I have worked with non-tech employees to set up Claude to help them do small tasks. I’ve helped to review and improve completely vibe-coded projects by such employees.
I’m not sure what my role will be, but I fully embrace that my traditional role of writing code is gone.
the honest truth is that maybe 10-20% of SWE (at best) are “good”. sure it is harsh but i won't lie. if you're good you'll probably relate.
the rest kind of suck.
i’ve never gotten anything lower than Exceeds Expectations in my career so I’ve seen how awful some engineers were. i’ve seen how amazing a tiny minority were and i made them my mentors.
these days i have a simple policy.
if they cannot think, they are fired. why waste resources (time and money) on someone who can’t use their brain? i’d rather give AI credits to someone who uses their brain.
thinking is the humans job. the ai needs to execute on what the human thought of, improved, planned.
and even at "good" companies you have people who can game the system to get in, and then they struggle to get anything done on time or be responsible for taking on and completing any initiatives bigger than a single task on a bigger scope.
I still think a single in person LC style (doesn't have to be LC per se, could be domain specific) logical thinking/reasoning exercise is useful. I want to ensure the person can actually put 2 and 2 together and think. This is just a fast filter.
If they seem like they can think, then I like to do 2-3 systems design interviews. I'll try to give them something related to things I like, such as graph structures, writing a complex query that needs to be dynamically generated, or something related to infrastructure or how they'd do something that I've already done. After all, this is MY project they're joining.
So far that has worked well.
Few more things -
I like to test if they are a humble type (they can work on a team putting ego on the side - the mission is our number 1 priority). if they say they know something that i know and asked, then they can be sure I'm going to drill them on it. if it turns out they lied, i'm not wasting more time. Thanks for your time, take care. This is very important to me. Just say you don't know, it isn't a big deal because ever since like 1994 that has not been an issue. You can just learn things online, and AI makes that even faster. I am never afraid to say I don't know something, and I've asked plenty of "dumb" questions (while doing some due diligence first) so I don't really mind.
Can they handle information overload? I am the type of person who has multiple branches in my head of actions I can take next, so while I may appear stressed I'm really not. Can they keep up? Our goal as software engineers should be to come up with solutions that solve the problem in a way that makes building on it simpler in the future. My goal is simplicity and effectiveness. So I'll see if they can keep up, and eventually reduce the work to be done into atomic pieces. This is a fun exercise because it is collaborative and we get to bounce ideas fast back and forth.
Finally, I like to let them use their favorite tools, including AI tools (codex, claude, some ppl have esoteric custom stuff which is cool), to solve a problem together. It might be code related, it might not. Really depends on my mood. I like to see how they work and what sort of output they can come up with. This filters out people who only ask AI stuff, instead of having some framework they've already developed to be effective.
Honestly I don't know how to scale this process. I'm not really going to feel bad either about firing fast, ultimately this is a business and I don't want customers to suffer because we have some issues internally.
At the same time, I wonder if I even need to build out an org with 100s of people. That was an inefficiency (look at all the layoffs), and it is traumatic.
If I can find a few great people who can be supercharged and turbocharged and electrified with AI, then they can take on & own bigger responsibilities. My number 1 goal is to ensure they're with me on the mission, and after that all things seem to sort of fit into place.
mass effect 2 is my favorite game ever. it is all about putting together the team, and ensuring you work with each one of them to get their whole loyalty.
each member is a badass, in their own regard. it’s also a video game and it’s linear unlike real life. but the mission is super important to me.
and when others have their own passions they want to express and carry out via fulfilling the mission, that’s super key imo.
so far it’s worked out fine. people get the fast firing thing. they know if someone isn’t onboard with carrying out the mission they also don’t want to be burdened.
like we are seriously helping people in an underserved industry. it’s insane.
i hate working with mids and bads, they are going to bring everyone else down. so i want to work with the best people i can get. they don’t need to be MIT grads paper weight types. they just need to be mission oriented and focused.
Someone who produces absolutely nothing and have no impact has cost, but is still better than someone who produces net negative. And the people who solely act as interface between LLM and whatever might fall to later category.
All my experience in trying to hire developers has been wading through an endless stream of people who were just useless.
Me: I want to represent a 2d grid, what data structure should we use? Them: A string?
This was someone applying for senior engineer. Others I've had filled their CV with SQL related acronyms. But couldn't explain what a foreign key was and then stubbornly insisted that at their current corp they would never ever use foreign keys in their SQL database!
I've had senior engineer when asked how to check if we had a 2d array with an item at x,y tell me if anything is on the same column or row, they couldn't do it, couldn't even verbalise how to approach it.
"Web Developers" who didn't know the difference between GET and POST. Web Developers that have never heard of PUT or what it would be used for.
The answer that almost guarantees I'll hire you is "there's got to be a library function for that, so I look in the manual". Almost as good is somebody whiteboarding how they'd convert ddd to mm-dd (and then account for leap years, etc.)
I get a disturbing number of people who say things like "I would communicate with the person asking for this to see what they're really intending blah blah"
My favorite answer was on a phone interview where he just hung up and wouldn't answer when we called back.
Sounds like they know this question is a “gotcha” question but just misinterpreted which direction you were going with it.
Some will ask a question like this expecting you to treat it like a puzzle and outline how you’d solve it as-is; others ask it as a way to probe how you’ll deal with strange or misguided requests (the case you noted as disturbing); and others yet will ask it to see how you’d practically solve it (your intention).
Seems like a bad interview question without context regarding kind of answer you’re looking for.
I prefer fizz-buzz as a question because it is obvious there isn't a library. It is also a problem you should be able to do in an interview. It has enough weirdness that there is no best answer, despite having several workable paths you could try.
If you would ask someone to write a piece of code, and a part of the problem is this conversion, then you would be right to expect they reach for a library, but even if they don't you would be giving them the opportunity to explain themselves, and judge the explanation, not the answer. Also, if your test is "does this person reach for a library at the right time", you could do a lot less esoteric and confusing by just asking them to add 10 days to a date. If you just ask this one specific problem, it is likely they assume you are looking for them to demonstrate the skills involved in actually solving the problem, i.e. leetcode.
This is also why some people give you the blabla answer, because it is indeed very unlikely that someone needs to do this legitimately. This is because its a toy problem. Someone's professional reaction to the problem in isolation should indeed be: this is weird, I've never been asked something like this, what's up?
Finally, even though the question is terrible, I would still rate the "whatsup?" response higher than the "leapyear" response. I would want a developer to triple check that this problem needs solving, before they would solve it themselves.
Finally finally, if there's one answer to one question that, when answered trivially in a way literally taught in most basic programming courses (use the standard library / a third party library), makes them a "guaranteed hire", I also have significant doubts about the level of talent you are bringing in, as any experienced interviewer will tell you that qualified people will get important questions wrong, and unqualified people will get important questions right.
I understand that this reaction might be quite harsh, and I know better than anyone that its hard and time consuming to do good interviews, but please consider that you are rejecting people who may be very confused and sad by this way of rejection.
Heh ... yeah well I wish I had it to do that.
However, you are asking gotcha questions.
Like talk to the person abit to avoid obvious toxic people and then fact check the credentials.
So yeah draw lots. Guessing right on whether you want a practical response for the gotcha or for them to solve the puzzle adds no value to the process.
The absolute worst is someone that tries to brand themselves as a 10x engineer by constantly using programming terms like "dynamic programming", "polymorphism", "recursion" and the like, but they're really a 0.1x engineer because they don't truly understand what any of those are and when they should actually be using them, and so try to shoehorn them in when they don't need them while also not understanding them, and end up writing low-quality crap.
Took too long for management to get rid of that guy.
We've already established that most of it is noise. You don't need to keep up with producing noise.
So the actual problem statement is not "how do I keep up" but "how do I correctly tune my filter", which is solvable.
The biggest challenge there I think is that many people are not prepared for just how sharp and uncompromising that filter needs to be, but that too is solvable.
feels like its becoming reality that we as a human don't need to this anymore