Superscript is hiring for two roles soon. A Lead and a Senior, both roles where experience matters, the Lead significantly more so. By experience, I mean the ability to judge without all the facts in front of you. Knowing which heuristic to reach for. Our existing take-home-tests were old, so I sat down to create some new ones.

The Lead test took some work. I added trap doors drawn from scar tissue. AI models struggled with those. Good.

But I couldn't get the Senior test to hold up. No matter how I shaped it, ChatGPT oneshotted it. Minimal instructions, trap doors behind trap doors, top marks. So we're moving to a follow-up chat for that role instead.

It raised a new question: should candidates be allowed to use AI in the interview? After all, the job expects it. If we run the interview without AI, we're testing a version of the role that doesn't exist.

Have you used or allowed AI during an interview? How did it work out?