AI, Interns, and the Changing Face of Frontend Hiring

Over the last five years, we've had many final-year college students join us as frontend interns.

We used various methods to shortlist candidates in the first round: take-home assignments, online tools to build a small UI in a few hours, and even pen-and-paper assignments (yes, you heard that right).

All of this worked well over the last few years. However, I wanted to move on from pen-and-paper assignments, as they feel outdated and make evaluation difficult.

I asked the talent acquisition team to assign a take-home project, give candidates two days to build the UI, and have them share the code via GitHub and host it online.

The assignments typically involve building something like a small calendar or a multi-step form.

Everything went well—until this year.

This year, every assignment appeared to be AI-generated. I'm not against students using AI to build their assignments.

The first problem is that, from each college, nearly 80–100 students take up this assignment. The majority of the assignments look the same: the code is readable and neat, so we end up shortlisting most of them.

The second problem arose when I conducted three interviews. I deliberately spent 15–30 minutes with each candidate to understand their knowledge of the assignments they had built.

I was disappointed with all three interviews, as none of the candidates could answer basic questions—questions essential for building those impressive assignments.

There are a few ways to tackle this.

Use an online interview platform with plagiarism checks to filter out some candidates. But this discourages the use of AI tools, which I don't want.
Look at the students' GitHub profiles or what they've built. This is a solid step, but so far, I haven't seen many quality web apps from students.
Go back to the good old pen-and-paper assignment—which feels outdated!

AI coding tools are definitely improving the productivity of experienced developers. But I feel newcomers are at a disadvantage. They get caught up in "vibe-coding" and build things without realizing how everything works under the hood.