Re-thinking candidate take-homes in the AI Era: transcripts over code

This totally aligns with my recent thinking, but there's a real risk in the model: If I'm yelling at Claude about something while I know the transcript is part of the output, I'm gonna naturally behave differently so I don't come off as some sort of asshole, so it's not the real ksmith's thinking. Nuanced, but it matters.

OTOH, an interview is basically the same thing: Here's me presenting myself as well as I can (even though it's probably not really accurate). But OTOOH, maybe I don't give a shit and just continue talking to Claude like I would do in real life, so who cares? It's not like HR is gonna pull me in for a talk about how I treat my teammates... yet.