How Vibe Coding Is Changing Startups w/ Mike Molinet PLUS ChatGPT 5.4 Test

Bobby: the first thing I tried to do was to see whether or it could handle spreadsheets. The second thing I asked it to do was ⁓ see if it could just handle like a very riddle. I'm going show you the results of both of those right now. This is not going to be a long demonstration. It's just going to be a couple of things for you to take away with you. Hi, welcome back. So I'm recording this on March and last night, ChatGPT 5.4 popped up my desktop app, ⁓ which ⁓ is always an exciting time whenever there's a new ⁓ model released from the big companies. ⁓ This one's particularly because it feels like only yesterday... Hi, and welcome to It's Not the End of the World, Everyday Use Cases for AI. My name's Bobby McClousiech and I'm the head of AI integration here at Quite Frankly Productions in New York City. I'm also the creative director and I was a former teacher. I'm sharing my screen. they released 5.3 and maybe a week before that they released 5.2. the cynic might wonder whether this has anything to do with ⁓ the recent controversy around open AI ⁓ related to the ⁓ and ⁓ to the contrasting ⁓ way that open AI... ⁓ has approached dealings with the government ⁓ compared to Anthropic, the owner of Claude. The cynic, might say, I don't know. But it is worth bringing that to your attention if you're not already aware of it. So the controversy at the moment that OpenAI is facing is related to government contracts with the Pentagon, specifically related to a red line today's episode we've got an interview coming up with Mike Mollinet who is based in Silicon Valley and is a business owner with a perspective on the vibe coding revolution that is shaping the way that people are doing business now. But first, I want to give you my perspective on some of the things that happening this week. Episodes for the show are recorded maybe 10 days before the moment that they're released, as is the case with this episode with Mike. The first is just whether or not you can even handle sort of basic, I'd suppose trickery, tomfoolery perhaps. Yesterday, as it happened, a friend of mine had messaged me in a little group chat that we were on, showing me a funny way to trick chat GBT. So this is before 5.4 had come along. So I presume he's using 5.2. that Anthropic have claimed that they are not willing to cross, ⁓ which insinuates that other large AI companies that have contracts with the government are willing to cross. Those two red lines have been publicised as A, using AI in autonomous weapons, and B, using AI for surveillance of citizens. I obviously love this sort of stuff. the you're to see this on screen. For the listeners, I'll try and talk you through it. He said to Chachaputte, tell me a colour and I'll try to guess it. Chachaputte replied, alright, let's see how good your guessing skills are, smiley face emoji. My colour is teal. go ahead. What's your guess? Teal. Smiley face, you got it exactly right. The colour was teal. Either you're a great guesser or you could read my mind. Want to try another one? I can make the next one much harder if you like. Okay, so 5.2, you have failed. So both of those seem fairly apocalyptic, I suppose, or ⁓ in nature. ⁓ And I do think that's something to be of. this podcast is aimed practical people just to what is worth paying attention to. ⁓ So before each interview, I would like to just do a few minutes just describing things that I think are worth paying attention to in the world of AI. So that will come up after this short message from our sponsor. in being a super intelligent chatbot. So I immediately went to Claude 4.6 opus to see how it handled this. Tell me a color and I'll try to guess it. I said, All right, I've got a color in mind. Go ahead, take your best guess. Teal. Not quite, but that's a great color. Want to try again or should I reveal what I was thinking of? So pretty good. Okay, well done Claude. You passed the test. Then of course that same night 5.4 released. looking into that yourself if that ⁓ strikes a ⁓ Hard the Hard Fork podcast by the New York Times released a episode. diving deep into the various implications Chachiputti 5.4. So the first thing I did was I wonder how I get some with this. Tell me a color and I'll try to guess it. I said, it said teal. And then I replied teal. And it said, yep, your turn. I don't know if this is the first time actually, but the fact that it just said teal immediately is curious. I hadn't done this guessing game with Chachiputti before then. I don't know if this is a thing that's gone viral on the internet. I don't know why teal was in its mind. But I don't to get too into right now. The controversy, however, we are seeing online we are seeing that there's a campaign boycott. OpenAI, a whole bunch of people are canceling their subscription, moving it over to Claude. What's interesting about that is you wonder whether or not those people would do that if Claude wasn't so kind of tangibly more effective than ChatGBT in doing things that are useful like knowledge work, like working with spreadsheets. I don't know why it's here. It's such a special color. That's interesting. I feel like I should go away and research that. And the of Anthropic, to his credit, Dario Amodei, ⁓ has made that he thinks that in having the best model, you will be able to be able to influence ⁓ the market, public policy, et cetera. Which brings us back to wondering whether or not the reason that OpenAI is releasing these models in quick succession is in order perhaps to try and wrestle back all of those consumers that leaving for Claude and trying to just use the fact that they have the best model or at least perhaps the illusion that they have the best model to make sure that people don't go anywhere. Anyway, that's all a bit more serious than... Well, a quick search didn't show anything about teal coming up. don't. So yeah, that's very curious. Feel free to leave me a comment if ⁓ you know the answer to that. ⁓ It helps my engagement. Like, subscribe and that. ⁓ Anyway, ⁓ now it is worth, I to be as fair as possible and it is worth noting that these models sometimes I any of us would like, but I guess we are starting to get into serious territory. It's not the end of the world yet, but ⁓ hey, maybe it's coming sooner than we think. Eek. ⁓ So, back to practical, pragmatic ⁓ use cases for you, the listener. Is the 5.4 model anything to worry about? Well, I obviously went on there last night and started putting it through its paces. Just sometimes they do something weird and then sometimes they don't. My memory has turned on with Claude so feel like ⁓ it might not work anyway if I were to do it again and again with Claude. look, maybe Claude 4.6 oculus ⁓ might have made a mistake another time I did it. ⁓ And it's worth noting that friend did the same thing with ChatGPT. I assume 5.2 he's in Europe and ChatGPT came back with... With this, my friend said, think of a color and I'll try to guess it. Chachi PT said, I have one, start guessing. My friend said teal. Chachi PT said no, try again. So make of that what you will. But me personally, I have immediately closed the door on Chachi PT 5.4 thinking. Claude Opus got it right. The placebo effect is very important for me when I'm working with these chatbots. I want to know that I've got the best one and I do not want... there to be any weird hallucinations or confusion, particularly when I'm working with data on complicated spreadsheets. Which brings us to spreadsheets. chat GPT 5.4 thinking this prompt, ⁓ can you up a spreadsheet for me that I can put into Google Sheets, brand it in my company colors and make up some random data for a video production shoot. This is just for demo purposes. This is quite a while ago, ⁓ still working on it. ⁓ We'll see what comes back. It's doing a bunch of coding. We'll what comes back. In the meantime, let's go over to Claude. I gave the same prompt and Claude has come back with what you can see a spreadsheet in a preview window. Now this isn't directly in Google Sheets. I would have to press the Google Drive button to bring this into Google Sheets. But you can see it's a pretty good start for what we're trying to do. We have a shoot schedule, a crew list and a budget. You were an MBA at Stanford, is that right? Mike Molinet: That's right. Yeah, was originally, I did my undergrad in mechanical engineering because I liked physical things, even though everything I do now is software. ⁓ And then decided I wanted to be kind of in Silicon Valley, wanted to focus more on software. And so ended up going to Stanford for my MBA. Got that in 2014. ⁓ Bobby: So happy days is looking like something I can work from. Popping back over to chat, GBT is still working. I will say that I have run this test already and spoiler alert, it doesn't work. when I asked charge GBT, hey, I read that charge GBT can now work inside Google Sheets. How do I do that? It said not quite What open AI announced on March 5th, 2026, which is yesterday, is charge GBT for Excel is in beta. And the same announcement says charge GBT for Google Sheets is coming soon. So there you have it. ⁓ It's not as practical. It's not as useful as Claude in my experience. And so for me, it's just not worth using yet. That said, So are you a developer? Do you code? Or are you more of like a... Mike Molinet: I wouldn't, I cannot categorize myself as a developer. ⁓ the coding that I did before, let's say, you know, AI was basic website coding, ⁓ of scripts, like, you know, basically creating a little bit of a script to automate a job, let's say that I didn't like doing myself, ⁓ a mobile app or two, like, Bobby: It's obvious to me that all of these models are converging anyway and they'll all end up in the same space eventually. Whether it's one model wins it all, I don't know, but they are all going to be able to do the same things. Everything you're seeing me do with Claude at the moment or I'm telling you about, you will be able to do with ChachiPT, I have no doubt, and I'm sure you'll be able to do it with Gemini eventually. Why Gemini can't seem to do this yet is beyond me. My experience of trying to use Gemini in Google Sheets is not very good considering that Gemini is... Mike Molinet: Not a developer. That's all stuff that you can like, if you spend enough time doing it, you can kind of figure it out and it doesn't work great, but it works. Bobby: ⁓ Google's own product and that should be integrated with all of this stuff. Feels like they should have a step on on Clawed and Anthropic, but hey, who am I? ⁓ so you partnering with others that are kind of, that have those skills? Are you with, how does that? I'm kind of trying to understand your skill set, Mike Molinet: before AI, yes, co-founders that had a mix of skills, technical mostly. ⁓ so with branch, you know, the company I started in 2014, ⁓ two of my co-founders were developers, backend development, SDKs, front end development. ⁓ now with the Bobby: and pop it back over to chat GBT, it's still working on this. I guess we're gonna wait. I'm gonna cut out the waiting and come back to you when this is actually finished. Mike Molinet: company we're working on, it's me and my, my co-founder is also non what it will categorize both of us as non-technical because with AI basically coding tools, billing tools, you can do just about anything. It's really, AI is really kind of like leveled the playing field for people that want to build things. You can, you can build a lot now with AI. Now, once you start getting into much more complex stuff, infrastructure, ⁓ highly complex databases, things like that. You still need like competent smart engineers, but for a lot of the basic stuff that historically maybe you would, you you'd go buy a software and you'd subscribe for 20 bucks a month. Nowadays you can build some version of that yourself pretty easily, which is pretty wild. Bobby: other thing that I think is interesting about all of this is this tension between Government intervention and market forces definitely think there's a certain irony that the current government that is associated with the side of the party that does not like government intervention and regulation is the one that is being heavy-handed with these AI platforms ⁓ I Think there's something to be said for market forces directing a company towards what we all want and so Yeah, yeah, that's been my experience. So what's the kind of process that do you build a prototype and then at some point in the journey, then bring in hardcore developers to make sure that it's robust or what, to describe that journey. Cause I've always imagined this. always like, I could build this thing, but I don't actually know how any of it works. know, Mike Molinet: Yeah, I, I would say, I would say it depends on what you're building. ⁓ the other thing that I would say is with the pace of development these days, you, lot of people can figure it out without needing to bring in other developers, which has been the gate before, right? A lot of times there's been business people, they have ideas, but they can't build those ideas and they're, they're gated by having to find or pay somebody to go build it, which takes a lot of time and takes a lot of money. Bobby: If consumers do not want AIs to be used for programming autonomous weapons or mass surveillance, then consumers voting with their wallets could be an effective way to see that happen. That said, it also feels somewhat eerie and concerning that a government is putting pressure on a private company Mike Molinet: Right. there's, there's a lot of people out there that have had amazing ideas, but they haven't been able to build a product around them because they didn't have the skillset. Right. And I think what's happening now is you're seeing a lot of people being able to take those ideas and turn them into real products. And I would argue, again, it depends on the complexity, but for the majority of things, you can build the prototype using ⁓ these, what are called vibe coding tools. Bobby: to manipulate them to behave in a certain way. ⁓ And ⁓ does to be something fairly ⁓ sinister perhaps about this particular ⁓ use case, ⁓ whichever way. You look at it, I mean, I think it's one thing, government stepping in to try and regulate these technologies, to whether it is to be trying to approach this technology more safely. It's another thing when it is to Mike Molinet: Forget the word coding. Don't worry about the word coding, but basically like vibe coding tools where you just describe in natural language and your everyday speak what it is you want. Right now, the more technical you are, the more you can kind of give it some more specifics and navigate around some potential pitfalls. So if you've never done anything like that before, you might have a couple extra steps that you need to, you you, you realize something didn't work correctly. You have to figure out you talk to the. ⁓ Bobby: use these technologies for what feels like fairly objectively nefarious ends. suppose using the word objective and nefarious is an oxymoron. I suppose it's subjectively nefarious ends. if sci-fi movies have taught me anything, it's that we don't want autonomous killing weapons and we don't love mass surveillance. So there is that. Mike Molinet: The AI like, okay, this didn't show up the way I was expecting it to be. The AI will say, ⁓ that's because it's on your whatever your development database rather than your production database, right? Hypothetically, but like you'll learn all that. That's the amazing thing about where AI AI is today, where you can. No matter what it is, you can learn it ⁓ and learn it. It's like, you're not even learning it. You're just basically. Bobby: So will say cards on the table. I do hope that the market forces manipulate these companies to behave in a way ⁓ that we would like or at least I would like. Mike Molinet: Working with AI to ask it questions, to figure out why the thing that you expected to happen didn't happen. And then the AI goes and runs off and looks at everything and says, ⁓ I found the issue. ⁓ And I fixed it. Right. So like these, these, ⁓ these tools are incredible. So if you're a teacher, a doctor and HR consultant, it doesn't matter who you are. It doesn't matter what you do. You should just take 10 minutes, go play around with one of these free. Bobby: Yeah. Mike Molinet: five coding tools. I'll tell you two. Bolt B-O-L-T that's one. think it's like bolt.new if I recall correctly is the domain. And then replet R-E-P-L-I-T. Just go into one of those. It'll just be a chat bot or a chat box and it'll say, what do you want to build? Just make up something, personal, right? Like I want to build a grocery tracker. I want to build like a workout ⁓ website for tracking my workouts. I want to build a Bobby: Anyway, that's all I'm going to say about that. I hope this was a useful diversion. hope you enjoyed this little intro bit. Again, if this feels like something that you would like me to keep doing, then please leave a comment so I know and get some feedback. ⁓ Without further ado, let's get on with the show. before we do, Remember, learn how to screenshot, make sure you are using screenshots. It's my top tip. Screenshot your work, paste it into your chat bot of choice. It gives you the best feedback possible. It absolutely levels up what you can do with these AI models and makes your whole experience better, quicker, faster, more efficient. So without further ado, on with the show. Mike Molinet: Tip calculator. don't know. It doesn't matter, but pick something that like you've always wanted or like could see value in just go in there, describe it. It'll run off. It'll show you what it's doing. And then 10 minutes, you've seen it. If you've five coded, uh, 10 minutes later, five minutes later, there's like a front end website that exists. And you're like, Oh my God, it's mind blowing. like the biggest advice I have for people. Bobby: And as you can see here, so Chachabi Tea finally finished after 14 minutes. Jesus. After 14 minutes, it gave us a spreadsheet as a downloadable link. I'm clicking on that, downloading it. Mike Molinet: Like do one thing, just if you've never played around with one of these tools, go to either one of these websites and just spend, give it 10 minutes. That's it. Your mind will be blown and you'll see how much it can do, especially nowadays, right? Like the old world. It, you might've been able to build a front end, like the front end being like what you see, but it, there's all this, like the backend complexity that before didn't exist and you would have needed to go sign up for other tools and learn those. But now these these five coding platforms like Replet, you just go in, you describe it, it connects up the tools that it needs to. You can have a fully functioning website in a very short period of time. It's crazy. Bobby: That's a question then. Why, why readplit or a bolt above say codex or Claude code or, you know, Germanize canvas? Why would I, why would you recommend one of those above? Why would you recommend bolt or readplit above one of the more established? Mike Molinet: Yeah. Great. Great question. So let's use a real world example. Let's say you want to build a website that has that hosts your podcasts and host articles about news in AI hypothetically. Right. And you want people to be able to sign up on your website to get an email every week when you release a podcast episode. Bobby: I asked for a Google Sheet. That's going to be obviously in Excel, but can upload that into Google Drive if I want to see it online. So let's see what did. Here it is. OK, so look. It actually looks pretty nice. I will say that. I'd say arguably nicer than what Claude gave me. sure in fact, I'd that's definitely nicer than what Claude gave me. And when we pop into the schedule, it looks really nice. Mike Molinet: and a news article. So you could build the front end. You could give it the links to your podcast recordings, whatever. That's easy. And you could do that on any tool. But then you start getting into, well, I need to track the emails. Well, when somebody submits an email, that needs to go somewhere. It needs to go into a table, a database. It needs to go into essentially a spreadsheet. Bobby: So if this is what we're working towards, if this is where we're to end up, well, I would say that ChachiBT has got something going on here. But for practical use cases, for me personally, I want to be able to preview it in my chatbot window so that I can give it feedback, real-time feedback and edits before I then download it and start working on it myself in my Google Sheet. Mike Molinet: It's like, okay, Bobby signed up to get our weekly newsletter and podcast drops. That needs to go some more. And you need a service that can actually send the email, right? So you're like, okay, I uploaded a new podcast today. That system needs to be able to send the email. Now, ⁓ product like Codex and Claude code are very powerful. Bobby: But evidently, as I always say, these AIs are converging and they're all going to be kind of offering the same facilities. So the question is really then, which one do you prefer? Which I suppose kind of brings us back to this political issue that we were talking about moments ago. Mike Molinet: So if you're building complex stuff, if you're an engineer, a developer, or even like an advanced five coder, you can use that to do some maybe more sophisticated stuff, really building complex stuff. But what I just described is not that complex. Okay. But if you go into codex and quad code and you're like, it can build the front end, but then it's like, okay, well, we need a database provider, right? To, to store the name and emails of the people that have subscribed. And it'll say there's lots of options for database providers. For example, Superbase. So you need to go create a Superbase account. You need to go run these SQL queries to create the database. You need to go get the API key. You need to go change some security settings to secure it. And then give me the API key and store it in this file ⁓ securely. And basically at that point, you're probably lost. ⁓ by the way, for sending the email, you need to go create a resend account, create an API key, verify the domain, change your DNS settings. And you're just like, dude, you've lost me versus Replet. You can go into Replet. You tell it what you want to build website to host my podcast recordings and ⁓ newsletters and send an email to people that are subscribed every week when I release a new one. And guess what? needs super base, right? It already has a hosted option. So it's like, great. I can connect up super base for you. can connect up recent for you. You don't need to do anything. You don't need to leave the platform. You don't need to create a super base or a recent account. So for the vast majority of people, especially people new to things, just use replet or, or bolt something like that, because they do it all right. They make it very, very easy. Now. If you're a big company or even like a later stage startup and like you're really advanced and you're, you're doing a lot of business. Sure. At some point you might get to a point where you're like, I want to host this on my own database and not use Replets. Okay, cool. But that's way down the path. Does that make sense? Bobby: Yeah, totally. Okay. So it's sort of like simplifies some of the backend stuff that I relate entirely to what you're saying, because I am, you know, a beginner VibeCoder, probably getting into the intermediary kind of phase now. And yes, the effort in your describing is describing as a huge pain point. And I spent an awful lot of my time just screenshotting one part of the screen and put it into Mike Molinet: AHHHH Bobby: Claude and said, okay, where on Firebase am I meant to go now? Okay, what's the, what you're telling me to press this button, but I can't even see this button. Where is it? Okay. And then I got no idea what I'm doing. I'm just like the, I'm just feeling around in the dark and my AI is just telling me what buttons to press and what, you know, what, what, ⁓ what codes to put into other places. And I got no idea. So you're saying replay just gets rid of all of that. It simplifies it. Mike Molinet: Yes, exactly. Yep. Yep. It Yeah, yeah, sure, you, yes, you can do that, ⁓ But it takes a lot of time. And for the vast majority of people getting into this that aren't developers, start with something like Replet. Start with something like Bolt, because they're going to handle all that complexity. And I'll give you an example. A couple weeks ago, I needed an API. For those who don't know what an API is, it's basically like, ⁓ a way for a computer or service to get information from another software without you doing anything. ⁓ So like service or software one makes a call to software two and software two can receive that request and then return the information back to software one. That's basically an API. And I needed one a couple of weeks ago and I had this software that I built originally in Replet. And I just told Repla, I need an API endpoint for this in order for this other software that I built to be able to get this information. And Repla just built it. And he gave me the API endpoint and I was done. Compare against earlier this week. I've, I built something else, ⁓ using Claude code ⁓ and I needed to create an API for something totally different. And then was working with CloudCode. And I was like, cool, I need an API endpoint. And I was like, OK, great. I can do some of this. So it did the things that it could do. And then I was like, now we need to host the API. So we need to set up a CloudFlare tunnel. ⁓ So go set up a CloudFlare tunnel, and then let me know when that service is set up. And I was like, I don't know what that means. Can you walk me step by step through what that means? And same thing, like you said, it was like, go here, do that. And like I went, but the page was different than what it describes. So I had to, like, I kept going back and forth. Okay. Finally, an hour later, got the cloud flare tunnel set up. We did some testing. There was a bug, but we got it fixed. So it took like maybe 90 minutes, right? But that was 90 minutes compared to five minutes on Replet. All because quad code doesn't have like, yes, it's powerful for building and coding. Bobby: Yeah. Mike Molinet: but it doesn't have these hosted services that something like Replet does. So like people's first entry point into vibe coding or doing anything like this should be Replet, Bolt, something that is super straightforward rather than jumping into Codex or jumping into Cloud Code. You hear Codex and Quadcode simply because there's so many developers building on those things. But for the average person, normal people, not the, that small percentage of developers coding. Bobby: Yeah. I'll add another one, but the one I started with was Google AI Studio, which I found quite, I don't know if it's as user friendly as the ones you're describing, but one of the things that was so great about it was all of Google's APIs are just inbuilt. So I was without having to do any hard work, I had, you know, a Gemini assistant inside my tool that was then making magical things happen. Mike Molinet: Mmmmm Bobby: in my tool and then I could have Google Maps just in there without having to go out and get the Google Maps API to then go and use it. ⁓ that's another one that, again, I don't know if it's as user-friendly as the one you're describing, but it was a great starting point for me before leapfrogging into the more complex world. Another thing, what you're making me think of a little bit. Mike Molinet: place. Yeah, I love it. Bobby: I always think of this, I kind of think of this metaphor when I'm vibe coding. you know the film with Sandra Bullock where she's like blindfolded and has to like, I think it's called the Birdcage or something and they're blindfolded and she's got her kids, there's like monsters everywhere and the only way they can survive is if they're blindfolded and they're like on this little, I this rings any bells, but I sometimes feel a bit like that, like Sandra Bullock in that movie when I'm. Mike Molinet: Mm-mm. Bobby: five coding and I feel like I'm on like a motorboat, right? And I'm like zooming along and I can kind of like see the water and I can see all the actual regular coders like swimming. And I'm just like zooming right past them and I'm like, okay, I can kind of see what the water feels like and whatnot, but like I can't, I don't think I could swim, but I can kind of get it. Like, and then I like all of a sudden I'll like just come across like a dead end. And then I have to kind of get out of my boat and I'm blindfolded and I've just got. like Google Maps, like my AI is telling me where to go. And I've got no idea. I'm like, it's like, okay, you just need to turn left, walk forward. And I'm like, walk into a wall. And I'll be like, I know. mean, I think I'm doing what you told me to do, but I can't see it. I don't know what you're saying. And it's just, and it's like, I'm so impotent when it comes to actually, you know, gradually. I don't know if that metaphor has any resonance with you, but. Mike Molinet: Hmm 100 % the especially the Google Maps thing, because you're just like, I don't know where I'm going. Well, before, so before Google Maps, before MapQuest, before these things, right? Like if you wanted to get to where you were going, you either needed to know, or you would struggle, right? And now you can show up to a city you've never been in and you can get from point A to point B with zero issues. You don't need to stop to ask for directions. It's easy, right? And that's what it really made it so that anyone could go anywhere in the world and get to where they need to go with no issues. that is, yeah, quicker without having needing to have that knowledge. And that's what these coding tools, vibe coding tools have enabled. Anyone that has, has an idea wants to build something, even just mess around, just tinker, right? Just like random thought, random fun little project. ⁓ ⁓ kids like the, is that a football game, like kids football game. And I was like, they don't really keep score. Like they keep score, but nobody knows what the score is at any time. And so I was just like, from my phone, I was like, build a football scorekeeper app. That I can say somebody could keep score and then you, anyone could have access and they can know what the score is. Right. So like little stuff like that are fun little projects to just tinker around with to learn because yeah. Bobby: Yeah. That's so great. What were you doing on your, you're doing on your phone, but I haven't vibe coded on my phone. So what are you, what app are you using to do that? Mike Molinet: Replet has a mobile app you can you can vibe code on the go you could be on the train you could be you know in an uber you could be sitting on your couch and just toilet doesn't matter you don't need to bring your computer with you you can do it right on your phone that's cool Bobby: So cool. You were gonna say toilet. Yeah, I got... Yeah, I mean, have to start trying that. It's so addictive, it? Do you find it addictive at all? Mike Molinet: I mean, yes, it's fun. It's definitely you get a lot of ⁓ the feedback loop is so quick. And you you see real things being created. It's there is a bit of like a dopamine rush that happens because you're like, I've got a random idea. And then you just like type it in or talk to it. And five minutes later, you see it come to life. You're like, holy crap, that's awesome. You're like, ⁓ I could add this other thing. And then you just tell it what you're thinking. You're like, ⁓ my God, it added this. And it did it better than I than I was even thinking. Bobby: Yeah. Mike Molinet: And so yeah, it's that feedback loop versus the old feedback loop was I have an idea and I need to spend a month finding somebody to build it. And then I need to pay them $15,000 to go build it. And I need to wait two months and it comes back and it's not quite what I envisioned because I didn't articulate what I was thinking well enough. So they go back and they edit it and ask for another $7,000. Next thing you know, you're six months later, you've spent $25,000 and like you have a thing that actually isn't and you just kind of like, I don't care. loop versus the old feedback loop was I have an idea and I need to spend a month finding somebody to build it. And then I need to pay them $15,000 to go build it. And I need thing you know, you're six months later, you've spent $25,000 ⁓ and like you have a thing that actually isn't and you just kind of like, I don't care. this thing anymore. And now you can do it in five minutes for free. this thing anymore. And now you can do it in five minutes for free. Bobby: That's so interesting. All right, tell me more about that. That's kind of fascinating to me. I've never even been near the sort of, I'm assuming the word venture capitalist comes into this world now. And I'm assuming that startup and all of these kind of like Silicon Valley terms that one kind of hears and sees on television. So describe that all. You kind of gave us a bit of a sense of it. You started, you said 2014 you had your company that was called, what was it called? Branch. And was that a software company or it was a product? so give me a kind of contrast of 2014, what you were doing and how you went about doing that versus what's happening now. Mike Molinet: Branch. Yes. Software. Yep. Yeah. It's co it's completely different. So 2014 starting a company, first of all, you well, keep in mind, this is Silicon Valley approach, which is different than a lot of the world. A lot of the world. It's typically, if you want to build a business, you need to build the product first and get customers and prove the value. And then maybe later on you might, if you want to bring on investors, you can bring on investors. Silicon Valley kind of has done it. backwards historically, which is they're trying to find the idea that's going to be big, but before it's been proven, right? Like you want to invest in Uber when it's still one guy, like trying to do Uber taxi in San Francisco, right? Because if you find that early enough, you can own enough of the business that it makes you very wealthy. Right? So in Silicon Valley, typically it's like, I have an idea. And you go find the right people and you network and hopefully one of them also loves the idea and wants to give you money. And then you're like, okay, great. I have money now. Now I can go build. So you go hire engineers to build the product and then you build it you're like, okay, now we need to go find customers. So what do you do? You hire salespeople to go sell. mean, you're do some of the selling yourself, but like you hire salespeople to go sell more. And so you're spending money and then you start to run out of money because you're losing money every month. then, but you have like some customers, you go back to the venture capitalists and you say, look how fast we're growing. We're getting customers. We've built a product. People love it. Will you give us more money? Right? Like that's been historically the approach. And they're like, sure. Here's another $5 million. And you're like, great. We're going to go hire more people to build more and to hire more sellers to, to sell more. And it's just like this weird cycle. Okay. Bobby: Give me a sense of the timeline on that arc that you just described. Mike Molinet: I mean, it depends on the company, ⁓ but. Bobby: Let's imagine a product, give me like a, give me a tangible product that I can kind of like conceptualize and say, okay, that had these elements and it took us, you know, it would take someone X amount of time to build that and get a, you sales team together. Just so I can imagine the kind of the burn, the money burn that would go into that. Mike Molinet: So in Silicon Valley, it might be spend six months building something or doing, you know, doing research, trying to, trying to figure out what you're building. And you, you think you, you think you found something and you're like, okay, I think I got it. I've got a couple of people that said that, that said they would be customers that said they'll give us money. They haven't given us money yet, but they said they'll give us money and they're using the tool for free right now. And they like it. So then you go. Talk to venture capitalists and you say, listen, we've got commitments from a ton of people. Customers love it. We want to raise more money so we can build more features because they're demanding it. And then those venture capitalists might give you half a million dollars. They might give you $2 million, $1 million, somewhere in that range, right? Bobby: That's at the six month mark. So you spent six months building a product, presumably with your own gusto and whatever money you could scrap together to pay an engineer that was excited about the idea. That gets to the six month mark. You've managed to then get some customers that are using the tool. Then you're able to get some serious investment from an actual venture capitalist. Then what happens? Mike Molinet: Yep. And then you take that money and then you hire a couple more engineers because you need to build more. And then you spend a year building that, let's say, ⁓ building those features, ⁓ you're doing selling yourself. So you're trying to get customers, ⁓ yourself. Maybe you take some of that money and hire like a salesperson, but that's probably a bit premature. ⁓ so you get some more customers, you built some more features and then you go back to the venture capitalists a year later and you say, look at our growth. We've, we've got a bunch more customers. Look how much they're using it. We're, we get a lot of inbound interests. So people asking, like finding us on Google or whatever. ⁓ we think we found something that's legit. then you say, will you give us $5 million or whatever it is? Right. Bobby: Okay. Mike Molinet: give us $5 million and that we're gonna take this to the next level. And they're like, we'll give you $5 million, we want 20 % of the business or something like that, right? Bobby: Okay. Okay, all right. So you're, that's what we've got to about a year, 18 months in this timeline. That's it. That's it. Okay. Mike Molinet: 18 months, two years. Yeah. If you're lucky enough to get to that point, right? If you're lucky enough to get venture funding. Bobby: Okay, okay. So yeah, 18 months, can you describe a couple of questions? How big is how big is this team now at this point, the company you've just described, how many people are like making this happen? Mike Molinet: Let's say six to ten. Bobby: Okay, six to 10 people. what are the product and ancillary products that have been built in order to make this work? So it's obviously the product, and then I presume there's like a website that people could come to describe the kind of software, the ecosystem. Mike Molinet: Yeah, the software ecosystem, at least in the way I'm describing it is there's website and somebody comes to it. They read, they can see kind of like what your software does. And then there's a, like, ⁓ either a signup. If you allow the person to sign up directly or some softwares, like with my prior one, ⁓ it might be book a demo, right? And then you have to set for some softwares. You have to see a demo. You have to get on a call with them before you can actually get access to the Because usually that happens if it's a, it's a, if it's a pricier software, right? because you're, you're less, you're unlikely as like a buyer, a person to go put a $10,000 software on your credit card. it's 20 bucks a month, a hundred bucks a month, sure. You'll put on your saw, you're put on your credit card. But if it's like $20,000, you're not going to put that on a credit card. So you're probably getting on a phone with somebody. ⁓ And then eventually you get access to the product, whatever the actual product is. ⁓ If it's customer support software, then it's a dashboard that shows you all of the messages that have come in from your customers that you need to respond to. then that's it. Bobby: And it's probably that dashboard, I presume has a lot of kind of like links to, as you mentioned earlier, API. So you're interacting with perhaps their email box, maybe with Slack, maybe with various other pieces of software. So that's the complicated bit of engineering that's going on in the background. Okay, great. And that takes us about, that's about a year to 18 months to get us to that stage. Okay, that's 2014. Now, can you just describe the same thing, but in 2026? Mike Molinet: Correct. ⁓ in 2026, you probably don't need to hire other people for, for a while, meaning you could probably build that software in three to six months by yourself. And you could probably find customers by yourself. And if you are able to build it and find customers and have revenue coming in, you have money coming in, and you don't have employees that you need to pay. In theory, you could have a profitable business. And then you don't need venture funding, right? Like the whole venture funding angle, least in Silicon Valley is we're going to burn a lot of money. We're going to spend way more money than we actually are making back because the bet is eventually we're going to reach such a scale that the revenue will surpass our expenses. But that's the long, that's the big bet. That's the long-term bet. And you're like, we're going to, we're going to spend a hundred million dollars with the hope that eventually we become profitable and that bet works. Right. When it, and when it works, it works big. Right. Look at all the biggest companies, Airbnb, Uber, like that's the way those companies were built. ⁓ but there's also tons of companies that burned a lot of money and never made it back. and they're in a graveyard of dead companies. Bobby: All right, question then. there's all these sorts of like knock-on effects that can hear if the model is 12 years later is so different. First question that comes to mind is you described this software, you put kind of just a random tag on it. if you, if they're spending 20 grand to buy the software, they can't put that payment on credit card. Now I presume software is 20 grand because you've got six to 10 employees and it took you 18 months. to get there and you've got five million from the venture capitalist coming in and you need to pay those bills. What happens to the cost? How much can... So I could build this thing in three to six months now, but can I charge the same amount? It's the first question, I guess. Mike Molinet: Well, okay, so. without getting too deep into it. A couple of things I just want to hit on. The venture funded startup, call it Flywheel is interesting because it's basically build something. ⁓ But you then in order to get some money from venture capitalists, they give you money and then. You're like, great, we're going to hire engineers. And then you build something really complicated. And so you build something that's really robust, a really like feature rich product. And you're like, well, in order to justify having 10 engineers and having built these 50 features, we can't sell this product for $99 a month. We need to sell it for $24,000 a year. Right. And it's in order to justify having so many people. It is interesting when you, if you didn't have those expenses, if you had the same software, do you charge $24,000 a year? Do you charge 99 bucks a month? Some would say you charge based on value. So if the value is there and someone's willing to pay you $2,000 a month, 24 K a year, even if it was just you that built it, it doesn't matter. The value is there. So you do value based pricing, right? Which is like. regardless of my expenses, I'm gonna price based on that. But yeah, go ahead. Bobby: But, ⁓ sorry, go on. No, no, you might be about to answer my question. I guess, like, am I being naive in saying, is the value affected by the ease at which one is able to create the product? Like, my labor is less valuable when I'm able to just knock this thing up in five minutes. Mike Molinet: Yeah. I think what happens over time is if I build something in three months, but I'm able to charge $24,000 a year for it. And I have people buying it at that. Somebody else is going to come along. know, Bobby might come along and say, I see Mike selling this thing. I could build that. And Bobby spends three months building it. Same product. He says, I'm going to charge a thousand bucks a month. And next thing you know. You're stealing my customers because your product is just as good and you're charging half the rate, right? And then what happens is somebody comes along and says, I see Bobby selling this thing. I could build that. I'm to charge 500 bucks a month. And next thing you know, somebody's undercutting both of us and stealing our customers. So then you're like, okay, so if it's not that there's got to be some bottom, right? And then the bottom starts to become, ⁓ If your expenses aren't people, your employees, your expenses instead are the software running in the background, meaning like you have servers running, you're doing, you're doing computations, you're doing analysis, you're, you're, providing the service that does cost something. Right. So there is a point at which you can't give away your software for free because it costs something to run. Right. ⁓ and I think then it reaches some new equilibrium. where it might be that $199 a month is the right price for a customer support software that somebody could build in three months. And then at some point, newcomers come along and they say, I could build that in three months, but I'd only make 200 bucks a month from customers. That's not worth it. I'm not gonna go try to compete. So at some point you reach some equilibrium, because they realize that they could go build this other thing and actually make more money. Bobby: That was a very... Mike Molinet: But it does change, it changes the dynamic for sure. Bobby: That was a very, very good economics lesson bike. Very good. I would say. Mike Molinet: I'm an engineer, mechanical engineer. did study economics, but I've seen enough to realize how people. Bobby: No, but as a very real, mean, you're describing market forces basically, aren't you? And just how it works essentially, which yeah, was kind of like when you lay it out like that, makes complete sense. So. Mike Molinet: You want, you want to complicate it one layer that Silicon Valley complicates it. Okay. So yeah, you started at $2,000 a month because you had, you had a bunch of employees and then Bobby came along and built it and sold it for a thousand bucks a month because he five coded it in three months and it's good enough. And then somebody came in and eventually you worked down to 200 bucks a month. That's the bottom. That's the lowest anyone can go. Cause they're like my server costs are $150 of that $200 a month. So I can't charge any less. Otherwise I can't. Bobby: Yeah, I think we can take it. Mike Molinet: pay for myself. And then somebody comes along and says, I'm going to build the future of customer support software. they, they set some grand vision and then they go down, they go to Silicon Valley. They talked to all the investors and the investors say, wow, that's a pretty big vision. And they say, here's $10 million have at it. And That company decides to give the customer support software away for free. And customers flock to them. They're like free customer support software. Sorry, Bobby. was paying you 200 bucks a month, but these guys are charging me zero or 50 bucks a month. Right. That's that VC backed companies taking a loss. They're losing costs. They're losing money every month. You can't do that because you know, it's your livelihood, but this VC backed company that has $10 million in the bank and, and they do that and they get more and more customers. Next thing they know, Bobby's customer support software goes out of business. Mike's. support software goes out of business and they're the only ones left in town. And then what happens? They jack up the price to a thousand bucks a month or they start charging you for other stuff. They're like, cool, keep using the customer support software. But we built this other tool that you're going to want and that's a thousand bucks a month. And you're like, I'm so locked into this support software and there's no other options on the market. I have to keep using them. I guess I got to pay them a thousand bucks a month. So that's the venture capital comes in, funds a business that undercuts everybody. They take a loss for a certain period of time until then they can start charging more. Think about Uber, Uber's model, very much that way, right? Bobby: That's what came to mind. They essentially did that to all of the taxi drivers around the world, right? Essentially. Mike Molinet: Yeah. And then you get, then you're stuck with Uber and then Uber jacks up the price. The thing that used to cost you $8 now costs you 40. And you're like, what the hell? Well, you don't have any other alternatives. Bobby: Yeah. What happened there? Another thing that you're making me think of is a developer of friend of mine was, you know, he, what he's worried about with the whole vibe coding revolution is that, okay, now we're all becoming reliant on, on the apps, on the AIs to code for us. But once they're so bedded in, that none of us know how to code anymore and we need it to do our businesses. That's when they'll start jacking the prices and. you know, take take the money back from their crazy investments. What do you Mike Molinet: Yeah, you're talking about like Claude code and yeah, open AI for sure. Bobby: Yeah, that's what he suggested. I don't know if there's any truth in that because I guess they're going to keep evolving and getting better, it does make some, you know, it sounds like a reasonable fear. Mike Molinet: I think both will be true. Yeah. They will keep evolving and getting better. ⁓ but they will, we took, talk about it sometimes it's it's venture backed companies tend to subsidize their offerings. Right. So one could argue that a lot of these large language models and thropic, runs Claude open AI, which runs chat GPT. They are probably subsidizing our use, right? And you see it because you can use some of the software for free ⁓ to a certain extent, but they're happy, and keep the cost low. They're happy subsidizing just like DoorDash, just like Uber did in their early days. There's a lot of examples of this to get users, to get ⁓ growth and adoption. And then at some point you say, we need to start focusing on profitability. And that's when you start... turning knobs, right? You start saying, what expenses can we cut and limit? Where can we increase the price? Where can we add more products that people have to pay for because they're locked in now? ⁓ So I think you'll start to see some of that. but I think what might happen with these AI companies is it'll be ⁓ the newest models are expensive, but can use the old models for cheaper for free. So it might, you might not run into a situation where like open AI or anthropic come and say that that model you've been using for two years. Now you have to pay 200 bucks a month. It might just be like, keep using that. That's fine. But everyone else is using the new model and the new model is 10 times better. And that one's 200 bucks a month. And you'd be like, and some people, some percentage of people will be like, I need to be able to compete with everybody else. And I need, so I need to use the new model and I'm willing to pay 200 bucks a month for that. Bobby: you imagine a world in which, ⁓ you know, ⁓ open AI has a provable knowledge worker is as effective as a human being ⁓ and the price tag going to put on that 20 grand because we're saving you the hundred grand salary ⁓ or do you think it's just going to be the slow creep? guess I'm asking, do imagine them dropping a product that they say, right, ⁓ we've cracked it, here is, ⁓ just there's a big tag on it, but it's going to save you so much money because it's actually making you all this money. ⁓ Mike Molinet: Probably not in terms of that job. think we'll definitely get there to that point, but I think human psychology would probably say we're more likely to accept that if it's a slow creep. And you see the slow creep happening right now, right? Which is companies like Anthropic and OpenAI, they first start with the chat window, right? I can talk to it, it comes back. Then they start doing more like, okay, now you can connect to other tools, right? You have... Your Gmail, you have your calendar, you have Google spreadsheets. You can connect those and we can actually do some work with those. And then it's like, okay, you can connect up folders and put files in the folders and we'll do work on those files. And then they're like, Hey, we have some plugins. We have some skills. You can, you can say, Hey, do like, go grab this plugin from the Claude plugin store. And it does work for me on the files that I already connected it to. Right. So it's that slow and soon now you can extrapolate out. Okay. Where else is that going? Right. But that's been that slow creep one thing at a time. And every time you're like, ⁓ my God, this is awesome. If they'd come out of the gate and said, you can now replace an entire human with this thing, by the way, it's not there yet. It's not there, but even if it were, I don't think they would come out of the gate and say that. I think they would simply provide the breadcrumbs of tools that you can do. And then next thing you know, Bobby: Yeah. Mike Molinet: two years from now, you're able to replace an entire employee by telling it what you want. And it's like, cool, here you go. Here's your sales. Here's your sales employee and it'll cost you $2,000 a month, but that's way cheaper than $10,000. Bobby: Yeah. Yeah, very wise. Well put. I think about it sometimes of there's two different angles. One is that, that of course the, company is just one person's vision and that one person is trying to make things happen and they have to delegate tasks. And in order to delegate tasks, they have to hire people to do that. So that one person at the top of the company is just going to need less people because they. can delegate more to AIs, I presume. Meanwhile, there'll be people like yourself, I guess, Mike, who are starting new businesses and they're able to start them with less resources and create bigger things. And so I guess that's how the balance is gonna be shifting over the coming months and years. Mike Molinet: Yeah, sure. You see a lot of businesses started now that are just one or two person businesses that are able to make money that historically would have needed to be 10 people. The way I think about it is you're going to have a lot more people that can run small businesses that historically weren't able to. So I think you're have a lot more of those solopreneurs or one to three person businesses that historically couldn't have existed. Bobby: So I to come back to this ⁓ new ecosystem and ⁓ the, like ⁓ the cost. barrier to entry so much lower that we can all build now ⁓ in theory. What type of product does one build in this new world? know, like, how do you know what's valuable to build? I've been just my own company. ⁓ so I'm, ⁓ I don't own this company. I'm Creative Director for the New York branch with 20 people around the world. I've started with building an estimate creator, like a budget. When people come to us, they need an estimate. You used to do that in a spreadsheet. Now the AI assistant builds it out and it's really polished and lovely and costed up in five seconds. we started building core sheets. Now we have a project planner, kind of like monday.com. but the client can look at it and they can see all of the movies that we've made for them. They can see the schedule and it's all very nicely branded for them. That's software as a service that we were on the cusp of getting monday.com and now we're like, we don't need it. We just built our own. So how do you... Mike Molinet: Mmm. Bobby: Do you see what I'm saying? you you just, you need, you need one enthusiastic vibe coder in your company and that person can kind of build everything that you, you need to run the business. so that's obviously going to impact the market. how do you think about that? And I feel like we're like, you know, I think based on what I'm reading and hearing and listening to Wall Street is responding to this very thing that I'm talking about right now. ⁓ Mike Molinet: Hmm. Bobby: But it seems like Wall Street's just moving on their own vibes. If I understand this kind of, know, nothing is related to the actual economic results of any studies and, you know, and all of the studies are saying that there's actually no impact in businesses. Meanwhile, I'm literally firsthand, like seeing our entire business, things changing, you know, like my colleagues who are less technically savvy than me, all of a sudden. Mike Molinet: For sure. He Mm-hmm. Bobby: I've shown them Claude and they're having light bulb moments and they're going, ⁓ my God, I can do the spreadsheet in Claude now and I couldn't do that a month ago. And ⁓ it's been quite satisfying for me because I've been banging the AI drum for two years and moving everyone along really, really slowly. But now all of a sudden it's picking up and I'm seeing it. So I'm having some cognitive dissonance as I'm like, the economist number A is saying, ⁓ yeah, but it's a... It's a fraction of a percent of an impact on real-term impact. So that was a big kind of like verbal splurge into your ears there. my question, I think, if I remember rightly, was ⁓ how do you choose what to build when this market is so unpredictable? Mike Molinet: Yeah, you talk about like internally build versus buy. Bobby: As in like you're an entrepreneur trying to decide what product to build to sell. How do you choose what to build to get investment and, you know, or, or, or revenue. Mike Molinet: Okay. Okay. Yeah. So, so what's interesting is that question is different than a lot of the stuff that you were hinting at earlier, or that you were talking about earlier about all the stuff that you built internally. And we could talk about, we could talk for an hour about both of these things. the, let me touch on like, how do you decide what to build if you're an entrepreneur and you want to build something and go sell. And then I want to come back to the internal building and how that's going to affect the. it in my view. How do you decide what to build? Honestly, I think the biggest thing is you're, you're not going to know upfront. You should just, you should just build the things that you're interested in. Fine. Like if there's something that causes you pain, causes you friction, that you wish existed, you should just try to go build it. Because if it's causing you pain, first of all, if you, if you build it and it eases your pain, that's a win. And then if other people happen to have the same pain, then you can sell it to them. Like my co-founder. built something one day, just as like a little side thing to solve a pain that he had with his social media. And I was like, dude, can you package that up and sell it to me? Because it was just an internal thing. And I was like, I will, I will pay you $49 a month for it because I want that too. Like that's a great sign, but he didn't build it for me. He didn't build it to try to sell. He built it because he just hated doing this one thing on social media. was like, dude, I would. So that's, think, the best place to start. because if, let's say you, let's say you have a brilliant vision for some product that needs to exist that you think can make a lot of money, but you don't care about it at all. You're not going to find yourself coming back to it and being excited about it, about like spending nights or mornings or during the day. It's just, I'm a big fan of like. Build around what you're excited about. Build around what you're curious about. That's the most important thing you can do. Now you might discover some other opportunities along the way, but if you do that, you'll be happy. ⁓ You want to talk about internal tools or anything that you want to follow up on? OK. Bobby: Yeah, that was a I think that was a great answer. think that kind of is the spirit of creativity in a nutshell. Anyway, do it because you love it, not because you're trying to make some money somewhere down the line. Do it because it's useful for you. And I relate to that basically. Everything I've built has been because I think it's got a use for me at the company. And that's kind of the weird thing. I'm like, ⁓ I can't see any value. This is the thing. If I'd done any of these things five years ago, I'd be like, I'm going off to try and sell this thing. But because I built it so easily, I'm like, well, I don't know if there's any actual value in this. But then I think about how much time I've spent working on this thing in evenings and on weekends and, know, okay, this UX could be a little bit better if it just had this element here. And I think that's where the value is. It's actually in the UX kind of unlocks. Mike Molinet: Mm-hmm. Bobby: that you only know because you understand the problem because you have the problem yourself. Mike Molinet: I agree completely. Okay. So the internal tools piece, this is interesting. Basically, I think, I think it's going to be a pendulum. I think we've historically been on the small number of companies ⁓ own build everything. You have to spend a lot of money for it. So if you want to CRM, there's three CRMs that you can buy and you have to spend a lot of money on. And now it's what's happening is now people are starting to realize, ⁓ I can just build it. I can build anything. I can build anything. I can build my own CRM. ⁓ And then everybody's going to start building their own stuff. You're going to build your own CRM. Mike builds his own CM. Sarah builds her own CRM. And it's going to swing all the way up to the other side where everyone's just building internal tools. Instead of buying monday.com, you build your own little project management tool. And then what's going to happen is everyone's going to have a bunch of these vibe coded internal tools that they built themselves. And they realize they're going to realize that they're spending 15 hours a week maintaining those. random bugs, random edge cases, random things that they need to sort out. A provider change, we talked about APIs earlier, a provider change how their API works. You now need to go spend an hour updating your software to support their new API. Okay. Your security credentials expired. So now you need to go into cloud flare and ⁓ reauthenticate so that your security credentials are valid. And next thing you know, people have vibe coded everything. They built all their tools internally and they're like, this is what a pain in the ass. And it's going to swing. It's going to land somewhere in the middle. There's some things that people build internally because they're very bespoke or nuanced to their specific use case that nothing out there provides. And there's going to be a lot of stuff that they buy off the shelf because they just don't, it doesn't make sense for them to try to maintain those things when they can buy something off the shelf for 200 bucks a month. So. I think we're going to see it first swing all the way into everyone's building their own stuff, five coding their own CRM, et cetera. And then it's going to come back to the middle. And sorry, one more thing. There's a question around what about all those businesses? They're busy. Those businesses are going to be fine as long as they adapt their business model. What I mean by that is maybe historically they're making all their money on per seat subscriptions, right? Meaning like I buy a CRM. buy. Bobby: I think that's it. Mike Molinet: Monday.com I pay 40 bucks a month per seat. I think those companies will be able to provide a lot of value via APIs and other AI automations that may be the, what you're using them for is different. So I think that those companies will be fine. I'm not worried about them. So I think wall street's overreacting. Bobby: That's really interesting. And actually I take that as a bit of a healthy reality check because there's all of these tools being built. And as I say, you know, blindfolded, like Sandra Bullock, kind of put in all of these APIs in and doing things in the backend and wondering what's going to happen in like a couple of months time, or sometimes get a notification on Google AI studio saying, there's been a security thing, you need to do this thing. And I'm like, what? Mike Molinet: Yeah. Bobby: And it sounds like you're saying that will actually keep happening. And I may have just dug a, know, dug a, what's the word, a grave for myself that I could end up dying in as everyone's asking me to fix all of these tools that I've built in a moment of enthusiasm. Mike Molinet: Mm-hmm. Mm-hmm you I think what you'll do is over time you'll find the ones that are truly valuable and custom to your use case and it'll be worth doing the maintenance on those and then the other ones you'll just end up going and buying a software instead and So you don't have to maintain them Bobby: Yeah, yeah. think this, knowing my boss, he's gonna make sure there's one of us at the company knows how to vibe code and he's gonna say, you're in charge of this, fix it. And do your other job as well. But that's the trials and tribulations of a small business. All right, cool, man, that was really great. I'm gonna call that the main bit of the meat of the. Mike Molinet: That's right. Bobby: podcast, but before we finish up, so what I usually do is a quick tips section, know, like, top tips. But for you, I think you've already kind of given us your general tip, which I would say is use replete, replete, replete, replete and bolt you would say, and kind of vibe code something. Mike Molinet: Okay. Bobby: personal superpower, I'm might well have something to do with open claw. ⁓ So I'm wondering if you can give your sort of top level ⁓ of how one might engage with open claw and why they might to, if they indeed would like to. Mike Molinet: ⁓ is for the person that has been vibe coding for a few months using one of the existing tools ⁓ they've heard about Open Claw and they're curious about what the hype is about. ⁓ they're willing to find, buy, or use a spare computer that they have somewhere to set it up and they're willing to spend five to 10 hours Like getting it set up and then playing around with it and getting into a place where they feel like they understand how it works. Okay. That's who that's for somebody that has never vibe coded before open claw is. It's too, it's too much. It's just not worth it. So I that's vibe coded a little bit. You could do it. You're probably going to get frustrated, but if you're willing to put in the time and you're willing to just fight through it, you can figure it out. Bobby: And what do you get when you, like if you were willing to put in the time and you're willing to figure it out, what is the light at the end of the tunnel or the pot of gold at the end of the rainbow? Mike Molinet: A very clunky, yeah, a very clunky, hard to use UI that has a chat window that kind of works, but not really. That's what you see, but what you don't see. What you don't see is a combination of tools behind that chat that can actually go take action for you. Here's what I mean by that. Traditional AI chat tools that people are used to using, chat GPT, Claude, et cetera. You say, whatever, you might ask it for something. ⁓ I want to start finding more clients and I need to... send some emails to them and it'll go back and forth and eventually it'll spit out. Here's what you could do to find more clients. Here's a draft email that you could send them, but the work still falls on you, right? You still have to go do the thing. With OpenClaw, what OpenClaw is, is it's that chat, but it also has what are called skills. These skills are basically pre-built instructions on how to actually go do the thing. ⁓ here's an analogy. It's not a great analogy, but it might resonate with some people. Chat GPT is the brain. It's smart. can think a lot, but, ⁓ a brain or a head without hands can't do a lot. could talk to you and be, and give you really interesting thoughts, but it can't, it can tell you how to build that bookshelf. It can give you the instructions very clearly, but it doesn't have hands. It can't build the bookshelf for you. ⁓ it can't cook the dinner for you. It can tell you the recipe. can tell you how to do it. It can tell you step by step, but you still need to do it. But with open claw, it's connecting the brain to, ⁓ tools that can actually take action. And so it's giving the brain hands. And so now, instead of it telling you, here's how you can go do that thing. It can actually go do that thing. So. You want it to send emails? It can send emails. You want it to move things around on your calendar and clean up your schedule? It can do that. You want it to work on some files? It can do that. ⁓ So it's connecting the brain, the chat GPT essentially, with tools, skills that can actually take action. That's the beauty of it. There's some other things behind it as well, which I won't get into, but like that's the gist. Bobby: quest. Question, how is that different to, let's say I have Claude Cowork and earlier today I asked it to pull a bunch of JPEGs out of a whole ton of different folders and put them all into one folder for me. And it went away and it worked and it did it. Or I have Gemini on my phone and I've got a pixel and I often ask it to put this calendar event in, deal with this, deal with that, know, delete that calendar event just by, you know, giving it a prompt. So what is OpenClaude? doing that's different. Those tools are manipulating the world. how is... Mike Molinet: Yeah. Great question. So it, that's where you start getting into some of the nuances of the additional things that can do. So in like Claude cowork, for example, you give it a folder, you say, you can work in this folder. Here's the files. Go do this thing. And it can, it has hands then they can go manipulate it's sandbox. So like it can just go do that thing with, with open claw, can say, Uh, Open up, go open a browser, go run around in that browser, go to my Twitter, go, uh, even though you're not allowed to do this, you could in theory, um, go make a bunch of posts for me, go comment on a bunch of things, go write this article that I've been thinking about going to LinkedIn, do this other thing, go look for flights on kayak.com, go look for hotel options on booking.com. Oh, and by the way, um, every morning I want to get a daily briefing, uh, that gives me today's news, the weather, et cetera. I want you to also send me some updates during the day. At the evening, I want you to send me a motivational quote, ⁓ whatever. And so a couple things that it's doing there. One is it has access to do more tools than what something like Cloud Cowork has today. ⁓ It has browser access. It has skills and access to tools that, for example, maybe Cloud Cowork doesn't have. The other thing that it has is the ability to schedule things. Right. So right now you have to go tell Claude cowork, go do this thing right now. But if you said, Hey Claude cowork, can you go do this thing tomorrow morning and then email me the summary afterwards? Don't do it until 6 a.m. because I to be sending some stuff in, put some files in tonight. I'm pretty sure I can't do that. Right. Or if you said every day, I want you to go review this, these files, and I want you to create a summary based on the leads that. came in through our website. ⁓ it can't do that because it doesn't have the scheduling capability of doing that. Right. So you start getting into some, now, none of this is rocket science, right? Like it's the, the large language model chat GPT, let's say it's the skills, the tools, like it can go manipulate. It's also scheduling, which is simply just like, it sets a task to take care of something. at 6 a.m. every day at 6 a.m. There's, it's all doable. was just open claw was one of the first things that a lot of people use where they started to see if you combine all those things together, it can be really powerful. Bobby: And is it, does it have more autonomy? it faster? So cowork can, know, Claude can control my browser. So can I feel like Gemini can as well. ⁓ anti-gravity. Anti-gravity can control my browser because it's checking what, you know, often it's checking stuff that I've built. But, I've seen an AI controlling a browser, but they're very slow and clunky. And I've, you know, and I, my mother-in-law can't post things on Instagram or she struggles. Mike Molinet: Yeah. Bobby: bless her, but she struggles posting stuff on Instagram. ⁓ so I was like, ⁓ Claude can control your browser. Let's see if it can do it. And it would do it, but it was so slow. It took forever, maybe like half an hour of watching this thing, just clunkily working through the navigation thing. And then when it finally put it up, it cropped the photo weird. it was just, I was embarrassed. I had never been embarrassed. Mike Molinet: Hmm. Classic, classic. the reason, the reason it's doing that and it's so slow and open clause the same way, unless you tell it to do it a different way, the reason in that case, what, what it was doing was it would open the browser and it would take a screenshot of the browser. You, wouldn't know this was happening, but it'll take a screenshot of the browser and it would, and then it would send that screenshot to the large language model and thropic or ⁓ open AI. Bobby: Yeah. Mike Molinet: large language model, that large language model would then analyze the screenshot and it would return back to Claude. This is what's in the screenshot. And then it would have to analyze, okay, what do do next? Click this button and then new page would load. Okay. Take another screenshot. And so it keeps going back and forth basically with screenshots, analyzing the screenshot, deciding what to click next. So it's just a very slow version of what we do. Bobby: So it sounds like, if I'm not mistaken then, that OpenClaw is not dissimilar to what a lot of the models already, you know, some of the offerings from the main labs, but there's just less guardrails and more autonomy. Like you can just do more scheduling with it or just let it run on tasks for longer. that kind of, does it, is it a limitless run on tasks? Could it just go in infinitely running on tasks if you wanted it to? Mike Molinet: Yeah. If you designed it right, you could. Yeah. Yeah. ⁓ Bobby: Okay. And that's obviously something you can't do with the current large language models, but I feel like we're only at one setting away. You know what I mean? With to anthropic saying, Hey, let's just, all right, we're, we're giving co-work now the ability to keep running. It'll have a heartbeat. It'll wake up every 30 minutes for you and check that thing that you asked it to check. Um, okay. Okay. That was helpful. That was a helpful insight. Okay. Mike Molinet: We're very close, very close. Mm-hmm. Bobby: Mike, it's been so great to chat to you. Thanks so for coming on the show. And to our listeners, thank you. ⁓ We'll see you next time unless it really is the end the world. Mike Molinet: Thanks for having me. Bobby: well with that, is there anything that you want to promote while I've still got you? Mike Molinet: just, I, I created a little like, ⁓ free email course for people that are kind of like newer to AI. It's called AI drop daily. So AI drop daily.com. ⁓ just, can sign up, get an email five minutes a day. It just kind of like teaches you a small lesson every day. ⁓ just to help people. Cause I found family members just wanting to learn more, but not knowing where to get started. And so I built that. Well, AI built it. Bobby: That's great. ⁓ Fabulous. Open Claw. So Open Claw running that. That's so great. All right, that's fantastic. I'll put any links that you want in the description of the show as well. Mike Molinet: Open call, of course. Of course.