What matters when building AI products

It's easy to look at AI and think of it as a black box, right? Either it figures it out or it doesn't. And I don't have too much control over it because, like, you just don't exactly understand how it works or it's not deterministic. And even myself, like, having been new to building these kinds of features, it's really easy to kind of feel intimidated or feel like it's not really my job to debug what's happening in the box, right? But actually, like, if you can take a much more proactive role in owning quality and saying, hey, this is not good enough, or actually, this is the kind of output that we want, it becomes an exercise, even with your engineers around, like, well, what are the prompts that could be, like, designed to, like, better produce that? Right? And just really understanding what are we feeding and what's coming back and what would we prefer to come back? How can we exercise more creativity around, like, what we're feeding in the first place? So I think there's just kind of this journey of realizing how to get involved in a way that I think a lot of our PMs went through as part of this process, including myself.