How to assess SWE in times of remote and AI?

Last time I hired SWEs, remote work wasn’t a thing, AI didn’t exist, and interviews were in-person.

What’s your go-to approach for assessing remote engineers? Take-home assignments? Live coding?

Do you try to separate AI-generated work from the candidate’s genuine work, or do you just care about the final output?

I thought about a take-home assignment and ask candidates to reflect / document how they used AI, followed by a live discussion to gauge their understanding of their own solution.