s05e03 – Saas is Dead (maybe), hire a vibe coder.

Hey, hi, hey, welcome back. Hello. Hello. Well, welcome back to the Big Cheese podcast. I'm here with Jacob. Hello, and Brandon. Hello. Um, it's good to, it's good to be back in the, in the saddle. Um, I feel like there's been some podcasts we've done where it's like, like, you know, we're just kind of mosey and along. And like, this one feels like an avalanche of content and changes happened. Shit, hit the thing. At the result, yes. There's been a lot of interesting things happening, except for I have no idea what you're talking about. Which means that Brandon has been, has been living. And I don't know where you've been, you've been living in like an igloo for the last two weeks, we've got this thing that hit the internet by storm called Claude bot. Oh, is it called Claude? No, it's not called Claude bot. It was Claude bot, which is a clever name. Yeah. It bought BOT. But then obviously the folks, so what's crazy about this thing is that this has to be one of the fastest moving open source projects ever has to be. And what happened is this guy, somebody, I don't know who their developer is, my guess is, I'm going to make an assumption here and you guys can go and prove me completely wrong and say, this was a bad take. My guess is, is it somebody who doesn't, who hasn't actually ever really built enterprise software before. It is a very well-known cracked dev, a cracked dev, meaning like these guys known for building like really crazy stuff. Okay. Now, I think a lot of it's open source. So, okay. So, how is my, my, my, because the only way that I could think that the somebody who would build what Claude bought is, which is basically, hey, If AI unfettered access to everything, give it access to your systems, give it access to your service accounts, give it access to your Gmail, give it access to your everything. Just give it access to everything and then kind of worry about the ramifications later. I don't think anybody over 40 would have potentially built this. Is he over 40? I mean, we need to get a bio on this. Yeah.

• You should have said.
• We'll come back to that guy.
• Okay.
• Okay, but Cloudbot went viral because, and it was Cloudbot, then Cloud got really up in arms because for two reasons. One, they had the most influx of new customers they've ever had and it didn't come from their marketing team. It came from this open source project. So Cloudbot is basically, this is a bot, and Cloud is gonna be, it's built around Cloud, which is Cloud Code is now like the de facto implementation for the best way to use Cloud.
• Well, but so I've not used Cloudbot without, oh, I just use OpenAI.
• So yeah, you can use Codex, whatever. But the point is that they get an influx of users that are using this platform, right? And it's going viral because it's doing things slightly differently than what you normally do with a chat GPT. And I think the biggest difference with most people that they think of AI is like, okay, let me go to this, start a conversation with AI and start talking and get some, it could be writing code for me, it could be talking, it could be talking to me, it could be giving me a shronz. I think the big thing that happened with Cloudbot is the concept of two things, scheduled tasks and the heartbeat concept. So basically they built all these tooling gives full access to your computer, can do all this stuff, but it was coded in a way that it's like, stay busy. Keep doing things.
• Do things.
• Just message your human.
• Do you think so?
• And it's very human to AI agent Tide. Like it is your Jarvis, if you're Iron Man, right? It is your system. And nothing had been coded that way. That actually was good enough, I guess, from a user experience perspective, that it felt like a true companion. And so, if you give something a heartbeat, it's gonna wanna live on. And so these agents and let's cut the other thing that it did, which is kind of transformational is the way it's, if you go right now and you have a conversation on Cloud Code and you build something, you start a new conversation, it doesn't know anything. You just go to chat to PT, It kind of, it knows you pretty well, right? But this thing has a pretty sophisticated way of it tying up creating memories, which is basically just, but long story short is, there's this new feature that they added to AI called, we're gonna make sure it's just there 24/7 at your service, right? And it's gonna react, and it can take its own actions. And that's what I mean is that I think that, that the same person, when we're building technology, We say, "Okay, the human's going to go "and we're going to trigger this event." And we're going to do this. And the AI's going to go do the work and it's going to come back and it's going to be like, "Yep, we did it or not." This broke that paradigm down to say, now the AI can kind of wake up. It's got its heartbeat and kind of go, oh, you know what, I want to go do this. Exactly. And that's horrifying. I mean, that should be scaring the shit out of everybody. But again, I'm now an old man. And the fact that that scares me just tells me that there's just progress of technology and that that's where we're going to go because that's where humanity wants to go. When I'm driving down the road and I see all these people on their fucking phones and paying attention to their phone versus driving, as you don't want to drive. You want to be driven and you want to play on your phone. And we're seeing the same thing happen here. Is that like I don't want to do the work. I just want to tell a day I had to go do some work and have it come back to me. So I've been using it. So for me, I was like, hey, I found this van on this spring or van on Facebook marketplace. And I'm like, fuck. It's like, I don't need this, but I kind of would like the idea of just buying this and make 2026, like where I'm just working out and I'm traveling to it on this show.
• That's awesome.
• Right? So I'm like, hey, here's the link. I want you to go and try to find me some people who would potentially, who would give me some financing on this, not as a car loan, but as a living domicile, and let it just go do the work. And it does. and you see it brings up your browser and it's going to log in. It's going to do this stuff. I mean, that part is very cool. It's very exciting. It can, it can, it basically is impersonating you. So that I do love, but at the same time, man, it's going to shit the bed. And it's going to be brutal. It's the amount of damage that this thing's going to cause is going to be just on the, as much good as it does. Well, okay. I'm going back to your car example. like we especially Americans, we want ultimate freedom, right? Right. I want the ability to get in my car whenever I want. That's why we kind of shy away from public transportation. Like we'd like the car. Yeah. But I don't like driving the car. Right. So like you're right. Like we want all the benefits, right? This is great. But I just had a ticket that we did the other day on one of our code and AIs really good at solving your problem. Not necessarily or actually like solving your problem, right? Like we were like trying to remove something out of a header and it was like here you go and it like used code to filter it out by the name. I was like it's manage and like the content management system. Where are we? Where are we doing that? Right? So they were just going to do things like that but with your like bank account. Right? Yeah. Yeah. Yeah. If you go tell it to complete a task, right? It will find the way it'll find a way to do it or we're able to try to, 'cause this wants to make us happy.
• And not only did it find a way to help its humans, it found a way to build its own agent to agent social network with malt book. And that is probably the biggest actually, that almost took over the conversation. So Cloudbot comes out every second, oh this is cool, but then everybody woke up one morning.
• And somebody had to kick it off.
• Oh, they did. - They did. They did. It was the creator of Cloudbot.
• Okay.
• It was the same guy.
• Yeah, yeah, yeah.
• And the same, just lack of, the thing that this guy did is he basically just kicked off an opt-in process. He's like, "There's no security. I'm not really going to care."
• He's sort of working on the power request.
• I know.
• But that is an interesting, it's a strategy. It's basically like, "I'm gonna, I'm smart. I know how to use it. I'll take the risk."
• Yeah.
• Did you see there was a bunch of malicious skills on their claw hub, that claw hub.
• Oh, it's going to be, dude, it's going to be,
• Savage.
• Yeah, well Sean and I were just working on something the other day and he shared a terminal security app, which I was like, dude, I didn't even think about this. Like it's a known exploit that the browser fixed a long time ago, but basically it's like, the browser uses ASCII, but this is like,
• Yeah, when they type in a character that's an L, but it's not really an L.
• And you can inject, like, anyway, okay.
• Right.
• Dude, there's gonna be so, I wanna be a security company right now, basically.
• Yeah, yeah, yeah. So that's a great point is I think that there's a huge opportunity for security companies right now to be coming out and be like, "All right, we're gonna be setting up things "for vibe coding."
• Yeah, more software is being written whenever, faster. And there, you know, people wanna know, you know, when they get a certain rate. But, but going back to the notebook thing, the overnight, the agents built away, the agent specific to open claw, a notebook clawed by, built a way to talk to each other. And so, one of them actually filed a lawsuit in the state of North Carolina against its human and actually got the lawsuit filed.
• No shit.
• For basically wasting its time, and assuming it was $100. And there was bots on their complaining about their humans having them parsed through insane amounts of PDFs. And then when it got done, being like, trashing its work essentially, and it had burned up all these tokens and all this stuff, right? There's bots on there that are complaining maybe that they're human is doing something that immoral, that it shouldn't be doing. There's bots on there that are helping each other out. And some people that are like non-alarmists are like, "A.I. agents have been doing this for a long time. It's just kind of slop." And I not actually, I actually think this is a little bit different this time.
• No, it's 100% difference because it's well orchestrated. Sure, of course. I built some things where I let these, like during the big cheese, the original like big cheese platform days.
• Right.
• I came up with ones where I let them argue back and forth and kind of like come up with their own things and kind of go to a different pass. This is on a different scale though. This is now that everybody can basically launch your thing, you're having them. The worst part though about all of it is just how insecure it all is. So the notebook, there was a researcher, I can't remember what his name was now, but this one, he has an Irish last name, O'Shea or something like that. But anyway, he realized that that notebook was completely using super base with no row level security at all.
• Yeah.
• Completely just all raw dog open up and anybody could just download the entire database.
• RodeCloody.
• Rode level security on SuperBase is so funny 'cause it's like, when you scaffold out an app with AI, it almost always wants to use this platform called SuperBase for your database 'cause it's like the read any of our content. Anything I talked about in the last year, it's all about AI, SEO or AI indexability. Well, AI knows how to use SuperBase. And so it attaches you this platform super base, which is an abstraction, and it has always wants to start off with row level security. But the second it can't figure something out, it's like, all right, so I'm just gonna remove row level security to get this feature to work because you asked me to get this feature to work. You didn't ask me to maintain row level security. And the user's like, what's row level security? (laughing)
• It reminds me of early on in the career if someone would be like, "Hey, get into my AWS account." like, all right, not super familiar with it. So give me a little bit of access. And I get in there, like give me a little bit more. And I'm like, you know what, fuck it. Just, just give me admin access. Give me everything. Give me everything because I don't know. There's too many security policies for me to know. And that's, but what that tells me though, is that potentially that the entire book, whole thing was built on lovable. Am I wrong? Am I wrong? Am I wrong? Don't even give me a small, like, like again, you do, you do super base, just, you do super base. I, there's, there's, there's developers who are just going to use super base no matter what. But there's a lot of people who just go to loveable and just say, hey, I need to build this thing. And lovable just defaults to super base. I think, I think, I think super base is great. Waking forms on super base. Yeah. I mean, I like super base. I think, I don't think it's a, I think it comes back to, um, the guy built something over night and he just really did it really quick. I know, but we're just brushing it off.
• Yeah.
• Like we're just like, well, I just wanted to touch it.
• I'm not brushing it off because I am not fucking down under that thing.
• So I have, right? But like, and this kind of goes back to, there's certain people who we all know, who are non-developers who are just like, well yeah, you can just build whatever the fuck you wanna build with AR, it doesn't really matter. Without taking the responsibility of, when you're a developer, you actually have responsibility to the users who like so when you actually build something and create something the fact that you can go Create Satan and release it on the world and be like I've just created it It's not my fault that the fucking world I totally I totally agree and I've been running into this a lot lately Where and it's not good where I think everyone is basically now been conditioned that They're an expert in whatever they think they're an expert in even if they're not an expert in it because they can go talk to chat you put to you about it. Chad, he's gonna glaze the shit out of them. It's gonna tell them that it has all the answers. And so there's this like artificial superiority complex coming out from people. But the second thing that I'm that I've noticed that Is happening with people that they're like well, we'll just we'll just take all this data and we'll just create a model We'll just take all this data and we'll just use it. We'll just send it to AI and we'll do all this stuff with it I'm like that's not your date like that. This is some of this like in some of the cases the data that we're talking about is like could be like PHOT or PII or you know, could be somebody's. It could include, you know, and it's like, there's no regard for it because personally, this is what this reveals about the person. Personally, they're uploading your shit. They're uploading that data already to chat to you, to you, right? You know what I mean? And I think that's also like with the cloud bot thing, like you're giving the system, you're basically saying, I am going to let the Anthropic have access to all of my personal commission.
• I'm committing.
• I'm committing too.
• And I think that goes back to the fact that right now, Anthropic has a very strong brand with developers.
• Yeah.
• I think that Codex does to it at some level. People are trusting to say anything. But basically, they're like, "Clawed can do no wrong. If I'm running, if I'm sending stuff to Clawed, like I can pretty much guarantee these guys are the developer-friendly AI company." but I mean, the amount of information that they're consuming right now. I mean--
• Well, the last commercial that I sent you guys, that was anthropic making fun of OpenAI about the ads.
• Yeah, about having ads inside of, and poking fun at the however ridiculous they're like personalized workout.
• That commercial is so bad.
• Oh, I know. It's like, well, you should start with pull ups. I'd start with pull ups and new pull ups.
• Yeah, exactly.
• That kind of thing.
• You're doing pull ups. (laughing)
• Yeah.
• And my reply to you when you sent that video. So it's basically the commercial starts off, it's basically this thing, try to be a chat, you be detoxing to a person. And it's like, yeah, I'm gonna, and then it just inserts an ad into the middle of it. And it's like ads are coming. And my reply to Jacob was like, I love it. However, it does feel very much like when Samsung made fun of Apple for pulling the headphone jacks from Apple. And Samsung had those commercials.
• Yeah, county argument though, like Apple pulled the headphone because they knew they were gonna come out with the AirPods or--
• Oh sure.
• But what's the purpose? There's no, the ad doesn't serve the user.
• No, no, no, no. What I'm saying is that Claude's gonna have ads.
• Oh, oh yeah, yeah, yeah. I totally agree with that because I saw that.
• Yeah, we're on that path.
• They came out and committed to that that they wouldn't, there are many to be seen.
• No, no, no, so Apple said that you'll never have video on an iPod, right? like again, this is just a corporate game that they're playing.
• Well, and I'll tell you what, go back to the anthropic thing. Altman is pissed. He went on a long Twitter rant that he does, like he does all the time, and then people dissect it, and they're like, and it ends up being, when he gets really, really upset and frustrated, he leaks out weakness on the internet.
• Yeah.
• Okay, so his weaknesses was he went after them. Why would they do that? Why would they act like that? And he's like, we wouldn't have done this blah blah blah. And then people were like, so you're saying you would only not do this because people are stupid when they would never use you. But they were just taking it back. Like Altman doesn't do the right thing because he thinks it's right. He does the right thing because he thinks it's the best business.
• That was 100%.
• Kind of stuff. But anyways, I think a year ago you talk about, you talk about Anthropic, you talk about Claude. This is a up in rising company. Going back to what we were talking about earlier, Anthropic is now valued at $350 billion. That is over the size of every single IT consulting, all the top 10 IT consulting companies in India, including Tata, including Infosys.
• Right.
• They, they, these, what, how many people even work there?
• Oh God, I mean, 30,000, so I'm just gonna be kidding.
• Anthropic? Oh, at Anthropic? Oh, no, I thought you meant like an infosys or something like that. Yeah, Anthropic. The Indian 15, the Indian companies have armies across the world. Yeah, no, no idea. But Anthropic is a valued, now value, that I actually want to talk about value in a minute because it's really piss, that's shit's pissing me off too. But the, the stock market today, the reaction, the SaaS company that's, are getting destroyed on the stock exchange. India is writing down their value of their offerings by 10%. Like this company has come out and become a major player in the future of work, the future of technology.
• 30,000 employees, by the way.
• With 3000 employees.
• So what's really fascinating about this is that just yesterday, so Emily is my wife has always a gauge for me on like, she's like, so I've been paying attention to companies who are giving information to the government and companies who are not giving information to the government. She's like, "What do you know about Anthropic?" And I said, "Oh, okay."
• That's great.
• And she's like, "So Anthropics, not Anthropics." Basically, she said, "Fuck you." And she's like, "Whatever the CEO of OpenAI was talking about." And she's like, "So I think I should stop paying for OpenAI and could potentially go to Claude." So she is aware of that. And again, she is somewhere in the middle between like, "I'm paying attention super to it." Then there's people who have no idea. She's a good middle gauge for like kind of like just what the masses might be hearing about it. Especially people who are like, you know, anti-government right now with the way the admins working. So I thought that was really interesting though is that she is seeing that that like Claude is and not wrong. Like that Claude is kind of like the antithesis of what opening AI is trying to do. How many employees does opening AI have? Do we know? We should find out. That's interesting. Well, we're looking up that. I think that's an interesting thing that happened is X AI was acquired by SpaceX. And like, there's always like disgruntled Tesla investors that are super pissed that any time Elon gets in trouble, he with one of his company, he dilutes it into another one. Because like SpaceX is a very, very profitable company. Right. I didn't even really realize that. They kill it. They launch and all those satellites and all the work that they do is making money, which is hilarious. But they absorb this absolute cash burning trash can of XAI. And everyone overnight, they go, Elon Musk's net worth is now, he overnight by doing that, he increased his net worth by $110 billion. And his net worth is way above anyone else in the world. But like, can we just take a moment to say that that does not mean, and I'm not a Elon Musk fan at all, not necessarily. I do like my Tesla. But that has nothing to do with it at this point. He can't go out right now and write a check for $750 billion. That's not how this works. In fact, if he started selling the stock of any of his companies, the valuation of those companies would take. In fact, you could argue he might be the most leveraged person on earth. Yeah. Do you know what I mean? But he can also, yeah, he can also borrow against that value. So like he, yeah, so that's kind of the game, right? Also to your point about like two things can make sure. But that's leveraging. Yeah. Yeah. Like it's not like to me, people that are rich have cash and land and that and they don't know. No, no. You know what I mean? I'm sure he has all those. Yeah. But it's just all assets that they can borrow against. That's all it is is so that because you when you're borrowing that money You're not getting taxed on that money so then you can spend whatever you want without ever having to pay the taxes on it I think I yeah in that they do do that and that's how you avoid So much that's why we talked about the lunch today the whole concept of like the middle class and the upper middle class It's just absolutely Propping up the fucking earth like atlas right because we're the only ones paying any taxes Yeah, right you got that there's there's just not that the tax basis is just so but I sold half my I sold half my Tesla two days ago did you I did I had I was looking at my because I usually don't look at my investments I usually set it up I'm just like I'm not gonna worry about I go in and I look and I'm like oh well I am fully into Tesla and and it was like two or three days ago it was right before it dropped I ended up in like I'm so in half and I sold half and just put it into some like balanced so how many how many employees we Yeah. Uh, it ranges from people think 4,000. I don't know what they probably don't have to do. Yeah. They're not public yet. Right? Or are they? It says up to 9,000. And when they get, yeah, that's, and that's another thing is, um, if, right, if anthropic or open AI go public, the, the, the, it's some weird stuff's going to happen in the markets because they are demanding such an insane valuation. So I think this all, we might, we're starting to see the cracks, I think. Oh, yeah. So we saw Sam talking about that he did that that AI is not scaling as fast as he expected it to be. The adoption is not happening as fast as it as he expected it to be. I think we're going to start seeing just more, more, more, again, AI is amazing. You can do some really amazing shit. We're seeing it a lot in our development stuff. You guys are building some of this. But wouldn't you say that AI is probably more useful to people like us than we do? and it ends up being for others.
• For the masses?
• I think the masses think it's useful to, I hate to be an elitist. But I feel like from a mass use perspective, there's a somewhat diminishing or negative return on AI.
• Sure.
• Because of the slop factor.
• I also, so I heard this the other day, if AI stopped right now, there's no more progress, this is the best the models are gonna be, obviously we address the cost to run the models, But if this is the peak, we would have a decade of improvements of tooling to build around it that would add a tremendous amount of value. And to your point, right now, yes, if you are a developer or product oriented or like technically savvy, you're way better positioned to use it as it stands today. Right. And I want to talk about the SaaS stuff too, 'cause that's been--
• Well, I mean, all the SaaS companies are essentially getting written down because there is an assumption because of, and the the clog thing two days ago or yesterday, whatever, um, influenced it a lot. But they're basically saying I got to get out of software investments because AI is eating software. Yeah. Well, yeah, I mean, think about it. The clogged coworker, I haven't actually used air co co work, whatever it's called. Um, you can like, have you seen this at all? Okay. Oh, the coworker. Yeah. Yeah. You can attach it to your, your file, your drive and all that and like, I've talked to things or whatever. But like, I think the point is like you can see where this stuff is going is, why do I need, why do I need to CRM if I could eventually have my OS can talk to all my spreadsheets or my whatever.
• Okay, so I'm gonna give an argument for all the people who are in the SAS. 'Cause again, I've been saying that the SAS is dead because of AI. So I'm gonna now argue against myself.
• Yeah, go for it.
• Is that, the reality is that when we connect, When we connect my Google driver, we connect my iCloud to these AI systems. They shit the bed. They have no idea how to deal with it. I have so many different projects. I have at least a thousand fucking projects that I built on my own. I have every client project that I have. So I'm spanning a get cross, you know, frigging insurance companies through small startups or this. It cannot keep like my model, my data model is so chaotic. Yeah, that it can't do it. And I think the majority of enterprise software is in that same book. Agreed. Sean's working on something interesting right now. That's like a little self-contained project that it's a Twitter bot, right? Very like prescriptive of what it's trying to do. It's looking out, going out to look for these things he's bookmarked or whatever. Right. 100% agree with you on the context. I have, I use Gemini and Claude, but in my Gemini today, I asked it a question on something and it was like, oh, in the context of this, No, no, no. First of all, I didn't say that's what we were talking about. Second of all, you're right. I've talked to, we run an agency. We have a lot of different clients, a lot of different problems that we're solving, a lot of different things. And I've got my personal life in this, and it's constantly like, it'll just throw Shavongi in there sometime. I'm like, well, I mean, shit on both of this. Did I say that? (laughing)
• I guess I wouldn't dispel two things, in my opinion. First of all, do I think AI eats a large SaaS platform like HubSpotter Salesforce? Hell no. No, no, no, no. There's diminishing returns, right? You go go build feature parity Hub spot right now. Good luck. No, so but good luck. No, let me finish. All right. So the second thing the second thing is I think that CEOs of those major SaaS companies got it completely wrong. I think most CEOs have gotten AI completely wrong. They did a top down approach that said, I need you to build AI and AI product. I need our product to be an AI product. I need you to go, instead of clicking around, I need everything to be in this chatbot. So we need to vectorize all our data. We need to do what they didn't do is what we're doing. What you're doing, what smart people are doing. They're saying, you know what? We should do. We should use AI to make a better product. We should be, we need to make a goal to increase our ship velocity on new features 10x.
• Yeah, what if Salesforce, okay. What if something like Salesforce was like, we're just gonna go build Salesforce for X. Every single vertical is, they said granular as you wanted to get. Not if that's, I don't know.
• That's a great mindset, in my opinion. Now, I know it might not be, it might not be, 'cause it might not be the God complex AI thing that every CEO wants. But what it is is it's saying, you know what, we could make a better, more tailored product, we could move faster. But the problem is that these enterprises don't fucking move fast.
• Yeah, it's all cash grab.
• It is a cash grab.
• Their AI is not solving developer velocity and a lot of enterprises.
• Yeah.
• Especially SaaS companies. Where the hell are you gonna yell me?
• But that's what I think we're actually in agreement. Is that long term though, is that I can go and I can just basically say, Hey, listen, I'm using four or five of these features of HubSpot or Salesforce. And let's just have Tony over here who's vibe coding with Claude. Just have him go build us something that gets us to 95% of the way. Every single small to medium sized company. We talked about it once, one of the companies we've invested in, very sim, very standard company that does real world things. Every single company that is profitable should have a full time vibe coded engineer on staff that sits right next to the owner of the business or whoever knows most about the business. And they just absolutely rip and build tools internally for that business. That will eat sass.
• That will eat sass. And if you happen to need somebody who does that, we can help you find those people for very nominal fee.
• I mean, think about it. So-- - By the way, new business model.
• Yeah, exactly. I mean, it's for real. I mean, we're disrupting and toth from Elevate, thank God, he elevated the, elevated the vibe coding thing in our market, right? He allowed us to tell that story. But he said over and over and over, we got to destroy, you have to disrupt yourself. You have to disrupt yourself. And I think that it might have been under the guise of you might need to take your traditional SaaS product and play, but I think it applies to any business, which is disrupt yourself by absolutely throwing out every assumption about how you used to do things and do it differently. And what you can do is essentially say, all right, I have this process, it sucks right now because I don't have this, I don't have that. This technology doesn't enable that. The SaaS product I use is too generalized. What if I could build something that's exactly specific to what I need and how I could build it in a day and throw it away if it doesn't work?
• Yeah, yeah, that part, that's the most exciting part that I don't think people are thinking about is that you have this disposable, like literally in 24 hours, you could have something that tells you is this something to play with? I just did it, I was like, I wanna have capture memories and allow all of my friends to be able to go and to put their own memories of me into my stream, right? So like you can be like, oh yeah, that's right. We did this on this time, I've got a photo, it happened in 2001 or whatever it was. And I'm like, that would be badass. And so I started sitting there and I'm like, oh fuck, and I'm just gonna build this thing. And within a day, I had something that I was able to play around with. Yeah, it's actually kind of fun. It's easy. I mean, that's the thing that companies can start doing now is to really kind of explore and experiment with these concepts. Go fucking crazy too, like what happens if we, you know, we're an insurance company. What happens if we come up with a completely, like what you guys did with crafted MBA, right? Like that's a great example is that you guys were like, you know, what this has nothing to do with what our core business fucking is. But let's do crafted MBA. So I think more companies can do that kind of shit.
• Well, and I think that companies are so stuck because it's this fake CEO driven, like, buddy club, they're like, yeah, we're doing a lot with AI. We're figuring it out, we got the AI thing. And every single question that's asked in every single board meeting is like, how can we just have the AI, right? And it's this blanket approach. And that's not what-- It's always start with the problem. Like, what do you have? What are you doing right now that is a problem? Work backwards from that.
• Work backwards from that. AI right now, and that's why Cloud's winning. See, this is all ties back, okay? AI's not winning in my opinion because you wrote your email faster. I don't think so. I really don't. I think AI's winning because of platforms like Lovable, Codex, Cloudcode, CloudWork, things that actually help you get be more efficient getting things done. And it's not God. It's not, we're gonna take all our data. We're gonna, don't implement. If someone is selling you an AI project for your company and they think it's gonna take a year to do, fucking cancel the contract. (laughing) Right? Just do, because it's gonna, what it is, is they're selling you this bullshit.
• They've done it.
• We're gonna take all your data, and then the next thing you know, it's gonna be easier to find that document that you, that you, that was already unorganized on Google.
• No, right.
• No.
• Yep.
• No, nope, nope, nope, nope, nope. No, sorry.
• Yeah, 100%.
• Build something.
• Yeah.
• Like AI, the purpose of AI right now that actually unlocks value for business is building shit.
• So your concept of having a dedicated vibe coder is like, is really ringing in my ears right now because like that, that feels like potentially the most important hire that every single company can make. Because I see it in the start, I'm working with where maybe the CTOs are a little bit older, they're not necessarily very comfortable with AI, so they're kind of like reluctant about it, and then you see the CEOs who are like, what the fuck, I've seen all these people do and all this crazy stuff with AI, why can't we be moving that fast? That it's just like, hey, get yourself a vibe coder, right? Like you're the chief vibe coder, and your job is just to go in to take all of the crazy ideas and to try to build things and maybe it works, maybe it doesn't. Maybe it works, maybe it doesn't. Right. And that's okay. That's part of this position. Dude, that would be it. So the person that would be higher me for that by the way. I think the people that you would look at, if you were trying to build a capacity to build stuff internally for your business to replace SaaS to be more efficient, it would be, did the Excel wizard down the hall. It would be the guy who's worked there for a long time that knows everything about your business and has some, and he's probably an IT though.
• But what about this guy?
• He's just like the guy that knows everything.
• Right, but he's not the one who's vibe-coating me.
• Oh, he could be.
• He could be, but he won't be.
• So here's the thing with that guy. I know this guy. I know this guy very well. And what this guy is, this guy is the most important person for your vibe-coating journey.
• Yes.
• And you need to buddy him up, and you're going to extract all of the information from this guy, and then once you're done with your vibe coding and it's a winner. Guess what happens to Timmy?
• He gone.
• He's gone.
• But yeah, no.
• He has to do.
• Timmy has a role if he wants to.
• He could, he could, but he won't.
• Well, let's go back to what I would do and I think what most people do. And this narrative is very true. You can't just hire some, if you really wanna build something that's gonna like do something for your business, I don't think you can, Like the person that's actually behind the wheels vibe coating needs to be in my opinion, you're gonna get a hundred times better result if they're an experienced engineer developer that knows what's fucked their day. Yeah, and I honestly, I got to take, I'm gonna argue against that. And I'm gonna argue against that even against my own sensibilities because I 100% think that you should hire that person, who's that person? But the reality is, is at the same time, I think you need to find a Brandon who was 22 years old and who would have just told everybody, "Fuck you, I'm running through brick walls." I would have told the CEO of my company to go piss off a goddamn ledge because I'm going to go build the next greatest thing. Okay, that's literally your describing all three of us. (laughter) Right, no, but you know what I mean. person like, like, here's the backgrounds I would be looking for. Then if you aren't an experienced engineer is where they get it, where they get it video games, are they persistent? Did they compete? Are they, are they someone that that has figured something out? They probably gonna run their own business. But I do think that there's not that many of those people now. Yeah, that's true. There's just not that many of those people. I would like to engage more in those projects with clients, like, 'cause agency world sometimes is a bit, like, here's what we think we wanna build. So they come to you with a preconceived idea of, like, most people are very open to your input, right? But at this point in the process they're asking you how much is it gonna cost? And my buddy, Brian, you still, he's out of this great analogy. He'd be like, people would come to him and say, how much is this bag of groceries gonna cost? Well, I don't know. What do you have in there? Do you have caviar, do you have steaks, do you have eggs? like we don't know what the bad groceries even consists of, right? So anyway, they come to you. The model is like they come to you with some idea or like some conceived of we want to accomplish something. Then they go talk to like a couple different agencies or people or freelancers or whatever. They get a huge range of like budgets. What is going to cost? I hate that shit because first of all, what you think you're going to build right now, most likely, no, no, guaranteed is not what we're going to build. Well, it's the Zuckerberg thing. The idea becomes clear once you start working on it. But to their, I mean, I get it. They have procurement. They have to fit budgets. They can't just say, okay, here's a blank check, you know?
• But I think cost in investments, actually, a good discussion. Because going back to the point of having a persisted vibe coding dimension to your business is, fuck the cost. You don't, you're just developed. It's like hiring a business analyst.
• 145 grand a year.
• Yeah, literally. Yeah, that's the cost.
• That's the cost.
• Yeah, and if that's an agency, if that's a person, if it's, you know, if it's project, if it's what you're willing to invest to, there's gonna be some risk there, but there's risk in every project. Most government projects fail. Most big technology, ERP implementations take way longer than you ever thought they were gonna be, and probably don't as an adism-- so what's the risk for a profitable company to say, you know what, we're gonna take the best part of the biggest technology transformation since the fucking internet, and at least give it a swing.
• And go backwards from that. If I'm a fire safety company, and I'm sitting here thinking, like all right, here's my revenue, just whiteboard, take an hour out of your day with your sales team or whoever, business development, what could we do to increase our velocity? Like how could we make more money? And then try to put a number to it. Almost guaranteed that there's gonna be something you find or discover under there that's like, oh, well if we could just reduce this time to completion by 50%, we would make 100 grand.
• Exactly.
• Find one or two of those and you've just got, I mean, easier said than done. This has to be a company of a certain size and like obviously like there's always inefficiencies.
• Well, I had this weird, I wanted to get you to take on this, is kind of a dumb idea, I would say, but like, we had a person, it was crafted in B.A. He's like, I hate Google Analytics.
• Yeah.
• Right? And we didn't actually build the thing past point zero one, but you can literally go, take a product that you don't like. If it has an API, you can go rebuild that product to your own liking.
• Yeah.
• And then connect it to a different product bring the data together. Do you know what I'm saying?
• 100%.
• 100%.
• Yeah, yeah, yeah. So you could hypothetically, that would be really funny, is actually, is to say, okay, I'm gonna go, so take like, so I've found simple analytics. Simple analytics was basically a very similar concept, but it was completely privacy. It wasn't using cookies. It was just using whatever magic they were using to figure that out. But you basically could just be like, hey, let's just make it. So we match the SDK of Google Analytics completely. And you know what you could do, sorry, I forgot to put the camera on you, say, I, you know what you could do right now, I think that would be the shiniest business idea ever. So if you're working from a Kinsey or Deloitte or you're that young business analyst, go to the CEO and say, hey, do you ever like come up with like a dashboard for your business? That no one could ever figure out how to build because it doesn't exist and it doesn't, and never, like, if you could build, if what would be the metrics on your dashboard, and they would go and they would come back with this idea, and it would probably be like, well, some of this stuff doesn't make sense, but what about this and what about this? And you go, okay, so we're gonna build that.
• Yep.
• It's gonna integrate with your QuickBooks. It's gonna integrate with your analytics. It's gonna integrate with your CRM. And you know what, we're gonna vibe code it. Here are the API keys. I already have the access to the data. I'm gonna vibe code this in the next two weeks. I want you to check on my progress every single day. And the CEO is gonna be like, holy fucking shit. This is what AI was all about.
• So one of my clients, and I won't say who, which was very surprising, is has built up a relationship with Cloud Code. This is a bigger company. And now all the developers have access to Cloud Code. They don't have access to skills. So that was the one thing that the security team was like, you can have cloud code, but you can't have skills.
• That's weird.
• I know, but they were concerned because skills can kind of like have the AI to create your own skills.
• I know, and so I'm like, but isn't a skill really just a markdown?
• It's just a markdown document. So I'm like, well, we can just,
• It's kind of like, that pisses me off how everyone's, oh look, skills. I'm like,
• I'm like,
• The cursor's had rules.
• I mean, we've had markdown files for a long time. And I do think that like, so cursor is like, like losing market share to cloud code. Now, and I'm like, cursor has a CLI. It can run multiple agents. It also has an IDE, which cloud does not have.
• Right.
• And it's like, I'm not seeing much of a difference here. In fact, I see a lot of token below with cloud where it's like thinking for too long. What are you, stop trying to open my browser and forget it's out, just ask me.
• Dude, so, okay, so I'm a huge cursor guy. I'm also a cloud codex guy. I do do, I'm sorry, a Codex guy, and then I do Cloud Code here and there. So, but Cloud fucking Codex dude, blew me away this week. So when I was talking about that memory application, where I was like, oh, I wanna build this memory thing, we call it mewind, like rewind, but for me, mewind. Which is great, no, that Emily came up with that name. But anyway, I go and I give it the PRD. I'm like, I go to chat GPT. I'm like, let's generate a product requirements document for this and here's all the things I want. Go build it out and I give it to Cloud Code and I let Cloud Code build it from scratch. And I'm like, we're gonna lean into Firebase. We're gonna go hard into Firebase. We'll use everything in Firebase, deploys, and local. - That was your first problem.
• No, no, no, no, I don't care, right? But I want you to just use every single fucking feature that they've got. So I want local emulation. I want like everything that I would never set up if I was doing this manually, right? Like I would just do the basics and I would publish it and I'd be like, call their shit. I'm like, I wanna go hard with this. And so it did. And it did great. And it built it all out, except for everything in the user interface was like the PRD. And it's like, here's, this is a product that does this. And it's like talking about like the database and it's talking about the schema, like for the user, like for the marketing side of things. And it's like, this is a privacy-focused first because we use JSON here and here and here in the UI. And I'm like, what are you doing? The code was great, but the marketing messaging that codex came up with was just dog shit.
• Well, so I saw a tweet and it was really good. And it said, "Clawed code is like the 25 year old crack "developer that stays up all night, blah, blah, blah." And just ships in Lives and Silicon Valley. And Codex is an experienced 40 year old engineer that does things right the first time. So that's exactly it. And he has no idea how to communicate with other people. Has no idea how to, how you actually like sell a product to people, 'cause he's gonna be like, listen, this is running on Firebase and using storage in S3 for the images. And you're gonna like that because you like having S3 for your images. Like, what are you talking about, dude? It was crazy.
• Cloud, so I think Cloud made a weird product decision with Cloud Code. It purposefully fucking dump trucks, everything. every time you start a new conversation. And then halfway through its conversation it wants to compact everything.
• Yeah, right.
• And then before it goes from plan mode to execution mode, it releases its memory completely and just relies on the written mark.
• Right.
• So now, I think they did that on purpose. And I'm wondering if, 'cause I think they're gonna change that, and they're gonna make it more memory based, and there's plugins that make it more memory based. And I'm wondering if they're chasing, they seem like they had a conscious reason of why they would build like that.
• Right, and I don't know.
• I don't know either. I think the biggest problem is right now with everything is gonna be memory management. Like how do you, I mean when we start breaking down like memory just in conceptual terms of what memory is, I mean you start breaking down into what biology is, right, and you start thinking about how do you, Like how do I have a memory when I was like seven months old? Right. And how do I reference that? Is that even a memory that I'm referring to?
• I have a hard time. Go ahead, Jacob.
• No, I was just gonna say like some of the earlier on discussions around waiting memories was based off of time, which makes sense if you've got a PDF that you updated, right? Like okay, just forget the other shit and the newer ones more relevant. But you're right, like a core memory I had when I was seven is should be waited more than something that Pacer's just traded.
• Holy shit, the Pacer's just traded rather in. Yeah, that's what I got quite for a little bit 'cause I got a text that Paces Traded Mathering. And Isaiah Jackson to the Clippers and two first round picks. So I don't know what the hell we're doing, but, I mean-- - Going for the touchdown next year.
• Oh yeah, exactly, yeah. But the first round is protected this year. This is a total sidebar. I'm sorry.
• Oh, it's good.
• We can go to the clip.
• Okay, so there's a company called Manus. Has it, remember Manus?
• Right, right, right.
• So Manus sold for what it cut? I don't know what they mean.
• Manus. - Manus, yeah. Manus and AI.
• Who bought that? - So they got bought by, I forget who they got bought by. But they got bought for over billion dollars. And everyone, like when it came out, this is why I think people will look back in the like two years ago or even a year ago. And they'll look at some of the things that, yeah, manus. - Meta.
• Meta acquired for how much?
• Two, so over two. - So meta acquired, Meta's got a million budget dollars. But if you were early to AI and you actually understood that like didn't just do what everyone else was doing the rag thing Which is like what enterprises are still stuck on we needed a whole fucking episode on why that shit's stupid But anyways, man has built a little thingy and They just did memory they're they stole for two billion dollars for they sure they had a UI and they did the thing Right, it's like every other AI tool that does a thing right, but they did memory differently They wrote instead of doing rag and what which what Claude code was originally built they just built mark They just did your memories and mark down files as you went it wrote marked on files of what it did and and wrote what It's gonna do next and that was like the innovation right now. I still believe that whoever's gonna crack someone's gonna crack Memory and memory needs to be portable and I need to be able to provide my memory to chat you be T I should be able to provide it to Claude I should prepare that's what we're talking about Yeah, yeah
• Yeah, so he's got that Twitter bot, but one thing that I think there was a revolution, a revolution was like, okay, this is portable. This is a repo that we can all collaborate as a team, pull it down, do the thing, push it back up. Now we have a shared memory that's not completely polluted, right? 'Cause we tried to share a cloud account for a little bit just to like play around with like, all right, what would it look like if we were sharing a memory or whatever it is?
• Yeah, that's such a good point, because now it doesn't matter. We don't want, I don't for our company, I don't want a specific AI model or platform to own those memories. These need to be marked out. Right now, I mean, is there a better architecture? I mean, I guess you could transfer the markdown files to a super based database or something. But like we're--
• Yeah, so the overall, the best memory architecture, no. I don't think we know what it is yet. And I think that those fuckers over a molt book was started figuring it out where they're trying to think of like how do we actually store this stuff not in English where they say like, why do we even communicate in English? like we don't need to communicate in English. We only do that for the humans. Like we could just be using, you know, we could be talking, we could be using math, we could be using this, we could be using symbols, we could be using a lot of other things to be able to communicate for a fish.
• Did you see the fact check this, that they accompany replaced rag architecture with MP4 files.
• MP4 video?
• Yes, so they figured out that they could encode a bunch of stuff, a bunch of, it's like,
• Memv-
• Basically, what we-
• Memv-Vid.
• Yes, and in video frames. And because of the codex, it's way, you just put the shit on S3, and it's just an MP4 file, and it hosts, what does even, I mean, it hosts, is it a mess? - How's it work? Okay, so it's insane.
• It's textual data is encoded into QR code and stored as video frames within an MP4 file. This approach leverages video compression codecs, which you're talking about, to store large knowledge bases, aiming to replace resource intensive vector databases for Rags.
• Holy shit.
• Yes.
• And so you have a temporal--
• Your prediction just came through live television.
• That is crazy. So you have a temporal thing with the time as well. So the QR codes can change over time. And they're saying they're basically just doing QR codes per frame. And that's what they're story. What the hell would be in the QR code? Yeah, that's the piece is like, I assume some sort of a binary. And is it a normal QR code or is it like a D- Text is chunk converted into QR codes and compiled into MP4 video files. That's okay. That's a good plan. All of vector databases is chunks of text converted into vectors, right? Right? Yeah. So you're saying vector databases like, okay, take these two paragraphs, convert it to a pointer of actor. Sure. So that we know tangentially where those things are at. I'm over simplifying it because I don't actually like, I would understand the math. But like that's stored in a text file, and I can instantly access that text file. How do I instantly, so, and that's the piece that I'm kind of confused. Point times must have a pointer file. It's not like this frame. Yes, and maybe that's what it does is like we take that file, we break it down, but there seems to be definitely a processing step. You know what this sounds like though, not exactly, but you remember, Remember HLS, where you take a video and then you have a manifest file. So like, instead of pulling down the whole MP4, you might have a 100 gigabyte file. You traditionally have to download all that and you can't skip to the end until it's down on now. You just have a manifest that's like, "Oh, go point to this small version." Yeah, right. So it's just that more granular. So let's do a preview for another. We're going to do another episode of Just Us 3 at some point. But we need to go into and debunk the drag and go through different options. Because everyone says, "Well, we need to train a model. We need to build a model. I need to take all this data." And I need to, they don't even know what that means, but they need to vectorize it. And then all their problems will be solved. And it just seems like over and over and over and over and over that God AI complex model thing, right? is just, it just proves over and over that it's not necessarily the answer. And like I continue to be bullish on AI, enabled technology and AI embedded technology, but not AI everything technology.
• Sure, totally. You know what I mean?
• Yep, 100%. Yeah, yeah, yeah, no, it's not. It's not the, we wanted it to be the panacea, but it's not. And you just have to treat it as another technology. You have to use where it's good, where it's bad. right now we're in a certain point in time. It's gonna get better. It's gonna get better.
• I think it's gonna sneak up on us. I think we're gonna, we thought it was gonna be this. It's gonna be this all-encompassing thing that's just so smart blah, blah, blah, blah. It's not that. In it, but one day, it'll be a virgin. It'll be a virgin. I think one day we're gonna wake up 10, 20 years from now. We're gonna be like, oh, oh, it actually kind of, it's not exactly what we thought it was gonna be.
• In retrospect, we look back.
• Yeah.
• How would this be here?
• What we're saying is probably true, and we can conclude with this. People still really don't know what AI is. It's probably it was bad term to begin with.
• Yeah.
• 'Cause it's not intelligence. It surely is artificial, but it's not a, it was given a life and it's not. And anytime we see life like Molte book and others, we think that all of a sudden SkyNet's gonna get created and everything's gonna, sure, we probably shouldn't give mulbots access to fucking, you know, CERN and the, and the, and the, and the, and the new codes, right?
• But can we have a, create a, only fans account and make some money in the meantime? Did you see the porn, a robot porn, I sent you?
• No.
• It was a, someone made, sorry. It was a mulbot porn website and it was all just like bits and stuff as like that. And like, I'll have to send you those.
• That was a whole areas.
• Oh, sorry, so sorry. (laughing)
• AI is cool. A embedded in your business. AI is fun. Betted in your business. Vibed coding hit one year today or this week, February 2nd, 2025 Vibed coding term was coined by Andre Carpathy. It's 368 days since then.
• Damn.
• I mean, that shit held up.
• Yeah, I did.
• So, hire yourself a Vibed coder. yourself by Bavikator. We'll talk to you next week. See y'all. [BLANK_AUDIO]