Hiring Software Developers in the Age of AI: New Challenges

I’ve interviewed hundreds of engineers over the years through technical screens, system design rounds, take-home projects, and pair programming sessions. A strong performance in the coding portion of an interview used to mean something clear: this person can think through a problem, hold context in their head, and write working code. That confidence translated directly into a hiring decision.

The last year has been the most challenging stretch of my career as an interviewer and, surprisingly, has nothing to do with the number of applicants. Since my aha/Eureka moment with generative AI in August 2025, I’ve felt my perspective shift. The signals I used to trust have been inverted. I find myself watching where a candidate’s eyes are moving. Research on eye patterns during active thinking suggests people look up and to one side when retrieving information from memory, and down or straight ahead when constructing something new. When someone is staring slightly off to the side of their screen, or their eyes are tracking somewhere outside the shared window, it catches my attention in a way it never used to. Someone typing quickly without the cursor following their keystrokes. Eyes drifting to a second monitor that isn’t being shared. Pauses that feel less like thinking and more like reading. It’s a strange position to be in.

I now ask every candidate at the start of the interview not to use any AI assistance, and most agree without hesitation. And then I spend the next hour looking for signs that they might be doing it anyway. There are tools out now that will listen to an interview in real time, interpret the question being asked, and surface suggested answers that the candidate can read and type out themselves. They run silently in the background. You’d never know they were there unless you knew what to look for. The act of using them isn’t obvious. It can look like careful thinking, deliberate pacing, and clean code on the first pass.

And that’s the uncomfortable part: someone who writes correct, well-structured code on the first try used to be a strong signal. Now it makes me pause and I wish that weren’t true. It’s an unfair burden to place on genuinely talented engineers who just happen to work cleanly. But the dynamic has changed, and pretending otherwise doesn’t help anyone.

Here’s the irony: the candidates I feel most confident about lately are the ones who struggle a bit visibly. The ones who say the wrong thing first and then correct themselves. The ones who ask a clarifying question before diving in, get partway through a solution, realize it won’t work, and have to rethink it out loud. That kind of messy, human problem-solving process is reassuring now in a way it wasn’t before. It’s a signal that they’re working through the problem themselves.

The way we assess candidates was built for a world where a coding screen was a reliable proxy for engineering ability. That assumption is shakier than it used to be. Hiring managers and interviewers are going to need new frameworks - not just for spotting misuse, but for designing interviews that are genuinely harder to game while still being fair. More focus on reasoning through ambiguity, explaining tradeoffs, discussing decisions made in past work, and adapting in real time when requirements change. Things that are harder to answer from a prompt.

There will always be people who look for the loophole. That’s not new. Every system has workarounds, and there have always been candidates who over-prepare for the specific format of a technical screen without having the underlying skills. As the tools improve, the ways to misuse them will improve too. It’s an ongoing adaptation, not a problem to be solved once and done. What I keep coming back to is that none of this changes what we’re actually trying to find. The goal has always been to identify people who can think, communicate, grow, and contribute to a team. Coding was one way to observe that. It’s a less reliable signal now, but the underlying thing we’re looking for hasn’t changed.

The people are what make a company great. Our job as interviewers is to find those people who have the right skills and know how to use the right tools. That job just got harder.