AI Coding Revolution: Build Empires, Not Just Apps!

[00:00:01] Speaker A: Yeah. Awesome. This, this works out pretty good. I'm very surprised, but hopefully it's better. [00:00:07] Speaker B: This is fun. This. There's. Dude, I was like the. Let's just kick this off with like the new X chat is stupid. I am super disappointed. [00:00:22] Speaker A: Yeah, I just got that too. [00:00:23] Speaker B: And like now, now I'm. I'm totally, I'm totally confused now. It's like, it only works my phone. I have like some conversations in like the old DM system. [00:00:32] Speaker A: Yep. [00:00:33] Speaker B: I have a few people that I'm complaining to on X Chat about how dumb it is. It doesn't work on a PC at all. And so I'm here on PC like doing this live stream and so I guess I could. I'm like, I don't even know how to like have people like message me anymore on this app. [00:00:50] Speaker A: It's all. [00:00:52] Speaker B: I used to think all the quirks with like X were kind of fun. And like, lately I've just gotten very frustrated. I'm just like, it's. There's things I like about this app, but like, there's things, there's things I don't like about it. I guess that's just the easiest way to frame it. It's like, yeah, there's some things that. But there's no alternative. Right? Like I, I looked at like trying to find a different chat system to like manage consulting and my community. [00:01:17] Speaker A: Yeah. [00:01:19] Speaker B: And like people are like, use discord. I want discord. And like there's just like. And just all kinds of stuff on there and there's just too much going on. So then I end up back on Slack and because all the, you know, startups and developers and like Silicon Valley space, everyone uses Slack. So it's got some ubiquity and you know, new and iron on Slack. Right. So I'm back there. But like the Slack, like pricing and like, it's so confusing now. Like there's Slack Connect, but then people can be part of your community. I swear to God, I spent hours like reading through FAQs asking AI questions, trying to understand Slack's, let's call it Byzantine pricing. And when I finally, finally figured out, and I'm happy for anyone to help me with this, like, please. But I finally determined was like, if I pay for my. The reason this got so messed up is like I have all these customers and they all have their own Slack communities. Right? [00:02:17] Speaker A: Yeah. [00:02:17] Speaker B: And so the problem is like I log in a Slack dude and like I have been invited to like, I counted the other day, 35 different Slack workspaces are available to me. Like, I look at like 35 workspaces. Like, and like, there's no tool or way to manage anything happening between workspaces. So everyone's like, got their own startup. They invite me as a consultant now I'm on that. [00:02:40] Speaker A: Yeah. [00:02:40] Speaker B: And so it's just a mess. [00:02:42] Speaker A: It's a complete mess. [00:02:43] Speaker B: So anyway, and then the pricing, it's like what I figured out was like, if I use Slack Connect. So you have your own workspace. I have my own workspace. That's Connect over Slack Connect. No, it's not free. You have to pay. You have to pay the like 7, 95 or 10 bucks a month for your own account. [00:03:02] Speaker A: Correct. [00:03:02] Speaker B: And then. [00:03:03] Speaker A: Correct. [00:03:03] Speaker B: That account includes Slack Connect. If you use the free version, don't get Slack. [00:03:10] Speaker A: Correct, Correct. [00:03:11] Speaker B: Oh, my God. And then I believe you can have as many Slack Connects as you want. Because I have a couple of Slack connects where like, they pay for workspace, I pay for workspace. And there's unlimited number. No, no, it's not unlimited. I think it's 250 or something. You got a certain number of Slack Connects, but there's a point where they. [00:03:28] Speaker A: Start charging for more. Yeah. [00:03:31] Speaker B: Who knows? The pricing will change next year anyway, so I don't know what a way to kick off the live stream today. [00:03:36] Speaker A: Oh, man. Complain about Slack. Yeah, man. [00:03:40] Speaker B: Just. Yeah. Do we have. Because then the tool that we're using here, there is like its own chat as well. Because I'm trying to figure out how people would even ask us questions, Right? [00:03:51] Speaker A: Oh, that's a good question. Maybe if they, if they see our post on X, they can ask us question in the comments. [00:04:00] Speaker B: That's what, that's what started. I was like, are they gonna DM me on the, on the new chat thing that's only on my phone, or are they gonna send like, oh my gosh, X chat? Are they gonna d. They going to reply or like, is it coming through Slack Connect? I like, I'm like, I don't even know. Like, eventually I'll find it. So who knows? Like, people are just replying to stuff, random stuff on Twitter. So that works for me. [00:04:25] Speaker A: Exactly. [00:04:26] Speaker B: That works. That works. We'll do that. We'll do that. The other thing I want to say was like, I'm here at this co working space I've been using in Seattle now. [00:04:34] Speaker A: Nice. [00:04:34] Speaker B: It's super fun. It's called nine zero. And so they're a. [00:04:42] Speaker A: That's a nice view. [00:04:44] Speaker B: Yeah, Yeah. I mean, dude, it's Beautiful. It's like on the 15th floor and wow. What's called the central. Central business district. Somebody's making fun of me. They're like, there's no central business district. I go on Google Maps like says central business district right here. That's, that's what I'm reading. I didn't like make this up. It's cool. It's like a climate startup, incubator, co working space. Someone here and one in San Francisco and this one is pretty interesting. Like I've noticed that. I've been to all these incubators in Seattle lately and a lot of them are like private public partnerships. Like there's some kind of funding. Yeah. Funding from either the state or local or some kind of public funding. [00:05:29] Speaker A: Got it. [00:05:29] Speaker B: And then some kind of private funding. And so this one, the University of Washington I believe owns this building and they have some. Yeah, they own a bunch of real estate in downtown Seattle. So they own the building and there's some funding that they provide for the space. And then there's other people that are involved and I think there's some federal money that was available for this. There's another coworker, he's basically went to same thing. There was like a state law that was passed. They got like a grant to open up a co working space that some venture capital firm is running. [00:06:01] Speaker A: Nice. [00:06:02] Speaker B: So make a long story short, like everyone should go and take advantage of these things because it's your money that's helping some of these things which yeah, at a high level I'm supportive. I think the government putting some money into these private public partnerships to help encourage entrepreneurship and startups and like just high tech ecosystem I think is good. [00:06:22] Speaker A: Exactly. Toronto has the same thing. Universities, there are incubators, there are like government mandates. Even though they themselves are like VCs a little bit, they putting money into your startup but you can use their space that's achieved. [00:06:36] Speaker B: Yeah. And I've had some customers in the climate tech area. It's not like my primary focus. I just came here because like it had this very fun bohemian grad school kind of culture and approach and I was like, I really like this. Someone took me to a much fancier high end club where there's like, it's like, I guess it's like a co working space that comes off more like an old. An old. [00:07:04] Speaker A: Yeah. [00:07:05] Speaker B: What would be called a gentleman's club which was kind of fun. I met some cool people there. But it came over here and it was a little more my kind of like Younger bohemian, right? Yeah, it's funny. Like, dude, I'm old and at this point, like I can afford to go to like the, the old man VC club if I wanted to. [00:07:21] Speaker A: That's right. [00:07:21] Speaker B: No, dude, I, I like the grad school bohemian guys walking around with like no shoes on. That was a little much for me. Like sandals and stuff. I don't really do, but. Yeah, but I did enjoy the like open kitchen and everyone's like making their organic salads or whatever. [00:07:39] Speaker A: Salads Kombucha on. [00:07:40] Speaker B: I like it. Exactly. It's been fun. [00:07:46] Speaker A: Shall we get into nerdy news of the week? [00:07:49] Speaker B: Yeah, we got so much cool stuff to talk about. Let's do it. We've been out for two weeks, so there's a lot to catch up on. [00:07:53] Speaker A: We do. I guess I can take the first couple of things. So. Okay. You know, AI coding is absolutely a thing this week. I thought it's super interesting that Cloudflare deployed or publicized a repo called workers oauth provider. So first of all, Cloudflare is doing amazing work. I. I'm super bullish on their company. They ship like madman consistently for the past like decade. They are now fully bought into the MCP Claude ecosystem. And what's cool about this repository is that a bunch of their commits, maybe the first 100 or maybe first like few dozens have the actual cloud prompt as the commit message which essentially tells you exactly what the developer prompted Claude to generate the code. I thought that's super interesting. I think more and more repos or code bases should do this. So one thing, one of the commit messages that I saw was it asked Claude to remove the backup encryption key. A little bit blurbs as clearly it's still important to do security review based for clots code and they actually just pasted in the entire prompt. I thought that's super interesting. [00:09:19] Speaker B: Okay. So we read this last night and I was like, okay. It kind of broke my brain for a minute. So let me reframe. Yeah, I figured it out. So what's happening is that. Let me even start from the beginning. Like GitHub is a website where developers go and they. Right. They upload all their code and it actually records every version of everything they ever do when they're writing code. Correct. [00:09:45] Speaker A: Correct. [00:09:46] Speaker B: Correct. That is the purpose of that website is it's like massive. [00:09:49] Speaker A: Correct. [00:09:50] Speaker B: Google Docs version control for coding. [00:09:52] Speaker A: Correct. [00:09:53] Speaker B: Okay. And so now people are finding ways to use it to track prompting. [00:10:00] Speaker A: Correct. [00:10:02] Speaker B: Which is cool. [00:10:03] Speaker A: Which is super cool because this is like the new paradigm in Software development now is beforehand you would actually look at the code itself to be like, okay, you review it, you talk about it. Now I think you're going one step above, which is you just look at the commit messages and instead of the commit message is the actual prompt to, let's say, CLAUDE or GPT. Then you, instead of looking at the code itself or discussing the in and outs of the code, you would just discuss about the prompt. That's so cool, right? Like, you'll be like, okay, how do I prompt this better? Right? Or what is, what is a prompt that we missed? [00:10:49] Speaker B: Yeah, I think, I think like everyone has predicted we are seeing the next or the evolution of coding into natural language, which has been like what a lot of people have said about prompting, that it'll just be a layer abstracted on top of like coding and you'll just use natural language to get what you want. And so now we're seeing that. And we're seeing that being documented in the existing tools that are used to write code. [00:11:15] Speaker A: Exactly. [00:11:16] Speaker B: The other thing I would throw out there is like a newsletter is. Is Martin on live stream? So Martin does a great marketing newsletter called Uncharted. And he put this rather like long and involved prompt in his newsletter. And I was looking at it and fooling around with it and it was really interesting. It's one of the first prompts that I don't know, I guess I've seen or that I was using where he like, he like wrote like it was kind of like a, a basic style program in the prompt. Like, rather than just like have, rather than be like, let's say open ended about what the output would be. He had specific things where it said, if the output is X, label it this, if the output is Y, label it that. Like, yeah, almost like putting variables and kind of things that are commonplace and you know, basic coding into the prompt. [00:12:08] Speaker A: Yep. [00:12:08] Speaker B: And it was the first time I've seen that like he had some way to kind of rank it and say, like, is it a 1, 2 or 3? And. And he defined what 1, 2 and 3 were and it was in the prompt and even had like, you know, what do you call it? Notes. Just the way you would in, in your code. [00:12:22] Speaker A: Yeah. [00:12:23] Speaker B: And I was like, wow, like it's really becoming like a code base, but just using natural language. [00:12:30] Speaker A: Natural language, Yeah. I do a lot of prompting in my workflows and for all of my system prompts. So like there's actually levels to the prompt, right. You have the user prompt and then you have the system prompt. System prompt is like the space prompt that you, that the LLM appends before your actual prompt. So for the system prompt you can define the output structure. So you could be like for every single of my user prompts, make sure it's outputted in a way like a table or set variables so that like some downstream LLM can interpret that easier or are you feeding to other systems? [00:13:09] Speaker B: Dude, it's. It's wild, huh? [00:13:13] Speaker A: We're living in the future. [00:13:15] Speaker B: It's really cool to see all this. It's really cool to see all the stuff emerging. [00:13:20] Speaker A: Yeah. And like the. [00:13:21] Speaker B: There's just so much to talk about. Yeah. [00:13:23] Speaker A: 100%. And like the power here is if you know programming or if you can thinking like a systematic way, it's still, it's still beneficial. Right. So yes, people can vibe code with the prompts now, but if you can think of input output, which all developers do and you can build really complicated, really complex systems with just prompting. [00:13:52] Speaker B: Cool. [00:13:53] Speaker A: Really cool. [00:13:54] Speaker B: All right, should we move on to Cursor? [00:13:56] Speaker A: Yeah. So big milestone for the Cursor team. They've reached the one point, the big 1.0 release essentially for version control or semantic versioning. Everything before 1.0 is considered as like a beta release. So they don't usually give any SLAs around bugs, performances Etc. So 1.0 is a huge milestone for, for these guys. The main release of this 1.0 version are. Well, really a few things. One is background agent mode for everyone. Background Agent mode is literally a way for you to just point the LLM towards a task and it will complete it in the background for you. So you could be like cold be a website. It will just spin that in the background completely go off on its own while you're coding on some other parts of the application, like the back end part. I think that's, that's super awesome. We're just like getting more and more rely more and more reliable agents. [00:15:09] Speaker B: Yeah, indeed. That's so good. [00:15:11] Speaker A: A couple other things is like MCP is getting more integrated into Cursor. Before you had to like copy paste a piece of JSON for you to install any MCP servers. Now you could actually just click a button that'll take you to their website where you can one click install MCP servers. So just becoming more and more user friendly. [00:15:34] Speaker B: Yeah, yeah, it's. I mean, dude, there's so much happening on the vibe coding side. [00:15:40] Speaker A: Yeah. [00:15:40] Speaker B: So my brother, who he's been an engineer for a long time and he was like really against all this stuff. Like, he was pretty conservative. Got a lot of interesting skills too. He's really good at like hardware, like making like making websites go fast. He just knows a lot about like how the Internet works and stuff. [00:15:59] Speaker A: Yeah. [00:16:00] Speaker B: And like works on a lot of like low level networking things. And so anyway, there's some mandate from the CTO about having to use AI to generate code and stuff. And so he finally started to use these tools and dude, he sends me a text. He's like, he's like begrudgingly saying, like, it's kind of cool. Like it sped up, it sped up some things that, you know, might have taken me like a long time. Like I could see he was like trying to temper his incitement enthusiasm. He's like, he's like, you know, I guess like at the highest level, that's what I found too with like any of the market things or I guess anything I've done with, with these tools is that it's got like massive find and replace type of functionality on steroids. And it's, it's basically like autocomplete. A lot of people described it as those two things and I think that like, if you, you think of even from that perspective, it's really powerful. Right? If you have like massively amazing autocomplete and the ability to kind of find and replace with like logic that's a lot more sophisticated than just like finding things. Like those two things can just make you really efficient. If what you're trying to do is go through a bunch of text and update it and keep it all maintained and keep it all consistent, it's very obvious. Right? Like that would be like a lot more. [00:17:20] Speaker A: Yeah, yeah. Code completion auto complete. Like it's actually really complicated. So like for, for programming you actually have to implement this giant lexical language parsers and like you have to run language servers. So really for a really long time, code completion was really hard to do. So you have to run these really, really clunky ides we call them, which is like developer tools to get the benefit of code completion. So for a really long time, for 15 years, I coded in a tool called Vim, which for a really long time I just typed everything out and no code completion. So like I kind of skipped a step where you get completion to AI. And that was just a magical moment. It's like, it was magical for me. Yeah. All right, benefits are there. [00:18:14] Speaker B: Cursor. So cursor 1.0 is out. And then, you know, the one thing I was just reading like some newsletter that was talking about, you know, the benefits of all these tools and like the, the story about my brother, they were like detailing out too like the areas where live coding is impacting the business. Right. And so at the highest level, my takeaway on like all these tools is that I think developers and writing code is going to be there that's going to be the most impacted by perhaps this wave of AI is the right way to frame it. Yeah, I think there's actually less impact to marketing. Although like at a high level, marketing has been maybe the most impacted by technology for the last like 15 years. So like. [00:19:06] Speaker A: Right. [00:19:06] Speaker B: So it's hard maybe to define it where like program has been fairly nascent or the changes in the fairly nice. Where it's been like fairly the same and like for a long time. Right. And now you've got these tools that can just really change the dynamic. And you and I have talked about this, but like there's some very obvious areas where these tools are having a huge impact. So building MVPs, prototyping, anybody that needed to create like models or show how things would work, that's one area that's been hugely impacted. Like there was a whole business and creating MVPs, designing them everything like it's not as big of a business anymore because basically teams or individuals can do it. I even saw some examples where internal marketing teams or creating tools that like with lovable, that would do the like naming conventions correctly or like all kinds of like really interesting things. And you know, so the prototyping and just that whole area is, is gone or I don't know if it's gone but like it's, it's easy for people to do themselves. I don't think there's going to be a big business in the future. Yeah, building prototypes. [00:20:16] Speaker A: Right. Like we talk about this as well. We talked about this previously. It's just like the, the flow of order. Things are completely different now. I think so before because building is so difficult, you would kind of start with a prototype, start with the MVP product, maybe raise a C fund. Right. And then you try to sell the thing. I think because it's so cheap to build now, I personally think the first thing that you should do is get 100 customers interested. Right. Put a dollar down, put $100 down, whatever that number justifies intent. Then you can build the thing. Because now you can literally make promises like give me a hundred dollars today, I'll deliver the goods in 30 days. In terms of software, this is Possible now beforehand, if you collect a dollar today, maybe you can possibly deliver. It will take you six months before you can deliver the software. So like with these kind of like shrunkage in time to deliverable, you should do marketing, you should do selling, you should get way people's interest, drum up buyer interest before you build your mvp. I think you have a pretty good term coined for instead of mvp, you would call it most minimal sellable product. [00:21:43] Speaker B: I think, yeah, I call it MSP. Exactly. And I think, yeah, 100%. I agree with you 100% that like engineering resources are no longer the bottleneck. You can create prototypes, easily workable prototypes. So the other area that's impact is entrepreneurs. Right. So big companies can make all the prototypes themselves. Entrepreneurs can make prototypes themselves. And this whole MVP idea of like, resources are scarce in the engineering side. We've got to go slow. And I think even the whole concept, and people love to argue with me on Reddit about this, but they're all wrong. The MVP means it's viable in terms of can it be coded, can it be created from an engineering perspective. There was nothing in that concept about can you sell this? And I think, and like, what you're saying too is like, sales is the test now. Like, we know we can build it. The point is like, can you get interest from people who perhaps want to buy it? And all the, like, entrepreneurs and founders I speak to, if they've built something and they haven't been able to get traction. That's my, that's always my question is like, like, why, why can't you sell it? Have you tried to sell it? Can you sell. Like, I like that is the test of a, of a startup, right? Like, it's just a concept or a hobby. If you're not able to sell some of it and selling some of it in advance by building a wait list or some type of, you know, beta, whatever it is, is absolutely what I think people should be doing. [00:23:21] Speaker A: Exactly. [00:23:21] Speaker B: Right. Yeah. [00:23:23] Speaker A: For spin up a stripe link, send it out to 100 people, see who, who actually put a credit card behind, behind, behind it, you know. [00:23:33] Speaker B: Yeah, yeah. It's a great way to test the idea, which also I think changes people's thinking. I just talked to an entrepreneur the other day and like, he was sharing his thing and he was talking about his project and he had, he got a bunch of users, but he's difficult. He was having challenges in getting people to buy. And so he kind of took me through a demo and I was like, I said at the end, like, how do I buy it? There was all this stuff going on. And then it hit me that what he needed help with was the core kind of product marketing and UI design that just made it very obvious and easy to understand, to understand that it was a product you could buy and that you could buy it. Like, I think both of those actually were not apparent in his current setup. It was like, correct. It was kind of a cool tool and it provides some value and it actually was quite interesting. It was like pretty, pretty interesting tool that I think even appeals to a group of people could absolutely afford it, right. And, but I just like going through the demo and I'm like, want to buy this, like, get some upgrade button somewhere. And so it was very clear to me that like he just need to present it as a like free version, paid version, like right from the beginning. So you understand, like, oh, if I like this, I have to pay for it. And the other thing I'll just throw in there. Like, my advice to basically everyone I speak to these days is like, don't have free users. Like, free users are just really difficult to deal with. They just increase all your costs. Like give people like a trial, seven days, 30 days or something. And if they don't want to pay for it, they don't want to pay for it. But like, having these free users that just like ride your, your product forever is not a good idea. [00:25:15] Speaker A: Yeah, it comes from like. So I, this is my opinion, but I think it's just a lot of like old ways gets rolled into one. So like free software came from the Freeman days, right, where it takes a really, really long time to build software and most people don't even know what software is. So you give software away for free, which is relatively stable. So you know, the free users will hopefully convert into paying users. [00:25:43] Speaker B: And. [00:25:43] Speaker A: But they don't really like, if they have a bug, they're like, oh, well, I'll just have to wait until the next release. In a year. They wouldn't like email you and be like, complain about the bug. Because really that's not a thing. And then like, so nowadays if you're building, if you're vibe coding, first of all software, you're probably going to have a lot of bugs. If you have a lot of free user, you're going to be inundated with bug reports. And now you're like under this amount, mountains of pressure to fix these bugs for free users. It makes no sense to me. Absolutely makes no sense. You should be like fixing bugs that your Highest paying user is complaining about. [00:26:26] Speaker B: Yeah, 100%. I think you nailed it. That like, the nature of like software SaaS and startups is changing and that a lot of the ideas are antiquated and they all come from like the lean startup era. Like, they all come from that. And I think you said something very important, that like at that point SaaS was relatively new and not a well understood model. [00:26:50] Speaker A: Correct. [00:26:50] Speaker B: Freemium was a great way to scale users and just drive awareness for the entire category. There's awareness of that category now. And so you don't need to help people understand how SaaS works. Like, everyone understands how it works. [00:27:04] Speaker A: Yeah. [00:27:05] Speaker B: And so it's, yeah, it's really changed the dynamic. And so I don't think like free, free products and freemium make any sense anymore. I also think what happened was like we trained users to go, hey, like, I'll just keep like lily hopping from. I'll just keep hopping from lily pad to lily pad to find the free version. I won't pay. Right. And now we're in a world where like you basically have to pay for everything. So, so, you know, you pay for X now you Pay subscriptions for ChatGPT. [00:27:32] Speaker A: Just like, yeah, there's a lot of these products fatigue. That's a real. [00:27:36] Speaker B: They're all paid for. And because like ad supported has like, let's say, you know, it's not the only model. I think in the beginning that was the only model that really worked at scale was like everything was ad supported or tried to make everything ad supported. And that has just kind of changed for a variety of reasons. I don't want to get into like ad support. It isn't the primary model anymore. Right. Subscriptions are what people are willing to do and they pay. Right. And even, I mean, it was for a say, like, seems like people are willing to pay anywhere from like 5 to $20 a month for like basically anything. [00:28:15] Speaker A: Yeah. [00:28:15] Speaker B: And then there are other tools. I pay for that. I mean, I pay 40. I pay the full price, the 40amonth thing for X. And sometimes I look at that and go like, why do I pay that? Like, I'm like, I'm not sure it's worth it. I don't know. I love it, but it seems like a lot. [00:28:30] Speaker A: Yeah, yeah. You treat it as like a business expense at some point. Right. Like for a lot of people, they're also posting on X. They get money back. Yeah, 100%. [00:28:39] Speaker B: Yeah. It's like, I think I pay more than I do annually for LinkedIn at that point. It's close anyway. It's like a thousand dollars between LinkedIn and Twitter for the year, right? Or 1200 or something like that. Yeah, yeah. All right, should we move on? We got, we got other stuff to talk about. So this, this fits nicely with like all this discussion about Cursor and Claude and Vibe coding tools. [00:29:04] Speaker A: Yeah. [00:29:04] Speaker B: So Anthropic, they actually limited access for Windsurf to their, to their application. So I guess like Windsurf, which is another Vibe coding tool similar to Lovable and to Cursor, they were using Anthropic as one of the models that they're connected to. [00:29:27] Speaker A: Yeah. [00:29:27] Speaker B: And Windsurf was like offered an acquisition by OpenAI, but it hasn't technically closed yet. And so within that time, Claude decided like, hey, you know what, we're not happy about all of this and we're maybe not going to let you use our model anymore. So it caused like a disruption. [00:29:53] Speaker A: Yeah, correct. [00:29:54] Speaker B: When Surf had a race around and kind of, and kind of change things or make some, make some adjustments, which they were not. Yeah, they were not happy about. [00:30:02] Speaker A: So there's, there's been a few pieces news regarding Anthropic. Right. So they were also sued by. Or Reddit just announced a lawsuit against Anthropic yesterday or two days ago. [00:30:18] Speaker B: Right. [00:30:20] Speaker A: I, I. For, for anyone that listens to all in podcast. Did you listen to the last episode where they talked about Anthropic? [00:30:30] Speaker B: Yeah, I think they even. [00:30:31] Speaker A: Yeah, they mentioned Anthropic EA and they talked about Anthropic as a company. Who's the people behind it, how it's. It gets political really fast. And it's interesting divide because Reddit signed a deal with Google and OpenAI to essentially allow them to leverage their data use training. I made a tweet literally today about a couple hours ago breaking down the source of all of the different LLMs. I thought it's super interesting. Maybe I can. Let me pull that up and real quick. [00:31:09] Speaker B: Okay, I didn't see that. [00:31:12] Speaker A: Let me just pull that real quick. Okay, so the breakdown of 10 most cited source for LLM platforms. Perplexity. Near half of their citation comes from Reddit. 46%. YouTube is at second place at 13, 14%. [00:31:30] Speaker B: So this is where they get all them. Look at your. But that's where they get the majority of the content that they cite. They actually link back to. [00:31:37] Speaker A: Correct, Correct. Okay, so link back to. Or essentially they're, I mean if they're linking back to it, that means they're aware of it. They're probably putting into their models. Right. [00:31:47] Speaker B: So almost half of everything on Perplexity goes back to Reddit. [00:31:49] Speaker A: Correct? Correct. [00:31:51] Speaker B: Wow. [00:31:52] Speaker A: Okay, so like for ChatGPT, Wikipedia is the top at 48%. [00:31:58] Speaker B: 47. [00:31:59] Speaker A: Okay, 47.9. Reddit is second with 11.3%. That's ChatGPT. But they also signed a deal with Reddit. I think that number will increase. Google AI, surprisingly, is that. Well, of course, YouTube is the first at 90% citation. Reddit beats out. Beats it out at 21%. So actually Reddit is the most cited source for even Google's AI. [00:32:33] Speaker B: Wild. [00:32:35] Speaker A: I think Reddit is at the center of Internet and especially in the space for AI. So, like, what I'm trying to get at is like Anthropic is the only company building AI tools that doesn't have a deal with Reddit. And I think what that means is all of these companies are going to go after Anthropic and they're essentially picking sites. So like Anthropically, after windserf, owned by OpenAI, is essentially a countersuit. I'm really intrigued by this. I think, like, you have to pick your side. Anthropic is going after, let's say enterprises or they're going after the developer market. I don't know what their play is. Claude is still the best tool for coding, but Gemini is really, really doing catching up with 2.5, with the recent 2.5 release. [00:33:36] Speaker B: Yeah, there's a lot of complexity here. It's more complicated than the classic OS model where when we talked about this, right, like platform risk and OS models, where you have an operating system, you're the operating system or platform operator, People build stuff on top of your platform. So this is, this is going all the way back to like Windows and Mac. You had two platforms, people build stuff on top of it, and then the operator of the platform can choose at some point to perhaps add new features in their platform and perhaps hurt the business of a developer on top of it. Right. This is like the lawsuit about Internet Explorer versus Netscape, that Apple has done the same thing to lots of other developers. This whole concept has been around, it's a long time. Right. So let's fast forward to like the AI world where you've got models and a lot of people say, like, models are similar to the os, but they're a little bit different, which is interesting. So if you're, you take this like, concept, like you've got this operating system and you can integrate stuff into it. Yeah, I think that's what's driving the interest in like OpenAI to buy Windserve or I think all these Vibe coding tools will get acquired because like, as we said earlier today, that appears to be the number one, like obvious use case. And let's say business model for AI tools is that developers, professionals, all kinds of people are going to pay money to get them to build Vibe coded whatever they're doing prototypes, tools, whatever. Right. That, that is like the business today. It's a massive business. They all want to be in that, but they have this interesting need that operating system I think didn't have, which is like they need access to other content repositories that Mac and Windows did not. And that's the other side of this equation that you're talking about with like Reddit and Maybe things like GitLab and GitHub and Stack Overflow and like there's a whole source or world of like data that LLMs are going to need. And so you've got this really interesting way that this is evolving that's different from how operating systems, I think worked. [00:35:50] Speaker A: Yeah. [00:35:50] Speaker B: In Silicon Valley and so much more complicated and difficult business challenge going forward. I think really hard to predict how it's going to play out. Like, will everyone make an agreement with Reddit? Only some people make an agreement with Reddit. Like will people try to duplicate. There's, there's a lot of, there's a lot of questions out there about how this plays out. Like, I don't, I don't know exactly how to. [00:36:10] Speaker A: I don't know how, I don't know how this will go either. It's really interesting because OpenAI for a fact, do not respect IP. So it's interesting that they're paying access for Reddit. So Reddit is probably doing a lot of things to block, block content or traffic from LLMs. And they're doing good enough a job that all of these LLMs are paying. So I don't know, I don't know what the. [00:36:42] Speaker B: I know there, there's all terms, right. There's Quora, which I don't think was quite as good. But like they do have a repository of content that is like, comparable to Reddit. And then there are all the publications, like, because it's, it's listed there like, right. New York Times and yeah, Journals Business. There's all these. Right. There's all these professional publications. [00:37:05] Speaker A: Right. [00:37:06] Speaker B: That I think they will license content to these things and stuff. And there's also other people out there thinking about it and copyright is, you know, Going through this really interesting time. I don't, I don't know if they steal it. I just think that, like, there's just a debate and there's questions about, like, what copyright means in the, in the future. Yeah, I think it's pretty straightforward that there'll be some kind of, like, blanket settlement and LLMs will just pay for access to different types of content. And there's a lot of different models that will, that will emerge. But from a business perspective, what's interesting to me is like, you've got these two sides to the systems that operating system didn't have. And so how that plays out is really interesting. Like, will. Will maybe, maybe a way to frame it is like, will the access to the raw material in terms of content be the most important thing? [00:37:58] Speaker A: Right. [00:37:58] Speaker B: Or will it be on the demand side, like access to revenue and people who want to, you know, pay to use your tool. Right. So you kind of got supply and demand here. And it's. It's hard to know, like, where this plays out. In the past, when I've seen, like, supply and demand types of dynamics in marketplaces, I usually believe, or I've come to, like, generally frame it is that you have to have supply to fuel the demand. So you need something to sell, and the quality of, like, what you sell or the quality of your business will be dictated by that demand. Right. So using like, I don't know, Zillow, like a real estate platform. [00:38:40] Speaker A: Sure, yeah. [00:38:41] Speaker B: The more house you have, the better those houses are. The more demand they are, the better the sales is going to be. There's not a whole lot you can do to juice sales out having great supply. And I think that all these things will end up like that. And my guess is that AI and models will end up in the same way that if they don't have great supply, they can't get the best raw material to train the models on. They'll struggle in terms of monetization. [00:39:07] Speaker A: 100% completely agree. We'll see. We'll see where this all unfolds. [00:39:12] Speaker B: Yeah, it's. It's super interesting. We did cover the MCP stuff. Yeah, I know. We had that in the document too. Oh, there's a lot going on. [00:39:24] Speaker A: Revenue. Should we move on to Reddit? [00:39:28] Speaker B: Oh, this was. Okay. We can do you want. We should talk about the revenue market was a really interesting. We should drop that link in the. On Twitter. That stuff was really interesting. We'll just go over. We'll go over it quickly. Like, you highlighted this yesterday because some of these numbers even blew me away. Like I knew a lot about this that like hey, you know, certain people in the software world, like they really spend a lot. [00:39:55] Speaker A: Yeah. [00:39:55] Speaker B: On Reddit, on, on marketing. I got red on the brain but it was. Wish we could. Can we show the chart here on the live stream? [00:40:07] Speaker A: I can. Yeah, give me a sec. [00:40:09] Speaker B: We can, right? Yeah, dude. Show the like Salesforce. At last. There's. There's a. So there's a. This, this article shows by percentage of revenue what certain companies at least report that they're spending on marketing. And so I can't remember if this report end specifically but you do have to report this in S1. So like a public company does need to put some kind of sales and marketing numbers out there and other people break it out so everyone can do it differently. Yeah, but it was wild that like Salesforce was spending what, 40. Hold on, how much was it? [00:40:52] Speaker A: Yeah, so Salesforce is spending 43% of companies total revenue on sales and marketing in as of 2020. [00:41:03] Speaker B: Crazy number. And then Asana was even higher. [00:41:06] Speaker A: Correct. So Asana is spending a whopping 78% of their total revenue on marketing. 78%. [00:41:16] Speaker B: So yeah, for me this highlighted that. Like I like to joke that with vibe coding tools and AI that sales marketing will be the only job. And so when you look at these numbers already like they're spending an enormous amount of effort, energy, resources and money on sales and marketing. [00:41:39] Speaker A: Yeah. [00:41:39] Speaker B: So I guess like my takeaway for like startups, founders, early stage companies is like I always say that they underestimate how much effort it's going to be to get users and drive sales. And so I think these numbers that for me they highlight like even a company that has a huge brand and reputation like Salesforce or Asana still need to spend like half of all of their revenue on sales and marketing every year. It's a crazy number. [00:42:17] Speaker A: I think I heard somewhere or there's a term where all product companies will eventually become like media companies. It, you know like I think that basically falls into this. Right. Like you start, you build a great product and now you have to drive usage adoption through marketing and sales and now all of your budget are spent into producing these media. So like I, I truly believe that. Right. It's getting so hard to compete for attention really at the end of the day attention is the only scarcity that we're going to have and the only way for you to get attention is by spending more on marketing. [00:42:58] Speaker B: Yeah, that's the other thing I think like I'LL say this about marketing too, is that there's no free channel. So I think there's this perception sometimes that like, yeah, some channels are free. Right. And besides advertising, and I would say that, like, they're not free. They, they, they have a cost associated with them. Like let's call them organic channels. [00:43:23] Speaker A: Yeah. [00:43:23] Speaker B: So they're typically things like search or social media, but there's still cost. You still have to like, hire great people to create interesting content and to go onto those platforms and say interesting things or. [00:43:37] Speaker A: Right. [00:43:37] Speaker B: Create great content and, and help you understand, like, how what the competitive landscape looks like when it comes to SEO or generative engine optimization and make sure that you rank and those things. Right. So it's not free. It's just that, like, the costs are different and just sometimes, like, as someone who's like, sold a lot of advertising and believes like an advertising is a model, the reason for me, like, why advertising survives and is still really popular with marketers is that it is one of the few channels where there's elasticity, meaning that, like, the more you spend, the more you'll get out of it. So, like, search, if you go to search and you're just like, well, I'm going to dump all my money in a search. Like, there just isn't that much demand in terms of, like, people doing searches. You can't, you can't get any more out of it. Like, it's limited to how many times people search for that. Right. And so I think sometimes there, people are confused by how this, how this works, that like, there's a lot of channels where there's just a finite amount of demand that you can get from, from that channel. Search is a good example where you know what the keyword search volume is. They're rather obscure. There's not that many searches for it. That's why, like a lot of big SaaS companies I worked at, we just didn't really put a lot of effort into it because we looked at it and said, not worth it. There's, there's, there's so few searches done in the categories that we know are the ones that really drive our business that we're just going to take that, take that money, take those resources, take that expertise, and we're going to put it into other channels. And generally that's been, if it's not advertising, it's trade shows. So spending money on trade shows, where we get out there, we meet our customers, we talk to people, we get in front of them, we do demos, we take them out to dinner. Right. If you're an enterprise or you have a B2B application that's sold to like a finite group of people, whether there's like 10 customers, thousand customers, 50,000 customers, 50,000 customers isn't that many if you think about like how many people are in the world and how many people are on social media platforms, right? So if you go to like 100 conferences and you have a way to scale that and reach them in a cost effective manner, it's more efficient. And so yeah, that's kind of like my take on, on, on all of this stuff, right? That like, yeah, that like, you know, they all have a cost associated. [00:45:54] Speaker A: They all have a cost. If it's free, if it's a hack, it's going to, the void is going to get filled out quickly and it's going to lose effectiveness. [00:46:03] Speaker B: Right? Everyone's right, everyone. When there is a low cost channel for acquisition, everyone floods that channel and then it becomes expensive. [00:46:10] Speaker A: Expensive again. That's the. [00:46:13] Speaker B: All right, all right, what's this Reddit thread we got here? [00:46:18] Speaker A: Yeah, so this Reddit thread, I'll share in the chat as well. I thought it was really, really good. [00:46:27] Speaker B: Oh yeah, this one was good. [00:46:30] Speaker A: Okay, let me pull that up quickly. Okay, so some Redditor posted a pretty catchy title saying, we helped the SaaS company go from 80,000 MMR to 340,000 MLR in 14 months. So a little bit more than a year. Here's what we actually did. I thought this breakdown was really good. So like some of the key points, month one to two, stopped doing product demos. Essentially what they're saying is they told their sales team to get on the call and just listen, just listen to what problems their customers are having, ask questions about their current process, what they're frustrated about, and what happens when things break down. And the conversions from first call to second call went from 23% to 67%. I think that's huge. [00:47:27] Speaker B: I think, I mean I have a, I have a strong, I have a strong opinion on like everything in this thread. This Skyro, I think it's all really good. But this one I think is really important. A lot of the time in the early stages with, I guess at any stage, to some degree, yeah, a lot of time. The demo is a crutch for not having a better sales process where they try to understand the customer, qualify them, they just jump right into a product demo. I've seen this all the time with like taking like Enterprise SaaS where they've got Some tool people, the team internally like has tool and like, like a lot of tools. Like it's kind of rough. Maybe it doesn't have all the features and functionality that you want. Yeah. And, and like it could do some of the things that the customer's request. So during the discovery process they just jump in and start like building stuff and doing things and customizing stuff and doing demos and prototypes rather than like listening to the customer trying to understand what they, what they want and qualify and qualify them. It may be disqualify some people who it's not a good fit for. [00:48:31] Speaker A: Correct. [00:48:32] Speaker B: And then the ones who is qualified for maybe come up with a way that you could maximize that relationship and charge what it's worth, 100%. [00:48:43] Speaker A: The second thing that they said was fixing their pricing strategy. They had one price, $99 per user per month period. No flexibility. They created three tiers and added annual discounts. The real breakthrough was adding a professional services package for complex implementation. Average deal size jumped from $1200 to $4800. [00:49:06] Speaker B: Yeah, it's the, the, the urge like we've got all these engineers lying around here. We'll start coding stuff for you. Oh, what does that cost? We don't know. Right. We just charge per user. So yeah. Some kind of services fee on top of it or just making sure that you're being compensated for the work that you're doing. Yeah, that part's. [00:49:24] Speaker A: It's important. So interesting because like, so a lot of companies that I work for or work with on my marketing service. So here's the, here's the real thing. Peek behind the curtains. They, they don't think about the onboarding process nearly enough. So as like a second. If you're not the biggest, if you're not the biggest product in your market, a lot of the times onboarding is a huge problem for your new customers. Right. They're coming off of a much bigger platform. So the first question they're going to have is like, well I already have data in this larger platform. Do you guys help me move all of it over to your platform? And like, so usually the marketing department or the sales department hears about this complaint. Then they take the problem to the engineering team and the engineer is like, well this is going to be a massive problem because we have to either integrate with it or do a bunch of manual process par CSVs. Yeah. As a developer that's all horseshit. Like that's all just utter horseshit complaint made up by the dev team because it's stuff extra on top that they have to do. So like, it's true. Like I, my previous startups, I did this and it's like I have a Reddit post about this too. So like one of the biggest unspoken fear of one of our customers with my previous startup was that they thought the onboarding process is to going. Going to take like three month. Like they're going to manually enter all the data, they're going to have to like upload thousands or manually enter thousands of data points or entries into our platform. So they kept like kicking the ball down the road. And then one day I was like, okay, look, what can we do to like get you to pay today? And like, I'll just sit with you and do everything with you. And they're like, well, we just don't have the time or resource to import all the data. I'm like, just give me all of your data, I'll import it in one afternoon. And that worked. They converted. So like, thanks. Yeah, the point is like, I think a lot of SaaS companies, they really need to think about the onboarding process. They don't spend. [00:51:43] Speaker B: Dude, I have a. With AI, you can easily reformat data now. And like, so like that was always a big barrier in a lot of tools, was like, it's in this format, we need a math format. Because I was doing this the other day where I had like three different CSV files and I realized like, I'll just toss these in a chat TBT and see what happens. Like, perfect. Combined them. I'll put it exactly as I was like, holy smokes. So that is no longer a barrier. So there's opportunity there, I think for people to create tools actually to migrate things over. But also means the barrier that a lot of these SaaS companies had in the past of switching costs was that the data was hard to manage and. [00:52:30] Speaker A: Move hard to manage. [00:52:30] Speaker B: And now the AI, you can do whatever, you can do it pretty easily. Yeah, we posted this thread right. Because this thing has tons of great advice about like upselling and just like maximizing the features that customers don't use rather than building new ones. It's a lot of just kind of go to market account management advice that I think a lot of startups would benefit from if they were to follow it. [00:52:56] Speaker A: Yeah. So please take a look, have an open mind, paste this in the chat. [00:53:04] Speaker B: Okay, we got our AI marketing idea of the week. AI first marketing idea. [00:53:10] Speaker A: Yeah, this is, this. I think this thread is a little bit old. It might be from a couple weeks ago, but it's from the boring marketer. I, I think overall he has some good tips about how vibe marketing is still in the very, very early stages. Just even talking to people about vine marketing, most people are like, what is it? It's true. The best way that I've been describing is just generating content in a autonomous way for now with workflows. [00:53:43] Speaker B: I tell everyone. I tell everyone. First rule of vibe marketing is can't talk about it. I think we, I think you and I defined it as like. And it's an AI first approach, which I think that's the simplest way that I, I frame it. And, and like, there's a lot of implications of that. So if you're thinking a first changes a lot of different things. [00:54:12] Speaker A: Correct. [00:54:13] Speaker B: And I think that like, at the highest level, we'll get into some other time. But like, I think that's what's changing startups. Like that people have these AI first startups now and it's really breaking the way things used to work. [00:54:25] Speaker A: Yeah, exactly. The people that you have. [00:54:27] Speaker B: All right, so what is this thing from the forum? [00:54:30] Speaker A: That's it. It was literally just a trend graph where he talks about where. [00:54:35] Speaker B: Oh, okay. [00:54:36] Speaker A: Where we might be in terms of like, adoption of vibe, anything. Vibe coding, vibe marketing. I, I'm, I see the exact same way, so. [00:54:50] Speaker B: Interesting. Yeah, there's a lot more gonna come. All right, we got a couple more minutes. What else is on our list here? [00:55:02] Speaker A: You want to skip to meme of the week? [00:55:06] Speaker B: All right. Yeah. This one was. This was funny there. This one did. Did the rounds with Magnus the chess master. Norwegian chess master. I, I honestly can't think of a chess master that's had more like, I don't know, press than this guy. Perhaps, like, perhaps some of the great, like, you know, Bobby Fiser and stuff who were like, caught up in like political scandals. But it's interesting that this guy seems to keep his name in the, I don't know, nerd press or something. But he, he finally lost. Right. And so did you watch the video clip of this? [00:55:51] Speaker A: I did, yeah. He spat video of him like losing. Yeah, he got really angry, smashed the table. But he was respective. Apparently there's a chess rule where you have to put your king back into the middle of the table. That's like, to signal to the. [00:56:06] Speaker B: That's a rule. [00:56:07] Speaker A: That's to signal that you conceded the match. So, like you have to put it back into the middle of the table. And he did. So, like people are like commenting on it actually. That's, I think That's a snapshot of him actually moving the chip pieces. [00:56:21] Speaker B: Oh, is that what he's doing? Yeah, yeah. I mean, like, dude, come on. It's got to be hard to like be him and like lose that. I mean, I had full sympathy when I watched the clip. [00:56:30] Speaker A: Sure. Yeah, it's pretty. Yeah, he's, he's, he's a. I like him. I watch some of his matches on YouTube because YouTube thinks I like chess or no chess. [00:56:41] Speaker B: But, but I'll give you my, my take on this. I actually wrote a couple of, a couple of jokes for this one, but I liked. [00:56:48] Speaker A: Nice. [00:56:48] Speaker B: When you realize 50 of a 10 million dollar company is more than 1% of a 250 million dollar company. [00:56:58] Speaker A: Right? [00:56:59] Speaker B: Yeah, I, I sometimes I think that like, as far as, as far as math goes, like maybe engineers aren't as great at it as they should be. Right. Because like sometimes you look at these. Value patience. [00:57:10] Speaker A: Yeah. [00:57:10] Speaker B: Like I'm like, yeah, you got to retain a lot of equity though, if you want to, you know, have the kind of ownership that maybe people would, would think is like, that's a good call out. I don't know, life changing money. Perhaps that's the best way to write. Best way to think about it. [00:57:28] Speaker A: That's a great. [00:57:33] Speaker B: Awesome. Any else you want to say today or. [00:57:36] Speaker A: I think we're good. I. We took a two week break, but I'm super. [00:57:43] Speaker B: Yeah, this was good. This live stream works really well. Oh, the only thing I'll say is like, hey, if people want to come online, stream and ask questions and kind of hang out, more than happy to like send out the, the zoom link. So to people. It's one thing we missed from when we did it on X. So I liked that. I want to try to get back to doing that. But like the live stream seems to be reaching at least, at least if you believe the numbers are in here reaching a lot more people than we were next. So that's cool. [00:58:08] Speaker A: Yeah, that's awesome. Maybe we can do it the OING podcast way where we have one or. [00:58:12] Speaker B: Two more people, get some guests and stuff. Yeah, that'd be fun. [00:58:17] Speaker A: Okay. All right, have a great weekend. [00:58:21] Speaker B: Take care. Bye. [00:58:23] Speaker A: Bye.