How to grow a protocol | Dan Romero (Co-founder, Farcaster)

Transcript

So, going to be talking today about growing a protocol. But, before we got started, I had two questions for the audience. So, the first is how many people, show of hands, are building a protocol here? Okay. So, a chunk of you. And then, um how many of you here have a Farcaster account? Okay, that's a that's a good thing. Um we we're not out of the woods quite yet with Farcaster, so I wouldn't wouldn't call it a resounding success, but we we have made a decent amount of progress, and so I'm going to share kind of some of the lessons we've learned along the way.

Um just my background, I was at Coinbase for 5 years, so it was a really early employee. Um did a lot of stuff in the kind of crypto fiat on-ramp world. Um so, a lot of the stuff that's happening right now in the news, uh I basically lived that 5 years actually 7 years ago now. Um and then, for the last 3 years, uh I've been working on Farcaster with my co-founder Varun, who also worked with me at Coinbase. Um the kind of general talk that I want to give today are just some principles for growing a protocol. So, I I generally have a a point of view that advice is not particularly helpful, and I wouldn't over-rotate on any one thing I say today, but it's kind of more, in my experience, things that if I could have

given advice to myself at the beginning, based on what I am trying to build, uh would have potentially been helpful. And hopefully, you might find that helpful today. So, uh I don't really have much on my slides, so that way you have to listen to what I'm going to say. Um but, the the first thing I think, whether you're building a protocol or or product, I I strongly believe in the concept of dogfooding. Uh and and it's just being your your own first user. Um I find a lot and I've done probably 100 angel investments uh both in crypto and and not in crypto, and it always puzzles me when I run into founders who are trying to build something for somebody else where they don't actually

feel that acute pain point or or haven't been in the position of the of the customer they're trying to sell to. That's not to say that it can't work, but in unless you have like a an extremely clear idea on on what you're going after, the ability for you to be your own user and and dog food the product um is just going to increase your rate of learning and the kind of uh the accuracy of what you're building, right? Like cuz if you, you know, and I'm sure this has been mentioned over the number of talks here, but Balaji Srinivasan has this term and then and Chris is actually the one who made it pretty popular with the blog post, the idea maze.

And if if you can't be your own first user, um it's it's going to be really really difficult to to navigate that idea maze. You can almost imagine it would be like if you can't be the first user of whatever you're building, you're you're playing a game of like yelling into the maze to have somebody else be the one that's actually navigating through. Versus the the closer you can get to the metal, the closer you can get to whatever you're building and as you being customer number one, the faster you're going to get feedback and and we're going to talk a little bit about like loops um later, but but that's like the first the first thing that I could say is like if you if you're not dog fooding, it's going to be extremely difficult. And and one lesson um that we learned at

Coinbase that I actually think significantly slowed down the rate of innovation at that company was somewhere kind of along the lines of like I think 2017-2018, you know, you had the big run-up in the price of Ethereum and Bitcoin, and then there kind of became this whole set of new financial rules that kind of got put in place at at Coinbase, rightfully so because the company was starting to get bigger, but one of the downsides is that basically employees couldn't do any type of trading, right? So, you have this kind of like traditional financial rule, but in terms of the ability to be using the product like most people at Coinbase went from kind of using maybe daily or weekly to maybe monthly and in this very artificial way. And and that I actually

think really really hurt the company and I actually know that Brian spent the last few years really trying to get to a place and at least in some of their products where employees had some amount of free reign in being able to use the product. Because if you're in the case of building a business where you're a fiat on-ramp and doing trading, if you yourself are not actually sitting in that seat, it's extremely difficult for you to kind of like have the right intuition on on what you should be building next. This toothbrush um comes from a an anecdote that Larry Page, I think, is is the one who said it at one point. Um but but good products, especially consumer products, uh meet the toothbrush test.

So, hopefully you brush your teeth twice a day. The idea that a good consumer product is something that you're going to use at least twice a day. Um now, obviously there's things like Airbnb where the frequency of interaction is is kind of maybe on a monthly or maybe just a couple of times a year. But, if you're trying to build an app, something that you think that someone's going to wake up and actually want to navigate cryptocurrency Web3 with uh if the use case is not two times a day, you're just going to have a lot longer feedback loop time. And that's not to say and especially obviously if you're building an enterprise use case, that might be a a little extreme. But, that said, if you just think about the products that tend to grow the fastest and you can actually kind of really get

to that level of of polish, uh you know, even an enterprise, Slack, Notion, all of that kind of stuff, it's it's products that you're using multiple times a day. Um and so, just if you can, thinking about what are those kind of small interactions that are going to be bringing you back to the an app daily if not multiple times a day. Um it it's just going to compress the the feedback loop cycle, which brings me to the the third point and and I'm I'm pretty sure Shriram might have actually brought this up. Um I always call it OODA loops, but I think it's OODA loops and this is a there's an excellent book on on the the person who came up with this John Boyd, uh but it it's kind of like a military framework for fighter pilots.

But I think that the simplest thing is you as an early stage entrepreneur and building a protocol, you're looking for these loops, right? So and these loops are basically feedback that you're getting whether you you yourself are your own first user or that initial set of five to 10 people who for whatever weird reason want to use your product, um they're giving you this feedback. And so the the sooner you can get to a place where you are actually uh kind of action producing information where you are out and and testing new products, having people give you feedback, having people tell you oh this doesn't work for me, that that's that's where you want to get to. And and so I think that those first two things I mentioned, the the dog fooding and the

uh the toothbrush is is about you potentially being that early first user, but but ultimately, you know, you have to scale beyond yourself. And and so getting to a framework where you're actually out there with people using your product and it it's constantly feeding back into how you're making decisions, I think is is a is like that that is the critical thing that you should be focused on. I mean, Paul Graham basically all of his advice tends to sound pretty trite and basic, but it's usually pretty right. And he's just like you should just be talking to users, right? Like that's the classic thing at at YC. It's just like build product, eat, uh sleep, exercise, and talk to users. And and so that is the thing that I did early on with Farcaster and and we'll talk about a

little later also some of the tactics that can help do this. But if if I think I did one thing right early on that actually also helped keep things going cuz there was a lot of churn at the beginning is Varun and I kind of split the the job of Varun was doing most of the building of product. I did a lot of the design in the front end. But I just spent 8 hours a day talking to users on on kind of Telegram, on Twitter, on just like Zoom calls. And so the kind of loops that I was able to start developing in terms of like product feedback and reasons why people don't use the product uh dramatically accelerated and informed at least how we were building and and navigating that idea maze.

Um I don't know if you guys have ever seen this uh kind of two-part photo. But this is how I think a lot of people in crypto and and this is going back to my time even at Coinbase. They kind of all think of this is this like they're going to develop this like beautiful engineered system. And then when they finally get to the car everyone's going to use it and it's just going to be phenomenally successful. And I think that there's like a little bit of truth in that if you look at something like Ethereum, like you have to actually get the EVM to actually work before before it can happen. But I think very very few startups can go navigate building in this manner. And the manner that I think is a far better way of

doing things is is a more iterative approach. Where at each stage of this journey you can actually move forward with that vehicle, right? The skateboard, the scooter, the bicycle. And then eventually you you get your way to a car. And so I think that's also a little bit of a trap that people fall into sometimes is especially cuz they they read something like the Ethereum white paper or the Bitcoin white paper and they think they have to have a complete V1 before they ship. So they spend a lot of time uh kind of on the whiteboard or or in in VS and a lot less time talking to users and actually kind of colliding with the market and getting that feedback.

Um and and to give you like a concrete example, this is the original kind of, you know, I don't have it all here, but the original Farcaster white paper. There's some irony here is that Blue Sky now has the AT protocol. We were actually originally going to call Farcaster ATTP, like at transfer protocol. Um and so we wrote this whole white paper in in October 2020, shared it with a bunch of folks, got feedback on it, and I I went back and read this before this presentation, and I think we got like 15% of it right. So, to think that like when you're building a protocol, the the 10 Commandments come down from Mount Sinai and like here's the white paper and this is how we're going to implement

it that it I don't think it's actually done like that. Um maybe Satoshi with Bitcoin is is actually just a a rare exception and and true genius, but I think in terms of actually building something, we we had an original architecture that looks like this, and seems like most of you are on Farcaster, so I can kind of maybe explain this a little bit. So, we had this this kind of vision of, okay, we're going to do the identity layer on chain, very small amount. Um we're going to build the first client, and then we're going to have this structured like the web, where content is hosted one place, and then in order to kind of have that be like a nice developer experience, you'd run an indexer, right? Because if you think about the web, you have DNS as the registry, every website's hosted on

a web server, Google is the indexer, and you have a browser. Um this architecture it it worked, but after a year of running a a beta, we realized that this developer experience, and and I know there are actually a few Farcaster developers in the room, uh it sucked. Because basically like we had done all the work to build our own indexer and all this magic, but for a new developer to show up, like it's just like a lot of overhead. And so, if we had kind of taken what we had written as our white paper as like, "Well, this is what our white paper is and so this is the implementation of the protocol." I actually think we would have been in a a much harder situation, especially when you're you're kind of trying to phrase it as a protocol and

other people are building on top of it um to be able to go and make the right changes for the architecture. Upgrading a live decentralized system like upgrading any system is really hard. Upgrading a live decentralized distributed system is extremely difficult and and we're going through this right now and we don't even have that many developers having built on on V1. And so, just to show you this is where we got to, which is still close to what we were doing before, but we we effectively came up with this idea of let's merge the host and the indexer into one server called a hub. And uh this has been um an immense amount of engineering effort. But, the other interesting thing is we actually if we if you're just going to take the the 2 1/2 years we've

been working on Farcaster, the first year and a half we probably spent 80 to 90% of the time working on our initial client on top of that terrible V1 architecture because we were really focused on getting users and kind of getting into that that loop of feedback from people. And only after we had kind of got to some threshold call it, you know, several hundred daily active users, did we then start work on V2, which we're just about to ship now. And we everything that we did in V2 was informed by real usage. So, rather than kind of having to try to come up with everything on on paper before starting the work, a lot of it was from the failings of of our V1 product. And and you'll notice this this client here and

the client down there is basically the same. So, for the end user they actually don't know that you're swapping out the back end, but for the developers who especially we're hoping later this year that that now are spending time building on top of the Farcaster protocol, this hub is just going to be an amazing developer experience. And and so I think the the key inside here is like get something out to market that people are actually using, whether that's end users or developers or or in our case we did both, as soon as possible because that feedback is going to be your most valuable map through the idea maze.

Uh another principle and and this I've become to really really believe and and especially having spent a lot of time talking to users of Farcaster is this concept of stated versus revealed preference. So Henry Ford, famous for saying, "If I had asked my customers what they wanted, they would have asked me for a faster horse, right?" And obviously the car replaced horses and and so and and Steve Jobs actually also has a a pretty famous line about we don't do market research at Apple. And I think it's it's important to kind of understand what that means because you could argue that like, "Well, talking to users sounds like market research. Steve Jobs said that was bad, so I'm not going to go do that." And I think that the nuance comes down to is and and this is I think a broader

psychology thing, when you ask someone uh what what you they think about your product, most people, you know, put sociopaths to the side, generally want to tell you something that you want to hear because they kind of want you to like them. It's just like a natural human reaction. And so even if they have like a more brutally honest like opinion of whatever their their opinion is of your product or protocol, they're going to they're going to kind of try to soften it on the side or some people just seeking approval are going to just kind of tell you what you want to hear. It's like, "Oh, yeah, I'd definitely use this.

Like, I I really really want to use decentralized social media." Revealed preference. I have 50,000 or 100,000 followers on Twitter and I spent 10 years building that audience. I don't care about your social network with 100 people on it. Um but they're never going to tell you that up front and and so you kind of have to as the the kind of talking to users extract that actual revealed preference. What what the person is actually going to do. And I think that that is where this kind of like when you're building a new product, actually asking people what they want is probably not what you want to do as much as A get something in front of them as soon as possible. See if they churn. And then

follow up and ask them why did they churn? They're probably going to give you a little bit of more brutally honest answer. And even then sometimes they're just going to give you something that's not that useful. So it's like, oh I just got busy. Well, it's pretty easy at least in my case because we're building something that looks like Twitter. I can go look at that same person's Twitter profile and I can see that they're either tweeting or liking tweets. So it's like, well, you weren't too busy for Twitter. Like you were you were too busy for Farcaster. And so you can really try to get to that core insight and it turns out it's like, oh well, I actually the the things that I'm interested in are on Twitter and there's not enough interesting stuff on Farcaster. And so you actually requires you to to spend a

little bit more effort in terms of navigating what people are trying to do to be nice and and get to the kernel of insight. And and and doing that at scale is going to give you a far better feedback loop towards like what you're actually trying to you know go and build. Um this is what I did at the beginning of Farcaster. Just like this is what I did for probably 3 months. This is like the 3 days in the middle of the week. I would just stack up a bunch of meetings. I would do a 15-minute, sometimes 30-minute Zoom call with people, show them the initial product, get them on-boarded, talk a little bit about the architecture, walk them through the whole product. So this is kind of like inspired by Superhuman.

85% of those people, maybe even 90% churned. And these are like people who were like my internet friends. I'd been in crypto almost a decade. Um they just like they didn't care like afterwards. They they were went back to Twitter to you know fight whatever Twitter battle they wanted to fight that day. But what I did get out of this is there were some oddballs who decided that Farcaster was interesting enough for them. And that was the kernel of people beyond Varun plus myself that we started to build the product around. So, one thing that turned out is like the people that actually retained tended to be a little bit more into NFTs. And so, when we started the product, like we were very much like focused, like okay, we're we're building a decentralized version of Twitter, like let's go and do that functionality. Um turns out that

like adding an NFT feed into the product not only got those people more engaged, but then those people were really excited about inviting or, you know, and and there was no even formal invite system. Like I was sending people a DMG link via Twitter DM to download I mean it was like really sketchy stuff in terms of like if you just kind of think about it, it's like random software you're downloading off the internet. But because that those people were so passionate about this like weird feature set that we had uncovered, we were able to kind of continue to grow things. And then what what happened was about a year later, we hit some number of people who were just using the app that now all of a sudden the people who wanted to use some version of a decentralized Twitter, there was like enough liquidity that

like like it became interesting. So, the initial approach didn't work, but we found some residual group of users which then kind of like we were able to build on, and then we hit some amount of scale that the original use case started to work. And so, what I'd say is like when you do 8 weeks of this in a row, like, you know, 400 onboards, it's like and most people are churning, that's like pretty demoralizing, especially if you've had success maybe somewhere else and and you kind of you you thought like oh like you know, I I I'm I'm so brilliant, I'm going to be able to figure this out. But you can either view that as a negative, like 85, 90% churn, or it's like, wow, I was actually able to uncover some 10% that was residual that this initial bad version of the protocol still resonated

with. Um Fourth principle. Uh developers don't care about your technology. I think this goes back to that kind of uh way of building an MVP. I think a lot of crypto people kind of tend to approach things as like, "Oh, I'm so so smart in terms of how I'm going to build this protocol and it everyone's just going to see the brilliance, the technical brilliance of this protocol and and go and use it." Peter Thiel in Zero to One has the the analogy it's like if you build it, they will come, if you've ever seen Field of Dreams.

Um and that's that's not the case. The thing developers want out of a protocol is users. And the the reason I have I have this guy up here is this is the original iPhone. I mean, some of you look pretty young, so I I like like this is uh 2007. Um the one thing to note is if you look at all of those apps on that phone, um one, there's no App Store. Two, every single app was built by Apple, including the Google Maps app and the YouTube app. So, even Apple with this magical device wasn't confident enough in a place where they were going to go have third-party developers. And if you want to get really nerdy, the the only way that you could actually build an app on the

original iPhone for that first year was this like weird local JavaScript thing, which now it's kind of weird with like React Native, it's like basically everything is Java- JavaScript. But, um there was no third-party App Store for the entire first year. So, Apple came with the approach of "We are going to will this thing into existence, do all the marketing to get this initial set of people to buy it." Um at the time it was pretty controversial, like there it was $600, uh there was no copy paste, there was, you know, no 3G, so it was really slow. But, and you know, Steve Ballmer has the famous thing where he's making fun of it, but they they were able to get 5 to 10 million people to go and buy a phone that they had built just themselves.

And then, the next year, they launched the App Store. And naturally, who were the people who built the first set sets of apps for the most part? Tended to be the people who bought the iPhone the first year, who tended to be pretty nerdy people who liked adopting new gadgets. And if you actually go back and look at the most popular apps from 2008, 2009, 2010, most of those apps don't exist today. Like, very, very few. And it's only when you get to 2011, 2012, 2013 that you start to like, oh, what's this app Instagram? What's this app WhatsApp? And and it it it actually took even a few years of the App Store existing for people to start to discover the the

native killer app for this application. So, there's actually a lot of early apps were very skeuomorphic. Uh if you don't know that term, it's just kind of like it it it's almost like when the early internet came out, people took magazines and newspapers and tried to make it look exactly like the physical item on a on a screen. And then over time, people figured out, okay, well, this isn't like how you build a website. Similarly with mobile apps, people were like, oh, well, we're just going to take desktop apps and try to kind of wedge them into a phone. And it took a few years for for effectively the feed to be discovered by by a a group of developers as this is the dominant way that people want to actually consume information on uh you know, a mobile phone. But but to So, if

you if you think about it like the the App Store is by far the most successful like developer protocol in the last decade, right? Put the web aside. And even then with with Apple selling all those iPhones and the adoption curve of mobile, it it it still took time. And so, I think the thing though is that no one went and built the iPhone app because like they were like, oh, the new technology. It was because they'd already been using the phone and Apple was doing all the hard work getting the users, right?

Like, getting in the App Store and featured on the App Store is is free distribution. And so, from our standpoint, um this is the Farcaster website, more or less has looked like this since the beginning. And I've gotten a lot of feedback. It's like this website looks sketchy. There's not like you know you've you've raised a whole bunch of money. Like why aren't you investing in a better website? Doesn't matter. Like the people who want to try out Farcaster kind of have found their way to my my Twitter DM inbox. And then the other thing and I actually got into like kind of a fight with Varun about this. Who he's usually right.

That we should we should at that that bad version that V1 version we should put API docs on the website. And Varun was like, "No. Like what will happen is we're going to keep building and if we get to a sufficient number of people using this initial like app and and protocol people will start just like figuring out like how to index us." And sure enough we we had a couple of developers who are just again early users. Who one of them just was like, "Well, you guys didn't implement a search engine. I'm going to build Searchcaster." And another person converted every Farcaster user into an RSS feed. There were no developer docs for that. That was a little bit of them like looking at the Chrome console and kind of reverse engineering our API

calls and kind of looking into seeing like, "Okay, what contract are they making calls to on the on the blockchain?" And so I think the the lesson at least from our standpoint is the thing to be maniacally focused on is the hard thing. And this is the kind of a related fifth principle. Uh daily active users. Right. And so um like I said I've done a bunch of angel investing. It always blows my mind when I get an angel investment update where someone reports monthly active users for a early stage product. Especially one that people are supposed to be using every day.

Because okay, think about this. I get you to open up one push notification a month. You're now a monthly active user. You don't use the app the rest of the month. Getting daily active users, that's what actually developers want. If If you're building it's like they they want liquidity, right? So, if you're building a D5 protocol, it's it's it's TVL, but like they don't want the total amount of TVL you had in a month, they want the spot how much actual, you know, liquidity is in this market. Same thing goes for for consumer social protocol. This is a super humbling chart. Like you This day you over here you you feel like you're doing really well, and then it just like comes back down, and every weekend it comes back down, and so it measuring the hard thing,

that's like actually colliding with reality, and and like you don't kind of get too high on your own supply that you're doing really well, because if you measure monthly active users and you have a decent amount of sign-ups, that number is going to look really good for a while, and you don't actually have to confront the reality that if this this number starts declining, your app is effectively dead. Like it just is on a lag. And so, if you are trying to build a protocol, finding that hard thing, the actual thing that matters for the protocol, and being intellectually honest about it, and and showing your team this, sending it out in your investor updates, being maniacally focused, this is the first thing I check every single morning when I wake up, that is actually what's going to drive

the growth of the protocol, and a protocol where this number is going up into the right, I can promise you developers will knock down your door, and and the protocol itself will will grow. And again, if it's D5, you can use TVL, and I think you can generally apply that to any app. It doesn't even necessarily need to be a protocol. I'll just leave you this last one. Um so, we technically started in 2020, we didn't have a client. So, there were There was a white paper and an idea and no client for the protocol. Um all of 2021, there were no other clients. And I kept telling people I was building a protocol.

And then in 2022, um we got to kind of two, and and there were other apps in the ecosystem, but uh DF over here actually I think was our our second kind of like client called the scuff and now coming into 2023 and the um you know kind of onset of hubs which we're really excited about we we probably have about six teams building what I would call like a full-featured farcaster client. Um that's a brutal amount of time if you think that you're building a protocol where basically no one else cares about your your protocol uh to basically have like year year and a half and then every time oh well it's a protocol so so other people are going to build it's like well okay well where are those other people?

But I think it it part of it is you you have to put the perspective it's going to take time. And going back to what I said before the best way to actually go do that is you're going to have to be getting some form of feedback loop hopefully from yourself but also from others as soon as possible. I want to end with a few tactics cuz sometimes the like advice or the frameworks it's like okay this is a little bit of like confirmation bias survivorship bias what what whatever kind of way you want to look at it. I'll I'll give you these three things that I think like all of you should consider whether you're building a protocol or not and they're a bit contrarian relative to most people in crypto.

So the first is I think Telegram is a is way better for building any type of early community than Discord. So Discord is A terrible on mobile and B it's extremely noisy and very hard to kind of keep in touch on everything right? So for a power power user maybe it works but I actually think if you have a simple one channel that you manage you your co-founder and you should be doing the management don't outsource it to a community person. I still do all the customer support for farcaster that's incredibly valuable source of of kind of customer feedback.

Um by keeping things in one channel and on Telegram most people who are in crypto or you potentially using Telegram versus Discord they They it on their phone and actually their thing that will bring them back to Telegram is one single chat. So, if every other thing is a dumpster fire on Telegram, but they have one chat, they're still probably going to come back to that app. I can promise you most people won't go back to Discord if just everything's a dumpster fire cuz it's not really a DMing platform for most people. You know, maybe maybe you're building something in Discord or related to Discord. In that case, take take the advice with a grain of salt. Second thing, uh I think I'm pretty well known for this at this point.

Um I probably have 25,000 DMs in my uh Twitter inbox. The actual website doesn't work for me when I try to use DMs, so I have to use like a third-party tool for it. But this is less about Twitter DMs and more about have a channel, and this could be Telegram, it could it could be Discord if that works for you. In my case, Twitter DM works where you can reach out to users that is not email. Because basically, what you're trying to find is when people go through onboarding, especially if you're kind of doing it maybe a little bit of a kind of slower release, um if if you have an ability to reach back out to them and just ask a really basic question of like, "Hey, so I you signed up, used the app, and then didn't

come back." Just that simple question and be able to engage with someone back and forth a couple of times is extreme amount of leverage from like customer feedback. If you go set up, you know, Delighted or Customer.io or whatever that sends some drip email that you never click on or respond to because you get annoyed that this is like a drip thing. Um it's not going to be that effective versus I can almost promise you and like there are a lot of like fancy names in that DM inbox that if I send a DM uh they're going to at least respond. It might not be the best response in the world, but like once you've established that channel, that's really, really valuable for for developing that feedback in those loops.

Last thing is I'm building a Twitter alternative. I tend to use Twitter to do the marketing. Um this can apply to all of you, but I I see a lot of people do very splashy big PR launches, which I at this point I I I just assume that most people should know that that doesn't actually matter. Like a TechCrunch article, you get a little bit of a bump, and then it comes all the way back down. What does work, in my opinion, is a constant stream of small updates. And especially if you're building a protocol, anytime you can highlight other things that are happening in the ecosystem, even if you think it's kind of like a little cringe or chilly, if you're doing it in an authentic way, that is kind of more your own brand,

it's going to actually resonate with people. And and so I I I kind of made this decision, and for a very long time I was very snooty about how my Twitter was, and I didn't want to be shilling stuff. And actually it was Mark uh Andreessen who was like, "Chilling is good, actually. Like, your whole job is to be doing marketing and sales, but just do it in a way that if you think this is cringe, like try to just be more authentic about it. And and people will see through that. And people actually want updates about things that are making progress. And so I think Twitter has been a really powerful source of inbound for us. And and yes, I have a decent-size audience, but it's not that big. And I do think that you can find a lot of early users, especially in crypto, on Twitter. And

because a lot of people have their DMs open, you can actually go and and just reach out to them and say, "Hey, would you would you want to check out this product? Or could I get a 15-minute slot for you and just give you a demo and onboard you?" And and you you're going to have a lot of no's, or just like ghosts, but there're also going to be a lot of people who are like, "Okay, this sounds interesting, sure." And and if you take the time to do like a thoughtful message, it it can actually be a pretty good channel. That's it. Hi, thanks for the talk. Um, what is the key metric that you use to measure retention?

I'm still learning this I because I didn't grow up in, you know, consumer social I like crypto and we very much focused on at Coinbase on daily transacting users or monthly transacting users. Um, so the the advice that I've kind of been given actually by Brian Kim at A16Z um, that we look at kind of three core metrics. So there's day one retention of like, okay, how many people sign up and they just never come back. Uh, day day seven just kind of say, okay, like that and then and then day 30. And I we we haven't actually quite figured out um, what number works best for us.

Day 30 is by far the hardest to to do. Um, but the other thing that one of the reasons I'm giving you kind of this evasive answer is what we found is that people don't churn forever. They actually there are moments where they come back. So a good example is like the Tali joined the network in November and there was like 50% of people who had ever signed up checked the network that day. Like and granted it was a small n but so I I think for us it's um, we're trying to actually develop a little bit of a better framework for us in terms of like what what do we think that the average user how many bites at the apple do we have before they really never come back?

And so far even two years in like most people and again, we're talking about 10,000 people in a beta so it might be different at 100,000 or a million. Okay, thank you. Yep. Hey Dan, thanks very much for the talk. Um, I'm always curious like what's what do you think is the end game or like what's the value prop of a decentralized social social network? So I think about it in two two ways. So the first is if if you have an audience in a decentralized social network that audience relationship is is direct between you and your audience. And actually, you as the creator, you don't own your audience cuz the audience decides if they want to follow you, right? Um so, but but there's no company that can kind of rug pull you on that,

right? So, it's like if I spent 10 years building an audience on Twitter or YouTube, they decide that they don't want me there, they shut my account off, there's no way to export those followers, right? So, I think that that's the value prop for a creator. That said, I don't think most are that focused or sensitive to it outside of maybe some political examples today. Um so, I think that the the more compelling case in the near term is for developers. And this is why we've spent so much time on on working on hubs is if we can get the developer experience to be just amazing. Um and and kind of like the ideal once hubs launch is someone can just go to GitHub, download a Farcaster hub, spin it up in a in kind of a Docker container, and within a

reasonable amount of time, you know, 15 minutes or something, they have the entire global state of the network in the same way that you would get with an Ethereum node for Farcaster. And there's no API limit like they could just do whatever they want with the data. And and and so, I think that experience um what we're excited about is if we can continue to grow the number of users overall in the protocol, and that's a little bit of like cobble together whatever we can and constantly change the the strategy on that. And we haven't quite figured out that, then you have a an addressable size of the the protocol that those developers will actually be the ones that start working on weird new ideas that end up being the the next Snapchat, the TikTok, the the

new social app or new app that is enabled by the fact that out of the box they get all of these like base level primitives from a social standpoint. And so, I I think where I'm most excited, to go back to the iPhone example, is if we can continue to drive the growth and adoption of the protocol, um I think developers will, you know, and you saw the progression, I think that there will be more and more developers that get interested. And then if you bet on the, you know, kind of aggregate number of developers who had now have true permissionless innovation, that I think is is the is the bet. And I'm old enough to have lived through the early days of Twitter.

And again, a lot of you look pretty young. In the early days of Twitter, the API was really open. And another important thing to note is Twitter did not have um its own mobile app. All of the mobile apps were third-party mobile apps. And so all of the main core Twitter features, at replies, hashtags, quote tweets, retweets, pull-to-refresh as a UI paradigm, that was all invented by by Twitter users and third-party developers. And so so some people sometimes think of like Twitter like nailed this like product. It's like, no, they actually they got the first primitive correct.

And that hit the the market at the right time and grew really really fast. But all of these additional features actually was just within the constraints of 140 characters and an open API, you had a lot of people um come up with that. And so that's that's the era that I want to bring back into social is where anyone with an interesting idea can just go do it and they don't have to ask anyone for permission. Oh, okay, I see. Yeah, that's that's super cool. Thank you. Thank you. Hey Dan, it's Emmanuel from Shield. Um you talked about DMing on Twitter. It's definitely something I like to do a lot of that. Um I feel like you can get a ton of user feedback kind of in a short amount of time by doing that.

Um but perhaps you can't you can't quite go as deep compared to like 20-30 minute conversations that you showed us you were scheduling with your users. Um so how do you balance between, you know, user feedback via text versus voice or FaceTime? Yeah, so again, a little unique in that I also get feedback like on Farcaster. It's like I basically have people yelling at me all day like on things that they want done, Um um which is a good thing by the way. So, I get feedback for my own products. There's a little bit of like the dog fooding actually gets a lot of that done.

I think that where DM works is if you've already developed like a pretty strong idea of like where the pain points are in your product, you can use it to confirm that that continues to be the case. And then if you start to find that you're getting new insights or things that you hadn't kind of understood before, that's actually useful sometimes then to try to get someone on a call or to spend a little bit more time getting the nuance. But I think at this point like I actually what I find really useful about um you know, the Twitter DMs is first of all, anytime the the like onboarding experience is confusing, I get that feedback. So, having done this enough like I kind of like now can reason about like, okay, I understand where that

could be confusing and and and improve it there. Um I I think it's also just like I can DM any user who's ever used Farcaster for the most part because I have their Twitter and I can just go there and again, it's just like it's a touchpoint. Uh and that could lead to a conversation. And so, I I had a call 2 weeks ago with someone who we we were going back and forth on Twitter DM and and they actually gave some pretty really interesting feedback and and I ended up, you know, doing 20 minutes with them just to kind of like make sure that I could fully understand their perspective on something and I think it's like may have changed our road map in a good way. So, I think I think it's like a first line and then you can kind of make

the call on on going deeper. That said, the first 400 people on the protocol I manually onboarded. So, I did a lot of that early on cuz I really really wanted to get all the nuance. Got it. Thank you. Hey Dan, Brian from Friends. Um how do you think about you guys continuing to get user growth and user adoption during crypto winter? And has your strategy like changed anytime during the last couple months? Yeah, so I had this joke with Shriram. Um in for most crypto apps, uh bull market is when you want to be be there because like people have a lot of money and then they don't have a lot of time so that they're using your crypto app. Uh with social media, if you don't have a lot of money in a bear market,

but you have a lot of time, like it's a perfect time to be growing a social media app. Um you know, maybe it's not as funny as I thought it was. But uh I think the the thing that really scarred me about Coinbase for the 5 years that I was there is we could never make number go up when the market wasn't there. And that's a function of at least where Coinbase is in the market, it's it's very much a financial product. And so convincing someone to do something with their money uh turns out is really, really difficult, especially if you're not like selling them a good. Like if it's like, you know, a an investment asset that that's actually very market-driven.

And so my advice generally is A, you should be building for crypto natives in in this market. If you're not, I could see the argument that you get everything in place that if the if market sentiment shifts and or, you know, kind of market is a little bit more uh crypto spring, call it. Uh then maybe more normies. But I But I generally think even if you're building for normies, if you can convince crypto native people to find use use case today, then they actually become the best ambassadors to your product, right? Like the reason Coinbase grew so fast in 2017, we had no marketing. There was no like nothing. It just It was all word of mouth. It's like, "Oh, if you want to buy Ethereum, buy it on Coinbase." And

it's like that that that is a a brand-building exercise. But that's because the crypto natives in 2016 who wanted to buy Ethereum, Coinbase was the only place to do that in the US, and so those are the people actually really drove the growth. And so I I think of it is like if you can if you can retain and deepen engagement with a smaller group of people during maybe a a kind of like more crypto winter period, that's going to set you up for A, you're going to be having feedback the whole way, so your product's going to continue to be improving, but at the same time, when the market shifts, like they're going to be the ones that are actually helping drive growth for you.

Cool. Thank you. Hi, Jay from Eth ID. It seems like a forecaster has two different groups of users, the actual end users themselves, as well as the developers. I'm curious when you're doing all those interviews, are they primarily with developers or with users? So, we took a and I wrote a blog post about this as product led protocol development is like how I like to frame it is and it's again inspired very much by how they approached the iPhone. Our view is that no developer is going to care about that V1 version of Farcaster because they would be like, oh, well, I'm going to go build my own version of it cuz it's not that hard or, you know, I'm going to do the more on chain version or

there's another alternative or why didn't you do this technology choice? That's basically for the first year and a half, I just had everyone criticize all the technical decisions for for Farcaster. And I actually I as little bit more like the the kind of like sales version of of it, I I I take all that like really personally and then Varun would be like, it is like, you know, doesn't matter. Like, we have to just get people using it. And once we get people using it, people will start tinkering. And that very much is the case. We There was not a single hackathon, not a single marketing thing, no partnerships. I I used to run business development at Coinbase. So, like, anytime someone's like, oh, you know, Polygon's BD team and and they're genuinely talented for a BD team. Like, the reality is like, you

can't fake organic traction with developers. And the best people to potentially be building are the people who start using your product early on. So, I like a good example is like Uniswap, really successful. Uniswap has a a product. 50% of, you know, some really large percentage of Uniswap's volume comes from uniswap.org. So, they started with a product and then that was able to turn Uniswap itself in into a bigger protocol. And if you look at most of the successful DeFi protocols, they all have clients.

And so, I think the yes, if you're building an L1, I could be you could convince me that you don't necessarily need that. But basically everything else, you have to be dogfooding and bootstrapping the initial set of users that then shows showcases the power of your protocol. Um in addition to if you're building on top of your own protocol, you're going to know all the pain points and the things that you actually want to change because you're in the the shoes of the developer before they even get there. Makes sense. And second question is what would you say would be the most surprising insights as to attracting developers?

Uh just goes back to I don't think developers care about a lot of stuff that people do that I think is like generally fake work. It's just like they care about is there a usage and if I go and build something here, will other people use it? And so, it's like if you just focus on that and and making that path to okay, I have an interesting idea, you have users and a thing that I can go build on, like compress that down as fast as possible. Um we've done, for example, a bunch of stuff like our our docs are in notion. Like we could have gone and invested in a GitBook implementation, but we knew that things were going to be changing around.

Far I mean, you can ask DF if we've done anything good, but like I think generally um like Replit or or StackBlitz or what like just like make things like really easy for people to go fast. And I think that one last one and and Varun had this insight is so, we have this technical white paper effectively for the farcaster protocol and he did a 45 minute YouTube series of five nine five minute YouTubes where he goes through each section he gets excalidraw which is like a whiteboarding tool and he does a screen recording and we didn't even use loom we just use the slack has a built in five minute like max video and he just walked through each part of the protocol and I think a

couple of them took him you know each time like one or two reps but then that actually became like since we added that like the number of people that like started to grok how the protocol was working cuz it was all there in the the white paper but like no one wants to read your white paper like versus like being able to watch something on YouTube and Varun is pretty good at doing that like I think like that also really helped like compress that that cycle time of like okay you have users something to go build on like how how how can I figure this out in the fastest time possible Thank you.

Hey Dan this is JP um so I was earlier user of farcaster I actually still remember the magic moment was you DMing me I I think after I joined or no replying to something liking everything I posted how did you figure out like what what was different between the people who stuck around versus who didn't I'm guessing some of those behaviors and some of the things I experienced were not accidents so like what did you find after you onboarded say 500 or 1000 users yeah so um my Twitter friends going into it who you know all stated preference that they were interested we need decentralized social media like this is an imperative they all just continue to say the same stuff but on Twitter cuz they have big audiences so that was a mistake in the

sense that and I had to go through that like I've like okay like if you have an existing audience like for something like farcaster like your likelihood of sticking around is is very low um Eugene Wei actually has an an essay called status as a service if you're thinking about social products. And it's just basically it's like every new social network mints a new set of and he actually uses Bitcoin proof of work. It's like a really really great essay if you've never read it. But um each new social network basically mints a new group of people who they didn't have a big enough audience on the old social network. So their next best alternative like, "Okay, I'm going to go actually invest in this new one and then grow with that."

Um I think one I mean this was kind of copied from Twitter is we put a recommended user list pretty early. We didn't try to make it deterministic. It basically was my taste of okay, who's active on the network and if you stayed active on the network, we basically gave you followers which tends to if you talk to those people, they were like, "Yeah, it was awesome. Like I started just basically getting followers." And it's like well, turns out like a lot of people who are on Twitter, what they're trying to do is build an audience. And so that's actually one thing that you can control but that comes down to if you have built the initial client and you're the primary way of doing the onboarding, you you can actually control that firehose.

So I think that that was important and then I think it was just I I spent a lot of time talking to the users uh who who stuck around for whatever reason and I started looking at their Ethereum wallets and it turned out like basically everyone who stuck around was like super into NFTs. And so we built an NFT feed and that was it that was just a bet. Like we we tried token gated posts, didn't work. We just deprecated those from the product. Like we tried five or six things and and NFT feed was actually the thing that helped us get from call it tens of DAO to hundreds of DAO. And then once we got to hundreds of DAO, the just like the network itself started getting interesting enough that people just stuck around even if they had no interest in NFTs.

That's super cool. How did you decide like who you know, you chose like the first 500 people to onboard? Mhm. Um were they the right people? How did you know at what point like what was your original rationale there and what would have had what would have been now if you were looking in hindsight? Yeah, if I was to do it in in I mean it's it's it's kind of hard with the counterfactual, I would say, but like basically I just hit up my Coinbase network and then I hit up every single VC that I know. They all wanted to like potentially do a seed round. We actually said no to all of them. Um but they at least would do a 15-minute call with me to go through onboarding.

Um that got me probably, I don't know, 80 people. And then even by then I would just ping people that's like the the few people that stuck around it's like I'd just ask them like do you have other people? And so like I remember one day randomly I didn't have no idea what South Park Commons was. Someone from South Park Commons was like, "Oh, like I have all these other people who are in the web 3, you know, this 2021 so it's like all, you know, hyped stuff." And so that became like 40 different people got, you know, onboarded via that and and so so so you just kind of just follow your nose of of basically anytime you have anyone who's excited or engaged, like you just like keep that person like really well protected and just like constantly making sure that

whatever they want because that's the kindling of like going from the the one person or the initial team to 10 to 100 to 1,000. 1,000. Thanks so much. Yep. Hi Dan. Thanks for the talk. This is Brandon. Um congrats on a phenomenal protocol. I'm a big fan of the premise of sufficiently decentralized social media. And I think Farcaster is a really great design. I guess my question is a little bit more philosophical. Do you think that web 3 introduces a new superpower when it comes to designing social media protocols or is it a little bit of a barrier because if I understand correctly with Farcaster, you know, we're using the wallet to sign messages and send them to data hubs, but you could sign those messages using another kind of cryptographic uh program

without having to deal with crypto. So, do you see it as something that's unlocking something fundamental there? Yeah, I think the way I approach it and the reason I got excited to join Coinbase back in the day is it very much aligned with how Chris thinks about it. It's like computers that can make commitments. Like I've always been more interested in the public blockchains, permissionless public blockchains are a new type of computing primitive, rather than the financial aspect. Like I'm friends with Balaji, like his million-dollar Bitcoin bet like interesting intellectual discussion, but that's like the reason I'm in crypto is like I want to build new products.

That's why I moved to Silicon Valley originally. And so the the way I view it is um a blockchain, in the case of social, the the core thing, and this is the sufficient decentralization of social network thesis, is if you can get the identity, you know, your pointer online that says like DWR is this address or this signature, out of any one company, country, individual, like, you know, no one else can muck with that, that to me feels like the future. And if you want to talk about the metaverse or any of this other kind of stuff, we we can't exist in a completely digital world if like people can be disappeared like that because of like a terms of service violation. And

so in in my view, like I'm not aware of any other technology that you can actually truly get to, because, I mean, some people would say Ethereum's not decentralized, okay. Um like you if you get something in Ethereum smart contract and you program it correctly, that that's a to use a term from Jacob Horne, a hyperstructure, it'll run forever as long as Ethereum is there, and that logic will will hold. And outside of maybe quantum, you know, cryptography, which it sounds like you guys had a talk on that today, but like I think um like it's it's truly a new type of institution as humans we can actually have. If if you want to make the bet these blockchains are basically going to run forever, which I I think more or less they will.

Yeah, I think you're absolutely right. I think the idea of having the the kind of a programmable permissionless world computer with like a verifiable data registry to look up entities and identities is is really really smart and that's why I'm a fan of Farcaster. I just wonder in terms of the message signing, does the mainstream consumer have to have a wallet to do that cuz you could sign messages without a wallet? So, actually in V2, we moved all messages so you you have to have a root wallet at the end of the day. Um but I think with like 4337 and and some of the other stuff is we may never even expose that to users like and just be like in a recovery scenario you can grab it.

Everything else on Farcaster in V2 is actually signed with an EDDSA key so it's a non-cryptocurrency address. So, you can think of it as like a decentralized OAuth token. And I think that's important for two reasons. One, um if you have to custody someone's keys, that's actually kind of stressful for you, right? Cuz if people throw money in it or something in it and then you lose it or it gets hacked, now those people are going to be really really mad with you versus like an OAuth token, they'd still probably mad if you got hacked but like not like come to your house and like be beating down the door.

Uh the other thing is that in the US it's regulated. So, the moment you take weird that we have different crypto curves that are like fall under regulatory but the state of New York if you host an ECDSA key or an Ethereum key in the cloud, the state of New York considers you a custodian. Uh you don't I haven't gone through that at Coinbase. I promise you don't want to have to go get regulated by the state of New York. And so, our view is if someone's building a social app, putting an EDDSA key in the cloud, it's no different than an OAuth token. You can't shove any crypto in it. So, that that's actually what but we wouldn't have known that until having spent a year building in the first version with the architecture that's assumed everyone's going to use an Ethereum address.

And so, that goes back to my point before about just like get something out there and start collecting feedback. Cool. Thanks so much. Appreciate it. How's it going? Great talk. Thank you. I actually onboarded to Farcaster yesterday. This guy over here onboarded me. Great. Um my question is just regarding you guys's like global like vision. Like how Are you I'm assuming you guys are starting in the US, like Europe-ish, like where does Latin America, like Asia, how do how do you think about like when do you acquire those users, like how do you speak to them differently?

Um just that that type of stuff. Yeah, so I ran international for Coinbase. That was actually kind of like my first job. Um I have a little bit of a contrarian point of view is like crypto is very uh developed country-centric for the most part. I think there are users everywhere, but I think in terms of like if you're focused on crypto natives and people who just like live and breathe this stuff, like the the classic crypto Twitter user, I can guarantee that a large percentage of those people are in developed countries. And so for us, I think initially we're not focused on like a global strategy, internationalization, all that. It's just like find people who actually want to use the product every day. Make that dial chart instead of being like this, go like that. And once we can figure out

what the the kind of like the core app experience, use case, features, all of that. Once we've nailed that, then then you actually go broad and think about like global expansion and stuff like that. Whereas I think um if we were to do that now, we can get a lot of like false signals, right? Like you can get a bunch of sign-ups, I don't know, in Indonesia. And now who who are you building for? Like the people that you got a bunch of people to sign up in Indonesia where I don't know anything about Indonesia like I've been once. Or no, I haven't even been. I've been to the Philippines. So it's like I So that that that's what I worry about.

Whereas if I'm building for a crypto native person like any of you in this room I have a much better understanding of the psychology, what they're looking for out of the experience. And so play to strengths, and then once you figure out like you have product market fit, then I think work globally. Hey um Varun asked me to heckle you, so I have I have two questions. Okay. Um the first one is uh I'm curious how you're thinking about uh DAO growth uh as as a builder in the ecosystem. Um it's kind of an evolving answer.

Uh I think we've been really really focused on quality cuz I think generally the fastest way to lose active users on a network like Twitter is if you have low quality whether it's a spammy reply, bot reply, a troll, uh you know, nasty negative person. And so I'm just very cautious about growing too fast. Um I do think we probably just need to move faster towards getting the network to be completely permissionless though. And so I think we're we're going to have to be thoughtful at thinking about our client Warpcast, but also for the developer ecosystem is in a world where anyone can go sign up. And so you

can actually grow at the rate of like what what the natural DAO growth in the network would be. How can you make it so the default developer experience you don't have to go rebuild like spam filtering, right? So if we spend a ton of time doing that, how can we think about that in some type of a decentralized way? Because I think that is the the downside to like going too fast on DAO growth is like the network quality goes really, you know, downhill really fast. And there's nothing worse than you to potentially get a little red dot that you got a notification and you open it up and it's just like some worthless post because someone's airdrop farming or, you know, whatever. And so I I think that that's the the tension. And so

hopefully in the next 3 months we have a solution for that. Cool. Um and the second one is um I I guess like the DAO growth could come from the ecosystem rather than your client, right? And and I guess that's what makes like Farcaster special as a protocol versus something like Twitter where it where it you know now with Twitter you can't really have third-party clients. So how how do you see that cuz like it's very hard to onboard new users to my app for example without them going through farcaster and warpcast first and that's like almost to the point where it's like not worth it for me to try and acquire users from outside the network and just focus on like making the experience of people who already use it better.

Yeah, so I think going back to the getting the network to be permissionless that enables any developer right like whether it's you or the jam team like you can try to grow the your app in any way you see fit. And so I think that that's important for both the neutrality like credible neutrality of the network and and the success of the protocol. But I think actually it brings up a point of like I think a lot of people also get very focused on it's like oh well the network isn't permissionless at the beginning it's like doesn't need to be. What you need to get to is like figure out like sort through the kinks first and if you're if you're actually completely permissionless or completely decentralized you're just going to move a lot slower. And so going back to the point of like how can we get the network

now that we we have a lot of learning under our belt to actually permissionless state it enables developer like yourself to grow at the rate that whatever you think is the appropriate way to do it. But I do think trying to see ahead of certain problems that will happen as second order effect of making decisions um that is also important right and so it's like I want to make sure the if if we if if the network can grow as fast as it wants there's going to be a lot of spam and then instead of working on delightful features that like make your users happy you have to deal with spam which is like basically like doing good spam for the network it doesn't actually make anyone happy like you don't get credit for it.

Uh versus you know like if we can actually be thoughtful as we make those changes at the protocol level everyone gets that basically out of the box for free. And so that that that's how I think I would think about it. Awesome. Thanks, Dan.