Lenny's Podcast: Product | Career | Growth
00:27:44 · Lenny Rachitsky, Adam Mosseri
[00:27:44] S1: Yeah. Mythos has gotten really good at being like, "I can't do this. Let's move on." There's all these limitations. [00:27:50] S0: Claude has always been a little bit of a jerk in a way that I actually appreciate. I appreciate it, I really do, because I don't want one that's just like, "Oh, you're so right. I'm so sorry I said that." It's like, no, no, hold on. I want the real intelligence. I don't want the pleaser. [00:28:08] S1: This point you made about people being excited about the strategy is such an interesting one. There's this idea that I read, I think Cory Doctorow wrote this. There's this kind of concept of a centaur and a reverse centaur. So centaur is a human body horse ... This is going somewhere, I promise. Human body horse ... Sorry, human upper part, horse, lower part. [00:28:28] S0: Horse body. Yeah, yeah, yeah. [00:28:29] S1: Horse body, where the human is in charge. And we prefer that we want to be in charge. Reverse centaur, which is what we want to avoid with AI, is where the AI is controlling us, and we're just doing its bidding. [00:28:40] S0: So it's a horse head on a human body. [00:28:42] S1: Yeah, exactly. [00:28:43] S0: It's terrifying. [00:28:44] S1: So in a sense, Uber drivers and DoorDash people, this is their life, which is not great. And this is the danger, I think, for a lot of people is if it's giving us the strategy and telling us, "Here's what we're building" No one's going to want to do that. So that's a really interesting counterpoint to, we don't want AI to be telling us the strategy, almost.
00:29:02 · Adam Mosseri
[00:29:02] S0: Yeah, no, I think there's a lot of things to be careful about right now. And I would certainly not just assume that because you might be able to outsource some workflow to AI, that you should. There are certain ones where I think it's really just a win-win and there are certain ones where I think the risk outweighs the benefits. Yeah. I mean, you do sometimes find amazing product leaders who are just idea machines, just prolific idea machines. But I do think a lot of the best have taste, have something about them that really makes really strong talent want to work with them, but end up sort of being curators, curators of people, curators of ideas, curators of technologies, curators of strategies. Because I don't really care if I'm hiring a strong lead for an area, if the strategy comes from them or comes from somebody else. I just care that there is an amazing strategy, and everyone has bought into it, and that we're executing against that strategy well. And so I think that some of the best product leaders, yes, have ideas, it's hard to be a great curator if you don't have some of your own ideas, but embrace the reality that they can't come up with everything themselves, and so they need to create an environment in which great ideas bubble up and are chosen or decided upon. And so I think it's not just about curating ideas, but it's sometimes about curating teams and people.
00:32:13 · Lenny Rachitsky, Adam Mosseri
[00:32:13] S1: I love that. I so agree. I feel like everyone's always joining a team, and they just want to do vision strategy, just not actually hands-on work. "No AI's coming in here, let me do the strategy." [00:32:23] S0: Yeah, exactly. [00:32:27] S1: And I love this point that there's so much power and value, and people underestimate just the need for just a really good curator of the team's ideas.
00:32:27 · Lenny Rachitsky, Adam Mosseri
[00:32:27] S1: And I love this point that there's so much power and value, and people underestimate just the need for just a really good curator of the team's ideas. [00:32:35] S0: Yeah. Sometimes it's not just who's good or what idea's good. It's also what is going to work given the broader context. So for instance, on team building, a huge thing that I'm always considering is not just like, is this person a really strong candidate for this role? It is how does this person fit into their leadership team? So for an area like trust and safety, I have an engineering lead, I have a product staff lead, I have a data science lead, I have a design lead, I have a research lead. I need to make sure that those five compliment each other. And that's about what skills each one has, what weaknesses each one might have. I also need to make sure that this is more art than science, have a good vibe. You need trust and rapport. A leadership team with strong trust and rapport can work through most anything. A leadership team without trust or rapport, anything can become an issue. And so that chemistry bit is, like I said, much more art than science. But that also matters. And so I think some of the best leaders and product leaders specifically also either do that instinctively or consciously, but they have a nose for building teams that are going to have good energy and good collaboration. [00:33:59] S1: Warm and fuzzy stuff.
00:33:59 · Lenny Rachitsky, Adam Mosseri
[00:33:59] S1: Warm and fuzzy stuff. [00:34:01] S0: Yeah. Well, the flip side is also true. I've had many times in my career where I've had two people who I think are amazing, and I even adore them and love them, and they just can't get along. And you're like, "This isn't a competency issue. This is just a personality issue," and you just have to just sometimes call it, and split them. [00:34:23] S1: I want to transition to talk about Instagram, the product, the platform, things you guys have learned there. Let me start with this question. What's something that the Instagram algorithm knows about human behavior that people may not realize?
00:34:38 · Adam Mosseri
[00:34:38] S0: One of the most common misconceptions is actually in the opposite direction. I think people assume that there's a much more detailed semantic understanding of everybody's interests and preferences in the algorithm than there is. Most of what's really driven the progress in the world of recommenders over the last 5, 10 years have been these large embedding models and these other techniques that basically produce artifacts that cannot be read by people. They're not legible. They're giant vectors. It's like, sure, I can show you the vector, but it's just going to be a bunch of numbers in a seven-dimensional space. And so when we talk about, does the algorithm know something? Usually we think in these more semantic terms, it knows I like surfing. And it's like, "Eh." It doesn't. It just has this big-ass number that happens to correlate with surfing. That said, I think that is starting to change. I think that one of the things that LLMs are enabling is they can describe in words, English or whatever language you prefer, what some of those previously-illegible artifacts are at least proximate to, if not mean directly. So this is like ... The thing I haven't really ... I posted about this this week, this thing called Your Algorithm. Basically the idea is we take a look at all of the stuff that you've interacted with, and then all of that is in an embedding space. You can think of embedding space as a map. You map a bunch of videos into the same map, and so videos that are close or similar. And now, we can just have an LLM just be like, describe that part of the map. And it can be like, "Oh, that is deep pour over coffee snobbery." And that's kind of amazing.