar situation last summer, totally unexpected
onnemk@reddit
Live code review works really well to me.
I made a small one page code sample with loads of bad practices and design choices on different levels. I explicitly did not put any bugs in it because then it may become a puzzle again.
But it always results in a good two-way discussion, and you can immediately see what the level of the dev is and also how to dev communicates.
Stefan474@reddit
Just for fun could you send it to me? I wanna see how much I can catch (if it's not company specific)
onnemk@reddit
It’s C++ specific
Stefan474@reddit
fair, in that case not useful for me, thanks
PLEXT0RA@reddit
seconding this!
Four_Dim_Samosa@reddit
+1 that. youre anyway reviewing work whether it be from LLM or human. I enjoyed such round and you can still get good signal
hannahbay@reddit
have been begging my company to add a code review round instead of just leetcode-style coding and system design. one of my favorite interviews I ever had as a candidate was a code review. it is a real-world skill in the job, it tests comprehension, you can see what they focus on in terms of following the code flow and understanding functionality, readability, finding edge cases, thinking through tests etc.
I would probably put a bug or two in, but this interview would be an hour, and I think testing that comprehension is pretty important. it would not be the only thing to index on though.
mirageofstars@reddit
I second this. A good dev can see it and then you can ask follow up questions.
SurfGsus@reddit
It's easy to forget so, just going to mention this - do **not** skip the behavioral questions. I learned the hard way by hiring candidates with excellent technical skills only to find out that they're total a**holes on the job. Political, stir the pot types that I regretted having to work with in the very near future. I was naive and, looking back, was too focused on just the technical aspects.
Unfortunately, I still haven't quite figured out how to suss this out during an interview (open to suggestions if anyone has them). But a good starting point are the somewhat cliche questions such as "tell me about a challenging situation and how you resolved it". The goal is to not only evaluate the response (which often can be rehearsed) but also trying to catch red flags (overly negative, everyone else was wrong, etc.)
ResoluteBird@reddit
All the biggest assholes i have ever worked with get jobs pretty quickly. Idk why people think behavourials work
kennpacchii@reddit
Not mad at all Michael :)
kennpacchii@reddit
Snowflake
ResoluteBird@reddit
;) u mad
Wide-Pop6050@reddit
The cliche questions work fine. Assholes often put themselves, but more importantly, good candidates give interesting, non canned answers.
DWALLA44@reddit
I fully stand by the idea that any solid, experienced engineer, can have a 60 minute conversation with someone about their career, their projects, their communication style, etc... and know if someone can do the job and get it right 9/10 times no problem.
Unless your working on life or death projects, or at NASA or something, we don't need all this crazy shit.
ecw3Eng@reddit
The direct answer that many dislike for some reason, is to interview like most other white collar fields do: - Degree required - Tell me about your last job: how did u do X? Y? How did you deal with constraint in this? How do u handle miss understanding with peers?etc..
The way we were interviewing in tech was bs and broken. Leetcode and live exams, why? I mean I already went through tons of exams at uni and NO I do not solve puzzles under 20 minutes while explaining my thoughts at my daily job (20 years exp). It was sadistic and quite frankly AI fixed this.
Concerning: “it is expensive to hire a bad engineer. “
Well, it is way more expensive and more dangerous to hire a bad nurse or doctor. Still their interviews are: degree checking + a few questions and then the performance on the job decides if they stay or not.
AI put an end to tech hiring sadism, and it’s good.
iegdev@reddit
This.
I have a BS in computer science and a verifiable employment history for the last 8 years. Why the hell am I being forced to do some coding exercise that literally has nothing to do with my day to day work?
YesNoMaybe@reddit
It's clear you guys haven't been on the other side of interviewing. I've interviewed will into the triple digits of candidates and the are very experienced engineers with masters degrees that somehow can't write the most basic of code.
ecw3Eng@reddit
100% my friend. You make efforts for 4-5 years at uni like all other white collar fields do, but then pay the price of some cheap ass ceos brotherhood who realized it’s the tech era and they need to keep the rates low, so they decided to flood the field with everyone with a pulse and their grandma to lower those rates. Imagine doing the same with the medical field? This was madness and the status quo was protected by those who were allowed in with zero skin in the game. AI now FAFO’d everyone.
Wide-Pop6050@reddit
Well doctors and nurses have certifications. Engineering could consider that
ecw3Eng@reddit
So do we, i have bachelor of engineering in software engineering. I took an oath similar to doctors upon graduation. The infrastructure is there, time to use it and stop this madness.
Wide-Pop6050@reddit
When you actually do coding tests you realize why they’re necessary. Plenty of candidates with the right degrees don’t actually have the skill we require. We don’t do exam style tests, I agree no one’s in school here, but the technical has been immensely useful and necessary.
ecw3Eng@reddit
That’s the whole point, plenty of nurses and doctors are not proficient, their lack of proficiency is way more dangerous than a software engineer’s yet they are not given tests after having spent the years to get the degree. This sadism in tech interviewing is only due to the bs that “anyone can learn to code in 21 days”. Well, now ai put an end to it.
Wide-Pop6050@reddit
Much cheaper to find out during a 1 hr, non leet code phone screen than by hiring them
ecw3Eng@reddit
If you really think u can assess an experienced engineer’s depth & architectural knowledge with a 20 minutes stupid puzzle that they must spend months regurgitating to memorize (because it’s not the norm to come up with a “new” algorithm on the fly in 20 minutes all the time unless you seen it before) - which is like playing Russian roulette - then I have a fancy bridge to sell u.
gjionergqwebrlkbjg@reddit
Most of the absolutely most garbage candidates and people I've worked with have degree too.
ReikaKalseki@reddit
If you mean specifically a CS degree, no thank you. Just as how someone having the degree is not a guarantee of skill - as you rightfully point out - the lack of a CS degree is not a guarantee of lack of skill.
You can learn software development experientially, and there are tons of people in the industry today doing a good job without having that very specific educational background. 80% of my department at work has engineering rather than CS degrees - myself included - and honestly some of the worst devs on the team are the CS grads because they tend to be the type to ignore practical realities in favor of endless refactoring, overengineering, and getting lost in minutiae like "we need to update this application because this one random library had an update, what do you mean there's a whole approval process for getting this work done?".
I switched to software (professionally; I was doing it independently for almost ten years before that) during covid, something I always say is one of the best decisions I ever made. I got a number of people who slammed the door in my face over the lack of a CS degree, and your proposal to make that the universal standard is not just galling but massively myopic.
ecw3Eng@reddit
If we parallel what u wrote into other fields, say in the medical field, here is what is happening here: A non doctor who suddenly became a “doctor” just cause some cheap hospital CEOs didn’t wanna pay dr rate, is telling certified doctors that he doesn’t think their degrees matter because he in his infinite wisdom decided doctors without an acknowledged training and degree are bad, but your random joe from the ballet department is better. That is where this field is at. In any case it doesn’t matter, AI put an end to this madness so it doesn’t matter what you or me think at this point really.
Expert-Reaction-7472@reddit
a good interview now looks largely what a good interview before looked like - a conversation with a peer.
Imagine you met me at a bar and we both discovered that we work in tech - we talk shop for an hour.
After the hour you should have a good idea of whether or not Im somebody you'd work with or not.
Over complicating it really is what makes them suck.
obviously some standardisation about things you're company is interested in will help but keep it fairly general and open ended - let the candidates show what they're good at rather than trying to assess them quantitively - it never worked that well in the first place so why try to return to a broken system.
Jennsterzen@reddit
Yes! Why does the software engineering interview process always have to involve BS tests. Your experience/training, and ability to communicate about it (as well as good personality fit) should speak for itself
YesNoMaybe@reddit
I've interviewed developers with 20 years experience that were great talkers and could talk at length about the architecture and design of projects they worked but couldn't code a for loop in a language they were supposedly experts in. This has happened multiple times.
A simple coding exercise weeds these people out. If you can't take a very simple request for a function that performs some basic functionality and turn it into code, it doesn't matter how good you are at talking about it. You're not getting an offer.
Don't discount his good some people can bullshit.
asstatine@reddit
“If you got a project you want to ship, but the prioritization is low for a team you depend on who needs to make a change how would you approach the problem?”
This is the type of real problems I face in software engineering. If I need an optimized sort function I’ll just use a library.
accidentallythe@reddit
For the life of me I just want interviewers to ask me about my experience and accomplishments. It's amazing how infrequently and late in the process this actually happens. A lot of staff engineers I know have patents, research papers, and have done all sorts of deep, complex work (listed on their resume!) that they can talk your ear off about. But instead they have to play four rounds of code golf before reaching anyone who actually asks them anything like that.
lphomiej@reddit
I think focusing on "tell me about your project" and finding out what the person finds engaging/interesting/energizing is super valuable.
lionelpx@reddit
Stop testing for how well candidates perform without AI, start testing how well they perform with it ? Or were you planning to ask them not to use AI tolling at all ?
CodelinesNL@reddit
Just do these technical interviews in-person. A lot of these issues are self-inflicted due to giving people take-home assignments. People always cheated on them.
TaskSetZero@reddit
Interviews can no longer be about recall, which is a change that most teams haven't really grasped yet.
It's not a meaningful signal if a model can respond to it. The purpose of the move is to assess thinking in action rather than prepared responses. Give a messy, real-world problem from your stack to the candidates. Allow them to clarify decisions, make trade-offs, ask questions, and adapt when limitations change. You're assessing how individuals prioritize, reason, and adjust to ambiguity rather than whether they "know." That's considerably closer to the real work and far more difficult to fake.
NeuralHijacker@reddit
This is so true. We're in the ridiculous world where our hiring policies want us to use stupidly hard tests and ensure no AI is used, to hire for jobs where engineers are expected to do as much as possible with AI. We're just not filling roles because the sort of people who can get 90% on leetcode are often not actually very good engineers.
TaskSetZero@reddit
Indeed, there is a mismatch in the system.
Using AI-resistant filters, we are attempting to hire for AI-augmented work. As a result, we wind up choosing candidates who can overcome fictitious limitations rather than those who can function well in the real world. "Can you solve this alone?" is no longer the signal. "Can you use tools, think clearly, and produce outcomes?" is the question.
Teams will continue to exclude the engineers they truly need until hiring reflects that.
jegsar@reddit
Close your eyes and describe your ideal distributed development framework.
Yes its a bullshit question, but all im looking for at this point is common sense that understands mutex and address space.
Wooden_Step_5691@reddit
Do whatever, then finish with an in-person technical interview
kirkegaarr@reddit
Same as always, just pair program with them. I'd let them use an LLM but you might not want to.
stedmangraham@reddit
It is literally impossible to make a leetcode question hard enough for a person but easy enough to actually solve. I’ve been saying this for years but leetcode doesn’t actually test anything relevant to software development. If it did, we would all be replaced by AI
Either ask questions in person and eat the cost of flying them out, or ask questions about their background. Ask what projects they have done, what was their specific role, what problems did they solve? What was the outcome? Would they have done something differently? What would they learn.
We don’t need to make the interview itself such a fucking death gauntlet. Experienced developers will appear more experienced. Bad devs won’t even know how to lie correctly.
If it turns out they suck, fucking fire them. Why does this all have to be so difficult? Other types of jobs don’t have this problem
Undercoverwd@reddit (OP)
I wholeheartedly agree about leetcode. We've never used leet code we used to just have them code a simple react component that updated something - and you'd be surprised how many people could. not. do that.
But this challenge was just a proxy for someones experience, we don't hand code components off the dome, but most experienced devs knew their way around a code editor and you could tell if they were full of shit.
We can focus on the verbal discussion parts, but generally, the folks who did great at the verbal tanked the code challenge and vice versa, but maybe verbal is the skill we actually need?
Unfortunately, I can't fly out or fire anyone.
iegdev@reddit
I always tank the code challenge. I just can't think under that kind of stressful situation with people watching me. My mind goes completely blank. Call it social anxiety, ADHD, whatever. Same thing with tests during school. It's just not how my brain works. And if that's all you're basing your idea of a successful candidate on, then yeah, I'm not going to be a fit.
But I must be doing something right if I've been gainfully employed the last 8 years.
forgot_previous_acc@reddit
Don't get me wrong. I am full stack dev with almost 6 years of exp and i still might fail at some java program or some programming question. I have created cicd pipelines from ground up, fix broken pipelines. Built scalable hlds llds and implemented most of it and this was all before llms and copilot. But i still might fail if you ask me to program in front of you. I think of myself as more of a problem solver you know, i might not know everything but i am confident that i will be able to figure it out. Don't know if i am making any sense or not.
zuilli@reddit
If they can confidently navigate a complex conversation about their prior experience explaining their reasoning for dealing with the challenges they faced it means more to me than being able to code well on the spot, specially for more experienced devs.
All that complex thought is the part we can't use AI for so even if they are actually bad at programming once on the job it's a lot easier for them to be productive than the other way around.
AchillesDev@reddit
Always has been.
This should tell you that it's not a good proxy.
shifty_lifty_doodah@reddit
It tests IQ
teknoise@reddit
No it doesn’t. It tests a specific subset of intelligence (abstract problem solving skills that aren’t generally related to the job, under unrealistic deadlines with intense scrutiny). Smart people slip thru the cracks because they’re smart at their jobs, not smart at rote memorization while being gawked at and judged)
shifty_lifty_doodah@reddit
Eh excuses
It’s not a perfect test. It’s pretty good. The smarter people tend to do better.
Maktube@reddit
I totally agree with your point, but I will say, that IS IQ. It's famously very culture-specific and not particularly correlated with actual intelligence.
PudgyChocoDonut@reddit
it tests how much someone wanted a role.
systemnate@reddit
Preach it. My first couple of dev interviews were just curious conversations with other developers and we never had a problem hiring someone completely terrible.
What's weird is that as soon as people really started making your average run of the mill software interview look like you're interviewing for a Staff position at a FAANG, they simultaneously decided that they shouldn't be quick to fire someone if they can't hack it. Like we simultaneously started gate keeping harder while being too scared to let someone go for poor performance.
Careless-One-6835@reddit
this reminds me of when our team ditched coding tests for real project discussions
Careless-One-6835@reddit
why did they choose that specific color for the background
wisconsinbrowntoen@reddit
I don't understand why companies don't just hire people with the understanding that it's extremely difficult for either side to know if it's a good fit before you're working together.
Like why are you spending $40,000 hiring an employee when you could just hire the first decent person you come across, and if after a week it doesn't seem like a good fit move onto the next one?
harraypottah@reddit
I have given prompts like this to an LLM ".. give me interview questions to evaluate a senior dev with experience in... and answers shouldn't be easily searchable/parroted from LLMs".
You get some very good stuff that you can test the candidates on, and the genuine ones seem to do well in these - giving more clear signals.
zubinajmera@reddit
I guess the "make it so hard bots can't do it" instinct is a trap. You end up with a process that filters for people who enjoy adversarial interviews, not people who can actually build.
The reframe that worked for us: stop trying to prevent AI use and start making AI use irrelevant to the signal. Put candidates in a live running environment — real API, real database, something actually broken — and let them use every tool they normally would.
The question isn't "did they use AI" it's "do they understand the system well enough to know when the AI is wrong."
Experienced devs actually prefer this format. No trick questions, no test theater, just real work. The signal on seniority is immediate - how they navigate unfamiliar systems, what questions they ask, how they debug when something breaks unexpectedly. None of that is fakeable regardless of what tools they use.
We built our whole screening/shortlisting process around this after hitting the exact wall you're describing.
happy to share more specifics if useful?
TwoSignal6882@reddit
One of the reasons I haven't applied for companies yet is because I haven't brushed up on (DSA) advanced leetcode questions that I know many companies will ask. I wish all interviews were like this!
Undercoverwd@reddit (OP)
Please do! Did you use a particular interviewing tool to set that up?
zubinajmera@reddit
https://utkrusht.ai/
headinthesky@reddit
I'll tell you what has been working for me. First, have a repo that has issues tests are failing, logic issues. I asked them to run it and explain what the issues are. Then I let them use AI to fix it. I put details in another comment, but basically I want to see how they're using AI. Are they just prompting it to say go fix this bug or are they first using it to explore the repo, make a spec of what needs fixing, etc. basically to see if they're the slop code vibe code kind of person, or they still do planning and are able to use it as a tool to augment their skills
new2bay@reddit
No idea. Nobody wants to interview me.
c0ventry@reddit
Same as a good interview has always been. Learn how they think about and approach problems. The tools evolve over time, but critical thinking skills are always in vogue.
fued@reddit
I just ask them 15-20 questions relevant to the project.
If they can wing the answers, they can probably wing the job.
The entire rest of the conversation is about culture fit, do i wanna work with this person?
sweaverD@reddit
“The entire rest of the conversation is to enforce my biases”
fued@reddit
Pretty much. I don't get benefits for the company doing well. So I'll hire whoever I enjoy working with the most.
SolidDeveloper@reddit
This is pretty much illegal where I live. How do you prevent discrimination in terms of race, ethnicity, gender etc.? Because I can easily imagine an interviewer being more comfortable with someone who looks like them, but then that’s basically illegal discrimination.
cheesecow007@reddit
Real
Nofanta@reddit
They don’t happen. If you tried to design one the backlash would be so great it would never happen and would damage your reputation.
Wide-Pop6050@reddit
Okay I just hired
It was much better to hire off recruiters or specific job sites, rather than LinkedIn
First round was a typical phone screen
Second round was a 1.5 hr technical. We asked them to share their whole screen, and said they could use Google/AI/wharever. The first task was pretty straightforward but the second was extremely open ended. Or at least there were multiple ways you could do it. This was a crucial round, really showed who knew their shit or not
Then we did a 1 hr panel behavioral interview with members of different teams.
Found 3 great top candidates and have hired one! We did get one candidate we interviewed who was a fraud. He was the only interview candidate who hadn’t come from one of the specialized job sites
protocolzed@reddit
What job site are you ya l referring to?
Wide-Pop6050@reddit
It depends on your specific industry and area. techjobsforgood is one
ninetofivedev@reddit
I'm going to let you in on a pretty big secret.
Interviews have always been bullshit. We've never had a good signal for detecting good candidates.
YMMV based on country and state laws, but the best solution is to interview based on experience, use your best judgement, and then fire someone's ass into the fucking sun if they turn out to be a dud.
The best company I've ever worked at, our VP had the expectation that within 90 days, you know who is good and who isn't, and they weren't afraid to fire anyone.
That guy we hired from Google who couldn't use git. He was gone. The bootcamp grad that learned kubernetes in 3 weeks and completely owned multiple product and feature realease cycles end to end? They're getting a promotion.
Merit exists. Despite what people say.
Wide-Pop6050@reddit
Yeah not being afraid ri fire makes a huge difference. Even in countries where it’s difficult to fire there is usually a probation period of 90 days where you can fire
Ok_Chemistry_6387@reddit
Google didnt use git really until recently. Seems an odd thing to call out.
ninetofivedev@reddit
He couldn’t figure it out. He couldn’t figure out basic git usage in the world of ai.
Ok_Chemistry_6387@reddit
Well thats a big difference hah.
KittenMittenz1@reddit
👏👏👏
Spimflagon@reddit
The best interviews I've partaken in have basically been a conversation, rather than question and answer. Rather than asking arbitrary questions and testing their memory with a pass / fail criteria, just ask them about their projects and engage with them about it. If you're having a tech test (urgh) make it something you can discuss their approach to afterwards. The interviewer's going to need some insight themselves, but really if you're interviewing for a tech role with a non-techie interviewer just stick all the CVs in a barrel and pull one out at random.
In that sense, designing an interview process might be sending you in completely the wrong direction.
rer1@reddit
We switched to a mix of code review and coding with ai.
Code review is relatively short (half hour) and very deterministic. Either the candidate found enough of the issues we were looking for, or they didn't.
The coding task is relatively simple, not leet code style. Candidates don't have to use ai but we very much encourage it.
You'd be surprised but many candidates still struggle with using ai. They either don't make it in time, fail to understand what the ai wrote, or fail to guide the ai to a more extensible design pattern / code structure.
StepIntoTheCylinder@reddit
People have been over-interviewing for a long time. It has ratcheted up to 12. Dial it back. People used to just look at the resume, talk shop with them, and maybe ask a few interesting "white board" questions. What was wrong with that?
HoratioWobble@reddit
Most of my interviews have just been conversations. Even when the companies have tests they accept my social proof (twitch, linkedin, live app etc)
The irony is, most of the tests companies give I would fail.
Even my current role, my boss is very happy with my work, happy with me but had they given me their standard test I wouldn't have gotten the role.
Have a conversation with them, if you like them and think they're competent either hire them. If you can't verify competency then give them the choice of - code review, - mini project - or showcasing their own project + code
None of it is going to tell you if they are good at the job - you need to give them the chance by doing the job.
Mini project is almost certainly prone to AI, but does that matter to the business? Deliverables and value are the important part.
drguid@reddit
I have an amazing side project yet can't get hired. People think the code I write at home is not the same as the code I would write in an office for an actual company.
What is up with idiot recruiters?
Also while I'm here I no longer do offline coding tests... the last 2 people to send me stuff to look at have been scammers.
bigorangemachine@reddit
Pair Programming & Code review your worst code base.
PracticallyPerfcet@reddit
I don’t know what the answer is, but the current trend of not sharing the interview format and making the interview too hard to pass is ridiculous. A company did this to me (a person that has never cheated with AI, has a graduate degree, and 19 yoe) at the start of March.
It was deeply insulting, and made the company look awful.
It was also the final bit of nonsense in this market that made me quit interviewing to focus on consulting and wait until someone in my network hands me a job.
jrodbtllr138@reddit
I just ran a bunch of interviews and it was very clear to me when someone was using AI for the interview.
For conversational, just be interested and ask them questions about project specifics, problems they ran into, the trade offs and other options considered, opinions they have on it, etc.
For technical I focus on how they do requirements gathering, adapt to changing requirements and expectations, and how they debug and refactor through an evolving FizzBuzz variant
I feel like we are starting to see the decline of the standardized interview and the value of the personalized interview, but the skillsets are different from what interviewers have been doing in the past.
Mobile-Scale4899@reddit
so basically a game of 20 questions then
jrodbtllr138@reddit
The only time it felt like 20 questions was when people couldn’t talk about their work (which is a bad signal that either they can’t communicate well, or didn’t actually do it).
Otherwise, it just felt like a conversation with a coworker who’s work I took an interest in
Playful-Wing-4711@reddit
what inspired the choice of words here
BanaTibor@reddit
I think first you have to declare what qualities are important to you, then design the interview process around them, including the technical part.
featherknife@reddit
Ok_Chemistry_6387@reddit
Ask about future past projects. Ask them about warts in the language they use. Ask them about the worst bug they have encountered. Ask them to outline a project they worked on. Ask them to debug something. Dont code ask them to just talk through a solution to a problem. Ask them about the tradeoffs in their design soon after. Anyone whos thinking will be able to answer quickly, or even have already told you. So many ways. They all require work though heh.
Maleficent_Tank2199@reddit
I like them to bring a project of their own, and let them give a short presentation about something they choose and think gives a good insight into their abilities.
Then in the discussion I want to know what they planned, what worked, what did NOT work and what they would like to do differently. I also want them to point out explicit tradeoffs, learnings and what they would do differently now.
The signal I am looking for is the ability to communicate decisions, tradeoffs and the ability to explain a shifting mindset over time. A hard negative signal is someone who is uncritical or his view has not shifted during / after development.
FatefulDonkey@reddit
Have you considered code readability?
No-Stuff-3951@reddit
focus on problem-solving and real-world scenarios maybe
dacydergoth@reddit
I can answer that ... for money.
sgs4b-nito80@reddit
make sure to start with "why do you want to work here?"
mirageofstars@reddit
You could try asking them why [bad practice or crappy framework] is good to work with. AI might say it’s great, the human would say it’s crap. Or ask them about something they definitely wouldn’t know (eg hardware edge appliances or something) — if they answer it easily then they’re cheating.
GlobalCurry@reddit
What if you get the one person who knows a lot about hardware edge appliances lol
jaxsaxsf@reddit
We do a combination of coding and system design interviews, same as ever. We do have to be on the lookout for folks cheating though. It's hard to be reading from a chat window and having a normal conversation with me and typing code all at once though. We're pretty good at spotting the people who try.
GlobalCurry@reddit
What happens if someone is reading off of their normal notes? For example, I used to bring a notebook to interviews all the time full of notes. Not only was I taking notes in the interview, but I'd also refer to them or use them to ground my thoughts (not necessarily for systems design, but when answering questions in general).
Abadabadon@reddit
Best interviews ive had are usually code reviews
RespectableThug@reddit
I see it two ways:
As someone who’s been interviewing for over 10 years, in-person interviews may be coming back. If you’re a remote company, you’d have to fly people out. So, it may be too expensive for many, but if you can control the environment, a lot of these problems are solved.
You up the difficulty. Accept that candidates will be using AI and create the interview with that in mind. Make the interviewers tougher. You’re testing their abilities to correctly apply their tool, not do what the tool can already do.