Think the technical interview is dead? Think again

Estimated reading time: 8 minutes

Key takeaways:

• Traditional algorithmic and LeetCode-style tests are slowly shifting toward assessing how candidates review and debug AI-generated code.
• Smaller companies are rapidly integrating AI into interviews, while major firms remain more cautious.
• Assessment criteria are moving toward system design, technical judgment, and the ability to handle ambiguity within large, unfamiliar codebases using AI assistance.

---

Over the past decade, a burgeoning industry formed around the promise of helping software developers pass technical interviews and nail exhaustive multi-round interviews at desirable, but elusive, tech firms. 

Now with AI reshaping the entire software development industry, the traditional technical interview – heavy on LeetCode style tests and algorithmic questions which test developers’ coding skills and practical knowledge – is becoming redundant. However, the coaching firms who built their reputation helping developers pass these tests aren’t feeling the heat.

Keeping pace

“It’s the most rapid pace of change that I’ve seen in terms of how interviewing is carried out by tech firms,” said Max Serrano, co-founder of coaching platform IGotAnOffer. “If you look back to 2015 to 2024/2025, things had always been the same.” 

Despite AI shaking up an interview format that’s been stagnant for years, this pace of change isn’t on par with the shakeup being observed across the software development industry.

While the coaching firms at the forefront of this shift say that reports of the technical interviews’ death have been greatly exaggerated, they are changing, and maybe not quick enough. 

“We see reports of top engineers saying that they literally spend 0% of their time on code now,” said Sophie Novati, CEO and co-founder of interview prep platform Formation. “And given how fast the role is changing, I would say [technical] interviews are definitely not dead, but [they are] very much evolving and actually a little bit more slowly than I would have expected.”

Some things never change

“We’re still seeing companies ask LeetCode questions,” said Jacob Simon, co-founder of interview prep provider Exponent. “Almost every company is still, today at least, asking those questions. Even AI companies like Nvidia are asking these data structures and algorithms questions. They’re maybe not weighing them as heavily as they were before.” 

Zero out of 52 interviewers who work for FAANG companies said they had moved away from algorithmic questions in a recent Interviewing.io survey. Almost half said that algorithmic interviews will be less prominent within the next two to five years, and over half said they had adjusted the types of algorithmic questions they ask.

Part of the reason algorithmic questions could have staying power is the divide between the C-suite and management. There’s positivity and pressure from the C-suite to adopt AI at pace, but a more cautious brand of optimism at the mid-management level, said Sarah Doughty, VP of talent operations at Talentlab. Managers want to be able to ensure developers have the baseline skills without AI, so they know when AI is veering off course, she said.

Exponent’s Simon suspects companies are also using the take-home test as a baseline to filter out candidates. He firmly believes the take-home test should be dead in the age of AI because of AI’s ability to solve those  problems in minutes. Instead, it’s thriving.  

Bringing a candidate into the company’s environment and sitting with them to solve problems has always been a better way to assess someone’s potential than a take-home test. The reality is that it’s expensive, requires a lot of organization and is time consuming, said Doughty, noting it’s a luxury often afforded to the largest and most profitable companies.

“Not all engineers are really game for it, so it can be difficult inside an organization where you don’t have a decent pool of intermediate to senior independent contributors who genuinely want to do this,” Doughty said. 

“You add to that two or three years of systematic layoffs, I don’t know that you’re getting an employee pool that feels very generous right now with their time,” she added.

There are multiple reasons the technical interview is lagging behind the industry’s adoption of AI. It takes time to recalibrate interviewers and many processes are standardized. Having AI in the loop allows for tests that are wider in scope and more ambiguous, but also require a fundamental adjustment to that process. 

Ambiguity and the change in scope makes it difficult to stack rank developers because it’s no longer a standardized problem and answer, which can be used to make  like-for–like comparisons , Simon said. 

Research from Cornell University and Fujitsu Research of America, which ran a user enactment study with 16 software engineers paired into 12 simulated live coding interviews using AI assistants, found that candidates and interviewers had different interpretations of what it means to productive with AI and a lack of clarity on this in the interview process  often led to misinterpretations, for example, some candidates would use the AI assistant quickly assuming that was a measure of productivity, but the evaluator was actually prioritizing task comprehension and code explanation. 

Startups get the jump

As you’d expect, smaller startup companies are adjusting quicker. “What you see is a lot of startups are changing faster than the bigger companies,” Simon said.

In Interviewing.io’s survey, 67% of the few startup respondents said that AI had meaningfully changed the interview process compared to 0% at FAANG or FAANG-adjacent companies. 

These changes included adding AI-assistance in the interview process as well as getting rid of algorithmic questions and take-home assessments. Ilya Tillis, VP of product at coding test platform Codility, says startups are the customers they are learning the most from because they are experimenting, turning off the proctoring features, which can be used to identify when candidates might be cheating, such as copy/pasting or changing tabs, and telling candidates to use whatever tool they want.

Of the bigger companies, it’s Meta leading the charge, having incorporated an AI-assisted coding stage into its interview process. In a conference presentation, Danit Nativ Navon, an engineering manager in Meta’s AI Infrastructure division, describes how this change allowed Meta to widen the problem scope and introduce candidates to a full code base to help find engineers who thrive in an AI-native environment.

“What really surprised me with the end result was that the assessment criteria didn’t really change,” Navon said. “We still need to measure how engineers do problem exploration, coding, validation, and communication. The only thing that changes is the way we look at it.” 

Meta is looking for code that is easy to maintain and validate, said Navon, noting developers shouldn’t assume the code is working, but question the output from the AI assistant.

“Speaking to some of the engineers that are now conducting these interviews, the amount of training that they’re getting on conducting these is shocking to me,” said Novati, who worked at Meta alongside her co-founder Michael Novati, who was referred to as the “coding machine” internally. 

Some companies, even the AI company Anthropic, are opting for no AI and want to test raw coding abilities. Firms that haven’t adopted AI in interviewing are still watching very closely, Tillis said. Top priorities include seeing how legislation evolves, ensuring interviews can be defensible and repeatable, as well as being in a place where the process can be designed to ensure there will be some longevity, he said.

What coaching firms are advising in the AI age

For the firms tweaking their interview process, the priority is reviewing, debugging and explaining the code written by AI. System design and judgment of technical trade-offs are increasingly important, Novati said.

The questions themselves are becoming much more ambiguous, which is a welcome change, as it’s aligning the technical interview more closely with the realities of the job. For example, Cursor’s final onsite interview involves giving candidates their codebase and telling them to spend a day or two finding something useful to fix and then present what they did, Simon said.

Candidates can now be asked, with the help of AI, to refactor a codebase as well as implement new features, or remediate security issues in an unfamiliar codebase, Tillis said. “Those are all real on-the-job responsibilities that you can do in an assessment that may not have been possible two, three years ago in 30 minutes,” he said.

As the interview process evolves, so do the coaching firms. Up until now, Formation has really focused on helping developers repackage their skills in a way that accurately reflects their skill level in a high-pressure environment, but now it’s also about ensuring developers can meet the moment by including AI-assisted coding and AI in a system design context as part of the coaching process, and developing training for those going from zero to one with AI.

“The number of people interested in AI training exceeds the number of people who are just interview prepping,” Novati said. “That’s one of the things that we’re looking at.”

To help developers make sense of the new landscape, Exponent is rolling out a living database where they can “get a look behind the curtain” of what’s happening in the interview process. With an uptick in developers interested in roles at AI-native companies, they are keeping a close eye on those companies’ processes.

“We need to now follow that trend, and so that means we need to create content about OpenAI and Anthropic,” Serrano said. “Our coaches need to be able to coach for these companies, so those are the main changes we’ve made.”

More disruption ahead

There might even be more demand, rather than less, for coaching firms in the age of AI, as AI levels the playing field when it comes to coding, and more roles may incorporate a form of technical interview. Serrano highlights how product managers now face a new interview round at Meta titled ‘product sense with AI.’

“We’re certainly seeing a mandate of assessing AI skills not only for engineers, but for everybody, increasingly coming from CIOs of various organizations,” said Natalia Panowicz, CEO of Codility. 

“An organization making a pledge to do AI transformation, then the next question is, ‘Do my people have the right skills? Can everybody be effective with AI tooling?’ And that spans across all the roles, and it starts actually to translate into non-technical roles and how these roles are being hired for,” she said.

As job titles change, technical interviews will likely continue to change as a result, Simon said. “There’s more to come because the models are still getting better,” he said.

Get articles like this in your inbox

Choose your LeadDev newsletters to subscribe to.