Lenny's Podcast: Product | Career | Growth - How to win in the AI era: Ship a feature every week, embrace technical debt, ruthlessly cut scope, and create magic your competitors can't copy | Gaurav Misra (CEO and co-founder of Captions)

Episode Date: March 27, 2025

Gaurav Misra is the co-founder and CEO of Captions, an AI-powered video creation company and one of the most successful consumer AI products in the world today. Previously he was a product leader at Snap, where he created the design engineering function and spent years helping develop features used by hundreds of millions of users worldwide. With a background in both engineering and design, Gaurav brings a unique cross-functional perspective to product development.What you’ll learn:1. Why the “ship a marketable feature every week” approach helps his team stay focused and the product stay top of mind for users amid constant AI breakthroughs2. How to balance rapid shipping with maintaining quality by cutting scope rather than compromising on timelines3. The “secret roadmap” strategy that helps Captions develop breakthrough features competitors never see coming4. Why taking on strategic technical debt is essential for startups to outpace larger companies5. How Captions accidentally ignored their most successful product for 1.5 years (and why it still grew to 500K users with no updates or support)6. How Snap’s unique product development approach—with designers functioning as PMs—enabled their success as the last major social network to break through7. Why AI video will transform marketing before other industries—Brought to you by:• Brex — The banking solution for startups• Paragon—Ship every SaaS integration your customers want• Coda—The all-in-one collaborative workspace—Find the transcript at: https://www.lennysnewsletter.com/p/how-to-win-in-the-ai-era-gaurav-misra—Where to find Gaurav Misra:• X: https://x.com/gmharhar• LinkedIn: https://www.linkedin.com/in/gamisra1/—Where to find Lenny:• Newsletter: https://www.lennysnewsletter.com• X: https://twitter.com/lennysan• LinkedIn: https://www.linkedin.com/in/lennyrachitsky/—In this episode, we cover:(00:00) Gaurav’s background(04:47) The exciting era of AI and startups(09:30) Staying top of mind(11:26) Tips for staying focused(13:14) Shipping marketable features weekly(19:03) Managing technical debt in startups(25:31) Snap’s unique product development approach(32:09) Brainstorming with AI(35:09) What Snap got right(41:06) Scaling with a small, agile team(49:33) The shift toward prototyping in product management(51:47) The product manager role(55:40) Snap’s mission and product decisions(01:02:13) The future of AI-generated video(01:10:20) Leveraging AI for marketing(01:14:37) Failure corner(01:20:21) Lightning round and closing thoughts—Referenced:• Snap: https://www.snap.com/• Captions: https://www.captions.ai/• Iron Man on Disney+: https://www.disneyplus.com/movies/iron-man/6aM2a8mZATiu• J.A.R.V.I.S.: https://en.wikipedia.org/wiki/J.A.R.V.I.S.• Cursor: https://www.cursor.com/• Devin: https://devin.ai/• Eye contact: https://www.captions.ai/eye-contact• Nvidia: https://www.nvidia.com• Descript: https://www.descript.com• Evan Spiegel on LinkedIn: https://www.linkedin.com/in/evan-spiegel-8ab74034a/• TikTok: https://www.tiktok.com/• Spotlight: https://www.snapchat.com/spotlight/• Building product at Stripe: craft, metrics, and customer obsession | Jeff Weinstein (Product lead): https://www.lennysnewsletter.com/p/building-product-at-stripe-jeff-weinstein• Patrick Collison on X: https://x.com/patrickc• DeepSeek: https://www.deepseek.com/• ByteDance Goku: New video generation AI model, better than OpenAI Sora: https://medium.com/data-science-in-your-pocket/bytedance-goku-new-video-generation-ai-model-better-than-openai-sora-56c017a320a5• Will Smith eating spaghetti and other weird AI benchmarks that took off in 2024: https://techcrunch.com/2024/12/31/will-smith-eating-spaghetti-and-other-weird-ai-benchmarks-that-took-off-in-2024/• Silo on AppleTV+: https://tv.apple.com/us/show/silo/umc.cmc.3yksgc857px0k0rqe5zd4jice• Severance on AppleTV+: https://tv.apple.com/us/show/severance/umc.cmc.1srk2goyh2q2zdxcx605w8vtx• Linear: https://linear.app/• Superhuman: https://superhuman.com/• Notion: https://www.notion.com• Perplexity: https://www.perplexity.ai/• OmniHuman-1 AI Video Generation Looks Too Real: https://www.youtube.com/watch?v=fY0KB516m-E—Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email [e-mail].—Lenny may be an investor in the companies discussed. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lennysnewsletter.com/subscribe

There's rarely a time like this where so much as possible. Even like five, seven years ago, it's so hard to start a company. Everything feels like it's done. Someone else is working on it. Suddenly, it's a time right now, which I've never even experienced, where everything you try just works. With people constantly hearing about all the things happening, is there any tools or processes or approaches you've figured out to help stay focused?

Our engineering goal is every engineer should ship a marketable product every week. I love just how wild that sounds. How do you maintain quality? and make it all cohesive. I actually think as a startup, your job is to take on technical debt because that is how you operate faster than a bigger company.

Bigger companies don't take on technical debt. They pay it usually right away or they're paying back technical debt from the days when they were a startup. Is there anything else in how you operate and the way you build product that you think is really unique and interesting? We have what we think of as the public roadmap.

This is basically what people have asked us for. There's all these surface areas we receive user feedback. But these are all features that every competitor knows about. If a user is asking us for it, they're asking everybody for it. It's not going to be a game changer in terms of winning against your competition. So we have a second roadmap, which we think of as a secret roadmap.

Today, my guest is Gorov Misra. Gorov was an early employee at Snap, where he led the design engineering team, which he explains in the conversation. He was also an engineer at Microsoft and a couple other companies. Most recently, he's the co-founder and CEO of Captions, one of the most successful and cutting-edge consumer AI products, which lets you generate and edit talking videos with AI. They have over 10 million users and have raised over $100 million. In our conversation, we essentially do an archaeology of how a modern AI-oriented startup

operates, including how every single engineer at their company ships a marketable product or feature every single week, why they have a secret roadmap in addition to a regular roadmap. We also get in depth about how Snap, as a product team operated, what he's learned about what it takes to build a successful consumer and social app, why they had no PMs, and how designers ran the show, which may or may not have been a great idea. And also what happens in a world where AI video is so good that you have no idea if it's real or not. This episode is for anyone that is building a product on top of AI. If you enjoy this podcast, don't forget to subscribe and follow it in your favorite podcasting app. And also, I just launched an insane deal for subscribers of my newsletter.

every yearly subscriber now gets a year free of notion, perplexity, superhuman, linear, and granola. Learn more at Lenny's newsletter.com. With that, I bring you Gorav Misra. This episode is brought to you by Brex, the financial stack used by one in every three U.S. venture-backed startups. Brex knows that nearly 40% of startups fail because they run out of cash. So they built a banking experience that focused on. on helping founders get more from every dollar. It's a stark difference from traditional banking options

that leave a startup's cash sitting idle while chipping away at it with fees. To help boundaries protect cash and extend runway, Brex combined the best things about checking, treasury, and FDIC insurance in one powerhouse account. You can send and receive money worldwide at lightning speed. You can get 20X the standard FDIC protection through program banks, and you can earn industry-leading yield from your first dollar, while still being able to access your funds anytime.

To learn more, check out Brex at brex.com slash banking-solutions. That's BREX.com slash banking solutions. This episode is brought to you by Paragon, the integration infrastructure for B-2B SaaS companies. Is AI on your 2025 product roadmap, whether you need to enable RAG with your user's external data, like Google Drive files, Gong transcripts, or DIR tickets,

or build AI agents that can automate work across your user's other tools, integrations are the foundation. But building all these integrations in-house will cost you years of engineering, time you don't have, given the fast pace of AI. That's where Paragon's all-in-one integration platform comes in. Build scalable workflows to ingest all of your user's external data

into your RAG pipelines and leverage Action Kit, their latest product, to instantly give your AI agents access to over 100 integrations and thousands of third-party actions with a single API call. Leading AI companies like AI21, u.com, 11x, and coffee.a.i are already shipping new integration seven times faster with Paragon, keeping their engineers focused on core product development.

Ready to accelerate your AI roadmap this year, visit useparagon.com slash Lenny to get a free MVP of your next product integration. Gorov, thank you so much for being here and welcome to the podcast. Thank you. Thanks for having me. Excited. I very rarely have early stage founders on the podcast, but I wanted to chat with you because you're at the center of so much of what is top of mind for a lot of builders these days. AI and video and just consumer and social apps. Also going viral and finding new marketing channels. So I think there's a lot that people can learn from the way you approach product, the way you've built product, and the way you just think about where things are going.

So again, thank you for being here. Appreciate it. Honestly, it's an exciting time. Like, I got to say, like, there's rarely a time like this where, you know, where so much is possible. Like, in normal times, if you think about, like, even like five, seven years ago, it's so hard to start a company. It's so hard to come up with an idea, right?

Like, it's just like everything feels like it's done. Someone else is working on it. Right? Like, oh, it's been tried three times and failed three times. And, like, suddenly it's a time right now, which, like, I've never even experienced, honestly in my career, right, where everything you try just works. There's so many possibilities. There's not enough people in the world to work on them. Like honestly, right? Like, there's more things that can be done than there's people available to do them, right? It's just such a rare

thing. And honestly, it's not going to last forever, right? Like, we are going to catch up to this, but just feels lucky to be, you know, part of that movement. It's awesome. When you said everything is working, I think what's an important distinction there is like the building. The building of the tool works, like the tech is now there to build all these things that have not been possible before. The thing that is increasingly difficult, and I want to get your take on this, is getting anyone to pay attention and stick with your thing because it's so easy to build stuff and everything is just awesome and interesting. It's harder to get people to pay attention and stick with your product. So I guess is there anything there you've learned? You've built

a number of successful products. We'll talk about Snap and what you're doing now, about just, I don't know, what you need to think about these days to get anyone to pay attention and then stick around. Yeah. I mean, honestly, it's a great point. And I think there is a lot of hype, obviously, right? And part of it, that's kind of what's driving a lot of this growth for a lot of companies, right? And I think from like user acquisition such marketing perspective, right, like in a world,

five or seven years ago, if you were making something novel and you went to users and you was like, oh, we got something better, right? Like, people are going to be like, well, whatever. everybody says they got something better, right? I don't care. Right. But today, and this is not, you know, probably the way you should do it, but you can go and just say like, we've rethought this thing with AI, right? And a bunch of people will just be like, well, how or maybe I should check this out, right? They'll just try it. Obviously, you have to deliver on the promises, right? Like, if you don't deliver, people will come in, they'll, like, play around a bunch and then just leave, right? But if you can truly

deliver on the promises, you know, there's great opportunities to require. users at scale. So I think that's slightly different. And I don't know how long that lasts, but it is definitely a different time from that perspective. I do think also, you know, at the core of building products is solving problems. I think a lot of people sort of get caught up in this, you know, well, it's cool. And people will come for the cool right now. People will come in and be like, well, let me check it out. It's cool. But at the end of the day, like, if you're just building a playground and people play around in the playground and then to leave after, you're playing around. It's not a business, right? So I think that is still key, right? You have to be

solving real problems. As you were talking, I'm thinking about like every day there's something, like something that would maybe a few years ago be like news for a year. Holy shit, this is now possible. Now it's like every day. Something like that happens. And then we're like, all right. So what I think about is like, we'll have AGI one of these days or super intelligence and everyone's amazing. And then, okay, what's for dinner? Isn't that already happening? Think about like in a way, like I kind of self-reflect on this sometimes of like you've seen Iron Man and stuff like they have the Jarvis thing and you've seen like interstellar

and they have like the Tars machine right and they're talking back and forth with these things like bouncing ideas and that is science fiction. That's literally science fiction. Okay, it's not perfect but it exists in a way that nobody could have imagined. That's science fiction has become reality. and I feel like nobody cares, right? In a way, like, you would have expected the world to be turned upside down, but it feels like almost in a way so slow

and people are like, yes, there's adoption is happening, but like I feel like it's almost a shocking development in a way. It feels like you guys have done a good job staying top of mind and continuing to get people excited because to your point, there's so much happening, how do you get people to continue to be like, oh, okay, wow, what their building is actually is interesting and continues to be interesting?

anything you've learned about just what it takes to stay top of mind and continue to pull people back and get people re-excited over and over? 100%. I mean, I think honestly, it just comes down to like not just AI for the sake of AI or AI for the sake of excitement or hype or novelty or whatever that is, right? It's actually effective AI, like AI that solves real problems, practical problems, right? And, you know, the fundamentals haven't changed, right? In a way, there's three steps to building products.

you identify a user problem, you apply some technology to solve that problem, but then finally, you have some mechanism to find people who have that problem, right? If you can do all three of those things, then in any environment, you can create great products. But I think right now what's different is so much is changing on the technology side that you can create products that could not have been created before and solve problems that could not have been solved before. And that's what's creating the opportunity, right? And for us, especially in the video space, it's truly endless, right? Like, we've just begun.

And like, our goal specifically for video is not to build professional tools, right? We're not building for professionals at all. We're building for the person who could not have created video before, right? They didn't have the tools, the skills, the means to be able to create video and now they can't because they're able to jump over that skill gap, right? Or that time gap. maybe they're business owners. They don't have time, right?

They want results. And honestly, a lot to solve there, right? Just tons. Solve people's problems. We've easier said than done, but it's a good reminder. Like in the end, that's all that matters. Something that I always think about with people in your shoes is just how do you not get overwhelmed and how do you know what to pay attention to you?

How do you stay focused? Any tips there for folks that are just reading every day a new announcement? and then just like, I just, how do I? What do I do? There's too much. It is the new problem of product development in a way, right? There's too many possible paths you can go down. There's too many ideas. There's too many, you know, things you could do. And I mean, obviously, prioritization is always an important skill set and has always been. But it's become an even more important skill set right now because you have to figure out what not to pay attention to. Our general framework for it is to look for user demand, right? And actually the easiest way to

check for user demand is to just see what has a firearm. Right. Usually what has virality, what people want to share and talk about, there's something at the core of it that actually is interesting. Now, it may not always be interesting in a way that's like, you know, maybe it's not, maybe it's a one-time use case. Maybe it's not something that people would do repeatedly. Maybe it's not something you build like a subscription business off of. But oftentimes there's something, some core sort of element of it that has resonated with people. And if you can identify that core and then, you know, mold it into fitting into your business. it's actually a great way to identify like what actually works. And we have these tools right now. We don't have to build anything. You can just kind of talk about it. And people will share it, share the idea, right? And you can kind of measure how well the product might be received even before you

build anything. Right. So it's a great tool we use for like prioritization. We spend a lot of time on social media. Obviously, like our app is often used for social media. So a lot of our employees will spend a lot of time on social media. We look at what the trends are, what's happening. And based on that, we can get a pretty good read of what might resonate well with people.

So as a leader of a company with people constantly hearing about all the things happening, is there any tools or processes or approaches you've figured out to help people continue, like stay focused, not get excited about every shiny object and, you know, actually ship things? I mean, honestly, it's all about incrementality in a way, right? Like, I think we do aim to ship every week, right? So like our engineering goal is every engineer should ship a marketable product every week. And so what's a marketable product is a product that you can show to users. And the user might subscribe or pay for the app just for that or come to the app essentially just for that.

And that's why like table stakes features like, you know, let's say we're talking about like a word processor or something. You know, if you had like auto format or, you know, just table stakes stuff like justify a or something, no one's going to come to your word processor for justify alignment, right? Like, you can market that, right? Because it's obvious, right? Of course it exists. But if you did something unique that nobody else has done, you can go and show that to people. And people will come to your app just for that. And even if your app doesn't have a lot of the obvious stuff, maybe it doesn't have justify alignment, right? People will jump over that just to kind of use sort of these new tools and new abilities that you might be building and marketing. So

we try to do every engineer, one marketable feature per week. And yeah, a lot of that stuff may not work, right? But a lot of it does work. And we can figure out, obviously, where to put in more effort. Things that start to work, we double down on those things, build more. People often complain because think about it. Like in one week where we're shipping, it's not complete. It's MVP, truly.

And we slice the hell out of it, right? Like we take the design and we cut, cut, cut, until we can really say that it's going to be useless if we cut anymore. We get that out and people come in and if things are going well, people will use it despite all the problems that it might have. And now people will complain and we'll have a list of problems and we know what to do next. So that's a starting point essentially. So as long as we're shipping one a week, we get a ton of volume of features and products

and directions we're releasing, cut a lot of that. what remains expand from there. So it works really well and it keeps people focused. I love the simplicity of that. I love just how wild that sounds for a lot of companies, I imagine. Every engineer ships a marketable feature or product every week. Yes. There's some people listening to this and are just like completely stressed out by this idea

and there's some people listening are like, this is exactly how I want to work. This is how every company should build. Yep. How do you maintain quality and make it all cohesive? I imagine that's the big trade-off. Just any tricks there for folks that want to maybe start operating this way. Quality is not something to compromise on most of the time, right? I think, yes, there's strategic compromises on quality,

but most of the time what you want to do is have a bar for quality where people should come in and if they're using the feature, like, it should work, right? Of course, right? And the way to cut down on time, and I think this is the mistake people make a lot of the time is when time is being pressured downward, a lot of times, engineers, PMs, designers, they will cut on quality rather than cutting on scope. And actually, you can cut on scope. It's actually, you know, the method that we use is we look at every element that's

going to take any time to build. And we just say, what if we remove this? Is the product still useful? And we keep repeating that until we remove whatever's left and we say it's going to be useless at this point. And that becomes the one-week project, right? Yeah, it actually really works, right? kind of narrows down to the core of what you're really trying to ask.

So for example, let's say we wanted to build something to add an image, you know, on your video or something like that, right? And this is like, you know, a really basic idea. I kind of just made it up right now. And you might imagine a design in which you import your image from your camera roll. But before it lands in your video, you might want to remove the background, right? You might want to, you know, change the hue and saturation or something like that. And you might expect, like, a designer to, like, build, design all those features, right? And you kind of land of design. But you really quickly realize that you can cut all of that stuff, right?

You can cut the background rule. You can cut the hue saturation. All you really need is pick. And then there might be a picker. We need a picker with a library with a lot of different, you know, what if you want to pull from the cloud? What if you want to pull from, you know, the drive or something like that? Cut all of that, right. And essentially come down the core, which is just like, you know, native picker from the camera.

lands straight in the video, no UI. And that is already, that should be useful. If that's not useful, then anything else built on top of that is also useless, right? So that's kind of how we might go about it. Now last sentence is so key to this. Like, it's the core idea of ship small iterative features before you invest a lot in something to first figure out, is there anything there?

Is this worth spending weeks on? Totally. And I think like the coolest part of this method is like, the first thing that the users will come in, they'll use the thing, they'll import images. And the first thing they'll complain about is what kind of bothers them the most. Is it human saturation? Is it background removal? Is it like picking from the cloud? Right. Like, you'll just get the most complaints about that thing. People will be like, and people will be honest about it. They'll be like, this sucks, right? It doesn't even have

background. What kind of image thing is this, right? And you kind of have to take that feedback, and just next week, you can ship in a single week all the things of the user's complaining about. And then they're like, wow, this team is shipping like crazy, solve the problem. So responsive. This connects a kind of a common sign of product market fit, which is when people are complaining about the thing, that means they actually care enough to complain. And that's a really good sign. If they're complaining that's something. It's very true, very true. If nobody complains, it's almost red flag, you know. I love it this turning into kind of an archaeology of a modern product team and startup. So I want to keep digging. This is not where I was planning to go, but this is

Awesome. I love that this approach of every engineer shipping, something every week that's marketable connects directly to where I started this conversation, which is how do you stay above the noise? And part of the answer is just ship stuff constantly and just continue to impress people like, here's a new amazing video feature. Look at this thing. Exactly. Yep. I think it's definitely key, right? And there is enough area and enough scope for that to happen. Right. Like, I think truly in normal times, it may not be possible to grow. create that much roadmap that quickly. But I think because there's so much innovation under underlying all this, there is that scope available, right? Like the roadmap almost seems unlimited,

right? Like, just truly. Okay. The other question I imagine people would be wondering is how do you work on longer term projects that take many weeks? There's also infrastructure, I guess, back-end stuff. So maybe answer those questions. How do you think about long-term stuff? And then how do you deal with back-end stuff that isn't a feature than anyone would care? Yep. So usually we'll dedicate time to that separately. So for example, usually Q4 for us is infrastructure quarter, right? We just like go and build all the infrastructure. Q4 is generally, you know, we've already delivered a ton of products and stuff. We're feeling pretty good about the rest of the year. Things are winding down, you know, obviously holidays and stuff coming up. And so we spend all that time paying the technical

debt, right? I actually think there's a unique thing to think here about like technical debt in general. and, you know, as a startup, your job is to take on technical debt, right? Because that is how you operate faster than a bigger company, right? Bigger companies don't take on technical debt. They pay it usually right away or they're paying back technical debt from the days when they were a startup, right? And they took on a lot of it, right? I mean, SNAPs, I used to work a snap and there was a lot of examples of that over there.

Right. So, and I'm sure it happens at every other company, right? And we think about it as like, well, is this a problem we need to solve today? Or is this a problem that the 50th engineer or the 100th engineer or the 500th engineer can solve? And if it is a problem that a future engineer can solve, we should use that future engineer now. Essentially, that's what we're doing. And we're saying we're going to push this to somebody in the future. And by the way, if the company fails, that engineer will never be hired, right? And all this won't matter anyways.

So it's kind of like financial debt in many ways. Financial debt is taken on to create leverage. It can be a good thing, right? Like if you're buying a house, you take on debt, and you can create, buy something probably more than you can afford without taking on debt, right? And it's the same thing. You can create products that you wouldn't be able to build

with a small team that you have by taking on strategic technical debt. It's very positive, actually. Wow, this is such a cool idea. And where my mind goes is that future education, engineer, maybe an AI agent engineer. Exactly. Just solving problems, just on technical debting you. Exactly.

Some engineer in the future, you know, 500th engineer, many years from now, will get a promotion because they solve this big problem that does really bad, you know, early engineers created. So obviously there's lying to this. Like, you don't want to, you know, there's like only so much debt you can take on before you become a big problem. Is there any thoughts on just that balance of just like how much is too much and how you know, if it's enough for a future engine error.

Yeah. I mean, I think generally the rule of thumb is, you know, every piece of debt that you take on, you have to pay interest on, right? So if there is debt that you've taken on, there's 1% or 2% of your time that is going to be taken away every day in maintaining bugs and issues and restarts and crashes and things that are happening with that because you did it the fast way, something's going to go wrong with it. Every day, 1% of your time will be taken away.

If you take on enough debt, you'll be paying 80 or 90% interest, and you'll not have any time to do anything new. You'll just be paying interest. That's all. And that's when you get into a mode of like, oh, we're just keeping the lights on. We don't have any engineers to do anything. We're just keeping the lights on. That's the failure case for a startup, right? So in a way, you have a technical debt runway.

Right. Once you run out, once you take on too much debt, and if you haven't delivered value in that time, enough value to hire the engineers to pay the interest or just pay off the debt, you will get in trouble. I love that. That's such a nice heuristic of how to think about when to invest in something.

I don't want to go down this too far, but just a thought I have is sometimes there's big technical decisions you've got to make that impact the way everything builds or is built in the future. I imagine those you spend more time on and take really seriously.

Definitely. Yeah. I mean, I think as long as it's possible for wherever it's like a two-way door, you can kind of do whatever you want. I mean, this is a classic methodology,

right? If it's a one-day, one-way door, it's worth thinking about and sort of doing correctly, at least as much as the one-way door would matter to you in the future. How much do your engineers use cursor and tools like that to build? How much is AI helping your team move? 100%. Yeah. I mean, everybody's using it. It's super helpful. I mean, even I'm using it, honestly, yeah, it's a huge multiplier for the team, no doubt. And is it a cursor specifically? Is there anything else that you guys feel useful? We are using cursor. Yep. We've tried all the different tools. We're using Devon as well, which is another, you know, that's more advanced, I guess. It's kind of solving bugs for you.

Yeah, Devon's basically, I think it's 500 bucks a month and it's like an AI engineer that you just chat with in Slack. Exactly. Yep. And, you know, in a way, these are the types of things that us as a startup can do it that bigger companies can't just, you know, they can't just pull in Devon, right? They have to, you know, get 30 lawyers in the room first before that happens. Yeah, and they're all called Devin, right? These like agents, like everyone's going to have like hundreds of Devin's working at the company. Exactly. You can have multiple Devens. I actually heard you can have a manager of Devin's who's managing Devin's. I left that managers are all getting layered, like unlayered and then they're going to have AI managers. That's the ultimate, the ultimate bait and switch.

Okay. Is there anything else that in how you operate and build the way you build product or set up the way you build product that you think is really unique and interesting that other people might be able to learn from. Our process is a bit interesting in that way. Like, we have a design team. We have a PM team. We were very early on those teams right now. And obviously, we have engineering.

And we have all the different surface areas. So iOS, Android, web, right? There's backend team, machine learning team, research team. So generally, like, when we're developing products, we may start off with like a PM-first approach where we're kind of, you know, finding some sort of overall issue that we want to take on some new area or pillar we want to take on and then creating sort of product specs from there. But a lot of times we'll also start the opposite way.

We'll first design something without even having any idea of what or why we're doing it. But we'll design a bunch of different things. And then we'll sit down with the PMs and look at the designs and just go over one and the next and the next until we find interesting things and ideas that kind of pop out of that. And a lot of times that leads to us discovering, right, like things that we wouldn't have discovered if we were just like to focus on the metrics and the numbers and, you know, things like that. So it's like almost reversing the process a little bit and starting with design first. But it can often result in like finding unique ideas, basically. I also think that we have a unique setup and how we create our roadmap. So normally you have a single roadmap, right?

And we actually divide a roadmap into two different roadmaps. So we have what we think of as the public roadmap. This is basically what people have asked us for. So like there's all these surface here as we receive user feedback. And we look at all that feedback and people will ask for features. Like they'll ask for like, I want background removal. I want like to undo and redo. I want to like upload longer videos, whatever it is, right?

A bunch of different features. And we'll just make a list of that and just like anything else will prioritize it. And we'll look at how many people it affects and what the possible markets are. and like just get it done basically, right? One at a time. But these are all features that every competitor knows about. These are public. Like if a user is asking us for it,

they're asking everybody for it. And every team has essentially more or less the same list and everybody is prioritizing it. And yeah, sure, you can win a little bit like extra nicely prioritizing it or winning a little in prioritization or execution or something, but it's not going to be a game changer in terms of winning against your competition.

So we have a second roadmap, which we think of as a secret roadmap. So this is a roadmap that nobody asked for anything on this. Like literally nobody has ever asked for it. And if a user were shown something on it, they might be like, I don't need this. I don't know what this is. But given our unique vantage point, our unique understanding of the problem set, the user space, and the technology, we've come up with some special ideas that we think will completely revolutionize how something is used, right, where we can truly change the behavior of the user. I think that's what it, is at the core of it, right?

It is like, people are doing things one way. If we're able to show them another way, and once they try it, they never go back, that's what a product is, right? That's success, right? And those are the types of ideas we put on the secret roadmap. These are things we never talk about publicly, never tell anybody about.

And we announce them and just give them to users, right? And see the effects. A lot of this we come up with through brainstorming. So we actually do quarterly brainstorming, company-wide, everybody's included, like, everybody from, it's not just a product team thing. It's like engineering, recruiting, everybody's included in. And we all come up with marketing, obviously. Like, everybody comes up with ideas.

We vote on the ideas, rank the ideas, and then the product team takes over from there and things about, like, feasibility and technology and what the different things could be. So this is a way where we can take all that noise that people are getting. Like, everybody's browsing social media, seeing all these, like, different things that are blowing up, these models and advancements, and we can get all that information together and provide sort of a unique internal roadmap, right, where how are we going to approach and create value out of all these different advances that are happening? So that's our general methodology. And a lot of times, the biggest wins will come from the secret roadmap, right? That's the game-changing

stuff, right? It's not going to be the user requests, usually they're going to do that. I love just how calling it the secret roadmap makes it especially extra interesting. Exactly. It's a secret. It's a secret. I'm not even going to ask you what's on that secret roadmap. You can't tell me. What's an example of feature that came out of that secret roadmap?

That's been a big deal for you guys. Tons. I mean, I'll give you an example from a long time ago. One of the first sort of AI features we added after sort of the app initially took off was this feature called eye contact. So this was a feature where if you're, recording something, oftentimes like, you know, people who are new to recording a video might read from a script or a teleprompter or something like that. And they might have that

like off-screen. So it kind of looks like you're reading. And, you know, it's not great from like the perspective of the video itself or the viewer of the video. So we had this feature where it basically shifts her eyes to look at the camera. And we were actually the first company to build this. We worked with Invita on this. It's actually really interesting because when we originally reached out to Nvidia about this, they were kind of not sure why we needed this, right? And they actually gave it to us pretty openly. And we're excited about some sort of partnership

of like how can we get this technology into like something that could be useful. But we saw sort of this creator use case, which was unique. And, you know, it was one of the ideas that came out of the brainstorm and we threw it on there. We launched it. It was a huge success. I mean, I'll be honest, like the video, the ad that we made, like a social media post that demonstrates this was so viral. It was made in basically every language around the world. It still till today, like gets millions of views.

Like we find reposts and reposts of that thing that other people have created that get millions and millions and millions of views because people are like, wow, this is like, this is a great idea. And now it's been like copied the hell out of like I think it's available basically on every, you know, every app you can imagine for good reason, of course. But that's one of the ideas that came out of it. You talked about how you come up with these secret order map ideas. I'm just intrigued by this. I'm going to spend a little more time here. Does your team ever work with an AI LM to help brainstorm? I imagine that's where things will go where you're actually jamming.

Like the AI agent is brainstorming along with you. Honestly, I would like for it to go there. It hasn't gone there yet. We haven't done that exactly because the problem is context. And I think just the context of understanding the user, the use case, it's so abstract. Even right now, like I feel like I understand our users, obviously. But I can't exactly verbalize why that is or how that is. It's a little bit abstract, right?

And I spend a lot of time with RPMs and designers, like imparting anything that I understand. And I've learned over the many years, you know, I've been working on this. How do I impart this to them? Right. But then it's a challenge because I can't even verbalize it myself. And so it's an extra hard challenge to figure out how do I put this context, how do we make it available to an LLM when I can't even put it into words exactly? And honestly, this is probably my own failing.

but, and I need to work on this, but there is something to it. I do remember at Snap, for example, right? I think one of the most unique things about Snap and the CEO, Evan Spiegel, was that he had an unmatched understanding of the user. I think years and years and years of the company's existence passed, right, like almost a decade. And nobody understood the user like he did, right? like he would come up with ideas that everybody would disagree with.

And we would launch them and there would be hits, just hits after hits, right? And nobody would understand why. Everyone would line up and be like, great, like, round of applause for everyone, right? But no one knew why. And, you know, a great example of that is like, and a lot of this was figured out in retrospect, too. Like, I think there was a point in which Snap declared that they're a camera company. And a lot of people laughed at that as a camera. Like, what are we making, like digital cameras or something?

or like, why is it a camera company? But I think at the core of it was this idea that Snapchat opens to the camera. And that was actually the differentiator, right? That was actually that small decision was holding the entire company against all competition. Because when the moment passes where your friend is doing something funny and you need to capture it, you're not going to open Instagram or anything else because it doesn't open to the camera. You're going to open Snapchat, right? because you can capture it right away.

And Instagram can never copy that because all their metrics are going to go down as soon as they do that. So, you know, that is a fundamental understanding. And I figured this is how much later, actually, you know, but it's such a powerful idea. I'm glad you talked about Snap. That's where I definitely wanted to go.

This is where I was going to start. So I'm glad we circled back to your experience at Snap. So the reason I am interested in this is if you think about social networks, seen, you know, this, like, Snap was basically the last social network to have launched and stuck around other than TikTok, which I don't think is a social network. I think it's just kind of like content platform. I don't think you're really interacting with people, really. And that's 2011 when it launched. So it's been like 15 years since the last social network launch that has

worked. And I think it's interesting also because there's rarely been a lot of insight into just how Snap operates. You were there really early early. You're a big deal at Snap. You're built a lot of really important features. So I wanted to spend a little time here. And it feels like a lot of things you learn from Snap, you're bringing into your company now. So let me just ask, I think you may have answered this, but I'm curious if there's something else here of just broadly, maybe other than Ev's brain, what do you think was core to Snap being a successful consumer social product?

There were a couple of different things that went well. I do think for a company like Snapchat or social network, there's a company. The core product market fit can be extremely strong, right? Like, essentially the reason that people are downloading it, the way that it's spreading, the way that it's distributing, the way that it's inviting friends or sending snaps or whatever it is, right? That product market fit can be so strong sometimes that it can be hard to actually build something because you actually can't tell if what you're building is what's responsible

for growing the thing or if it's actually hurting it. and, you know, it's growing despite what you're doing, basically, right? And I think because of that, it actually sometimes teaches people the wrong things. It teaches people that the, you know, contrarian thing that they were doing was right when it was actually just wrong and the company just grew despite it, right? And I think some of the things that SNAP did well and it needed to do, really, you know, was to continue innovating, right? Because for a company like Snap, it has a ton of competition.

Social networks are monopolies by nature. And there's a lot of reasons for Facebook or any other social network to stop the growth of Snapchat. And they tried. They tried really, really hard. And the way that Snap was avoiding that was by innovating. I think at the core of it was the setup that they had, which was very unique. Like, I've never seen anything like it. I've worked at a bunch of different companies.

But obviously there's a CEO, and the CEO was very product-led. his designer himself, right? But he surrounded himself with the design team, right? That was sort of the central team in the company. And the design team was like 10, 12 people, basically pretty small, even at, you know, 5,000, 6,000 employees, it was that small still. Wow. At 5 or 6,000 employees, the design team was, you said, how many 5 or 6 people? 10, 12 people. 10, 12. And to add to that, there's no PMs, really, like, for a long time. For a long time. Yeah. Initially, there were no PMs at all. PMs were introduced with monetization. Once monetization was a big sort of element,

that's kind of where PMs came in. Today, I think there's an adequate number of PMs across the company, but there was a long period of time, especially when the innovation was happening, when, yeah, there were a much, much smaller number of PMs and it was very designer-led. But at the same time, I think that's slightly misleading in the way that these weren't your sort of average designers,

right? These were designers who were actually PMs as well. that's kind of what the secret sauce was. They were able to not just design, but also do the PM part, which is a big responsibility. It's a lot of work, especially for that many employees, right? But it gave the CEO a way to sort of have granular control over what exactly was being launched in which part of the app at all times, right? Because he could meet with a set of 10 or 12 people

and know every change that was happening that was user impacting. A lot of changes were being worked on that were like infrastructure and like, you know, types of things that kind of keep going on in the back end we're improving ranking and whatever that might be, right? And performance and things like that. And those were not usually his concern. He was concerned with what UI are we adding where. And if you needed to add UI to the app, you needed it designed.

And if there's no designers in the company except for a handful who talked directly to the CEO, you kind of create a very granular control over what's being launched in the company. Right. So everything needed to be approved by Evan. If you hadn't approved it, it's not going out. So the design team actually held a lot of power in that. This is awesome. So what I'm hearing partly is, like, I don't know if this is true,

but it feels true that to make a consumer app that is successful and breaks the route, you almost need like a singular mind that continues to stay in the weeds on everything. And the way Evan did that is stay very close to the design team who basically ran product. That's very true. Yeah, it's very true. And he was able to keep the context of the end. entire app right in his head at the same time. He knew the interdependencies and what we're doing

and why we're doing it. And so that gave him just very granular control over the company's product roadmap. It makes me think about Brian Chesky and like Airbnb is a consumer product. It's not a social network, but I wonder if that's just an interesting insight. Just for consumer products, they will generally do better if there's one person with a really the right sort of combination of experiences, insights. And just they continue to. run and own every detail. Definitely. And also the ability to bring about change, right?

The ability to truly, you know, energize an entire organization to do something that's not just incremental, but fundamental, right? Founder mode. Exactly. I mean, that's what we're getting to, basically. Never heard of it? Okay. And then you said that these designers, so I know it's like famous that Snap had no PMs for a long

time. Designers were PMs. At this point you made about the designers were, PME is really important because I think a lot of people look at this. They're like, amazing. We're just going to hire designers. We don't need all these damn PMs, slower to thing down, just tell us what not to build.

Can you just talk about like the level of these designers, like what allowed them to be as successful as they were without any PMs? Yeah. I mean, I think what was expected from the designerist now was not just the ability to design like, you know, the skill set of designing, which all of them were icy designers, by the way, right? And there were no reports, right? So they weren't allowed to have reports, actually. And so they were designing everything themselves.

But they also had to have the leadership skills, right, to go, you know, figure out the roadmap, write all the documents, work with the different teams, figure out shipping schedules, and, you know, just know everything, not just sort of the technical and the engineering part, but, you know, the U.X and the UI and the product needs and wire. are we doing this, you know, the roadmap, there's just a ton to keep in mind. And that means that it was a job that was, you know, just highly, it was very high workload, no doubt, very high workload, right? These people work really hard. And they were paid highly too, I remember what it's worth. They were paid way higher than you would expect designers or PMs or engineers would be paid,

right? With quarterly bonuses and all kinds of things. That's interesting. And it reminds, you know, people always say, why don't need PMS? There's like someone has to do the work that a PM does. They're not sitting around, you know, doing nothing. And it's important to note the person that will take on the PME work. They have to be good at it and enjoy it. And a lot of designers don't want to be doing, writing docs and organizing stakeholders

and getting alignment. 100%. 100%. That's why it was so hard to find those people who were, like, able to do two things. I actually think there's a insight in there. is innovation between, you know, when you're kind of merging craft, right, between two different functions. And I do think there's something special about one person doing two different functions, or at least being able to do. And I think a lot of, like, unique insight and innovation can come

from that. I actually think, so on sort of my personal side, like, I eventually joined the design team. I started at Snap as on the engineering team. I eventually joined the design team, you know, over the last two years that I was at SNAP. And a big part of what I did there was create this function called design engineering. And that was actually a different combination, right? It wasn't the designer PM. It was a designer engineer, right?

The person who can think of the UX design it and also build it and launch it, all those things. And we saw both the ability to take designers and teach them engineering and take engineers and teach them design as part of that, right? Obviously, you know, the reason that we created that function was very different. It was actually to continue innovating as the company got bigger.

One of the problems that we identified was that as the company got bigger and bigger and there's like 500 engineers, 1,000 engineers, 2,000 engineers, 3,000, right? Suddenly, it just becomes very difficult to do everything. Like, everything is a six-month project or a one-year project. Every product is a massive investment of like 500 engineers and, you know, a lot of time. And so you really have to pick your bets, right? If you get it wrong, if you are innovating and trying to create new products and you spend 500 engineers for a year and it doesn't work, it's a big problem, right?

You're going to be in trouble, especially if you're coming like SNAP, where everybody was copying what they're doing so they had to constantly innovate, create new stuff and push the bounds, right? I think Evan's philosophy was always like, he didn't fight the things that were getting copied, right? Stories got copied pretty much straight up. A lot of things that SNAP created got copied. But he was more of the mindset of like, let's expand the pie, do that.

something new and push the boundaries, we'll keep innovating, basically. And so to do that with that scale of a company, it becomes really hard. And so we had this idea of like, let's create a small team where we can go and pre-test a lot of these ideas because we had a lot of ideas and, you know, we can't go and build all of these things. So the idea was create a small team of these design engineers, people who are able to do the entire sort of product design engineering process in their head and can put together early very very very very, very different. versions of the product, which we would actually bake into the Snapchat app itself. And we were able to even test, for example, run a test in Australia, see how it's performing, you know, run a test

in a couple of high schools, just a couple of high schools, see what's, you know, how people behave. And that way we already have data on how this might perform in a real world environment, but we haven't built it to production level, right? It's like, it's a prototype, essentially, right? It's how a startup might build something. The same idea of, you know, what we're doing at our company now, right? Go to fast, get it out there. Right. Get feedback, understand whether it works or not. And then work with the engineering team to build it at a scale. Once we understand the product and the dynamics, then it makes sense to put on 500 engineers for six months to build it. So that was like a big part of it. I think the nice thing that came out of it that was completely unexpected, but actually kind of transformational for me in a way was, you know, obviously in big organizations, alignment is a big issue, right? How do you get everybody on the same page? And a big part of a PM's job. is actually to create alignment, right? And it can be a lot of work because you've got to talk to all these stakeholders

and get them on the same page. But one of the insights that we had, which was unique, was as the company gets bigger, you can actually create alignment by causing internal virality. If there's enough people in the company, it actually starts acting like a consumer-based might. If you share something interesting with someone, they will share it with somebody else because they think it's interesting. And you can actually create virality inside a company. So one thing that we would do is we would create these prototype products, right? We would just go into an area,

redo a bunch of stuff, create these prototype products that didn't exist in Snapchat normally. And then we would just share the build, right? And it would explode. Like, it would just go viral inside the company. Like day after day, we would hear from, you know, engineers, then managers, then VPs, and eventually from Evan being like, oh my God, like everyone's talking about this. Why am I the last one to hear about it? Right. And so it would create like instant alignment across the company of like, this is exciting. This is something that we want to get behind.

And everyone would be asking like, when are we doing this? Like, when is this happening? I see someone's already working on it. Right. So it was a great way to do that. And once we really understood that the product actually had good sort of dynamics and we had tested it. It was a great way to sort of get it out in front of everybody and create this idea of like, hey, we're all working on this. This is sort of the future, right.

Today's episode is brought to you by Coda. I personally use Coda every single day to manage my podcast and also to manage my community. It's where I put the questions that I plan to ask every guest that's coming on the podcast. It's where I put my community resources. It's how I manage my workflows. Here's how Coda can help you. Imagine starting a project that work and your vision is clear, you know exactly who's doing what and where to find the data that you need to do your part. In fact, you don't have to waste time searching for anything, because everything your team needs from project trackers and OKRs to documents and spreadsheets lives in one tab, all in Kota. With Kota's collaborative all-in-one workspace, you get the flexibility of docs, the structure

of spreadsheets, the power of applications, and the intelligence of AI, all in one easy-to-organize tab. Like I mentioned earlier, I used Kota every single day, and more than 50,000 teams trust Kota to keep them more aligned and focused. If you're a startup team looking to increase alignment and agility, Kota can help you move from planning to execution in record time. To try it for yourself, go to Kota.io-slash-Lenny today and get six months free of the team plan for startups. That's C-O-D-A-O-S-Lenny to get started for free and get six months of the team plan. Koda.io-slash-Lenny. Another thread I want to follow up on is, is prototyping.

It feels like that is where a lot of PM work is going, is getting straight to a prototype versus design or versus PRDs. And it feels like that's something that you did and worked super well. Like here's a, like basically it's a team to prototype ideas that in theory now, you can just build really quickly with AI. So I think that's a really interesting seeing where the future is going. 100%. Right.

Like getting things in people's hands, trying it out. oftentimes, like, unless you truly try it out, like, you know, in design, it can, in theory, look good with like all the perfect conditions, right? But when you actually use it, you realize it's actually not that useful, for example, right? Or when you give it to users. And some of this is like intuition, honestly, right? Like, just like anything else. But there's nothing like getting something in the hands of users at the end of the day. I love how many of these things you brought over to your current company. And I'm trying to think about one is this idea of just constantly innovating. It feels like that's informed and tell me what I'm missing. But that feels like that's

and form the ship of market will feature every single week. This idea of getting like design, starting almost with design versus PM a lot of times. I'm curious why you don't even go straight to prototype in those cases. Is it just the tools aren't there yet? I mean, I think our shipping process is fast enough that within a week we can get it out anyways, right? So that way we just get user feedback, which is even better. Okay.

And then the other really interesting thing, I'm trying to visualize like that triangle of a product team that triad of PM engineer design. It feels like you guys at Snap took like the, corners, not the corners, the line of that triangle. And like, you have design engineers. You have design PMs. Yeah.

I imagine engineers were sort of PME already. They're like very product-oriented PMs. Did you have a function called design PMs? Probably not. I mean, it's interesting, right? Engineer PMs. Yeah. I mean, engineer PMs should be a thing, I feel like.

Or every engineer should strive to understand the product. Yeah. A lot of companies operate that way, like Stripe. I think they have hundreds of engineers before. where they hired the first PM because I think the engineers were doing what they did at SNAP. Right. Do the PM work. So it feels like at your company, you don't operate that way.

It feels like PMs, engineers, designers. Talk about why you decided not to approach things that way. I do think PM is a very valuable function, right? I think it may be actually, and, you know, maybe I'll get roasted for this. But I think at the end of the day, not hiring PMs at SNAP might have been one of those decisions where it actually succeeded despite that. And because like someone needs to do that work, right? If you don't have enough people to do it, then nobody truly owns it. And then it kind of doesn't really happen. Or if it doesn't happen, no one's responsible, which is not the right

structure you want in our organization. So I think though that being said, there was something unique to be said about what if a designer had the PM mindset, right? It's actually the same idea as what if an engineer had the PM mindset. And then you get, you know, even crazier, what if the PM had a design and engineering mindset? I think all we're talking about is everybody truly understanding all the functions that they're working with, right? Having a fundamental, broad understanding of the functions they're working with.

And at captions, we're actually saying, going even once a step further than that, right? Why shouldn't the PM understand marketing? I think that's actually the biggest, you know, opportunity for PMs to understand. It's like, how do we actually find the users who have this problem? Right. I think that's a big part of solving the problem.

I have a unique take on this in terms of, you know, I actually think PMs should own all the way to marketing in a way. And the reason is that if you think about marketing, it's expanding the surface area of the product, right? It's like search marketing is just placing a button to your product in Google. Facebook ads is just placing a button to your app in Facebook, right? It's almost like you work at Facebook, right? You work at Facebook.

You have a button in the app somewhere. You make a specific thing, and people show up. The funnel begins there, right? And you have all the metrics all the way from the beginning, right? All the way from when the user tapped on the button in Facebook. And then they went down all the steps, and then they landed on some onboarding screen. And, right, they did the thing. They used the application.

that's where the journey begins. And all of that is like, in a way, it's a product. It's the same skill set. Understanding users from that point on is like, I think that's fundamental, right? How do we not do that today? We should be. So that's kind of how we think about stuff, right?

But I think the core idea is that every function should understand every other function deeply as much as possible. And maybe even to the level where they can operate in that function. and that just increases the likelihood that all decisions being made in the company at the micro level will be optimized for, you know, all possible, you know, parts of the funnel that different people are essentially looking after, right? So that's something we think about quite a bit. I completely agree with that take.

It's interesting that Irving B, Brian, was famous for changing the titles of all product managers to product marketing manager. Yep. for exactly this point because he's like, you should be doing the marketing. You shouldn't just be building the thing. Right. And like to me, it's always, I've always assumed as a PM, your job is for this thing to grow and to get adopted and be loved. Of course.

So it's interesting. People don't already think of it that way. I agree. But obviously it's hard to learn the skills of being awesome and paid growth and SEO and product marketing, messaging, positioning. But I completely agree. That's such an important element of building a product. You're not just building a thing. Hope it works.

Goodbye. So I love that that's how you think about it. And so I guess when you hire PMs, it sounds like you look for marketing instinct and some experiences. 100% right. And at least the ability and instinct to be able to learn it, right. Yeah. Okay.

So I'm going to share one other thing that I thought as you were talking that I think is really interesting. And it comes up a bunch on this podcast. And this connects back to Ev and what we can learn from his success. So Patrick Olson once tweeted this. tweet that's really stuck with me, which is it was around user research. And the way he described it is user research isn't. Go to user research that informs what you build and then you build that. It's instead, you do user research, it informs the mental model you have as a leader, a product

builder of what your customers need and what pains they have. And you adjust that model in your head, and then that's how you decide what to build. And it feels like Ev is very much that. Like his head was on learning what people need teens in particular, and it just worked. Yeah, I think it's very spot on. I would say, though, Snap didn't like user research as a function for the longest time. Like, I think there was one user researcher in the company until, again, 5,000 employees, like post-IPO, basically. But I think the people that were making a lot of the product decisions and the CEO himself,

of course, were very steeped in sort of the how the user behaves and how they operate. Like, they understood that. I do think, like, Snap also had a unique way of thinking about how to determine if a product is within scope or out of scope of what their mission was, right? And I think a lot of companies use this type of framework, and we try to try to as well. but essentially the idea at the core was that they want to enable like private sharing and in a safe way, right? So if, so I think that kind of makes it clear

that certain things just are out of scope for SNAP. It's actually one of the reasons why Snap wasn't the company to discover, quote, short form video, TikTok style stuff because it was just against the nature of the company to even try something. It was against the mission of the company. Public sharing means possibly, bullying and bad behaviors, which is exactly what Snap was trying to avoid. We don't want those behaviors to develop on the app. So, for example, on Instagram stories, you can share somebody

else's stories to your followers, right? You can take, I can take your story and share it to my followers. You can't do that on Snap. And there was a discussion where, like, should we do this? No, because it can enable bullying, right? I can, essentially, I don't, you know, you're not consenting to your thing being shared to my followers, right? And that's essentially bad, right? So a lot of it was done based on this type of pillar-based thinking of like, you know, this is our mission, this is what we're trying to do. Does it fit within or is it outside?

If it's outside, we don't do it, no matter what the cost of it is, no matter how exciting it is. So, and even on Spotlight, the big challenge was like, how do you take something like that and put that inside the Snap Mission? So that was something we worked on quite a bit. Yeah. I mean, I think there's, yeah, there's tons of stories about earlier versions. I mean, Snap almost had essentially what is TikTok earlier than TikTok existed.

And it kind of died out because it didn't align with the mission essentially, right? But happy to get into it. Yeah, that'd actually be really interesting because it's interesting that it, these things are important. It's important to have these clear values and the mission of the company and to not focus on things that are outside of that. And then you hear these stories of like they had like they had TikTok. potentially. So yeah, whatever you can share there, that'd be awesome. Yeah. I mean, I think,

I don't know if you remember this, but there was this product called Our Stories. And essentially, it was like my story, but it was a public story. And it started off with this idea of campus stories where you can post your campus and other people can see it. And that actually started creating a lot of virality, right? Because essentially people would post, there was like, you know, viral moments truly where people would post stuff like, oh, like, you know, I think two people fell in love on it or something like that. Like those types of things like really went viral. And it had really good engagement.

But at the end of the day, the problem was that we were against like algorithmic, essentially ranking of those types of things. So there was a curation team that was looking through every single one so that there's no, you know, negative behaviors happening essentially on the app. And that was just not scalable. Even though it had really high engagement and was doing well, it just wasn't feasible to have a person looking at every single thing posted to determine whether it's appropriate or not. And so it kind of ended up dying out, but it looked like what was

an early version of TikTok, you know, before it had launched. So I think in a way, though, it was a good thing because I think Snap does have a mission and I think it is solving a problem. I do think, you know, there is a bifurcation of social media at this point. There is what you traditionally think of as like social networking where you share things with your friends. And by the way, like, remember the days where that used to be the way that apps would go viral? You would share things with your friends and then they would share with their friends, right? And everybody was worried about like friend sharing and how do you send to a friend and,

you know, can I text message my friend or whatever, right? That was, that time is over. virality now happens through a completely different mechanism. It happens through essentially algorithms, right, that are deciding whether your piece of content is worth showing to an arbitrary number of people. And this is the new age of social media, right? It's TikTok, it's, you know, YouTube shorts and Instagram reels and so on. And I think actually is changing the fundamental nature of how people interact, fundamental nature of how things go viral. and I think these, I actually think from a regular hair perspective, we should be thinking these

as differently. On one side, you have, you know, something where you're deciding, you know, where who sees something. And then on the other side, you have something where the company is deciding, which means that it's kind of semi-curated, right? It's actually the company's voice. And so, yeah, I don't know, like should Section 230 apply to that? I have no idea, right? Maybe not. Maybe we're thinking why that's the wrong way. So it should be interesting. Wow. All right. Well, I'm out of my depth on the legal legality decisions. So I'm going to

follow that thread. But I imagine there's something really interesting there. Actually, so you've been talking about this just like how much things are changing. Yeah. And I just wanted to follow that thread. And specifically, you guys are at the cutting edge of what is possible with AI video.

Yes. It feels like we're approaching and maybe we're there, this world where you have no idea if it's real or AI. I'm curious, first of all, just how far you think we are from that. And second of all, the implications on the world where you can just generate any video you want. It's fundamental. At the end of the day, like a time where video, images, audio can't be trusted, actually hasn't existed for a while.

Like, if you think about, I mean, there was a world in the 1800s where there was no video or audio or images, right? and everything was proven by he said, she said, for the most part. And it's possible that if everything can be generated and anything can be created and it looks just as real as if it were real and there's no way to tell, then we might actually return to that world, right, where there's no way to prove anything besides, you know, physical evidence or he said, she said. And I think that's kind of scary, but also possibly opens a bunch of new opportunity for someone to figure out how to solve this problem.

right? I think it's going to be a big problem. I do think today we are almost there in terms of creating absolutely photorealistic video. I mean, the very recent model is a very cutting edge is just about like it feels like a few centimeters away from achieving it. But I do think to fully get there to the point where it cannot be differentiated at all, it's still a couple of years away. I also think that it is use case driven in a way. Like I think thinking about captions for a second. We take a unique view on what type of video we want to focus on.

Video generation and text to video generation. If you look at it today, it's all silent video, right? There's no audio. And it's often what you think of as like stop video or B-roll, right? You can actually make a movie with B-roll, right?

And a lot of a movie or a TV show or a social media post or an ad actually is dialogue or monologue. That's actually what it is. is people talking to each other, to the camera, right, interacting. That's actually what makes true story. B-roll is sort of like supportive elements, you know, that are showing up to, you know, set the scene or something like maybe, you know, before the scene opens,

you see a few shots of New York City or L.A. or something, right? And then you jump into the room and, you know, now two people are talking. So our goal is to solve the talking video problem, right? How do we create video where people are delivering, dialogue or monologue or, you know, things like that. And that's what we focus on purely. And there actually isn't a lot of work happening in that area today, right? And it's not a solid problem. We're getting there. We're getting closer and closer. But today's models actually bifurcated a little bit. So there's a set of companies today that are able to create these types of what we're

talking about is like avatar videos. They're using this technology called neural rendering. it's actually not a technology that's affected by the transformer and diffusion model revolution or the large model revolution essentially. This is a technology that exists separately, and it doesn't have anything to do with the AI sort of growth happening right now. It just happens to produce semi-realistic outputs, but it actually kind of stops at some point because it's not clear how it becomes generalizable in every situation, right? You can't, you know, it has to be trained on people individually.

So you might ingest a little bit of video of you, and then you can generate you. And so it's a different technology and different outcome, essentially. And a bunch of companies using this type of model, a bunch of companies are doing general text video with no audio today. These are a large sort of generative models, and they have the capability to do more, but that frontier just hasn't been reached yet. I think there's no doubt in anybody's mind on the research side that it, it is 100% solvable.

It's just like somebody has to go do it and we haven't gotten there yet. Nobody has had the time to go and do that yet. So that's kind of where we're at, essentially. We're working purely on large generative models for talking videos, right? So that's like our core focus. I do think, though, from a safety perspective,

we have a unique framework or how we think about it. So generally, videos divide into two categories, right? So for us, we think on one side of what is like documentation. So this is the type of video that it could be a personal video where you're taking a video with your friends and you're hanging out, you're at a restaurant, right? Like it's documenting what happened. You had fun. Whatever it was, it's for your memories, right?

And there's like a non-personal version of this, which is like, oh, it's like a reporter like documenting a crime or, you know, something that happened or whatever it is, right? And who was involved? What, you know, where was it? maybe it was a national disaster or something, right? And this is for history, like, we want to see what happened, right? And there's actually no benefit to Azure video in any of this, right? Like, actually, all of this is just negative.

It's all negative, right? Like, if we are generating fake versions of reality to full people, like, there's just nothing good about that, right? And we want to stay away from that, essentially, right? We want to design products and build products that make it difficult to use for that particular use case, right? for anything that falls within that. And on the other side, you have what we think of as like storytelling.

Now, this could be ads. It could be social media posts. It could be TV, movies. Like, all of these things are storytelling. They're designed for entertainment. They're designed for fun. And nobody believes, like, if you watch a Geico commercial, right? Like, you're not thinking that the gecko is real selling insurance somewhere out there.

Right? Like, you know that this is fabricated and it's for entertainment. And same with reality TV. even, right? It's called reality TV. It's definitely not reality. And, you know, social media, ads, you know, all this stuff kind of falls in the category. And if we can enable more people to tell stories and entertain other people and get their message out there, like, that is pure positive. Like, this is where we want to focus. And a lot of our effort in the product and design process goes into how do we design products and build products that specifically make it really hard to use

on one side and really easy to use on the other side, right? And that's, That's the real challenge. That's really helpful. Something that I'm really curious about as you're chatting, as ByteDance just released a really amazing model, is actually just looking at it, where you put a photo in, I think, and it just creates a video of this person talking in all these different ways.

Where does that fall amongst then buckets you just describe? I think that falls exactly in the area that we're in, right, which is talking people. And that's what they're going after as well there. So that's actually one of the first examples of a large model that a larger company has released where it's able to do sort of these dialogue or monologue videos, right?

And I mean, you yourself, I mean, you've seen it, so I'm not going to describe it too much, but as you know, it's like highly expressive, right? Like, it doesn't look like an avatar video, right? It looks like... And that's because the technology that's used is fundamentally different.

It's just like, this is using a true large diffusion model is what they use. Whereas, like, most companies that are working on avatar technology or actually using like, you know, something pretty basic in comparison. How long has it been since that Will Smith spaghetti video, just to give us a reference of how fast things are moving?

Oh my God. It's been so fast, right? Amazing. Or is it like two years? I think it's probably like about a year and a half, two years, right? Wow. We'll link to that video and then you could tell basically that video is the state of the art of AI video, one to two years ago.

And then we'll link to this other Omni something. I forget what it's called. I'll just showing what it's like today. geez Louise Okay Final question And this is around something That I know you

Have a really interesting insight on Which is that you see marketing Using AI video Basically is kind of the final frontier Of how people will experience AI Is marketing, is seeing it in marketing channels Talk about why you think that's the case And just what that looks like

It kind of comes back to what we were talking about before Where you know the reality is that no matter how interesting, advanced, and amazing a technology is, like, you know, science fiction has become reality. We were talking about this, right? Like, what was literally science fiction on TV is real now. And most people still don't even know about it, to be honest, right? Like, my parents live in India, and they are the only ones in the neighborhood that know

about chat GPT, and they write these, like, amazing, like, notes to the community, just like with all these words. And, you know, people are just like, how did you, like, how did you get so good at, right? right and they're not telling anybody but you know there's still a ton of people who don't even know that these advancements have happened and so adoption is actually much slower even for the most exciting things right of course in tech circles everybody's talking about it but the reality is like it takes a while to get out there and I think for companies that are going to succeed they're going to have to figure out how to market these products so that they can be the ones to reach you know

all these people that have the problems that they're now able to solve and we think about that every day. So on that note, like as a consumer product, we spend a bunch of time and money on marketing our products and we often use like performance channels and all kinds of things. But about a year ago, we would run AI video in ads and things like that. And we would get all these comments of people being like, oh my God, this is so fake. Like, you know, don't show me this. And around that time, the technology got just about good enough that suddenly those comments stopped happening, right? And suddenly you could, you know, get performance that was even better than actually recording with a person because you could just try more things.

You could just generate 30, 40 possibilities and, you know, one of them would win, and it would win more than the one created you can get from a person. And more interestingly, when you think about localization, you're going to go do that in every language. Like, once to discover winning creative, right? Now you have to go localize that in every market and rebuild it from scratch. It's just a ton and oftentimes it doesn't perform as well because it's been like rethought essentially.

But we found that just translating it with AI was able to get performance almost as good as the original in the original language, right? So this is gonna flood the entire market, right? I think wherever there's dollars to be made, saved, right? Like it was, it was, Well, it's inevitable, right? It will be consumed and it will very quickly be a lot of social media.

I mean, you could imagine a social network of the future where, and this is dystopian, by the way, so watch out. You could imagine a social network of the future where all content is generated. None of the people are real. I mean, the algorithm isn't tailoring whose content to show you, but it's purely generating content that, you know, is completely catered to you, right? with people and everything completely catered to you, I don't think it's out of the question.

It almost seems inevitable in a way. But that's not, you know, that's not too far away, I think. That's actually very possibly real in five years or something like that. What I'm imagining, because it's hard to imagine like a social network where it's people, because you usually want to know who these people are. Like, I don't care of random sharing status updates, but I can see a TikTok that is all like I generated. Exactly.

Just like content tuned to your loves and entries. Exactly. And just random videos. Wow. Yeah. Because do you know like you see a TikTok feed? You don't even know who's real or not today, right? It's not like we.

Right. That's how I would approach it. I would just join TikTok and start uploading videos that are AI generated. Exactly. And then build a whole network of that. Oh my God. The future is wild. Let's go to failure a corner.

Something that I try to do with this podcast is share moments where things. didn't go well. There's all these stories of everything's going great all time. All these founders killing it, building a billion dollar company. Oh, so awesome. But they don't know when they, all the things that go wrong. So let me ask you, is there a story you can share of when things didn't work out when they failed? At the beginning of the company, we actually had a bunch of time where we spent figuring out what we wanted to do. And I think it's like kind of a unconventional story almost in a way because we started off the company. The first thing, we did was build the Captcha's app, right?

Like, we launched the app. That was the first thing we did. Took two days to build it. We put it out there and it immediately took off, right? It was absolutely shocking because I built it on a weekend. We put it out there. I call my cold town around Monday and like, it's at the top of the app store. We're getting like 600 videos a day. Top of the app star. Yeah. And we didn't do anything to enable that. It just kind of happened on its own.

It was almost anticlimactic in a way because, you know, we thought it would be a lot more time spent figuring out the product before that would happen. And so it felt like, wait, this can't be it, right? It can't be this fast, right? How did this happen? Right. So we got distracted because of that, because we were like, oh, okay, well, maybe, you know, this is cool.

It will kind of work. That's great. But, you know, we got to figure out what the product is, right? And so we spent at least like a year, year and a half thinking about building social networks and all kinds of things when we should have been working on captions, right, because there was product market fit there. And how we figured that out is captions was sitting on my personal account. So I wasn't checking that a lot. And about a year and a half into the company as we're working on other projects and stuff, I went back to a personal account,

just opened it. And I saw that there was like $500,000 in there. And I looked at a chart and it was just like growing. The revenue was just growing completely on its own. No employees, no releases, no bug fixes, no customer support. There was like 2,000 open support tickets that were unanswered for a year and a half. And great reviews is just, it's just growing completely on its own. And so that was like a clear sign to me. It was like, oh my God, you should have been working on that, right? Like that product works. And so we immediately had a meeting. I mean, it was tough to kind of figure out what the right path was at that point because we'd invest so much time in other things as well, but reset and we kind of, you know, got back on the traffic captions.

And literally, as soon as we started releasing the first features into it, it blew up. Like what looked like a vertical line at that time became a horizontal line. And the vertical, the new vertical line was so vertical that the old vertical line became a horizontal line, essentially. And so, and that's kind of, you know, it's continued since then, which is crazy. So we basically wasted about a year and a half. I love that new way of thinking about a hockey stick moment where not only is it going vertical, but the rest of the chart is now just flat along the bottom of the axis.

For people that may not know what captions is, we'll try to describe at the beginning and we'll link to it and stuff. But basically, like, the reason you thought it was nothing is it just adds captions to a video that you report automatic captions. Exactly. Yeah. So I think, you know, we wanted, like our thought was like we're going to build a social network, but first we got to build a creation tool for the social network, right? And we knew that

we wanted to use AI to create video, and it seemed obvious that, oh, speech to text, a solid problem, we should start with that, right? So that's why we decided to start with, you know, captions, because it was a solved problem at the time. What was funny is that once GPT and stuff started coming out, like a lot of the things that were unsolved became solved very quickly, right? So timing was almost perfect. And that aligns is something you should earlier, just so many of these problems that were not yet solved are now possible. And the companies that are in the right place at the right time benefit greatly. I've been just waiting for this part. The other thing that I think is interesting about that story is you try to build a

social network. I think it was around like high schools and things like that. As we've seen, it's very difficult to build a new social network. So let me just get your sense. Do you think it's possible for somebody to come around and build a new, the next Facebook, the next nap, the next whatever. I think it's definitely possible. I do think, let me tell you something kind of crazy, actually. The social network that we had at the time, we actually removed it from the app store, so it's not available anymore. But till today, there are people, there are thousands of people that are using it, posting on it, and all the different things, which actually kind of speaks to like, you know, the power of the social network in a way, right? It is hard to create and hard to kill.

I mean, I think X is actually a great example of that too, right? A lot of movement kind of happened there and it continues to, you know, work, I guess, somehow. So, you know, testament to that. The power of network effects, especially. Someone he once described this so well, they're like Twitter slash X, they took, they changed the brand. They changed the team building it. They changed the URL. Like, everything changed about it except the network effect of the people in it.

It's true. And now it's true. I just saw a story that they're like making billions of dollars. Like he's actually turned it around. It's actually becoming a really profitable company. Wow. Yeah, I just came out the other day. So Elon did it.

Well, with that, we've reached our very exciting lightning round. Are you ready? I'm ready. Let's do it. What are two or three books that you have recommended most to other people? I have to say here that I actually don't read books. It's actually something that I decided on purpose where I decided I don't want to build my skill in reading and I want to build it in listening and watching instead because I think that's the future. I love how intentional that is and I love how it's like a really cool way of saying I don't read books.

The future isn't reading. But I love that you have books behind you. I do. Yeah. The one that I didn't read. They're back there. That's funny. Okay, cool.

I want to ask more questions. but I'm going to keep going lightning round. Do you have a favorite, speaking of watching and listening, do you have a favorite recent movie or TV show you've really enjoyed? I like Silo and Severance. I mean, obviously, I think everyone's watching these. There's a book around Silo too. I read that.

I read all of them. There's three of them. There are. It sucks to watch the show because you know all the tricks that are about to happen. And I'm just like, why am I watching this? I know where it's all going. I mean, for what it's worth, it does seem like the show is going on a slightly different path. It is.

That was also what annoyed me. And just like, what the head? This is made up. All is made up shit. I don't like that when I watch a show. Two reasons I'm not watching it. Don't worry, by the way. I didn't actually read the book.

My wife read the book, and then she told me the story. Okay, okay. I was worried. I was worried. Okay, cool. Yeah, and severance. Okay, great. I love severance.

Next question. Do you have a favorite product you've recently discovered that you really like? My favorite product, honestly, is linear. I'm not going to lie. Just because it's so well designed and it's so easy to use. I also like superhuman. I mean, these are obvious answers, but I do use these things.

every day. And, you know, it's hard to create products that you use every day and don't hate. So props for them. Cool. I haven't announced this on the podcast yet, but this is a good time. Whoever's listening right now is I just launched a bundle where if you become a paid subscriber to my newsletter, you get, listen to this, a year free of linear and superhuman and Notion and Granola, which is incredible AI app for note-taking and perplexity. for flexi-free $2,000 in value for the price of my newsletter, $200. Damn.

That's real value. It's an unbelievable deal, and it's a no-brainer at this point to buy a subscription. But this isn't an ad for my newsletter. I'll keep going. Next question. Do you have a favorite life motto that you often find yourself coming back to sharing with friends or in family and work or in life? I actually learned this because someone else told me that I keep repeating this thing,

but I have this sort of framework of how I want to operate at work, basically, right? I think I love to compete and to win at the end of the day. And I think that to win, you have to be the best. But I also think the easiest way to be the best is to be the first. And that actually is key. And so is the motto the easiest way, is that the, is that the, that's it. The easiest way to be the best is to be the first. first.

First. Interesting. Okay. I have to resist following threats here because I want to make this lightning round. Okay. Final question, just for fun. Is there, what's like the coolest most wild AI video you've seen recently? Is there one that comes to mind of like, wow, that was something?

I mean, honestly, I got to say the omnibumous stuff was pretty cool. The bite dance video that we talked about. Yeah, exactly. I mean, the broccoli talking. I don't know if you saw that one. There was like a little broccoli, like kind of delivered. delivering a little speech. Yeah, it was, it looked like it was animated by like an animator, you know?

Just imagine being a kid these days and just seeing stuff like that. I think you're probably just used to it, right? You're just like, this is just normal. It's just like we were saying AI is just going to come around. Exactly. All right, cool, what's for dinner? Cool, that's great. Yep.

Amazing. Gorop, this was incredible. So insightful in so many levels. Two final questions. Where can folks find you and what you're building if they want to learn more? And then how can listeners be useful to you? Awesome. Yeah. I mean, definitely find me on LinkedIn. That's where I live most of the time. My DMs are open, et cetera, et cetera. So feel free to send me a message. And I think what will be useful, I mean, we're building out our early product and design team. So if AI video is interesting, if consumer apps are interesting, now is the time to join. We're really small, early. We work together across the team. So there's going to be no better time to join, basically. And you get to ship a marketable feature every week.

Exactly. I mean, that's a PM's dream. Think about it, right? It's a PM's dream. Yeah, like, I like that that's a filter. Like the people that get excited about that, great fit. The people that are stressed out by that, not the place to be. Exactly.

So awesome. All right, Garov, thank you so much for being here. No, thank you. I appreciate it. Bye, everyone. Thank you so much for listening. If you found this valuable, you can subscribe to the show on Apple Podcast, Spotify, or your favorite podcast app.

Also, please consider giving us a rating or leaving a review, as that really helps other listeners find the podcast. You can find all past episodes or learn more about the show at lenniespodcast.com. See you in the next episode.