Reading this, anyone give great advice on formulating take home projects that work really well in the AI age? I ran an interview cycle where I gave ~1 day, and tried to hold out some technical questions for a follow up design interview after they had delivered the take home. I felt like the project was very hard to define given AI lets people chew through most of what you throw in front of them very quickly. Giving a bigger project felt like it required drawing them too deeply into our domain, where I was trying to provide a slice of it that we could onboard candidates to in ~45 minutes. Given that the authors here are getting multiple weeks to do it, its hard for me to imagine an appropriate scope.

Give them a take home and then talk about its implementation details? If a machine did the work they won't understand it.

For me personally, the goal of a take-home was always to see how the candidate walks through code, discusses code, and how well they take feedback.

This is why I'd ask candidates for previously written code they were proud of, rather than telling them to code something trivial.

I spent years working on financial integrity at a large ads company and this isn't novel at all! The same resale markets are at play for the last generation of internet giant's products. Highly sophisticated actors, able to cobble together impressions through abuse of the billing systems, stolen financial instruments, taken over accounts, etc, create massive markets of discounted impressions for resale. It was very interesting to compete against them as we hardened our defenses and they invented new ways to exploit them. I imagine the same defensive tools and techniques are being deployed by my former colleagues who moved to the labs.

Yes, and I'm reminded of Bitcoin miners relocating for subsidized electricity or even stealing it outright.