How would you approach building proactive AI?

"I got into the process through a referral, but DeepMind recruiting was separate enough that I had to connect directly rather than just apply through the normal Google portal. The process started with a mostly informational recruiter conversation, then two 30-minute hiring manager intros that felt more like early team matching than screening. After that, I had a four-interview final loop, and despite the different titles, most of the rounds were really just AI product cases with some overlap, especially around how I would think about proactivity in Gemini. The weird part was how little the interview titles matched what actually happened, and even the recruiter admitted they were still figuring out a new recruiting system. My biggest takeaway is that this loop felt much less standardized than places like Meta, and they wanted me to go all the way to an actual proposed solution and walk through the UX, not just frame the problem well."