Anthropic Engineer Declares the Coding Interview Dead
9 152An Anthropic engineering manager says he has stopped using traditional coding interviews because AI can now outperform candidates on the kinds of programming challenges they typically contain. "There's no more point in giving you a little React challenge than there is in quizzing people about assembly; Claude is now better at both," said Felix Rieseberg in a blog post. "Our most impactful builders are successful regardless of how good they are at these tests." The Register reports: "I don't think the coding interview is going to come back," said Rieseberg. "It's easy to get sentimental given how many copies of 'Algorithms and Data Structures' were bought in hopes of unlocking good jobs and a better future -- or how the death of the coding interview is another milestone in software engineering reaching a new level of abstraction."
At the same time, he's not certain what will replace it. "I'm probably not the only manager trying to figure out how to interview for people who build well with AI," Rieseberg wrote. "I haven't quite gotten there yet -- but I'm scanning resumes and portfolios for little prototypes that exhibit taste, empathy, and principled thinking. If you have figured out something better, I'd love to hear how." In a response posted to social media, Jaana Dogan, principal engineer at Google, said: "I'm not doing a ton of coding interviews nowadays but I'd be interested to interview how people use coding agents," she wrote. "That itself gives me more signals about candidates' coding skills."
"We are not seeing tech companies abandon skills assessments widely across the industry -- at least not yet," added Nicole Bachaud, labor economist at ZipRecruiter. "In fact, based on our October 2026 Employer Survey of 1,500 employers, we see the tech industry (Information sector specifically) is actually the most likely to test for skills, with 62.5 percent of employers requiring formal skill tests on initial applications."
"The change at Anthropic might be reflecting a shift away from technical skills towards more human, soft skills," said Bachaud. "As AI becomes fully capable of writing code, tech employers may be pivoting away from rote technical testing and heavily toward human-centric soft skills." She argues that while AI may be able to code well, it can't replace critical thinking. "A staggering 61.5 percent of Information employers now report that 'critical thinking' is a primary skill lacking in today's candidate pool," she said.
9 comments
From one dev manager to another (Score: 5, Insightful)
by ZipprHead ( 106133 ) on Wednesday October 28, 2026 @06:07PM (#66398626)
Your process for interviewing should not change. You need engineers that can think well to judge the output of an LLM well.
Re:From one dev manager to another (Score: 5, Insightful)
by Brain-Fu ( 1274756 ) on Wednesday October 28, 2026 @07:12PM (#66398716)
I have worked with junior level devs who just take a dev task's requirements, hand it straight to AI, put that through basic testing (only the main intended cases) and then offer that up as a finished solution. The solution was, of course, full of holes, some of them quite serious, just not obvious. The holes were in the edge cases and how it handles and recovers from errors. These were exactly the kinds of mistakes I would expect a junior level developer to make (in at least one case I am thinking of, the task shouldn't have even been handed to someone so inexperienced, but the people making that decision didn't really know how advanced the task was and anyway the dev wanted to get real experience).
So, our process worked because we had senior level human code review, which caught these. But the real risk here is that without such experienced senior level people to do that review this would have seemed like a "slam dunk" where AI saved a whole lot of time and eliminated the need for expensive veterans on the team (only to fail catastrophically in production as soon as something hiccupped). This is what I am most concerned about. Companies will fall for this. Especially new companies where leadership is just as inexperienced at running a tech business as the cheap vibe coders they hire.
So if we are hiring people who will be using AI to do most of the coding work, I think we need to test the thoroughness of their thinking, their ability to detect unspoken problematic edge cases and know to focus on them when reviewing the AIs output, and the ability to recognize dangerous vagueness in requirements and ask for clarity. Also, the ability to mentally translate between "business speak" and "tech speak" is extremely valuable.
Though I still don't feel right about skipping a proof of coding skills. I just don't think we are yet at the point where it is safe for someone who can't code to be relying on AI to get it right. The aforementioned senior developers doing the code review are all people who have been coding by hand for years, and in order for such people to continue being available to do code reviews in the future, we need to make sure we hire people who can do this.
Re:From one dev manager to another (Score: 5, Insightful)
by tragedy ( 27079 ) on Wednesday October 28, 2026 @08:35PM (#66398800)
Your process for interviewing should not change. You need engineers that can think well to judge the output of an LLM well.
Quite right. The logic the manager from the article is presenting is incomprehensible. They're arguing what? That you don't need coding problems in interviews because an AI can do them instead. Classically, they have been stock problems that are pretty much just mild variations of standard stuff from textbooks. So if you don't need to bother with them now because AI can do them, then you never should have needed to bother with them because the problems had already been done before, thousands of times, and written down. What does this guy think, that the coding problems in interviews are vital new developments in computer science or something? Now, some companies do actually present those as problems just to see how the interviewee approaches them, and there are also some stories about unethical companies that have used long-form interviews as a way to get programming piecework done for free by suckers, but that's not what's being discussed here. The point is that we're talking about a basic way to make sure that interviewees actually understand the concepts themselves, not asking for the interviewee to demonstrate that they can accomplish some amazing task that no-one else on Earth could perform.
Still an Extremely Good Signal for Senior Eng (Score: 5, Insightful)
by hutkept ( 10503251 ) on Wednesday October 28, 2026 @06:08PM (#66398628)
Coding interviews test for fluency. You don't give coding interviews to someone to see how good they are, i.e., variation in speed is fine, you are looking instead for how well they decompose problems, abstract and structure things to make them tractable, and yes, whether coding is second nature to them. People who do extremely well in coding interviews are also going to do extremely well coding with AI. Is the converse true? This is where AI has made things tricky. If the engineer is supposedly experienced and senior, then yes, i.e., if they have not become a good coder by now, something is wrong, and you don't want to bring whatever that is into your project. However, if the engineer is very junior, than not necessarily, they may have chosen to spend their time on more important things than "learning assembly". Therefore, ironically, it is the senior engineers and more experienced folks who you get the best read on by doing coding interviews, but you do still get a good read on them this way.
Re:Still an Extremely Good Signal for Senior Eng (Score: 5, Insightful)
by XopherMV ( 575514 ) on Thursday October 29, 2026 @01:21AM (#66399080)
The majority of interviewers pick their favorite algorithm from the data structures class. The stereotypical question is some variation of a graph problem that damn near no one uses professionally. So, the most senior of engineers haven't dealt with it in 10-30 years unless they grind Leetcode. That gives the advantage to the junior engineer fresh out of university who studied it in depth just a couple years ago. Giving advantages to junior engineers in your hiring process isn't how you get the best people.
Further, most interviewers tend to care primarily that the candidate solves the problem in the time allotted. That used to mean writing pseudo code on a white board. Today, that means writing code that compiles, runs, and passes all the tests, including the hidden tests. If a candidate doesn't actually solve the problem, no one cares about the structure of his solution. Any time spent understanding and structuring the solution is wasted.
Secondary consideration is provided to all the traits you listed: decomposing problems, abstraction, structure. Good luck getting an accurate read on all that given the limitations of an interview. Your candidate is working in a high-pressure, high-stress, time-limited situation that's completely artificial and alien to most developers who nearly always have access to their IDE, the internet, and now AI.
You've made a completely artificial situation and think you get genuine results? The thought is laughable. Well, except that you're likely not hiring good candidates due to some BS interview process.
Raw milk enthusiaist declares pasteurization dead (Score: 5, Funny)
by Anonymous Coward ( None ) on Wednesday October 28, 2026 @06:36PM (#66398660)
NT
Not surprising to me (Score: 5, Insightful)
by jenningsthecat ( 1525947 ) on Wednesday October 28, 2026 @07:13PM (#66398720)
"A staggering 61.5 percent of Information employers now report that 'critical thinking' is a primary skill lacking in today's candidate pool,"
I'm an old fart so I may be totally out to lunch on this one, but I get the impression that most of today's schools really don't teach students to think, question, analyze, and synthesize. I think this applies across all subjects, not just in technical ones.
Part of me thinks that the ones who DO know how to think and analyze are tilting the table against subsequent generations by under-funding education, healthcare, and nutrition. If you look at today's parasite class - many of whom are in the tech field - you see that their political attacks on taxpayers getting any of the benefit from the money they've paid are increasing. The broligarchs want it all for themselves, and I think they see AI as a stepping stone toward making the bulk of humans now living on Earth expendable. I think that it's in their plans to expend a whole lot of us.
Re:Not surprising to me (Score: 5, Insightful)
by gweihir ( 88907 ) on Wednesday October 28, 2026 @10:03PM (#66398896)
It has always been like that. In fact, most of those claiming they want to hire critical thinkers actually are not critical thinkers themselves and do not want to hire any. Critical thinkers have always been rare, they are inconvenient, they tell you that you are doing things wrong. There are very few people that want to hire such people and these are critical thinkers themselves. But these people tend to get not promoted and eventually removed from an organization and hence the circle of mental incapability continues.
And replace them with what? (Score: 5, Insightful)
by evslin ( 612024 ) on Wednesday October 28, 2026 @07:27PM (#66398744)
If you don't know what to replace a coding interview with, then you aren't a "thought leader" as they like to call themselves on LinkedIn - you're just trying to make the rest of the industry dogfood your product for you while your company loses amounts of money equivalent to the GDP of small countries every year.
In the meantime, I still need to know if an applicant knows how to code, because if they get an LLM to write something for them and it's shit, I'm not yelling at their LLM, I'm yelling at them. So, the coding interview stays.