I recently interviewed at RB Global for a Sr Full Stack developer role and my experience exposed an interesting problem with how companies are interviewing developers in the AI era.
Before I get into it, I should acknowledge that the individuals I spoke with were pleasant to talk to, and I got the sense that they had a lot on their plate. Two were senior managers and my second-round technical interview was with a Principal Architect. I have nothing personal against any of them and this definitely isn’t a hate letter of any kind.
When I spoke to the senior managers, they grilled me on my experience — from working on IE6 back in the day to legacy migrations, hooks, memory and performance optimizations, leading about 15 frontend developers, and my day-to-day working style. Inevitably, the topic of AI came up.
As a disclaimer: I’m hardly an AI skeptic. It has become a substantial part of my development workflow; my Cursor usage alone approached one billion tokens in August.
I mentioned my situationship with AI to the senior managers and the conversations went well enough to lead to a second-round technical interview.
The task was simple: I was given an hour to solve any or all of the three problems outlined in the assessment repo’s README.md. I was to work through the solution live with the Principal Architect. I was allowed to use AI, but wasn’t given any indication that demonstrating an AI-assisted workflow was part of the assessment, so I chose not to use it to solve the problem.
I spent the hour working through the problem while talking through my reasoning. At no point during the hour was I nudged or asked to use AI, much less demonstrate an AI-assisted planning or development workflow. The Principal Architect mentioned that the assessment contained roughly five hours of work. After spending the first 10 minutes or so recalling some of the specifics of the SSR problem I’d chosen, I found my rhythm and wrapped up around the hour mark. At most, I used AI or Google for an isolated lookup, but I didn’t use AI to plan, implement, or reason through the solution. Given the format of the interview, I understood the goal to be demonstrating how I approached and reasoned through a technical problem.
I didn’t get the job. The feedback I received was that I hadn’t used AI during the assessment.
During the technical round, there was no obvious attempt to understand how I use AI, what I delegate to it, where I don’t trust it, or how I validate its output. The choice to use AI was left to me, and I chose what seemed like the more useful path for this particular format: solving the problem myself while pair programming with the interviewer. I was ultimately penalized for that choice.
Pair programming demonstrates one set of capabilities. AI-assisted development demonstrates another. If AI usage is part of the evaluation, say so.
Like with AI, it’s best to be explicit about what you require; otherwise, you end up with slop.