AI's Distribution Shift: The Land Grab Ahead - Unsolicited Feedback S3E8

Speaker A: Hey everyone, we have a crash course for you in the next great distribution shift. I'm Brian Balfour, I'm founder and CEO of Reforge. I'm joined by my co host Fareed Mossavat. In this episode we pull back the curtain on why AI's real battleground isn't the tech itself, it's the fight to be the next distribution platform. Fareed and I dissect the playbooks and cycles that crown Facebook, Google, Apple and LinkedIn as the winners of their categories and turn them from open platforms into complete toll booths. The key part is who's going to be next and what do you need to know to play the game. We cover my prediction on which LLM player will create the platform first and exactly how to ride that wave before the gates slam shut. We also cover how startups need to play the game differently versus larger companies, and candid debate on platform moats, memory versus action and whether PLG just made a ah, roaring comeback. If you build, invest or obsess over AI products, this 45 minute sprint will hand you hard truths and the hidden opportunities shaping the next couple of years. Plug in, level up and learn how to play the game before the game plays you. But one of those keys will be how do you stay ahead of the curve? And that's exactly why we built Reforge Insights. As distribution keeps tightening and competition is increasing, your critical lever is how you listen to and adapt to your customers at scale. Reforge Insights aggregates every customer signal from surveys to user interviews, sales calls, public reviews, support tickets and a lot more. It turns it into crystal clear data, back themes segmented by Persona, customer segment, feature usage and more and allows you to set up personalized views for every team so they get exactly the feedback and customer input they need around what they own. The team has been on fire lately. We just launched AI powered surveys that essentially draft themselves, identify the perfect cohort of users to survey and a conversational AI agent embedded in the survey itself to ask follow up questions to ext more insights out of the user just like a human researcher would. Don't get blindsided by the fast moving shifts of your customers. Be the first to see it's coming. Check out Reforged.com insights for more info. We hope you enjoy the episode.

Speaker B: Brian, you wrote an awesome post this past week on called the. What is the exact title?

Speaker A: Sparked from our last conversation.

Speaker B: Sparked from our last conversation uh, about the next great distribution shift. And this has generated I think a ton of conversation, ton of posts, reshares Questions, comments. Even though it's built off of our last episode, I feel like I'd love to just dig in on this more with you and like, hear what the reaction has been, what your core thesis is and sort of, um, you know, what, what you think it means for companies building in AI or building around AIs.

Speaker A: You know how it was a good post. Hmm. It's when, uh, it's when the naysayers and the trolls come out. All right, you get a little of that portion. You know, you wrote, you know, uh, you wrote something decent.

Speaker B: Yeah. If everyone says yes, I agree, then they probably didn't read it.

Speaker A: Yeah.

Speaker B: Or just like clicked on the title.

Speaker A: Yeah, I'm going to lay out the thesis. I should talk about my personal core prediction. Um, and then there was a bunch of good follow up questions and comments where I think we should go a lot deeper. Deeper, because there are some pieces that people I think misinterpreted or start to kind of point to the next space. Okay, so the premise of the whole thing is, um, so Casey Winners wrote a pretty good post. It was about a year and a half ago now. Right. And his main post, the, uh, point of that post was we have had this technology shift without a distribution shift yet. And that actually the most powerful shifts are when both of those things are combined. But his big point was 18 months ago, they don't tend to happen at the same time. If you actually look at history, the technology shift happens first and then the distribution shift lags. That lag period has started to get consumed. Right. And it's been something that's been, I've been thinking about constantly. And I think we are now in the moment where we're about to see the distribution shift happen as part of this technology shift.

Speaker B: And I would just like to add and point out it's actually the opposite of a distribution shift. That distribution has actually gotten substantively harder as a result.

Speaker A: Which is, which is, which is definitely different. Uh, which is definitely different.

Speaker B: And we've been over it a million times, so probably don't need to, for those who are listeners, go through it. But SEO is harder. Store distribution is harder. All ads are less effective. Like they're all of the things you could rely on in the past have gotten substantively more difficult.

Speaker A: That's correct. But I think it's also what is going to make the conditions perfect for this moment. Because everybody is looking, right, and it kind of gets to the point that we'll talk about, which is the cycle is accelerating. And I think we expect this cycle to be much shorter than the previous ones. If you look at all the previous, uh, distribution shifts from Google to Facebook to Apple to even some smaller distribution channels like uh, LinkedIn, they all follow the same exact cycle. The first is that there is essentially a step zero. It's the competitive conditions that emerge which is that the, there is consensus that a specific category is going to be monstrous, but the winner is not clear yet. And this is true. Like when Facebook was started for starting emerging, there was MySpace Orkit High 5 Friendster, right? Like a lot of people forget this, right? Um, or when Google was emerging there was like Yahoo and all of the other alternatives. And when Apple first launched the iPhone right there, how many different phones were on the market? We forget this. But that, that competitive condition of consensus that the category is going to be huge and multiple big players going after it is what creates the environment for the real first step, which is it forces all of these people to think about what is the moat going to be for this specific category. And um, and it tends to take some time for people to truly figure out what that moat is. Of course in the social realm it was about aggregating the social graft. In Google and Search it was more about uh, the data cycle. There's different elements to the moat, but once the moat is identified then one of the players tends to what I would call like opening up the gates as a platform in order to acquire more of that moat to hit the escape velocity across the others. Right? So they create these open ecosystems and usually the exchange is some form of hey, come onto the platform. I'm not only going to enable you with a new capability, but I'm going to give you some form m of organic distribution as an incentive in order to come onto the platform and either extend my functionality or give me more of the moat to hit escape velocity. So that's the second step which is opening the gates. And of course everybody rushes in, new businesses are created. You know, it's, it's awesome. But then inevitably what happens is the gates start to close, the platform starts peeling back all of the pieces typically in order to monetize. And they do this once they feel like they have escape velocity on the moat and by that point there's users have no choice to stay. And so they feel comfortable in changing the rules. And they change the rules in a number of ways. They either build first party versions of the most popular third party applications, they will tax directly by taking some percentage of revenue, or they tax indirectly by artificially Constraining your organic distribution to push you towards all their paid advertising mechanisms. And this cycle repeats over and over and over again. And so in the post, of course the first one I started with was Facebook because I was a personal master,

Speaker B: because we were there.

Speaker A: Yeah, right. My first company was a casualty of this. So I had PTSD from it. And that's why, uh, this is so visceral for me, which is that they were had a competitive landscape between them. MySpace, High5, Friendster, Orkut, there was a bunch of regional players like Bebo, Studi, Valzette, uh, there's a Russian one that I'm not even going to try to pronounce, uh, 20 in Spain. And so Facebook, they identified the moat which was like the social graft and that direct network effect. They opened up the platform in 2007. They were like, hey, like we're going to give you this canvas to put things. We're going to give you a bunch of organic distribution. And uh, once they really hit escape velocity, they started closing for control. They started clamping down the organic distribution push people towards ads. They went all the way to the point of absorbing what they felt like were the, the best applications, uh, into their first party ecosystem, essentially shut down the entire platform. But this repeats. Apple did it with their App Store and I think a bunch of mobile developers are probably shaking their head of like, oh my God, the 30% tax, all that kind of stuff Google did as well. It took longer, it was over the course of 20 years. But like do a Google search Today, something like 60% of the real estate is covered by ads and zero or

Speaker B: first party or first party stuff. That's right, they did a combo answers, et cetera.

Speaker A: Yep, that's right. And the one that's been most painful for me personally lately has been LinkedIn because, because they really did an emphasis over the last few years of okay, we need to move beyond this like digital resume thing to more of a true content and engagement platform to collect more data to sell to recruiters, salespeople, you know, target job postings, all those pieces. And so they went through a couple year period to really attract organic content creators to their channels. And distribution was amazing. It was way better than X and in some of the other platforms. But then the canary in the coal mine started to come when they launched thought leadership ads, which is you could sponsor your own ad. So they created a monetization mechanism, you know, for this. And it's super clear over the last four to six months they are artificially Constraining organic distribution of folks now to push people towards thought leadership ads. I've seen it anecdotally on my own personal stuff. We're in this Slack community with dozens of other professional creators who have all seen it. And so you know, they're, they're essentially going through, they went through that cycle with company pages, they're now going through it with personal pages as well. So that's the cycle and those pieces. But let me pause there before I get to like what my prediction was and in some of the advice.

Speaker B: So do you think this is malicious generally on the part of these platforms, the result of natural incentives, or do you believe that the, the stance that a lot of these companies take, which is they are, they need to defend their platforms in some meaningful way? For instance, Facebook's argument on a lot of the throttling on the Facebook platform was that it was degrading the experience, right? Yeah, some of it is that. I think the most common one is the absorption of the most popular first part third party use cases.

Speaker A: Right.

Speaker B: Music, photos, videos, et cetera on Facebook is a great example. Apple or Microsoft and the PC era, building things like Excel, Word, the Office suite to sort of take out the biggest players directly compete with the biggest players in their ecosystem, um, et cetera. Some of those, you know. Yeah. How about, I don't know, like when you first read this, it sort of sounds like the evil bad guys will steal from you eventually. But I don't think that's exactly what's happening. Right.

Speaker A: I'm not personally expressing judgment of whether this is, this is good or evil. I think everybody's going to have their own interpretation of that. I think the reality is this happens because of capitalistic and competitive pressures. You see instances where Twitter shut down like Periscope and Vine. And so a lot of these platforms are like, well, look, I don't want to enable my Disruptor through my own platform, right. And so like that's one version of it and the other version of it is just all these companies, like they have to grow, they have to keep growing over a long period of time. And so that's what you saw with Google is they just found more and more ways to grow those ad dollars by consuming more real estate. I think it's a result of essentially those two pieces. So I think that for developers and creators, the thing to keep in mind is to understand the cycle and play the cycle. They're playing you, so you play them. Right. And that's business.

Speaker B: Yeah. I think another piece to Recognize in this pattern is that there are coordinated interests between the platform, let's call them developers, but publishers also, who are supply side of the marketplace on these platforms and the users who are the demand side of the marketplace. There are things that are coordinated. I'd like my iPhone to do more things because I carry it in my pocket all day. That's pretty well aligned across all three of those players. Right. But I would like a consistent purchasing experience, you know, that I can trust and easily manage from a single place. Means is good for Apple because they can, you know, build their payment platform and take a cut of that and charge a high rake because of their distribution and the simplicity of it. And it's good for users, but bad for developers. Right, because the developers lack control, independence and ability. It is hard to build complicated monetization models on the Apple platform still today because of not just the tax, but the restrictions on the different ways you can monetize there. The same is true for Facebook. What's good for developer flooding channels full of notifications and content might not be good for users, might not be good for the platform. So I think some of it is the cap, is the need to grow, but some of it is the inherent tension in these platforms that not everything is aligned across all three players. And so there's an incentive misalignment that you have to understand. Often the supply side can pollute these marketplaces by overdoing it because the distribution is so good and that's attention there. And that certainly happened on the Facebook platform. It definitely happens on LinkedIn.

Speaker A: Uh, you know, that's right.

Speaker B: But also then the revenue and the business interests of the platform are also a piece of the puzzle. So it's sort of interesting to see that when you look at these dynamics, this same pattern emerges time and time again. It's almost like a law whether or not anybody wants it to be this way. This is the natural state of things.

Speaker A: It is what it is. Like, that's my mentality, like I get frustrated, but it is what it is. You just need to know the rules of the game so that you can play it, play it wisely, which kind of gets to this next phase, right? Which was my personal prediction. So if you follow the pattern, kind of that step zero, the competitive environment, I think we're at that ideal phase, right. It is 100% clear that these like AI chat experiences, like a ChatGPT or Claude or Gemini, is going to be an incredibly big and important category. There is no clear player and you've got Four or five major players vying for, for that uh, for that category. You've got OpenAI with chatgpt, you've got Anthropic with Claude, you've got Google with Gemini, who the hell knows what Apple's gonna do. And, and you get, you know. Yeah, yeah. And you got meta in the game, you know as well. And so the competitive environment is ripe and to our point earlier added to that it's destroying other distribution channels. So everybody's looking for a new place to go as well. Which I think plays into the mix. I think the. Then the first piece is identifying the moat. I think the thing that more clear now than it was like two years ago is that the mode is going to stem from this combination of context and memory. And so I think they're starting what

Speaker B: I thought it was which was size of model and model intelligence.

Speaker A: Exactly.

Speaker B: I think has while is still important and requires high capital is something all of the players sort of are chasing at the same time.

Speaker A: Right, Exactly. So now that that's identified then the question is is who is going to accumulate that context and memory the fastest and how do you do that? And I think that's where the platform plays in. The distribution channel starts to play which is my personal pred prediction is that OpenAI with ChatGPT will be the first to it. And the reason is they're. They're showing clear signals around that one is like they are probably the furthest along on the memory side of things. They've been doing a ton of plays around integrating context. They recently just launched all the connectors with deep research. They're actively soliciting more platform integrations and they're starting if you look through their job descriptions, they're hiring for a bunch of roles around like agent, platform and infrastructure. And so my prediction is that they are going to launch some form of a more open platform and there's going to be some value exchange where it's like we are going to give you distribution and other capabilities through our platform and exchanges you integrated to it. It's going to drive more context and memory accumulation for them.

Speaker B: How do you think about. I've always. I think context and memory. We talked about this in the last episode are really, really important. Do you think action, another moat here, meaning the models that can take the most actions and do the most things will emerge? So I'm thinking about ncps, I'm thinking about tool use for agents, et cetera. Or do you think that's all in service of context?

Speaker A: I think it's going to all be in service of context because the agents and the tools, in order for them to work properly and do amazing things and do more and more amazing things, is they need access to, to the context in memory. Right. And so there's, there's a bit of a loop there which is like I plug my, my agent that I built or into Chat GPT. I'll probably get access to some form of context and memory to, to make my agent work, which will drive additional usage on ChatGPT, which accumulates more memory that's like stored on ChatGPT. Right. And like the whole thing kind of fuels itself. And so, and so I think that's a bit of the, I think that, I think that's just part of, it's going to be part of the exchange. That's my hypothesis at least.

Speaker B: All right, so what's step three?

Speaker A: Well, step.

Speaker B: Does it matter? Like we're going to open, it's going to open. I think one thing you could say if you, uh, believe this law, is you shouldn't build on these emergent platforms because they're just going to go away. So you should try to be independent.

Speaker A: Right.

Speaker B: Uh, I think that was a mistake in past, you know, past eras around us. But I think what you're really recommending is there's a time and an opportunity and you have to know that step three is coming. Is that sort of the thesis you're making here?

Speaker A: I think the point that I was trying to make is there is no opting out of the game. And because if you look at it in isolation, like the integration that HubSpot did with Deep research, right. You look at that in isolation is like, well, why would HubSpot essentially want to become a database to ChatGPT's interface and have all the usage that makes zero sense in isolation. But we don't operate in isolation, we operate in a competitive environment. And that starts to lead to the prisoner's dilemma, which is just if my competitor is going to integrate with ChatGPT, ah. And then all of a sudden my customer sets start to love and like that experience, it essentially forces me to play the game as well. So if you're going to be forced to play the game, like, be early, be smart about it and just play it wisely. And I was trying to make the point, don't be naive to what will eventually happen, which is that they will at some point start clawing whatever they give back, uh, and start to shut that down. If you have that in mind as the inevitable Destination, then I, uh, think you can start to make much more deliberate moves, much smarter moves of how to play the game wisely. One thing I want to be clear about, which I didn't realize until I got a bunch of comments on the post, was a lot of people assumed I was actually referring to ChatGPT search.

Speaker B: Okay.

Speaker A: And there's actually two distribution mechanisms that may emerge out of ChatGPT. There's the search experience, right. Which is more of like the replacement to the Google experience. But the big point I was trying to make is that uh, there is going to be a second piece here which is more of like an app or agent platform.

Speaker B: Yes.

Speaker A: That all of the tools start to integrate and other products start to integrate towards both consumer and business. Right. I think it's viable for both. So there's actually going to be, I think, two different mechanisms that might emerge out of distribution mechanisms that emerge out of the cpu.

Speaker B: I've seen this a little bit with some of the experiments they've run in the past with like custom GPTs, for instance. This desire to create an ecosystem hasn't been particularly successful, I don't think so far, but to allow people to build apps built on top of the GPT ecosystem to be able to do specific tasks more effectively, not just through the API, but actually in the ChatGPT interface. Right. Yeah, I think that's a good point, that there's that version, but there's also the how do I get discovered version. Because I asked ChatGPT a question about what the greatest restaurants in Berkeley are that have been opened in the last 30 days.

Speaker A: I think on that point, if you're like a developer or creator, like, and then you know, the cycle, like the first reaction is what you said, which is like, screw that, I'm not going to, I'm not going to play that game. And part of that is a, uh, lot of questions around, well, like, you know, why isn't something portable is just going to emerge where I can just like kind of, you know, port all my memory and context from platform to platform. And look, there are some initiatives around that which may have some viability in parallel with a platform like ChatGPT. Like one's called MEM0, but, but you know, there was a very much the developer mindset of I'm going to port this from one platform to the next. First of all, a platform like ChatGPT, once they especially go into the close mechanism, they will not make it easy to port memory from plat to platform that that's accumulated on on their platform that that's number one. Number two, even if it is portable, this very much feels like the argument of all of the plays in the social networking days of trying to make your social graph portable. And at the end of the day even the develop, the developers wanted that, but the consumers did not, they did not care about it enough, enough in the market, did not care about it enough because it's not what created the best consumer experience and those pieces. So there were so many plays that, you know, making your social graph portable back then and just, just none of it, none of it worked. Right. And so I do think consumer experience wins out here.

Speaker B: Yeah, I, I, I agree with that completely. I think there's two reasons why it didn't work in the social era. One is people don't care enough except for a small subset of early adopter or like very privacy focused type people. And second, actually, the fact is, and I think this is true for memory and context as well, is that it's actually not universal. The same way my graph on Instagram and LinkedIn and Facebook are inherently different. And actually the portability is, is a bug, not a feature. Right. I want those graphs to be different. My suspicion is, is that you're going, is that people are going to want the context that's available to different tools to be different. And you could say, oh well, you need a tool to be able to manage all of those things. That might be true in an enterprise context or a business context, uh, across the context, across a whole business. But I don't think individuals will think of it that way. I think they will use different products in different ways and the context and personalization will be important to them in those different endpoints. And it will be actually very hard to reason about what portability even means.

Speaker A: Right, right. Yeah, I think, I think that's right. Yeah, yeah. Certainly in the enterprise context there's going to be tons of friction to demanded by the customer, the enterprise themselves to make this part. They do not want that, um, for a whole host of security and other reasons. And I think it's actually important for people to understand so that they can think about how to play it right. There is a difference between platforms and distribution channels. And the reason I say that is because the reason it gets mixed up is oftentimes a new platform comes with a distribution channel, but sometimes not and vice versa. You can have platforms without distribution channels and you can have distribution channels without platforms. So a platform is something that extends the capability of when somebody integrates them, it extends the capability of the platform. And you're usually getting access to some capabilities as part of that exchange. A lot of times with those platforms they also have a, uh, distribution channel, an organic distribution or paid distribution channel to incentivize people to come onto the platform. Right. To get that distribution channel. There are certainly platforms without that distribution channel. And that's why the key thing to watch four is more what the value exchange is around the distribution than it is necessarily around the specific capabilities that. Because that's the value exchange that they will open up and claw back around those pieces. And so anyways, because like a lot of people pointed to things like Shopify and other things and I was like, okay, well those are platforms, but those are platforms that didn't really come with like a core distribution channel for their customers. And that's why I think you've seen those platforms probably be. This cycle doesn't play out in as aggressive as a way as other ones.

Speaker B: Yeah. So I think this is an interesting point. I'll, I'll, I'll give an example and what's necessary ingredients for something to be a distribution channel as well. One is the relative size of the platform and its user base needs to be an order of magnitude bigger than the average thing being built on it.

Speaker A: Right.

Speaker B: So versus peer sized, if that makes sense. And ChatGPT certainly has that advantage. It is huge, growing. And then second is a highly engaged, retentive user base that is like that's

Speaker A: the more important explore. That is the more important piece for sure.

Speaker B: App Store or additional functionality. And I think third, and I'm kind of just riffing here, but third is clear. Touch points to tell you what's available on the platform. Now Apple has mostly done this with an App Store, but if you think about the Facebook platform as an example, in the early days of the Facebook platform, those touch points were social notifications and feed posts.

Speaker A: Right.

Speaker B: Without those, I don't think Facebook apps would have been all that successful. I don't think anybody was like, I want to go find apps on Facebook. Let me go see what's in the music category. And Discover through browsing had the size, but they also created the distribution touch points. So designing those well and having them integrated into the core retention or habit loop of the product is the other important thing. And I do think like Apple kind of gets this because the App Store is on the home screen and your home screen is the distribution touch point for new things. So plus other virality other things that are built in. As an example, the Slack Platform was a, is a great platform, you want to integrate stuff into Slack, but sort of like struggled with a couple of things. One, a lot of the things people most wanted to integrate were things that were as big or bigger than Slack at the time of the platform. So just it wasn't necessarily driving distribution as much as it was driving shared functionality and engagement. It was an engagement oriented platform. And second, while there was some distribution for some smaller players, there weren't good distribution touch points inside the product. There wasn't a great way to be like I magically know that this, I could add this thing by doing X. There are little bits of this that you started to see where like if you shared a link it would say hey, I see that's a JIRA link, do you want to install it? But those were later. And so I don't think it would turned into a massive distribution platform for a subset of app creators. And I'm only using that example because I'm familiar with it, but it's super highly valuable. But it is not a distribution platform in the same way that some of these other ones are. But ChatGPT has those things and could be. And because of the open endedness of their core loop about just talking to you with text, there's a lot of opportunity for distribution, right?

Speaker A: A lot of people looked at this like, well, Google has access to not only a ton of distribution but your email context and other pieces natively. But the big thing that came out and this, this first came from uh, a VC at Menlo, his name's Deedee I believe, told some retention data about all these players which is in the post. And uh, you gotta remember when you go back and look at the social or the search shift or all these shifts, the ultimate winner was never the one that had the biggest distribution in that moment. Facebook did not have the biggest distribution in that moment. They were smaller than MySpace and multiple other players. But the thing that they had was higher retention and deeper engagement. That was the leading indicator of who the winner was. They talk about this all the time in reforges. Retention is the God metric. It's the thing that determines the category winners. And the reason is it's the engine. It's what fuels all of the other loops around acquisition and monetization and defensibility. So while we only saw the retention curves where the retention curves were shifting up as well as they were starting to smile, they also did the category comparison which is that ChatGPT was flattening off at higher levels than all of the Other players. Now the missing piece of that data was the depth of engagement which is a, which is a different metric. But my hypothesis is even if Google has access to all of that distribution and similar level of monthly active users, at least when I look at my own usage and some of the people around me, it's flyby usage, right? It's because they have these random AI buttons sprinkled in across their things. Honestly half the users are probably people just act accidentally hitting them. It's very clear right now ChatGPT, even if it has is smaller on a monthly active user base, is doing better on retention in depth of engagement. And I think that's the leading indicator to say who's going to be the ultimate winner here.

Speaker B: Yeah, I mean imagine and the distribution touch points again because of the open endedness of the experience are really really strong. Like imagine a world, let's use a business case. I ask it about doing some sort of retention analysis. How would you go about doing this? It gives me an answer and it says would you like to connect your amplitude? And I, if, if you know click this button to connect and I will run this analysis for you. Like there are a lot of simple ways for them to drive distribution and engagement of connected apps and platforms if they do it right. Um, which I think creates a really meaningful opportunity. Now all of these products could do that but again the one that owns the most consumer mindshare in terms of this is the place I start has the biggest opportunity to win. I think the concern, the question I have here is okay, for instance not to use another Slack example but this was a pretty big one. In the last week Slack uh changed their API to disallow storage of messages over a certain timeframe. In particular it seems to be shut down A lot of the things like glean from being able to do like long term search uh, out of the platform. I think the bet here is that they can build their own consumer touchpoint and they don't want to give away the data. It's a defensive move. It's a defensive move and that might be the right thing in the near term. But does that actually hurt you because the likelihood of you being the end all be all place where this type of LLM um oriented search discovery behavior happens is low. So you are better off being available where it's happening or, or do you make a worse consumer experience for your customers by making it unavailable so that you can win in that space? So I do think a question mark is can you win?

Speaker A: Right?

Speaker B: Yeah, uh, about Whether you should, you should fight or join.

Speaker A: Yeah, I mean I think you see this across like I interpreted the stock move as we know this is important. We don't know exactly what to do with it yet, but things are moving so fast that we've got to take a defensive measure until we figure it out. Right. That's how I interpreted it.

Speaker B: Yeah.

Speaker A: You see this in a lot of places where either people are either locking down their data or they're starting to tax to disincentivize data out of the system. All those players realize the value but don't know exactly what to do with it yet. And as a result they're creating a bunch of friction around it, which once again sucks for other developers and startups

Speaker B: haven't figured out the incentive trade off yet. And we've talked about this ad nauseam about SEO and the incentive trade being I index, you get traffic. I think we don't know what this is for data systems of record yet. Like what is the trade off? And I think it's the reason these platforms need to exist and need to figure it out is because they need that data as much as those tools need the capabilities. So whoever wins at this is going to figure out the incentive trade, at least in the early days, better than someone else does.

Speaker A: Yeah, I think that's right. Or, or they're. Yeah, or they're very generous. Right. And this doesn't just happen on the big platforms either. I was riffing with somebody on this and so Udemy as an example, which was, which is like a course marketplace where all of their courses are created by third party creators. I, uh, you know, I learned this only like last year. But their incentive, in the very earliest days they shared something like 80% of their revenue with their creators. Take guess at what they share with their creators now. 50 they are, they have a plan to get it down to 15 over the night. Like that's the change. Right. And like that, like I said that was the value should come onto our platform, create a course. We're going to give you a bunch of distribution and some rep share and then they basically swung the pendulum in the other direction because they have to monetize, they have to get profitable like all of these other, all of these other pieces. So it, it happens all over the place. Big, big platforms, small platforms.

Speaker B: All right, so your prediction there will be my prediction platform of some kind in the near future.

Speaker A: Yeah, my prediction is ChatGPT is going to be the one that does this. But I think in the post I, I Also lay out some reasons why I could be wrong. Things that could leapfrog it. Does regulation, uh, get in the way? I'm doubtful on that. Regulation tends to lag so far behind the creation of these moats that they're so established at that, that point regulation is always too early or way too late. It's never on time.

Speaker B: Right.

Speaker A: So it doesn't, it, that kind of doesn't matter. There's the whole Apple wild card. It's what you and Aaron White said on the last episode. There's, there's an interesting entry point there, but they seem to be playing more conservative moves. I think the prediction matters because depending on the size of company and stage of company you are at, depends on how you play the game. Right. If you're a later stage company, you probably have the resources to not have to bet on a winner. You can kind of bets then, you know, whip things out. You can make multiple moves at once. But if you look at the history of most startups, startups emerge because they take a uh, bet on one and the bet is right. It's higher stakes, higher rewards.

Speaker B: Agree for sure.

Speaker A: So depending where you're at, you've gotta, your moves right now are a little bit different. Startups can't afford portfolio theory. I think that's the point is like portfolio theory doesn't apply. Uh, you kind of have to be the like wild man at the table.

Speaker B: Yeah, go all in. As an example, in the mobile era, it's hard to remember this but you know, there was a Apple, Android, Windows Phone, some others while it was. If you bet on just Apple, you could be a winner like Instagram. Right. If you bet on just Android, you could not. If you do Apple and Android you would grow faster. It turned out because both those platforms are dominant and building on Windows Phone was a huge waste of time. And the thing that looks uh, clear, which is, oh, let's build on Nokia or whatever the thing is because those are the biggest distribution in just terms of numbers. This goes back to the ret attention and engagement and that's the clarity of how the platform will drive distribution are most important. That was a losing move too, Right. Like even though the numbers were way bigger. And so I think that's a really good lesson. The same was true on the Facebook platform. If you built some people did okay just on MySpace in the early days,

Speaker A: but all of those developers shifted to Facebook. Right.

Speaker B: And there were a million other networks and there was a lot of encouragement from investors, from thought leaders, et cetera at that Time that like actually you should build on all the networks and um, you know, you should build as much distribution across them and try to pull people to your own platform. And it turned out the winning move was built on the Facebook platform. Right. So I think this is good advice which is you're gonna have to pick a winner and you're gonna have to pick.

Speaker A: Right, that's right. That's right. I, I think this gets to one of the other things that I've been thinking about which was a ramp published this Data that showed GitHub Copilot going from 100% market share of the AI code editors to looks like 45% in just a year and cursor overtaking them in terms of market share. And Alex Immerman tweeted this, this data with a uh, quote which was the battle between every startup and incumbent comes down to whether the startup can get distribution before the incumbent can build the innovation. Uh, this is sticking in my head because as we talked about, the incumbents are moving faster, they are copying faster. And so you have this interesting dynamic where if you believe this, this law is true. It's why you see the only big winners in AI right now be these consumery prosumer PLG type motions that can get to the distribution way faster than other types of distribution mechanisms. Because if you're in this sales LED model, your motion and the time to get that distribution is, is inherently slow and, and constrained by the human adoption process. Selling, getting the awareness, the main motion that HubSpot invented and developed around inside sales AI has actually slowed those cycles down. There's more procurement, there's more security, there's more data review. So those cycles have slowed down even more than they used to be. And meanwhile you have this other pressure which is that incumbents copy faster. So even if you innovate and you've got this sales only LED model, I feel like it's dead territory because you're moving so slow that the incumbent is probably going to copy. But um, I was interested in your thoughts on this because the data was, it was staggering. I think a year and a half ago everybody would have been like, oh my God, GitHub copilot amazing product market that no way that that thing's going to collapse.

Speaker B: My guess a year ago would have been yeah, Copilot plus VS Code plus Microsoft's advantages around that would be insurmountable. That this is a uh, sustaining or accelerating innovation, not a disrupting one. And it would be very straightforward for them to use it. I Don't know a lot about how Cursor has gained so much adoption. Um, I don't know if it's a model quality thing. I mean they use other people's models, right? If it's a UX thing, if it's something else. But I think you're exactly right that, and, and this is riffing on another article that we shared with each other from the folks at Tidemark Capital, uh, talking about bottoms up adoption of AI tools, but that it, it focuses on the hero work and is, is driven by adoption of the individual people doing their core job. And so I think if you solve a problem for someone that's not back office, that's not secondary to their role, that's not just assisted, but rather helps them do the primary thing of their work better, they're willing to pay out of pocket for it and they're willing to adopt it on their, uh, individually, on their own. And the fact that these aren't team products yet may actually be a benefit than a detriment. Which looks kind of weird, but they look a, they are prosumer versus SaaS and I think it allows them to win very, very quickly through word of mouth, et cetera because people are chasing productivity improvement and simplicity and awesomeness and they'll just jump to it and it's swappable. Like I think the next cursor could beat Cursor also in a year. Which drives a bunch of questions about the durability of the revenue of these things. Yeah, PLG is back baby. The winning tool in every market is going to be the one that owns the end consumer experience. That is the winner. Because if you own the touch point, you own the context, you own the interaction, you own the data. It is not going to be the thing that just stores the data in the background. Whereas in the SaaS era it was the system of record that won. If you were just a tool on top of that you were replaceable, the platforms could suck it up. But I do think there's a way to win by owning the most important work in a narrow way and owning the end user experience in a meaningful way. That's where you're going to have the most opportunity. And it's exciting to see that because I think it means better products for people.

Speaker A: Yeah, I'm interested. Just poking on the uh, the cursor 2.0. Can somebody come and disrupt a cursor just as fast? In Cursor's case they have such escape velocity that feels insurmountable. But I don't know. Maybe that's not true. Maybe we're still in this delicate phase where yes, they have escape velocity but haven't built all of the moats yet. So there's still a window. Yes.

Speaker B: Are the moats there? I don't know. Like.

Speaker A: Well, they were. They're working on. They're working on an agent platform too. So I admit.

Speaker B: Sure.

Speaker A: I mean they're working on. So I imagine there's going to be some similar dynamics to what we talked about before. There's like a lot of team, like cursor rules, like context. Ah. It's probably a lot of similar things but in a different way.

Speaker B: But it's not there yet.

Speaker A: It's not there yet.

Speaker B: It's not there yet. So there's still space.

Speaker A: Yeah.

Speaker B: Writing code is one of these things are really, really good at. When you think about first party use cases, the model companies are most likely to eat. I think in a lot of these spaces there's this idea that moats don't matter. We talked about this a few weeks ago. Speed is the only moat. It's. While I agree that speed is super, super important, but the long term, if you're building a decade company, you have to have something that gives you an unfair long term advantage. And I think modes are still really, really important. We are still so early here.

Speaker A: We're early but I think we just said there's a window but I think that window is much smaller than previous shifts. Then the cycle is of accelerating, um, and going through it faster and faster. All right, that's a wrap on today's deep dive into AI distribution dynamics. If you enjoyed the episode, let us know. Give us a shout out on LinkedIn or leave us a review. It's super helpful for us to know what's resonating with you so that we can keep doing these episodes. Before you jump to the next joke, here's one last thought. The fastest path to out executing every move of the platform is to know in real time what your customers are actually telling you. And that's why we built Reforge Insights. It hoovers up every morsel of feedback, support tickets, survey responses, app reviews, you name it. And it turns that chaos into a ranked list of revenue blocking issues and high leverage bets. No taking marathons, no opinion wars, just evidence you can ship on. If you like this conversation, I guarantee Reforge Insights will give you the signal that you need to win the next one. One check it out@reforged.com insights and let's turn your customer fire hose into an unfair advantage. We'll see you on the next episode.