Lenny's Podcast: Product | Career | Growth
00:33:32 · Lenny Rachitsky, Peter Sellis
[00:33:32] S0: I keep trying to move away from Snapchat, but new questions come up as you share more. Something that came up in my chat with Evan about this is seeing everybody copying Snapchat for 15 years, or however long it's been, his takeaway basically is software is no longer remote, you have to focus on other things. He described that's why he's spending so much time on hardware was his answer. And also just the ecosystem and the network effects of a product. Thoughts on that? [00:34:00] S1: I have tremendous respect for Evan, and I think he's essentially been right on almost everything, just early on some stuff. Maybe early on everything actually too. Even stuff like avatars, AI, essentially all of the on-device AR that we were doing was, at the time, it was more just called machine learning, but employed some form of machine intelligence that was super avant garde at the time. So, I think when he says something like this, there's probably some truth to it. I think a lot of people are frustrated with Snap because it doesn't have the financial profile to be able to support the type of hardware development that he wants to do, it seems, and that there still isn't a great answer on the monetization front, for which obviously I bear a lot of responsibility. But from that kind of moat perspective, like what he said, I watched the pod, yeah, I think he's right, but that maybe investors or the world doesn't want to give him the space to be right in some ways. [00:35:08] S0: It's challenging.
00:35:08 · Lenny Rachitsky, Peter Sellis
[00:35:08] S0: It's challenging. [00:35:09] S1: Yeah, yeah. It has artist vibes, rejected artist vibes, or on the outskirts artist vibes, which I think there's still so many things... It actually reminds me, there's so much talk these days about taste, and things like that, and so hard to describe things like that. But one of the things that I learned at Snap is I think I saw Evan personally say no to better ideas than I've seen almost any other consumer startup launch. And it kind of made me think of this, my goal as a product leader for founders at these high growth companies was always to turn them into almost the best museum, the richest museum curators in the world, where whatever's in the product at any one time is just a small slice of the overall collection. It's whatever that curator thinks is right for this time and place in society. And I think if one way you can get at somebody's taste is when you're interviewing them is just ask them about some of the things they built that they never shipped. Not that they tested and got bad results for, but that literally they held it in their hands, the prototype, and they're like, the world's not ready for this, or this thing is not ready. And if they have a lot of high quality ideas that are on the shelf, exercising that muscle of saying no, of restraint, is probably one of the easiest... One of the few muscles of taste you can exercise. If a museum has all of their collection on the walls, then the curator hasn't done anything. And so, yeah, I think about this a lot because Evan is probably the best at it.
00:36:51 · Lenny Rachitsky, Peter Sellis
[00:36:51] S0: That is such an interesting insight, especially these days because AI is very bad at removing things from designs and from products, it always adds, it doesn't say, "Hey, maybe we should cut this piece." And so, it's so interesting how this connects now to the... We talk about taste all the time with AI, and this is an interesting element of that, of being good at understanding what to remove as AI becomes a bigger part of the building process and designing process. [00:37:20] S1: Yes. Yeah. This will be hard, I think, this is a hard skill for AI to replicate. I think in general, it seems like right now it's obviously at the code level, it's getting to the design level, in some cases at the video and the photo level, it'll eventually maybe even get to physical level as well. But it's really going to empower anybody to be a creator, and we've seen this in the past. Lowering the threshold for creativity, not just enables the existing creators to do more, but it enables people who were otherwise bounded out or restricted from entering the arena to enter it. But this will put a lot of pressure on essentially creating stuff that deserves human attention in a way that I wonder if there are some analogs to this in the past. We'll have to think about it. [00:38:22] S0: And it just feels like an interesting new skill for PMs is pushing back on things, AI is adding to your products, engineers are like, "What if we had this?" So, it's not even on AI right now, it's on the product teams basically to cut stuff. Anyway.
00:38:22 · Lenny Rachitsky, Peter Sellis
[00:38:22] S0: And it just feels like an interesting new skill for PMs is pushing back on things, AI is adding to your products, engineers are like, "What if we had this?" So, it's not even on AI right now, it's on the product teams basically to cut stuff. Anyway. [00:38:36] S1: Yeah. Yeah. Well, it'd be pretty funny if the conclusion of every Lenny podcast is basically, yeah, the world might just need more PMs. [00:38:43] S0: Yeah.
00:38:45 · Peter Sellis
[00:38:45] S1: Crazy. I subscribe to the Eric Seufert side of the world, of ads are the product market fit for the internet. And I still think that from a monetization standpoint, there's a number of reasons, again, mathematical, for why that is true and why I think subscription products and things like that are overrated versus ads from a consumer side of things. When I look at what OpenAI is doing from an advertising perspective, they are speed running all of the ads infra that is exactly what you would want and do. At Discord, I call this last mover advantage because we've learned so much and we're at a certain place with online advertising... And really Meta and the team there from 2010 through 2020 really deserves, I think, most of the credit for this. What drives a modern ads marketplace and auction is essentially a solved problem, and OpenAI is speed running all of the components of that. And I do think that the mechanism, the mathematical mechanism, which is essentially a modified VCG, I think Vickrey-Clarke-Groves auction mechanism, this mathematical mechanism is the correct way to build an ads auction in even a chat environment, in ChatGPT. And what I mean by that is I think where the innovation came from Meta was not just that the highest bidder, and I'm oversimplifying like crazy, but not just that the highest bidder wins, but that the organic value of an advertisement and an advertiser to that user should be factored into their bid, such that you essentially have to pay more as an advertiser to reach people who may not want to be reached, or may be harder to reach, or may not necessarily be as relevant to you, and this also is great because it allows you to mox both content, organic content and advertising content, on a unified auction with dollars as essentially the ultimate denominator. And the reason that I say that this is good for chat is not necessarily that organic value is what's important here, like how much you like an ad or not in a content feed, but in chat, it probably feels like this could be replaced or augmented with essentially trust. If you boil this down to something very simple, it's just like, "Hey, what X should I buy? What mic should I buy?" And in that moment, we have to decide should I give you the objective answer? Should I give you a set of answers?
00:42:24 · Peter Sellis
And how influenced should I be by a set of advertisers that want to get in front of you? And I think the mechanism by which how influenced I should be should essentially be how much trust I'm trading off. And you see us failing at this a little bit probably... And I really don't want to pick on anybody because I know how hard it is to build these types of things. But a little bit on Amazon search, in terms of the ads, the sponsored experience there, it's not threatening the product, Amazon is such a great product, but what is the great product of Amazon? It's the selection and the next day shipping or whatever. It is getting a little crazy when you search for something and you have 16 ads before you get to the first thing, and you're like, "Oh shoot, is that subtly influencing my trust in the system?" And so, that's why I actually think OpenAI is speed running all the right stuff, and where the push is going to come to shove is how they instantiate it mathematically in their auction mechanism to maximize trust over the long term, because trust maximization will probably be correlated with user retention, if not causal, and with advertiser value. And then on the other side of it, it's probably a format question. What does this format look like in chat? What if it's voice? We don't know. So, I think one of the coolest jobs right now for a PM and designer would probably be ad formats at OpenAI. That's a societal level job in advertising right now. And so, yeah, I think it'll be really interesting to see how they work on that. Yeah, maybe one other thing I'll say is I was disappointed that they jumped right to an ad-free experience for the top tier, because I feel like ads could actually be really additive to every part of the OpenAI, not like coding and things like that, but actual chat. Like, let's set aside work and coding, and lopping off your most valuable users so early. I was a little disappointed, but I understand why people do that.