OpenClaw Has Taken Over Our Businesses

0:00Welcome back to the Building Public. And in this episode, we're going to be covering all things OpenClaw. From agentic workflows, from managing your own business [music] to how you can use it every single day. I brought on my buddies Paul and Brett who have been diving deep into OpenClaw, figuring out how it's working for them, how did they set it up. So, let's get right into it. Why do you think OpenC is having such a big like virality moment right now? Why is it changing the landscape? How are you using it? uh just kind of break it all down, you know? >> Yeah. So, I think that it's what everybody thought Siri was going to be whenever they first released Siri, whenever Apple first released Siri, like, "Oh, we have this thing that will get to know me and I I'll have this

0:45thing that can take actions on my behalf on my phone or on my computer or in my case on a headless browser inside of a Docker container inside of a Docker container on a VPS." [laughter] Um, but I think that back whenever I first got into AI, I was really interested in like what the people at Lang Chain and Crew AI were doing because they were like the first ones to really dip their toe in the water and say, "What happens if you have an AI return a tool call and then we execute on that tool to tool call and then feed the results back and then what what happens if we do all of that?" And I remember being like, "Oh man, these things are like actually going to do things for you." And then the models I think at the time just weren't smart enough. So I think it was there's always

1:31been this like end up energy of man I really hope that we can give these AIs tools that give them access to all the things that are meaningful to humans that use computers for work. And then the models got smart enough and the harness got good enough to remember things. And I think that's why there's such a big moment happening is first it started everybody wanted these things. Everyone's been working on it. Then the models got smart enough and I mean honestly a lot of credit to Claude Code for showing us what is possible with a harness. Um so that's that's what I think. >> No, I love it. And and for you Brett, you've been kind of implementing it for these last few weeks. Why do you think it's going to have such a big impact

2:17over these next few months in PBS where it flows? And >> I would love to go back. I've got to see if I can find a text to Paul from like a year ago where I was going I think I'm trying to get this agent I don't even know if I was using the word agent yet at that point. I was trying to get Claude to completely go end to end on a process. Come up with an idea, research it and build it. And I remember it was like a five or six hour process where it would run and at the end when it was done you might get like a you might get a semblance of a web app, right? but not not even the stuff that we can oneshot today. And that was 5 or 6 hours. So, I think it's exciting because like this concept has been kicking around for a while now. And it's the first like aha moment for everybody where you go, wait,

3:02I can just keep I can just ask it to do more work. And you know, I think that this was just the perfect combination of things that existed really. the tool calls, cron jobs, local file storage, somebody just took the time to orchestrate it really well and make it so like with one command line prompt, you can have this thing up and running. Now, obviously, we saw a lot of blowback from people like, oh, it's a security nightmare. It's like, well, yeah, he basically just gave everybody the nuclear codes. It's up to you to use it to create renewable energy, not to implode your system and delete your database and give all your customers records away. So, I posted something on X about how like we were given the fastest uh you know, fastest vehicle in the world and no helmet. And so, I think

3:47that's what you see. There's a little bit of skepticism, but I am all in the optimistic camp of like, of course, it's dangerous. It's new. It's uncharted. It's wild. I think that's the only reason that this succeeds is because it's open source because no company is going to take the liability for this. M but giving the tool to the people and letting the people decide what they want to do with it is not only awesome but I think it's unprecedented right like he basically undercut he leapfrogged all of the AI companies that can't that haven't done this yet they're probably doing it behind the scenes but he was just like hey you know I I had some time here you go see what you can make with this and as a tinkerer and a just avid explorer I was like oh my goodness like gimme gimme gimme I got to have this so >> yeah it's it's pretty unreal And as fast

4:34as this space moves, we know that we're feel like we're falling behind. But now with this autonomous agent at our hands, >> yeah, >> uh being able to help us in our businesses, personal life, um it feels like we can now like at least digest a lot of the new technology that comes out a little easier. You should see my Twitter bookmarks. I have like 150 in the last few days. I >> think if you're an early adopter for this too, you have to take with it that the learning is really the output here, right? Because again going back to like oh the security thing it's like okay what we're learning right now is how to work with this type of agency how to orchestrate it how to keep it on the rails and keep it safe and keep it on task and I think a lot of that is handled by openclaw and I think a lot of

5:20that will be improved through updates but just us as the forefront the people on the forefront of this technology using it will be able to say hey you know company X this is how you need to structure your agents so that they can work well with each other or so they can share ideas or which is all novel and new. So I think that's exciting too and that's the part that while while software and things tend to are going to start to dwindle the need for software packages will go away then the it'll be replaced by the need for people who can orchestrate and build good systems. So I think being on the forefront of that right now is smart. >> I I just found the text message. Sorry Bo, I just found Oh my goodness. >> From Brett, [laughter] >> you were [clears throat] talking to me

6:05about Remotion video, Brett. And then you >> also ahead of the curve on Remotion video. >> Yes, you were. You were like four months. Four months. Uh I had it build the idea of the day from idea browser. Would you believe I hit my model rate limit for the night? This is crazy. The app now has Working Gemini integration for determining the optimal lineup for the kids soccer team. This was less than 30 minutes to a functional app on November 20th, 2021. >> Yeah. That's actually even post the original workflow. But yeah, I was doing uh I'm doing a kids soccer subbing app. So, as a as a soccer uh coach parent, one of the things that's hard is matching the kid's talent levels. So, I built an app where I can just give them candid comments about like this kid is fast, this kid is not, this kid can kick with his left foot. And then I go, hey,

6:51I need to sub out person A and I'm going to get suggested from who's on the bench. this person a good is a good like for like swap for them or if you swap this guy also swap this guy so that it balances out. >> Taking it to a whole new level, man. But anyways, yeah, that's that's funny that that pro football focus has all those metrics that they track. You're doing it for >> I'm just taking it to U1 soccer, baby. I'm gonna we're gonna dominate the league. >> That's a hyper niche, but I'm sure there's money, you know. >> No, but it's interesting like going back to the idea of the day thing. That's that's one of the types of things that the open claw gets me excited about is this idea that you can have an agent that's constantly out looking on the internet for the for a great idea and then vet itself. You know, you could

7:36even have multiple agents try and agree and pick the idea apart, find blind spots. When you have a winner, develop the MVP and tomorrow morning send it to me as a message, as a Slack message and let me check it out. If I like it, we'll spend some more resources and some time. if I don't, I lost nothing because these agents were running during idle hours during the night. So, I think this idea of idea mining is going to take off for a while. Um, I've been experimenting with an agent right now and I can show you guys a visualization of like the workflow, but essentially it's finding local businesses near me. I'm just doing it locally for now that have a poor like a bad website or a non-existent website. It's drafting up the communication. I'm manually approving all the emails for right now because I don't trust the the system yet to not send out obvious AI

8:21writing and I don't want to be disingenuine when I reach out to these people. I do want if it's going to be an AI generated message, it needs to be in my voice. I don't know. I'm particular about that. Um, so anyways, I'm going in approving the email. It's sending it, but then it's watching the send replies. If they reply, because I'm telling them in the email, I've already built your new website for you. And if they reply, the agent will build it for me and send and publishers it to Cloudflare and sends me a preview URL so that I can send it right back to the client and say, "Yep, here's the site. Take a look. If you like it, I'll help you deploy it. It's built." But this idea of just like ondemand building is, I think, pretty new and going to be really prevalent going forward. Like, as this gets into the hands of more and more lay people, like the less technological people, they're going to be able to say, "Wait, I can just ask for like a new banking

9:07app. like you have one account over here and one account over here and one account over here. Why don't you just ask it to make you Bose's banking balances from your three account? You know what I mean? Like you'll just be able to ask for things like that and that's insane to me. So that's what you know say like why are people excited? I don't even know if people are excited for that yet. That's what gets me jazzed up is that uh we're in this age of just ask for it and you can have it. Yeah. >> Yeah. Yeah. You could build your own dashboard, your own uh your own instantly or go high level, your own Photoshop, your own thumbnail maker, your own emails. When I'm sending out emails to people, the agent was like, "Oh, we should get a service that handles cold email outreach to make sure we're warming up the inbox, we're spacing out the emails, etc." And I

9:52said, "Is any of that stuff that you can't do?" And the agent was like, "No, I don't think so." I said, "Good. Then build all of that logic into our email sequencing system. take everything you know about cold email outreach software and include that in our stuff. Yeah. Like that's it. Like that is a great suggestion, but the the data it was trained on is outdated. I no longer need to go pay for that software package that that is a middleman to do that. I'll just say yeah, adopt that to our system. Now I have that. So yeah, the moat of the moat of like, oh, I need to pay for instantly or I need to pay for whatever the service is because I don't feel like learning how to code and I don't feel like learning how to do the stuff necessary to build instantly. It's like you don't have to learn because you have super intelligence at your fingertips.

10:37>> Yeah. Yep. >> It's starting to feel like what do we exactly do here? >> Exactly. I'm telling you, just it's the orchestration and I' I've run into some limits with mine. That's why I still think we're not fully So I'm I'm going as deep as I can to get this as autonomous as I can as quickly as I can. And I'm still getting things where I'll say, you know, it'll say, "Hey, here's the subject line for that email." And I'll say, "Oo, I'm not really I don't really like that subject line." And it'll go, "What subject line? Sorry, I forgot what we were talking about." I'm like, "Well, that was one message ago." So, I wouldn't bet my company's reputation on are you only going to email them one time? Are you gonna accidentally email them 600 times? Like, are you gonna get my domain flagged? Are you going to upset potential customers

11:22and give me a bad reputation? So, as much as I'm bullish on this, and I love diving into it, I'm still doing human in the loop activities because I want to I want to test it as hard as I can behind closed doors. And then anything that's customerf facing, I still want to be in the loop on until the behind closed doors activities are just fine-tuned. and and oh my goodness then when you to your point like what am I even doing here I check in you've already done it I'm still playing a heavy role in terms of putting the right systems in place and honestly it's a funny thing like just giving it the the nudges that it needs like for example that email sequencing thing it doesn't think like oh I can just build that for you it's like hey there's software out there that does it and I have to say well can't you just make that and it's like sure yeah

12:07>> um so yeah I I think being involved in that process still matters for now And you know, will it matter a year from now? I don't know. But I think being in this early adopter group, I'm not afraid of that because I'll be on the next thing. >> Yeah. >> That's emerging. But I do think this is a a wave that you've got to ride or you're sitting on your board waiting for the next one to come in. >> Yeah. Yeah. >> My Yeah. Yeah. And it's so exciting. Um but I want to back up here for a minute because you know, you guys have been using it for a couple weeks now. And Paul, you know, you you got other businesses going on. I want to know like how is it serving you? Like what do you how how do you have it installed first of all? Cuz there's two different ways, locally or on the cloud. And how how did you set that up? And and what are you

12:53having it do? >> So my um my original goal was to have it take the administrative role of someone in my service business and do the entire administrative role. Um, but I I didn't want to use the software that we currently use for that business. So, I'm rebuilding that software, the the system of record that we're using for that business. I'm completely rebuilding it. Um, and I'm rebuilding it because that software doesn't expose an MCP, a CLI, or an API, which in my opinion, brief side note, is just like the easiest way to light your business on fire today. Is like if you don't expose an easy way for

13:40AI to interact with your software on behalf of a user, like you're just you're you're like knowingly staying behind. And interestingly enough, I could build an API on top of this other company's software, but it's not a public public API. It's not maintained. It's not managed by them. It's just the end points that you can see inside the network logs. Um, and the Chrome DevTools MCP Brett told me about months and months ago helped me figure that out. So, I I was going to use Cloudbot to uh run an administrative role in my service business, but then I realized that I didn't want to build my own API on top of non-public

14:27um API endpoints. So, I decided to build my own software using cloud code. Um and that's going well. It's it's just something that's like on the back burner right now because the business is running well and I'm not like putting all of my time and effort really into it. So the the Claudebot, I'm going to keep calling it Cloudbot. >> Yeah, exactly. the the Cloudbot. I transitioned from that role of administrator in that business to just being my personal COO because I feel like I have a terrible time keeping everything that I'm working on in order because I'm working on so many different things because I can work on so many different things with unlimited super intelligence at my fingertips because I spend 200 bucks a month on the Claude

15:14Max plan. I feel like I can work on anything and do anything. So, it's the personal COO. I'll I'll rack this up, I promise. Personal COO to me and my life via Slack. And the way that I [clears throat] decided to install it was on a uh Digital Ocean droplet. Um, and I needed to have browser access, but I don't want it to have browser access to the browser on my computer. I want it to spin up its own headless browser inside of a Docker container. So uh openclaw exposes a dockerized installment of openclaw and then you can install another docker container inside of that docker container where openclaw is running that's effectively a tool and that docker in docker is the openclaw

16:02opening a browser. So, OpenClaw runs inside docker number one and then inside docker number one there's a second docker container that holds just the browser volume um so that it can effectively reach in and say hey let me see this web page and it can go through the entire web page and navigate the web page all 100% in the cloud without ever having to have it on a local computer. Um, and the way that I set that up was I [clears throat] gave the docs for OpenClaw to uh claude code because I don't know how to set up Docker in Docker. I didn't even know how to SSH into a VPS. Like I didn't know what what that meant at all. I didn't know that I had an SSH key on my machine. So I said to Claude Code, "Hey, I want to set this up in the most secure way possible.

16:48Every single step that you take needs to be done in the most secure way possible." and it sshed into the droplet. It installed openclaw. It installed the docker inside of the docker container. It communicated through a through the slack mcp with my with my openclaw instance and set it all up for me without me ever having to lift a finger. Now, I went back in and like did a security audit of like what does this mean? Why are we doing it this way? Is this the most secure way of doing it? yada yada. Um, but yeah, I had Claude Code set it all up for me by sshing into a digital ocean droplet that has the Docker with the Docker in Docker uh installment running and it's fantastic.

17:33I can go into more on like what it actually does for me, but that's a very long answer to your question. >> Mhm. No, I there's multiple ways to do it. You know, everyone's buying Mac Studios and Mac minis and it's it's not the only way. you know, people are you running it from their cell phone, you know, so it's just yeah, multiple ways to have access to this technology. And Brett, you kind of went a different route. You dusted off the old laptop. Yeah. >> Uh, and how's that working? Do you find >> So, I've got a uh an old MacBook. It was probably a 2012, 2013. And they the only issue I ran into was it was too old to be put on the latest OS and all of the software involved for OpenClaw required a version beyond what I could get. Uh, so I actually just wound up putting Ubuntu on it. I'm not a big Linux user,

18:19but um I I wanted to go with something where I knew a lot of the command line tools would be supported and so stuff that works on Mac and Unix is uh going to work on obviously Linux. And so yeah, I I haven't had any problem with that at all. Honestly, it sits over here. This is so funny because when I first set it up, if I would close the lid, it would shut itself down. So I open the lid, it boots up. I messaged my open call instance and I said, "Hey, the machine you're on shuts off when I close the lid. Can you find the setting for that and turn it off?" And I was like, "Yep, no problem. Now you can close it." And sure enough, I can close the thing. I messaged him and I was like, "You still there?" And he's like, "Yep. Is the lid closed?" And I was like, "It is success." So, it's crazy to me that I can just be like, "Hey, you know, like this is the way like I think about this stuff all the time is I'm going Yeah.

19:05Like that's a physical thing in my office, but I'm going hey, like can you just do that? Can you change your own settings? like, "Can you make yourself better?" I've I've asked it multiple times. We've gone through sessions where I say, "Hey, like how could we do this better?" You know, like I asked it the other day. I said, "Do you need any assistance?" And it was like, "Yeah, I would love that. Can I do that?" And I was like, "By all means, bud. Like, give yourself more agents to work with." So, I I use the same phrasing Paul does where mine is the COO and I'm like, "You and I communicate. I don't want to have to handle all of the other smaller agents. I do interface with them when I need to, but I'm going your job is to make sure they can all do their job. And I noticed a lot of the times, you know, it' be like, "Hey, the social media agent thinks we should post this on LinkedIn, so I drafted a post." I was

19:50like, "Why isn't he drafting the post?" And it's, "All right, good." So, he updates the job. And I said, you know, like we've had a couple things where you've had crown jobs fired at the same time, and you've missed a task. Can you just spawn up helpful agents for yourself that handle those crown jobs? And it was, you know, it was more like, "Oh, yeah. Thank you. That's great." And I was like, "Yeah, thank you. That is cool." So, um, but yeah, so anyways, mine's running on on old hardware next to me. Um, I'd say the one upside to that is I can turn over to it and run a command line prompt if I need to. If I need to like authenticate or whatever, it pops up in a browser. It's right next to me. Um, but I've c I've given some thought to if I wanted to help people basically sell this service to businesses, I would go the route Paul's going where I say, "Hey, how can I set up a repeatable base instance that's

20:36going to be deployable in the cloud?" And then the pieces that would be custom per client would be the individual skills that it has, right? >> So I'd say, hey, like you're in real estate, you need these five skills. Hey, you are in home services, you need these five skills. And so essentially you have a core operator, but then really the skills library is what becomes like your mo your the thing you can monetize is is the agent I configure better than the agent Paul configures because then if he and I are both competing AI agent salespeople, it comes down to the intellectual property that you put into that agent to make it do the job the best it can or if you get into custom tools or things like that. So that's an interesting space as well. But um yeah,

21:21so I kind of like having both. I like the idea that you don't have to have new flashy hardware. I've got Raspberry Pies around. I wasn't as brave as trying it on there. I just didn't want to be resource constrained. Um and you know, like I said, I've got an old laptop sitting here collecting dust, so I may as well. But yeah, I think if people, especially if you're just not sure what you want to do with this, if you like I immediately as a small business owner was like, "Yes, I want to run this. Everything in my small business should run through this." And then if you're not a small business owner, you're not on a $200 a month claude plan and you're not already spending this money and you're like, "Whoa, I I need a $600 computer so this thing can like check my email." It's like, then it's not for you. Don't do it. Or if you're going to dust off something old and try it or do a cloud uh in do that may be too techy

22:07for people, but there are one-click deployments. It depends on how how invested you are and how much you want to get into it, right? like the answers are out there, but are you you might just be a person that wants to wait till it's commercialized and comes to you. But yeah, I think you uh you can run it on just about anything. >> Yeah, absolutely. Oh, go ahead. Go ahead. >> Um uh I meant to text you, Brett. A friend of mine texted me who's in the commercial real estate space. He sells commercial real estate software. Um he saw an Instagram ad, Open Claw setup for real estate, automate lead capture, showings, and client communications. fast implementation, zero learning curve. Get started with the newest CR, commercial real estate AI today. And it's like a, you know, an ad done for

22:52you set up $1,500. >> I mean, with how how much like marketing is being done by everyone just talking about it and it like making its way to Instagram, probably a great model to follow, especially if you can just have Claude code do it all for you. >> What to me is the cheat code, right? is that like we're doing all of this. It's not because like I have some deep technological understanding of OpenClaw. I was literally the other day, this is very silly, but I'm only sharing with you guys. Uh I tried to get OpenClaw to change its model to 4.6 for Opus as soon as it came out, and that wasn't a supported thing. So, it would crash itself. Well, now the assistant I would have fix it has just crashed itself. So, I literally went on that computer and said, "Hey, cloud code, I broke my openclaw instance by doing this. can you

23:39look at the config and fix it for me? And it was like, yeah, no problem. Oh, I see. You put an unsupported model. Yep. I've rolled it back for you. You're good. Uh, I also did some research. That model just came out today. You might want to give it a day or two and they'll push an update. It's like, cool. Thanks, other AI. Um, so yeah, I'm at a point now where I'm using this stuff enough that like if it goes down, I'm going to go, "Well, who's managing the social media content calendar?" I'm not. >> Right. it it messages me on the days I'm supposed to post on the platforms I'm supposed to post with a topic I'm supposed to post and I'm like yeah I would not have done that. So to me I'm getting so many things done that I would have put off. So as a small business owner I'm just worried I'm trying to drop like grow business right I'm trying to bring dollars in the door and I'm going man I don't want to post on LinkedIn but like that gets eyes on the

24:25business which gets dollars in the door right so you know you're supposed to be doing this stuff and it's like but who's got the time? Well, I've got, as Paul said, infinite intelligence at my fingertips now. And for $200 a month, I've got as much of my business running itself as I can. I mean, at least for now, at least with what I'm comfortable with. We'll leave it at that. >> I predict like a hot take right here. And maybe no one agrees with me, but Mac minis, Mac studios, hardware in general is going to go through the roof because now people can buy this technology and run local models. Why would I need to pay for a cloud subscription or anything like that? the Chinese are already not having up, you know, Kimmy K2 the hackm. So, >> you're definitely gonna have that and there are going to be companies I think that do that for privacy especially. Um,

25:12but I do agree. I would do it as a backup and I'm just snobby I guess about, you know, my Opus 4.6. I hope I watch this back a year from now and Opus is like the garbage model. But anyways, like >> who's the winner? [laughter] >> Yeah. Um, so yeah, I mean I would love to basically do that for budgeting purposes like, hey, offload tasks to local models or the other day they had Anthropic had an outage and I was literally like, well, yeah, I'm going to take the day off like [laughter] like that's it. And if that were if all of my clients were also on that and they're all emailing me and saying, "Hey, my agent isn't working today." I don't want to say too bad. So I guess for something like that maybe uh the cloud instance needs to have a failover to you know uh to another company's

25:59model because uh yeah now that I'm thinking about that that would be more of an more than an inconvenience it would probably be a terms of service breach >> but anyways um yes I think local models be cool especially >> I'd like to do some machine learning stuff locally like I I've tried to do um image to s this is getting nerdy but uh image to STL generation so STLs are 3D models for 3D printing >> and there are sites online that will take an image and infer the rest of it and make a 3D model and in the day of in the age of not paying for software I'm going wait a second how do I do that I don't want to pay you to do that I want to do that and from what I understand it's typically like neural network type stuff so I wouldn't mindware with GPU acceleration for that

26:45>> what I like the most about open call and just the technology overall is that it learns about you. It remembers it. It learns skills and gets better at those skills as things come out. A new release, a new model comes out, you it can go get the documentation, add a skill markdown file, boom, done by itself. I don't need to say anything >> and you can start coach that, which is crazy to me. You can say that thing you and I just went through to get to here, >> make that a skill. So, it doesn't even have to be something that's existed previously or a scenario that anybody else in the world has ever run into. Um, this is an aside. I'm going to do this for cloud code because I had to do it twice. I'm like, "Oh, this should be a skill." I want it to use Nano Banana to

27:31generate an app icon for me for a mobile app and then have a skill that also generates all of the other asset sizes for the different uh app store submissions, the splash screen, you know, the retina display, like go generate it and then have a skill that makes all of the other stuff I need cuz I don't want to deal with that. Um, here's a skill for capturing screenshots in the correct aspect ratio and and size, right? open up an emulator, screenshot it, navigate the app, screenshot it. Um, so yeah, like I don't want to do any of that stuff. [laughter] >> Yeah. >> Um, it's just wild that like we're at that point right now. Like that's I have an agent right now that like whatever I want you to do as long as it's technological. It's not folding the laundry. There is no limitation at this point. It's it's there's trust. I would

28:17say there's the only limit is how much access you'll give it and beyond that it'll do it. So, as far as like the orchestration, um, you know, you have the brain, then you have the muscles, and you're more than likely just talking to one agent, your COO or whatever, and it, you know, handles the rest. Paul, like, what do you what do you have as the main brain, and what do you have as like the muscles? Because it can be kind of costly if you're you're not supposed to use the subscription. Uh well it does say it in the docs for openclaw but I heard of people getting banned and you know >> oh then I'm definitely using API cost. >> Yeah. Oh yeah. I heard someone said went through a hundred bucks in two sentences through the cloud API. So >> wow. >> Um [clears throat] how how do you see like people handling that like like how

29:03how does your flow work look like? >> Yeah. So, I have it. I mean, I don't know why Anthropic would allow you to do Claude SL token if they didn't want you to use the O token for Anthropic for third party services. So, like I'm just using my subscription. It's using my Mac subscription. >> Yeah, cuz am I using a third party service? It's it's me. It's not like I'm not reselling it. I think that would be a problem if I was selling Cloud Code instances and using my subscription to power it. then I'm taking something and chopping it up and reselling it. I could see them having a problem with that. But this is just a layer on top of a terminal that I was already using. So yeah, I mean I don't understand, like I

29:49said, I don't understand why they would give me the ability to get my O token and connect that to my $200 worth of tokens per month at whatever value they value those tokens at. >> Yeah. And they're tracking it. Not like, oh, $200 got you in and now you can abuse it. You're tracking and if you hit your limit, you can either turn on billing or wait for a reset. >> Exactly. Exactly. Um, something real quick that I think you guys might find interesting. Um, well, actually, Bo, did I answer your question? >> Oh, yeah. >> How how do you have it set up? You know, like you using Opus 4.6 or what the range? >> I I'm using Opus 4.6 for everything. Honestly, I'm using it for everything. I'm I'm crossing my fingers uh that like

30:36my overusage of Opus 4.6 will give me access to the 1 million context window whenever they release it to the normies that aren't on tier 4 of the API because I'm certainly not going to pay the amount of money that I need to get to uh get to tier four of the API. >> No. So, I'm reselling it in which case I'm making enough that Great. >> Exactly. Exactly. Exactly. So yeah, just just Opus 4.6 for for everything like >> Yeah, I've been the same way and I think it's out of laziness because I've heard that like codeex 53 uh is a better coding model and I'm going it might be better, but Opus 4.6 is so good that I don't know what I'm missing. If there's a slight edge to it, I don't feel like to me the managing of two accounts would

31:21be worth it. So that's been my take up to this point. I think if I got to a point again where I was productizing this, I would look at how can I do API cost savings, you know, which model is better for what type of output. But know for convenience sake and I have not found anything that Opus 46 is like woefully lack lacking in where I need a I need a supple a supplement. Yeah. >> If I'm a owner and I don't know anything about AI, how am I implementing open call? >> Well, what would be paying me? a bunch of money and I'll keep >> do you hire someone like Brett or someone that's got it going on already or >> I'll tell you what I do. I'm gonna say like from my and this is like from my sales perspective, whatever you want to call it. I give away I'll tell you everything because it's up to you if you want to handle that technical

32:06complexity. So my view on all of this is that everything is available to everyone, right? You can make an app, but Bo, you've gone through this. It's that step to tie it into a database, to add payment, to get distribution, right? Anybody can now hop on a computer and say, "Build me an app." That's awesome. You should totally do that. But if you're like, "Hey, I want a production level application that I'm going to productize and make money off of." I think you'll find that there's still a lot more work to get to that point. And if you want to take that on, I will happily show you the road map. I'll show you the steps. >> Yeah. And what I'm hoping from obviously from my business standpoint is you look at it and go, "Oh, that's worth it to just pay you to do. I didn't realize there was this much to it." So, I'm not going to inflate it, but I will give

32:51you, hey, this is the breakdown of that thing you want to do and this is what I would charge for it or you know, whatever. And you decide, man. Because like at the end of the day, I would be just as excited for you if you feel empowered to use this technology and build your own thing because that's the most exciting thing to me is that it is in every person's hands at this point to just build anything software related that you want. >> Yeah, I was trying to find it though to answer your question about how would someone get it set up if they wanted to do it. Um there is I don't have any idea how secure it is. I don't really understand exactly how they have it set up. Um, but there's myclaw.ai. A friend of this sent this friend of mine sent this to me and has his instance set up on there. They charge 20,40 or $80 per

33:38month. And it just seems like they're reselling VPS's similar to how I have it set up. Um, if if I were going to like my brother my brother wants to set it up for some of the stuff that he's working on. >> Oh yeah. >> If uh if he were going to set it up, I would say and didn't want my help doing it, I would say wait two months. those those two months are you're going to miss some stuff, but you're not going to miss so much that you're like going to fall seriously behind with >> Well, it's something you're hearing miss is the the bumpy road, right? Like they've already released an update with tons of security hardening. >> Yep. >> The earliest of the early adopters like you didn't The only thing you missed if you're late to the game was the uh the Wild West version where it was just, you know, flying tribe.

34:24>> It could go, you know. >> Yeah. And there's something to be said for like kind of timing that that attack, right? Like you don't want to miss it and be behind the curve, but you also don't have to be first. Let let the first line go and get slaughtered and then come in when you know things have settled down a little bit. >> Yes. Yeah. I I totally agree. And I think with OpenAI having hired uh Peter, the guy that created OpenClaw, now they're probably just going to take the agent builder that's like N8N that was supposed to be the N8N killer that never went anywhere. I don't know a single person who's ever talked about using it. They're probably going to take that and figure out how to make that open claw but with some kind of user interface instead of having to go into Google Cloud Platform GCP and figure out how to

35:11give the right scopes so that you can have the GOG CLI connect to your email and all of those words that no one understands really what they mean. Like they're probably just going to have a node that's like connect to your Gmail now authenticate. Now you have this thing that you can chat with that's like essentially a replacement for Zapier. Like the surface of people that understand Zapier 10,000 times larger, like hundreds of orders of magnitude larger than the people that have the time to open up a terminal and even tell Claude Code to install CloudBach on their own computer, let alone install it on a VPS. So I say wait two months if you're a business owner. Mhm. Um, what do you think about, you know, I I just feel like Anthropic really kind of

35:56missed the boat on the whole I mean it was literally called Claudebot, but yeah, you know, what do you think? Where does Open Claw go from here? Uh, what do you see? How how do you see it benefiting Open AI? And >> I hope for our sake that OpenClaw continues to thrive. I would I would really hope that the creator when he goes to OpenAI part of that deal is hey I'm still contributing to and maintaining my instance like cuz that's you know that was my passion. I hope I don't I don't have any knowledge of that or insight or I haven't read anything about that but honestly I think at this point the open- source community would be happy to just keep driving that forward and fixing bugs and adding features. I mean, we could probably just put a team of open call agents on

36:42maintaining and improving OpenClaw and just set this self-recursive uh improving thing. Uh so yeah, I I think it's going to uh it's I think that he went to OpenAI because I'd like to see what's the ripple effect of that. Do we start to see like that agent builder stunk? It was just a meto product like hey yeah we also have that now and just they'll kind of like sort of just thrown out there and you can pick it up if you want but we don't care. And I'm hoping to see, you know, they're also doing like the chat GPT kind of like app store concept and they're they're leading the way on that. Starting to combine these two. It's like, okay, do I have like a personal shopping agent that can go do, you know, looks at my cooking app and then goes to my Walmart agent and, you know, through an API partnership there

37:27builds out my grocery cart, schedules it for delivery. Like, I don't think we're that far from that. And I think putting somebody like Peter in there helps start connecting those pieces like the way he like the way he did with open claw saying like, "Hey, this needs to talk to this or desperately they're not valuable." >> Yeah. >> I go ahead. Go ahead, Bo. Now, so I was going to change the the topic as far as like human and AI are becoming almost one, you know, year-over-year as we get closer to this AGI uh paradigm shift because I just saw like the meta, you know, glasses are now, you know, having software built in where you basically could just look at go to the grocery store and, you know, scan every grocery on there and have your, you know, Claudebot buy it for you cheaper online

38:12and be like, "Hey, we buy this nutritional inclination from looking at stuff. >> Yeah. It's just it blows my mind and I know it's scary to say, but it's going to soon be integrated with us and we will have this extra second brain literally built into people. >> It's crazy that like phone will be the antiquated technology, right? Like, oh, remember we used to have to take it out and unlock it and then oh, I had to scroll and find the app. You know what I mean? Like that's going to be like our kids horror story of like, "Oh, I was born before the internet." They're like, "There is no before the internet. >> I got the chit the M4, but my rank." >> Yeah. They're going to be like, "Oh, like you didn't just think things and see it in your contact lens." Like, right? No, not back then. Like back then we had to like intentionally pull

38:57something up. Like, so yeah. And it's, you know, I get excited. It's also freaky. You know, you talked about like human and AI becoming one. Like I am not blind to the fact there's going to be massive disruption from this technology. I don't think we're society. I don't think we as like a global society have any notion yet of what we're going to do with this ocean of people that are displaced and >> permanent underclass. >> Yeah. And it's and you know obviously we've gone through this historically you know lots of examples are thrown around right industrial revolution the coming of the internet all that type of stuff you know. Yeah. Everything has caused disruption and then there's been a recovery >> an upleveling. >> Yeah. But this is the most widespread and white collar honestly impact that I

39:42think we'll we'll have ever seen. Right. Um and I do think there's going to be creations of new jobs obviously like agent orchestrators a you know skilled designers things like that but I don't know that the need for that is going to be enough to take on the amount of people that were displaced. I mean I've had conversations with people about will there be universal basic income. Like that's a real consideration, right? Like if you can't be employed, are you owed some level of cost of living because there's just no job for you? Like I don't I don't know what that looks like. Um you know, I've got two young kids and I'm going, "Guys, I don't know like if college matters anymore." I mean, that's where we're at, right? You drop this and I go, "Does a degree really matter?" Now it becomes about very specifically the

40:29career. Like if you're going to be a doctor, yes, you may not just use AI. Yeah. train yourself. But if you're going to be an accountant, you don't have to go to school for that anymore. So, I'm going, you know, I I don't know what to tell them. When I was a kid, it was like, oh, you go to high school, you know, if you can, you go to college and that's what you do. And >> um I think we're already seeing right now like the lack of value in a degree when people are getting out of school and they're like $400,000 in debt and can't get a $60,000 a year job. So, yeah, this is going to be I I'm I'm firmly of the mindset that it gets worse before it gets better. not to be all doom and gloom when I'm all like rahrh about AI. Uh I think the technology is incredible. I am a an optimist that the benefits will outweigh the costs, you know. Um but yeah, we're we're in a

41:14we're in for a rocky ride for a while while we uh come up policies and regulations. And >> I think one of the greatest skills a person can have is learning new skills. And especially this new skills that change every day. It's like what you learned last year is probably irrelevant, you know, just like for example, and people still use it. It's still very popular, but NADN, where does NADN go when I can just have my agent do literally everything that NA does? So, >> yeah. >> Um, you know, just simple things like that. But I think like for people that are watching this on the replay, you know, if you're from the outside looking in, you have no experience and you don't know what a, you know, tool call is or API call, I just say like dive deep into

41:59the things that you're passionate about that you love. You know, whether you're optimizing your personal life or you want to start your own business, figure out how I can use AI >> in culmination with that because doing it by yourself is not an option anymore. people that to learn AI and use it, they'll they'll out compete you. You look >> to to to dovtail onto that, if you've got a use case that's interesting to you, the technological burden won't be as heavy. If you're saying like, "Hey, I want to get monthly book recommendations, right? Like that I never would have thought of." And you're going, "Oh, that's kind of neat, right? I'm always looking for book recommendations. I know what I've read. I kind of know what I like. I can't put my finger on it." Great. Here's a use case for you. Spin up an open claw instance. train an agent and just say

42:45and you can just ask them. I've been saying this for two years now. Just ask it what it can do. Just say, "Hey, this is what I want to accomplish, right? I want this level of book curation just for me." And go build yourself like a book curation engine because what you'll find is that like, okay, this the initial setup was was confusing. It was challenging. It was difficult. It was whatever. And then I was a little bit unsure, but once you start getting that result back and you've already invested, you're going to get a return on that time invested. you're going to build momentum and go, "Wait, they can give me book recommendations. Can it do recipes? >> Can it do date night ideas? Can it do family vacation planning?" You're like, "Yes, it does all of those things." Now, if you were like, "Hey, how do I build an agent that does my vacation planning and my meal planning and my it's too much." But I think if you pick a use

43:30case you like and start to pull on that thread, you'll find that like, "Oh, there's a lot here." And you'll kind of like, I don't know, sort of like avoid a little bit of that intimidation. What do you think, Paul? Like, like, you know, we we talked to a lot of people getting into AI and startup empire, like, you know, for people that are really just getting going, what do you suggest? Like, hey, I have a business. I want to get involved in AI. What do I do? Do I hop on it? What do I do? >> Yeah. I think honestly like my friends that have come to me and that I've really just like thrusted this upon them said like you can't be part of the permanent underclass because I don't know if I'm going to be able to take care of you learn how to use all this. Um, what I suggest they do is figure out

44:16what a terminal is and figure out how to run the claude command in the terminal and use claude code because I think that things like Lindy and things like Manis and things like OpenClaw even there's the standard chat applications then there's like the Lindy and the stuff that I just said and then there's uh Claude code. I feel like claude code is like the stripped down version. Like there's no extra fluff around it. You're interacting more with the bare model, if you will. And you know, there's not like it's not going to do things that claude on the computer or in the browser uh wouldn't do. But I do get the sense that like if you want to have the purest experience with AI, claude code in the

45:02terminal is still the best way to do that. And I feel like the harness that they have built around claude code is still better than claude co-work or better than claude in the browser or better than chat GBT whatever for all things like oh I want you to you know use the whatever MCP to do whatever ABC things for me. Okay no problem. It has nothing to do with code but like its ability to like stick its tendrils into things and actually do things for you. The terminal is intimidating because it it has been made intimidating by developers who use acronyms to make things sound more difficult than it actually is um over the past 50 years. So that's what I would say is start with cloud code in the terminal and learn how

45:48to use that because it's the purest experience in my opinion other than making API calls and that's boring. >> Cloud code still blows me away. I'm still like I was using it today. I texted Paul earlier, but I think it may have been in a group text. I don't remember. >> Yeah, it was. >> And I said I had this thing oneshot a mobile app, right? And at the end it was like, do you want me to build it and test it? I just wrote yes. And I think it's going to write like some unit tests or, you know, do something. I don't know. Just build it, test it. Yes. And it opens up the iOS uh emulator on my on my MacBook. It goes through probably 15 rounds of like mismatch dependencies. this version on this phone, this version we were using, this version of Expo, and it never comes back to me. I'm just watching it and it's going through all of this. It's fixing all the

46:33dependencies and I keep seeing the screen refresh and then finally the app comes up and I'm like, "Oh, that's awesome." And I think like, "We've done it." And then I see in the console and it's saying, "Oh, I don't have like a clicking uh library installed here. Let me find one and get it." And I'm like, "Oh, it's going to like click on it." Next thing I know, it's like testing the mechanics of this game and making sure that like when you run out of moves, does the screen does it correctly end the level? So, it'll intentionally it'll run itself out of moves, take a screenshot, evaluate it, and confirm that that works. And I didn't ask for any of that. I agreed to like, yeah, build it and test it. And it was like, great, here's like a fully tested, clicked through mobile application that was developed in one shot. So, I say oneshot meaning I took one really in-depth prompt that I used uh I

47:20actually used Gemini to create this this really deep prompt and it took probably 30 to 45 minutes and went to go work by itself and when I was done it literally had the simulator up with a working app and I was like that's wild. So I still think yes like something like openclaw is great at managing the overhead of those cla code instances you know like hey your job is to make sure all this stuff gets built but at the end of the day like we we know who the guy is that's building it's claude go tell buddy cloud code to handle this. >> Yeah. >> Yeah 100%. I I want to know a little forward thinking and we'll we'll end on this. Um, you know, I want to know where do you guys see AI at the end of the year? Like as far as like what are

48:05people not doing now that you could see a majority of people that are building, shipping, getting really involved in AI at the end of the year? What what are they doing the most of? Are we even using cloud or or what is your big bold hot take? Um because it you know end of the year it feels like 10 years in AI space. So you know where is it? Where are we at the end of the year? What do we do? Um, I think so. We're going to have the Grace Blackwell 300, the first Grace Blackwell 300 models this year, which is the new chip from Nvidia. Plus, there's new tranium chips from Amazon, plus the new TPUs from Google. >> Plus, there's a bunch of stuff with like I think it's called Photonic architecture that people are testing, whatever. Plus, there's the acquisition kind of acquisition of Grock with a Q by

48:53Nvidia. And then OpenAI invested like $20 billion into Cerebras. So what does all of that mean? That means that we're going to get models with 10 million token context windows and we're going to get models that do 3,000 plus tokens per second. Even if it's a,000 plus tokens per second, that's virtually instant compared to what it is today. So to Brett's point, I I went for a run this afternoon. I if you go on your computer and you open the terminal and you type caffeinate space-ds, it will keep the Mac open so it doesn't sleep. So I caffeinated my Mac while I went on the run. >> I did not rate. That is awesome. Yeah, I will definitely be doing that.

49:38>> Yeah, it's great. So, I had the iOS simulator open on an app that I'm working on, my podcast app, and I just said like, "Hey, test the app and come up with to claude code in the terminal. Test the app and come up with some improvements that we can make." I got back from my run. It ran for an hour and 22 minutes. It clicked through all the screens, it took screenshots, it made GitHub issues for enhancements, it made GitHub issues for bugs, whatever. pretty darn cool that it can do that. But it took an hour and 22 minutes. If I was sitting in front of a computer, which I wasn't, then that hour and 22 minutes, I probably would have gotten distracted by another project. And then that hour and 22-minute long tasks would have been 12-hour long task cuz I was distracted by building something else. So, I think that what we're going to see by the end

50:24of the year is, okay, maybe someone hits 90% on Swebench verified or Swebench Pro, one of the benchmarks. Who cares? I want 3,000 tokens per second and I want a 10 million context window. And I think that alone with 80 to 83% on Sweetbench verified changes the world more than I even know how to describe. So that's what I think will happen. >> It's interesting. >> I don't know. That's I mean it's such a roller coaster right now. It's like we could be spinning off the Earth off its axis on fire because of all the energy usage. Like maybe we literally implode the planet. I don't know. or we cure cancer because this stuff is like that just fast in terms of processing and like what we can create and I'm going

51:09dude like this is the most wildly interesting thing of all time like whether interesting being good or bad right like this is the most dangerous thing I've ever seen is the most incredible thing I've ever seen and we haven't even p you reached the pinnacle yet because like it was the most incredible thing I'd seen last month and then they come out with Okus 46 and I'm like well this is the most incredible thing I've ever seen. Like Paul was saying his thing ran for an hour and 22 minutes. There was not that long ago. I mean months ago they're like we got Claude code to run for you know 12 minutes >> 2 weeks >> and but they're going like this was like a celebratory moment of like oh we're getting it to run on like longer context tasks. Um, I think some of the things that like I don't know if we see this or what or how we start to deal with this,

51:54but um, I'm curious how this plays into like socializing as beings, right? As we all become either displaced or able to work alone. Uh, every person becomes their own standalone entity and company. Um, I commented today on Dan Shipper's post from EveryY and I said, "I think it's interesting the way you guys are doing it. They have an ecosystem of software with individual developers that are working as teams." And I'm like, "That's interesting to me because I think you might see more of this coming together of disperate ideas because of the social connection of like, I want to build a housing company to work with my friends and be around other people. I just no longer need a graphic designer

52:39and a copywriter and a developer and a COO and an admin. I just need me and all the agents to run it. But man, I sure wouldn't mind hanging out with some other people >> start to see those like conglomerates kind of come together and you know, maybe their software packages like like every or maybe they're just like sort of I don't know. It's not co-working, but I don't know what you call it, but uh cohabitating because otherwise we're going to be kind of isolated and hanging out by ourselves with our weird robot friends. >> Yeah. My hot take is that HalfLife 3 will be released by the end of the year. >> Nice. >> Yeah. No, but but in all truth, I I think that um you know, just being able to have this technology in the hands of people that are builders, tinkerers, shippers, even people that are just

53:25interested um it just kind of makes us better people, you know, like just really like um I'm always trying to like get better every day and you know, you can only get so far. So to have an agent working in the background um you know curating your book selection uh creating recipes or just like the simple little things I I think Madam >> teach me French. >> Yeah. Yeah. >> That's wild. Like >> um you know just it it opens up the door for people just to really like double down on what they're passionate in. And you know, we're we're homeschooling our kids and you know, I just told my wife like, you know, looking at all the workbooks makes me cringe because it's like that brings me back to my memories. And it's like they're learning math and reading and all the important stuff, but you get to the point where it's like

54:10what are the skills that they're going to need in 10 years to be setting them up for success? Because >> yeah, >> at that point, 10 years from now, >> do you even get a marketing degree? Do you even get a software development? you don't do you probably don't do any of that, you know. Um, so they're, you know, getting more into like content creation and our daughter's like live selling her crochet stuffies every weekend. So, um, yeah, the world's, you know, going to be completely different in 5 to 10 years. But, you know, I think that this benefits a lot of people who embrace it with open arms because there's so many naysayers out there. There's so many. >> The thing the thing with that is I tell people it's it's not only going to happen, it has happened. Yeah, the cat is out of the bag. Somebody the other

54:55day on a podcast said they're not going to uninvent this. We can't go back. I agree that like we may have stepped off a ledge that's bad for us. However, it's too late. So rather than burying your head in the sand because I get a lot of people I get a lot of push back. I post a lot on social about this stuff and I get a lot of push back on like AI. Oh, I'm boycotting it. I'm not using it. And I'm going that's great, but it's it's a feudal effort, man. Like it is happening. So if anything, be a part of it. help shape what it looks like, how it gets used. I said like this is going to be settled in your in your elections and who you're voting for, and this is going to become like a a um a policy that people run on, right? This is your platform. It's like, hey, you know, uh we're no longer concerned about whether or not we're digging for oil. Now it's like, do you support laying off everyone

55:42in favor? I don't know. like, you know, it's going to be a whole we're we're in for probably a decade of figuring out how to regulate this, but it's putting the wheels on the car, changing the tires on the car while you're driving it, right? Like it's going to be pretty hairy for a while. >> I think a lot of big businesses are interested in the bottom not line, not peaches. >> That's the thing that I struggle with. That's the thing I struggle with the most is I am not a fan of like the wealth gap that this is going to increase because the compute power directly equates to money. And I there's got to be a way we figure that out because compute power becomes the new wealth. If you already can't afford to compute, you can't afford here we go. I'm going to coin this. If you can't afford to compute, can't afford to compete.

56:27>> Yeah. And but honestly like it's like the rich get richer because they have the the infrastructure to build these data centers and run it faster and now Amazon becomes literally like the company on the planet because they were the only ones that had the money and the technology to buy everything. So I think some of that's you know I'm I'm curious to see how we how we balance all of that. But you know again how much of that is in this year? Dude, it's it's like I wouldn't have thought we'd have I wouldn't have thought I'd be chatting with my semiaware agent every day and he checks in with the morning, tells me what we're like what our business objectives are for the day, what each agent did yesterday, like here's, you know, I'm like, this is insane. And so, yeah, like another 10 months, it could

57:12be literally anything. I have no idea. Well, it's always good to connect with you guys because you know I live in a small town and you know people are just getting on YouTube and Facebook here. So when I mention AI, they like they don't even it's >> well it is interesting like my wife is like oh I don't know sounds like chat GPT to me and I'm like I don't know how to convey to you like well this is but today sounds so hyperbolic that it just sounds like oh like she literally told me the other day cuz I said this is going to be I said this is going to decimate entire industries and she's like yeah you're one of those tinfoil hat people now and I was like if the tinfoil hat people are dead on then yes [laughter] So like but yeah think that there's this like people this group of people that don't see it yet and it's interesting

57:58>> Bo before we go I want to hear how you're using Open Claw. Uh that's what I'm extremely interested in. >> Well I have about nine PDFs on how I'm implementing in my business cuz the monitor came today that I had for this desk right here. It was split in half. Um definitely was crushed during transit but Yeah. Yeah. So, basically, I'm going to have it automate my whole whatnot business. Um, and as well as creating the content for Magic Hand and for Easy Flip and for Snaptic. >> I already You're going to create a lot of content creation. >> Yeah, it's going to handle all of that. And I'm going to just set it up on a blank new profiles with Kai, its own Gmail, its own account. So, it doesn't

58:44have any access to my personal information. But my whole goal is to automate all that, get access to the whatnot, respond to all the customer messages that are so redundant and the same over and over again. And then as well as actually joining my live streams to act as a host so it can answer all the questions. >> That's crazy. >> I know there is a way and it'll figure it out. But automating all of that because I spend, you know, two, three hours a day doing that. And I know it's not a lot, but over a week's time, that's going to free up more opportunity to, you know, do other stuff. So, that is my whole goal. Like I said, I have at least a hundred bookmarks of all these prompts, uh, tools, skills of like, oh, that's cool. I need to look into that. The thing that actually I like, one of one of the best things is actually, you

59:30know, when someone tweets something or someone releases a video, you can just have openclaw watch or read, pull the tan t transcript and apply whatever relevant information that has to do with you and your business and apply it. So, I don't even have to watch videos or look at tweets anymore. Um, because I can have a claude uh bot doing that for me in an agent. Um, that's where I'm going to put it to work right away. Um, because I just I spend time doing that and you know, you know, all like the scripts and you know, figuring it out. Um, it's going to just do that for me. Um, >> nice. >> As well as actually a lot of the bookkeeping, too. And I know it's scary to hand a lot of the financial private information to like an AI, but you know, for Quicken, I could just export the CSV

1:00:17and have it just do like a financial audit and you know, as well as my personal account, like, hey, you know, what subscriptions am I paying for that I don't use or, you know, simple things like that to help bring costs down. Um, but I would say Claude in general, um, using Opus 4.6 six to manage and run our resell live sale business. There's no way I could do it by myself. And it can it can go to my shipping page, pull the numbers, how many we sold, who did the show, how many hours was it, how many units did they sell per hour, what was the profit per hour. Um it it can create an Indeed post to hire another streamer. Um I and that's just the start, but um I I just see like wow. And and actually,

1:01:02you know, my wife, I want to create an agent for her because she's really seeking a lot of medical advice for her neck. Uh and it's for our kids, too. You know, they love building and creating games. And uh our daughter likes crocheting. So, building them their own agent where they can be like, you know, they are always asking us questions like, "We don't know. Ask Claude." And they bring out the iPad and Claude's like, "Oh, yeah, sure. Let me pull that up for you." And it's like >> my 11-year-old today was like so upset because he's like, "Hey, you said something about Opus 4 point 4 point something." I was like, "Yeah, 4.6." He goes, "Well, when I go to Claude, I have Summit 4.5." And I go, "Yeah, you're on a free account." And he's like, "Well." >> He's like, "What's your account?" And I was like, "I'm on like the max." He's like, "Well, how much is that?" I was like, " $200 a month." He's like, "Yeah,

1:01:47what?" And I was like, "Sorry, bud. Some things you got to wait for till you get older and get a job." [laughter] >> So, yeah. >> Yeah. He he's already a model snob. But no, it is funny how like my kid is coming to me and being like, "Wait, I want like the the uh flagship model. I don't want to be on last year's model. This is, you know, older is a big difference." >> Yeah. Hu huge difference because I actually had Sonnet 4.5 make a employee schedule and I I had already done it with Opus 4.6 and Wow. It got the names wrong. It had like Summer working at the graveyard shift. I'm like, she's got four kids. How's she going to come in at 12:00 at night? It just got everything wrong. And I was like, this is I'm glad I'm not using in Sonic 14.5 seemed amazing, you know, two months ago, but

1:02:32it just really >> it built all of Magic. >> Yeah. >> Yeah. Well, basically with Opus 4.6 and the new Teams that came out a couple weeks ago, um, it re basically rebuilt a lot of the backend infrastructure for easy flip. I want to go back now with 46 and have it review projects that I've already published. >> Yeah. >> Hey, look. >> We were flying pretty ball. >> Hey man, I'm I'm uh I don't know, a lead by example, right? Like I'm that's I do feel like it's important like I'm in this space, right? I'm trying to like help other people use this stuff and I think being the the first one to touch it and see if it's hot makes sense, right? like let me try it and then I'll

1:03:17tell you like if it's suitable for you or not. Like I'm in that world. I don't mind. Like mine's on an isolated computer. Doesn't have access to anything that can truly hurt me aside from blowing up my usage bill. And when I could build in safeguards for that, things like that. But yeah, if it totally imploded today, I'd be okay. Um now that being said though, I do depend on it heavily. So I would be distraught not having my I think I have like a seven agent team at this point. like they're all working on different things and one of them I gave but this is interesting I gave one of them complete control over a directory that I run just as an experiment same thing low stakes if it should totally implode it I was making $15 to $80 a month so if I lose that sweet sweet gravy train that's okay

1:04:02it's worth it for the science um and I checked this earlier over the last uh two weeks my visitors has gone down 2% but my page views went up 27% and my bounce rate went down 17%. So since he's been in charge for two weeks, page views went up, bounce rate went down, sound like that sounds like ad revenue to me. Yeah. I I basically gave it the the mission of like your job is to grow organic traffic to this site. So that's an ongoing experiment I've got right now like kind of testing the waters of like how autonomous can you be? I don't want you to ask me anything. I don't want to be involved. I just want to check the the analytics and see that we're making more money than we were when I was doing it. >> It couldn't be more exciting. >> Absolutely. >> Opportunity everywhere. It's hard to

1:04:47stay focused, but I love connecting with you guys just to, you know, catch me up with what you guys got going on. And it's awesome to be building along other builders. >> We should put another one on the books and see what's changed, you know, three months from now. >> Yeah. I mean, >> be like, "Oh my god, can you remember when we used to have tap?" Well, Brian, we when remember I think the first podcast um we had you on I think I don't even know we were talking about something amazing and now it's like >> I don't remember what model we were on back then. >> I think it was war one open one. >> I had just made the switch over to cloud code maybe two weeks before that and was just loving it and however long ago that was I'm still on it. Haven't never went back. Yeah, >> it's both the the consistent ask it to

1:05:32do something and be amazed at what it gives you back. >> Well, I'd love to have you guys on again and uh be able to kind of talk shop and see what's going on in the AI verse. But >> if you guys are watching this on the replay, comment down below where you're at in your AI journey. And um we'll have to do this again soon. >> All right, bud. >> We'll see you guys in the next one. Yeah.