Are we doing more spicy takes? I'll give you one: this is just not a thing. Unless you're operating at a very large scale, I think you can simply ignore this problem.

If your team is at all well run, and your work-sample testing mirrors the team's work, it's not a very good scam to run: it'll be immediately obvious on the job.

We thought we'd put some energy into this back in 2021, and we have yet to reach a point where thinking about this makes any sense (except for tech writers, where spotting GPT is maybe somewhat useful).

You’re not but it already reduces the sample size by a lot. There is no 100% effective way to detect liars. I’ve been a hiring manager in multiple companies (including currently) and fortunately for me they were so far rare (maybe because Germany isn’t attractive enough as a job market for programmers to get the bulk of people putting a lot of effort into fraudulently getting a job).

Either way it is customary to start new hires with a 6 month probation, and their long term contract only kicks in after that. That’s long enough to detect if someone is incompetent (I’ve had to fire people during their probation when finding out they weren’t actually capable altho I don’t think they outright lied).

The one place I interviewed that did this had me email my program for them to run, then come in and answer questions about it and describe which parts I was proud of, which could use improvement, etc.

Just have people pair with people doing actual work. It's better signal for the candidate, for the existing team, and for the organization.

maybe a balance can be struck? audition/take-home to select potential candidates, a follow up interview with each to confirm they wrote the code and are a good team fit.