OpenClaw basics, current setup, my primary use case

Yeah. Okay. Perfect. I am live. Hello and welcome to a very different edition of Toby on Fitness Tech. So, let me go back here. Let me see if I can move myself to this. Okay, that moves. I'm still learning how this works. And if I can here, let me move myself down and make myself smaller. Okay. That's a little better. That's what I was looking for. So, hopefully you can see my screen pretty well and that's big enough. But what I wanted to do is I wanted to show off something that I've been working on. I talked about it a little bit in my last video, but I wanted to actually go over what I've been building and why. So, I've seen a lot of people out there making open call videos, but the one criticism that you see all over the internet is that these open call videos are by influencers and people who, you know, that that's kind of their business model is hyping things, right? Let's just be honest. It's it's people in the AI community have made a bunch of videos and then there is a set of influencers right who have also made videos and the use cases that they show are all their influencerbased social mediabased use cases. So that's why I wanted to show mine off because it's it has nothing to do with that and it it shows you kind of the power of this technology outside of that space and and I'll tell you the the basic problem. So, my problem that I've had for years and years and years, and this actually got exponentially worse with the tonal and the speedience whenever I got those devices, and it's both of them, by the way, and the reason why it's both of them is they don't have a web app in any way. So, I used to be able to correlate data through web apps, and Whoop's web app has gotten worse and worse and worse. And then uh and then a bunch of these things don't have web apps at all anymore, namely the expedience and the the tonal. And so I got into this pattern of not being able to actually own the data for these devices. And they do allow exports, but not all of them, right? I don't think Spedience does and I know Tonal didn't because it was a huge criticism. So these these devices are kind of like walled gardens, right? And the problem with it is they're walled gardens and I I wear both a Garmin and a Whoop to bed and then I have an eight sleep, right? So I have three sleep tracking technologies and they don't say the same thing every night. So I wanted the ability to correlate all of them. And then I also wanted the ability to take my strength training load into effect because that's also not being calculated. Although it looks like now that the speeds writes so there's a major update that actually happened recently they they had promised it in the past and it's actually there now for the speedians and it it doesn't sound like a it's a big deal but it's actually a huge deal and I didn't realize it was a huge deal and so I started working out and all of a sudden my whoop strains are more accurate to how hard I worked out and the reason being is because Whoop is actually taking into account the calories burned I think or something like that in in the Apple Health data and modifying the strain score of the workout based upon something that's in Apple Health and I think it's the calories which doesn't sound like a major deal but realize my old manually entered speed workouts counted for no strain and now I'm getting 10 plus strain off of them. And that doesn't sound like it would. It makes a big difference though in your strain score for the day, which is the way that Whoop actually calculates how well you've done progression-wise. And so my values were majorly off up until just recently. And that was one of the things that I was trying to solve by building this. So let me go and show you actually my fitness reports because I think that'll make more sense. Instead of showing you the dashboard first, I'll show the fitness reports. So, if I click here, I go to my fitness report and then I can go to this morning's fitness report. Now, this is going to be terrible because I just got back into sound and we were actually flying in overnight last night. So, that's where you'll see these Whoop total, street sleep, and Garmin total sleep. So, those are 1.5. This I didn't investigate, but this 4.8 8 is probably because it was counting my daughter's sleep cuz she was sleeping on my side of the bed and I didn't get to bed till way after then because I was actually working on getting all of this back up and running because when I went away on vacation I had everything up and running and then it broke while I was vacation and I'll go over all of that in great detail here. So what what I'm looking at is I have all of this data and then I have corlary graphs. I have all kinds of different pieces of information and you'll see like uh eight asleep data is missing, right? It's just broken and that's because I was away and um and then you'll have the lifting volume information and all all kinds of different stuff. I was actually still running while I was on vacation over the weekend. So I was trying to get ready for a race that's coming up. So, I'm training this year. I'm training. Last year, I decided I was going to do it without training. And I was in I thought really really good shape in terms of like not necessarily uh obviously not cardio cuz I found out the hard way that you actually have to train that. But I mean, I was like 200 lb and I was shredded at the time. Now I'm 230 lb and not nearly as shredded. But the difference is I'm actually training this year. I thought I would just be able to go in like I did in my 20s and just run a four mile race with no training with without doing any running whatsoever in the leadup to it and it was an absolute disaster. So this year I'm training up at 230 and then I'll probably lose a little bit of weight going into the run at the end of the month. So the the plan is to to get a little bit leaner going into it, but not significantly, you know, probably 215, something like that. At the minimum, at the minimum weight, I want to stay as heavy as possible this winter. But yeah, and then the the calorie stuff. This is actually something I'll be working on. So I was using Chronometer for all my calorie tracking, but over the weekend I used SnapCow, but the problem with SnapCow is there's no way to get the data out of it that I could find yet. um they have an API. It's not for it's for using their services for your thing and not for reading the data out. But they write to Apple Health and so I'm building a connector that will send the data to Apple Health onto my computer and then go from there. So that's actually a tenative plan for how to get that data from this weekend and correlate it with the chronometer where I I'm doing chronometer again today. And I don't know if I'll do Snap Health, my own app, Chronometer. I don't know which one I'm going to choose in the future. That's something that I'm going to make a decision on. So let's talk about the setup for this. Right? So I had to build this graph and I had to build a set of them as a morning report and a nightly report. The morning report tells me how hard I should train essentially and the nightly report tells me how well I did that day at training. And it sends me a message. It it posts it to this uh well this is the local host version, but it actually posts it out to GitHub for me to actually view all of my reports remotely. So, it uses GitHub pages and it will post it there for me to see these and I guess anyone online can see them. You can see my training status. Not a big deal. I figure if you want to see my training, I mean, go for it, right? That was my kind of thought. I'm pretty open about the way I train and everything like that. So, that's that's going to be there. Um, but yeah, it has all of these connectors. It connects to Speedience. It connects to Chronometer. It connects to Garmin. It connects to Whoop. It it it built all of these connectors for me. That's the biggest thing about open claw. I gave it what I wanted and it gave me the connections to do that. And every morning it runs all of these syncing processes and every night it runs all these thinking process processes to actually make this happen. So, this is actually doing all of this for me. And I figured I would go here. So, that's that's the open call control panel. So, this is where I was I was trying to actually have it make the thumbnail for this, but instead I just went and went into Nano Banana real quick and and made something. But, but anyways, that that's this piece of it. And then the actual reports. So this is the dashboard. So this is something that I had my open call build, right? And everyone's talking about this. All the influ I guess I should say specifically the influencer that caused me to look at openclaw and that's this guy. That's why I had him pulled up whenever you first his name's Alex Finn. Um I mean he's the goat on this. Like he's definitely the the um I would say the person now he he's pretty harsh. realize like he's pretty uh he's pretty interesting in the way he phrases things, right? He's very hype, you know, kind of hype man type individual. But what's funny is he hyped this up so much and I'm like, I'm going to use this thing and think it sucks and is a waste of time. And what's funny is I first loaded it onto an AMDbased Windows PC that I had because that just happened to be the machine that was sitting there kind of not used inside of a virtual machine inside of a Windows machine. And honestly, it was hot garbage. Like I'll just be honest, the setup on it was an absolute nightmare. It literally could do almost nothing. Like I thought it was an abysmal waste of time. But then everyone said, "Oh, you have to load it on a Mac Mini, right?" And you know, there's all this Mac Mini craze. I happen to have two of them, right? So, I happen to have an M1 Mac Mini and an M4 Mac Mini, right? because that's the way I mean that is by the way the cheapest way if you want to be super cheap and get great computers you can buy a new Mac Mini every 3 years and get substantially more compute power for way cheaper than ever buying a single MacBook as long as you can be tethered and guess what I can carry the Mac Mini downstairs if I want to so like I just have a screen upstairs and a screen downstairs and I'll just carry the thing downstairs It's not, you know, I don't need to be untethered moving around with the thing because it it doesn't move very often. And now I have two of them. So, like, it really doesn't have to move. Well, I have three. I'll get into that in a second. But, you know, I had two of them, right? And so, what I ended up doing is I took my M1 Mac Mini and I threw it on there and I first I wiped the drive and then I threw it on there. I'm like, you know what? We'll give it a shot on the hardware that has connectors into the operating and see how painful the setup is and see, you know, how well it works. And it was amazing. It worked right out of the gate. I had everything up and running and I was able to actually use it in almost no time at all. So, kind of shout out Alex Finn for his promotion of using the Mac as the way to go. And I know he says, "Oh, you can put it on a Raspberry Pi and you can put it on like to try to get other people into it, but your experience is going to suck. Buy a Mac Mini. If you can't afford 600 bucks, then you know you're going to miss out on one of the greatest gamechanging technologies I've ever seen. To be honest with you, that that's my thought of this after running it on that Mac Mini. The first time I told it, hey, install this, like it said, oh, I can't do that." And I'm like, figure it out, you know, figure it out. In install, get it done. You know, I want this. It was specifically with the Speedance because I wanted it to go out there. And there was someone who built the unofficial speeds manager and I wanted to go out there, download it, install it, look at how it works, decompile it so that I can add features to it, right? To be able to do what I want to be able to do with these reports. So like this piece of the report. So if I go in here, all of this lifting stuff. So you can see I fully recovered because there was no training over the weekend, right? But all of this data is coming from the speeds and that's not something they support natively. Like they don't support grabbing all the data out of this, showing my workouts here and um and and showing, you know, I I'm building basically what they have on the screen. I'm building my own version of that here, right? And you can see my systemic muscle fatigue is almost nothing because because of the lack of training. Now, that's not true. I actually did hit up the gym while I was on vacation, but I didn't count anything. And I was there with my family and it was one of those if you've ever watched my Lily Lifts videos, those Lily Lift videos are me having fun with my daughters and making exercise fun. And so I took my whole family to the gym and it was probably my favorite experience. Well, one of it wasn't my favorite experience of the vacation, but it was definitely up there. And it was just as simple as my littlest one doing lat pull downs and really actually doing it with pretty good form was seven pounds, which doesn't sound like a lot, but you know, for a kid that young to be able to do that, do it with great form, I was really impressed. I mean, she's five. >> She's not, you know, uh, you know, she's not she's not the age where you normally have them working out. And by the way, during their workouts, I'm watching them, right? Cuz we're in the hotel gym and I'm watching every single thing. We did some medball throws. They actually didn't have What was cool is my wife, my wife and I were doing medball throws. They don't have a medball heavy enough for me and my wife to effectively get a workout with medball throws, which is pretty cool. You know what I mean? like to to for us to both be that strong at this point in time where it's like this is not worth it. And also their dumbbells only went up to 50. So my role is in any gym I go into, I have to lift the heaviest dumbbells. And I thought they might have had hundreds, which is a real problem because I can't lift that yet. Close, but I don't think I could do it. And I especially couldn't do it like, you know, jetlagged and everything like that. and not expecting to actually go and then going there and you know but anyways it was it was it was a fun experience. I would strongly recommend I know this isn't particularly openclaw related but I would strongly recommend because I'm more on the fitness end of things. I would strongly recommend working out with your family, making it a family thing and making it fun. So those aren't your workout sessions. Those are about teaching your family that it can be fun, right? And it doesn't have to be as grueling as what it is whenever I'm up here working out by myself. But on a side note, yes, so I have all the speeds data in here and uh what happened is it went pulled down that unofficial speed manager and then someone in the Reddit post had posted the URL for the ability to pull down your workouts. So your completed workouts, right? they had actually done this, you know, that that piece of it where you can pull down your your uh workouts and that just all I did was sent that into my open call and said, "Hey, check this out. Can you actually pull this data for me?" And it started pulling it and sending me the information via Telegram that came out of that data. I mean, that's how revolutionary this technology is. And the key part of that is that whole time I was using a totally free model. I was using my Mac Mini that I already had that was literally just sitting there collecting dust. So it was an M1 Mac Mini and then I was using from this list I was actually using just the Google APIs. So, I was just using the G Google maybe the CLI and not their API key, but I was using uh Google's version. So, this list, by the way, is super useful. So, this is a a list of free LLM API resources. And these are free. Free as in like free to use up to a limit. Each one of them has their own limit. And I wanted to go over real quickly the ones that So, the one I used to begin with was Google. And then I quickly moved over to Nvidia. So this Nvidia uh with using their Kimmy K2.5 is actually the best of these free ones when it works. So when it works, it's phenomenal. It it's very good at reasoning. The one criticism of Google, and I don't know if this is across the board, but it definitely makes major mistakes. Like I'm talking major major mistakes. like uh one time it deleted itself, right? So that was pretty fun when it just decided it's going to delete its soul.md its heartbeat all of those files that it needs to run. It just was like ah I'm going to run a get clean and like wipe itself out because it and it decided to put a git repo at the root of the project that I didn't ask it to do. So it put it put a git repo at the root of the project and then deleted itself. So, you know, you want to talk about an absolute mess, right? It's But I don't know if that's specific to Google AI Studio. I can tell you Nvidia's also did a pretty I don't know if it was actually running on Nvidia or running on Google at the time, but I was using one of those two over the weekend and it made the decision to set the context window to 4,000 which then bricked itself. So the worst is I went into the weekend with three working bots. I went out of the weekend with only one working bot. And let me go back to my openclaw dashboard to explain that to you. So I have three actual open claw instances each of which are running on their own machine. Okay. Each of which not only have their own name but are actually their own physical entities. So those three are Arya, Claude, and Bob. And they'll eventually have a segregation of duties. So Bob is going to be my assistant. So Bob's Bob's think So Bob has a second brain with a set of reminder system that I'm already building up. That's Bob's second brain. So if I click on that, you'll get to see he has one pending task, right? And that pending task is the uh Alexa media player setup. We still need to do that. That's a me and him together thing where I want to get it to do some home automation stuff and that just requires me to sit down and do some setup with it. So that is its one pending task, but it also has reminders, right? This one is um to sign her I have to sign my daughter up from some stuff. So I took a picture of it and uh you know and I put March 2nd as the date. So that that's my only reminder right now because it's pretty far off and I I know I'll forget. So I wanted to take a picture of it and put it in there and I'll I'll continue to update this with more and more reminders and then I'll build the ability to delete into there whenever I whenever I need it once we hit March 2nd basically is is going to be the first delete of a reminder most likely. But that's that's the second brain for Bob. So Bob will have a second brain. He will also be doing a whole bunch of home automation stuff. He he will have his own kind of like dashboard separate from from the other two. Arya controls this dashboard and also this conbon board. So I have a conbon board of task. I can add my own tasks to this con board conbon board and and put them in the to-do status. And what I figured what I found out very quickly is it it is not trivial for it to pick up this to-do task and start executing on it. The easiest thing to do is to take things from to approve, mark them with the assigne you want and then tell Arya, hey, take a look at these and get started on them. Right? It just seems like it's a lot easier from a build structure. So that's that's the conbon board that I have and these are all works in progress. I mean these are so far behind something like what Alex has created. I mean he spent way more time with the platform. I just started a week ago with this. So realize that I'm I'm I'm a Johnny come lately compared to to someone like him. But I'm also using it for a very different use case, right? So this is all you know this is all my fitness stuff pretty much. Um, you know, Bob's going to have secondary more assistant type of a role. And then if I go back here, Claude is actually going to be kind of a backup to Arya and also its own set of its own set of instrumentation to be able to double check our basically originally claude which is actually the M1 Mac Mini that has the most rigorous setup in terms of its its configuration. So, it has all kinds of different it has from this list, it has everything. So, if I go back to this free LLM list, it has almost every one of these. So, it has I I think it might actually have all of them set up. So, it has all of these free providers set up and it can cycle between them. It can call on multiples. it it's it's the most versatile of all of them in terms of that. However, I did decide because I was getting very frustrated before the weekend because I tried to build apps with it. So, I'm trying to build a couple small apps. One, one's a nutrition tracker and the other one is a jiu-jitsu related app. And I wanted it to build these small apps for me. And honestly, it was failing right and left on these three models. It just wasn't doing a very good job of building, you know what I mean? And also coordinating, right? And even and my local model was failing in all kinds of different ways. And it's because the prompting from what I'm figuring out is it's the prompting from the model you're using and the latency especially on some of these, it would just stop working every now and then. So, what I did is I switched over to the uh Claude Max. You know, don't don't don't ban me. Anthropic, you know, I'm not using the Claude Max plan apparent, you know, the Claude Max $200 plan. I'm definitely not using that, you know, so don't ban me. An enthropic, but yeah, I'm I'm I'm using it and and and I'm using it through the through the API key, by the way, not using the relay. The relay is slow and is as slow as one of these free free free ones if if you by the way decide to go that relay route. There is like a an open AI relay that you can set up on your machine where it bounces them over to it. And I can tell you it seems like it uses slower models, doesn't use Opus 46, and it's just not as sophisticated. I it's it's just you might as well be using these free models if you're going to do that. So I did decide to do that. But the problem with it is you can see this if you see my usage. I'm at 46% usage. And 40% of that was from like one day. So that was from one day just trying to get this this BJJ tracker to this thing that I'm working on. And you can see it's still very early. It's very very early on with what it's doing. You know, it's pretty uh bare bones. you know, it's very very bare bones, but it's got the basic core tenants of of what I'm going to be building. And what I plan on doing, by the way, is building it partially with OpenClaw and actually building it with codecs. Also, so that's one of the things that I found out if you're doing small edits, it is not a good idea to do those in open. If you want to do like a really small like a good example of a small edit is if I I I wanted to change this icon to the correct one because it was wrong and I wanted to change this description. You know, that sounds like it would be super trivial, but for me to do that, especially because I moved off of my right now I am not on my uh Anthropic Max subscription. So, I'm not actually using it right now. Anthropic. So, I'm not using your subscription right now, but I do have it the capability to use it, right? Which is important. And that's actually down here, right? So, I have the capability to use it when I need to, but I do not use it by default. In fact, I'm using Miniax M2.5 because I really wanted to see. So, Alex Finn makes all of these suggestions, right? And I wanted to see which of his suggestions hold water and which ones don't. I can tell you the ones that in my opinion don't are like using a a Windows machine, using a Raspberry Pi. I don't know if he's ever tried to set it up on those, but the setup experience sucked in my opinion and it was not fun. And maybe it's gotten a lot better, by the way. It's been a couple weeks since I tried it. Like that was two weeks ago that I tried it there and then I gave it a couple days and then I retried it on a Mac whenever I you know whenever I'm like okay I watch this another one of this guy's videos and he's hyping it up. So I figured I would watch you know I figured I would just try it on my Mac and I was blown away with what it could do just installing everything it needed to get the unofficial expedience manager up and running once I said hey install it all make it work you know it just went through and this is on free models. This is before hooking it up to Enthropic. Once you hook it up to Enthropic stuff, it can do anything. Unfortunately, you run out of tokens real quick. It can do anything once it's on Opus 4.6. I can see why he uh, you know, rants and raves about how awesome this thing is. On Opus 4.6, it's it's a gamechanging technology. Unfortunately, you know, like I said, the token use is brutal. And even if you take Opus 46 and I actually hand it off to the one that I call Clarity, which is actually a local Olama Quinn3 coder, um that saves on tokens, but it's still not enough. I mean, just doing all of the kind of handshaking with all of the other. So, if if Miniax is replaced with Opus 46, which it will be whenever I'm doing hard work again, like I might kick specific jobs off to Titan, but I might actually just switch Arya to it temporarily if I want to do something really really complicated. That will that that'll be the way I end up going. if I want to do something complicated that's multi-stage that has multiple layers of integration. By the way, Miniax 2.5. So, I paid for that. I paid this is the $10 cheapest plan and that's what Alex suggests. And honestly, I strongly recommend if you're on a budget and you can't afford this max plan, you go out, you get a Mac Mini, you pay for this $10 plan even, and then hook up all of these other free ones. By the way, don't don't skimp on hooking up these free ones, but hook up this so that you have a reliable model as your base model. And then you can go and send off all kinds of tasks to, you know, to Gemini, to Mstrol, to uh to Grock, to Nvidia, you know what I mean? You can send off, you can swarm. I call it the agent swarm. So, what I'll do is I'll have a task and I'll just have all of these agents running in parallel doing different pieces of the task and then and then Opus will put it all together, you know, like a jigsaw puzzle. And it it it has to save on tokens. I mean, I'm sure it saves on a lot of tokens because I'd have burned through all my tokens in under an hour had Opus been doing all of that work, but with what I was trying to have it do at the time. But it is it is not, you know, it's not not using tokens. And that's where something like miniax comes in. I was really worried that I would burn through the tokens on miniax like you do on clopus. You do not, right? It I can look at it right now. I'm at 36% of my usage and it restarts in two point in two hours. So in two hours it'll restart and I I'll be back at zero. And the reason why that's the case is because it's just so much slower. You know, I've never seen something being slower as a benefit in the past, but it is because it's reliable and slow. The problem with Nvidia's Kimmy K2.5 is it is way faster than Miniax when it's working well. And then other times you'll just not get responses. It'll hang and then you'll wait. And I don't know if that's because I'm hitting the rate limits. I don't know. you know, you can't, you know, you can look at the logs, but it doesn't always show you. It just kind of hangs. It's almost like we're not getting responses from Nvidia. With the Google stuff, you definitely hit the rate limits. You're just hitting the rate limits. You can see it in the logs. That that's the case with Google stuff. And and this Grock one, you'll also hit one. I don't think I've hit minstrels, but I very rarely use that one. always use on demand. So, this was actually the the base of my app for a while and so was Gemini. It was actually Gemini 3 I thought that I was using, but this one only has Gemini 2.5 installed now. I had to rebuild the the config all together and that's one of them. I had to rebuild the config on Bob al together when I came back from vacation. I don't know what happened, but it lost its ability to connect to Nvidia. It lost its ability to connect to Google altogether, it was saying that there was no it was rate limited and it it couldn't connect to any of them. And once I, what's funny is once I installed Mini, once I paid for the $10 plan for Miniax, got that loaded on there, I was able to get all of them fixed. So, I was able to get Nvidia fixed. I was able to get you can actually just ask your open call to fix this stuff and it'll normally fix it. This was broken in a way where you actually just had to remove the context limit which unfortunately required direct access to the machine and I didn't have that. So this one So this one I broke over the weekend. This one I broke over with this one worked the whole time. And I think it's just the sheer volume of the amount of providers I had that I didn't have any and I didn't have it trying to do all kinds of optimizations and all kinds of other stuff like I was trying to do with AR at the time. And that's one thing where this is addicting to talk to these things and tinker with them. Like it is definitely a tinkering type of project. It's so cool in what it can do. But yeah, that's that's the state of this right now. I have all so all of these are so these are actually uh registered agents like the agents.list agents now. They used to actually all just be in-memory sessions up until today and we'll see how that goes. It definitely we'll see. It created workspaces for each of them. So now all of them have their own workspaces. They have their own tool. They have their own all of that kind of stuff. They did not have that before today. They were just sessions that were named. So, but that means that that's loaded into the context every time. All of those session names were actually being loaded in. So, all of these names in reference to what they are were being loaded into memory every time instead of now where it unfortunately the way I did it, I still have to load the names into memory because I want the ability to say go use your sub agents and do this task and you delineate the work. I don't want to have to think about which one of these does the task except Titan. It's not allowed to use Titan unless I tell it to. But other than Titan, it can pick any of these other ones and it's actually supposed to bias itself towards clarity because that runs on the machine. So that's how I have it programmed. Like bias yourself towards clarity and pick any of these you need except Titan to run to run whatever work I tell you to do and then go out and do it. So Miniax isn't actually doing all of the work, which is probably why I'm not hitting the token limit at all. Miniax is slow enough and deliberate enough to hand out all the work that it needs to hand out and then just get the results back. But yes, so I've been working on this. I have six different connectors that come in from from different fitness places, different fitness, you know, apps. And I'm going to continue to iterate on it. You can see that it needs a lot of work. I mean, it's still a work in progress. The conbon board, the dashboard, there's going to be a whole bunch of home automation that I'm going to do in there. But it's it's an exciting piece of technology. I think that'll be all for me today. If you have any questions for the next live stream, shoot them down below and I'll try to actually just answer them on the live stream. Going to try to do another live stream on Thursday at 6:30. So, this one was a little bit early, but the next one I'm going to do Thursday at 6:30. And this was kind of a test live stream. wasn't actually I mean it's live but I I I uh just wanted to see if I could get the technology working if I could get the camera working if I could get it so that you could see my screen. So I had a whole lot of questions about can I get this working and will it actually work and it it does look like it's all working. I put, you know, I put about a week's worth of work into this open call thing and I can tell you I I was shocked that, you know, Alex is hyping of this. It's it's real. I mean, it is an amazing technology. I haven't been this excited for something in a long long time. And I know people say, "What about the security?" But you know when the last time I heard what about the security was? It was the internet. Like I was around for kind of the internet revolution and people saying, "Oh, it's not secure. You have to use fax. You have to use this." Only idiots use fax. And by the way, I'm not going to say the largest organization that uses faxes, but they're idiots. You know, nobody uses that. It's horrible. It's a horrible technology in comparison to the internet. And yeah, obviously there were a lot of security problems. Obviously, there's still a lot of security problems. The internet wasn't built with security in the forefront, you know, regardless of what everyone says. I mean, it was literally built so that everyone could see everything and and edit it because it was built for universities at first. And, you know, the last time we had a technology this revolutionary, I think, was the internet. Realize I think Open Claw is as revolutionary as the internet was. People compare it to the iPhone. I think it's way more revolutionary than the iPhone was. Way more. The iPhone was just an amazing piece of technology. This is going to be a life-changing worldchanging piece of technology. Alex Finn is right. He is absolutely right. Now, whether it's going to be the survivor in the end, especially with OpenAI buying them, I have no idea. I have no idea what the state of this technology is going to be in the next six months. I don't even really want to speculate on it. I just I just don't know. I can tell you the visionary from the original creator. He was just on Lex Freriedman's podcast and it's probably one of the greatest Lex Freed me. It's three hours but I was watching it. I was listening to it on the plane back. I didn't sleep at all because I was actually enthralled by the conversation. Lex Lex is probably just as monotone as I am. So hopefully there's a market out there for monotone individuals because uh you know or maybe Lex has it cornered and there's no uh no options for a secondary monotone individual like like that. But I can tell you that interview is amazing to hear someone's personal beliefs on agentic coding. I know Alex Finn was into the vibe coding scene. I was never into that. What's funny is now I'm using, you know, I'm using codeex, which is like the vibe coding tool because I don't want to have to edit HTML ever again. To be perfectly honest with you, that's that's what I'm using it to do is edit the HTML. I don't ever want to have to single line of HTML in my life again. I'm just going to be honest with you. I started editing these things by hand when when it was actually the fitness reports. So before I put codeex on the machine, I started editing them by hand because I realized it was faster than telling OpenClaw to do it is to just go into the Python files, change the HTML I want and fix it that way. It's just faster than to try to have OpenClaw do minor small edits. not what its specialty is. But oh my gosh, once you get something like Codeex, you cannot even remember the file name that you're trying to edit. You say, "Hey, I have this piece of text. I want to replace it with this piece of text. This I want to like a good example of what I I had it do today." That would have taken me quite a while just because I would have had to find the file is if I go back and you see how this is centered now. This used to sit over here and not be centered. So, it was left aligned. And I didn't remember what file built the dashboard. I just said, you know, I rightclicked, inspected element, I grabbed this span, I pasted it into codeex, and I said, "Hey, I want this to be centered." It found the file, replaced it, and fixed it for me. And boom. I'm like, "Oh my gosh, I can't believe I never have to do this again. I can't believe I never have to build vertically centering logic again in CSS to not have to do some of these mundane trivial and to rely on these tools to do it for you. I'm never going to say the word vibe coding as a thing. I think it's stupid. I think it's stupid. I'm not going to be like the vibe coding, you know, whatever. I I just I just don't like the name, I guess, which is kind of trivial to not like a name, but I just think it it under software developers do. And it under means what even we're doing in today's day and age, which is trying to build better tools. And that that was one of the things that really came out in the Lex Freriedman podcast. There's a lot of criticisms that Open Claw isn't anything special. Open Claw isn't anything new. And it's like, yeah, that's it's true. I mean, all of the the pieces that made Open Claw already existed. It's just putting it all together in a package that's usable and transformable. A good example is this Miniax 2.5. This is what makes Open Clone more powerful than anything I've ever seen. So, Miniax I installed on Bob and I got all of these other well technically I got all of these other free providers working then. So, I installed Miniax. I paid the 10 bucks and I got all the other ones working right. I didn't want to use I could have put the Enthropic subscription on there and Enthropic do it but I would have been burning even more of those tokens and I just didn't want to do that. And I figured I would I would try out Miniax and see if it could do it. And the answer is it definitely can. That $10 plan is actually I think the best deal in Open Claw. Like it's if you want a place to start, that is the way to go. And and I still think you go connect to all these free connectors and you just have form of of power to be able to really power any change you want to make and and pass it through multiple models have check but that so that that's kind of the general gist of it. But yeah, I put that on there. It went out and it it fixed up all of its all of its, you know, all of its other connectors. And then Arya, I wanted to get off of using Opus 46 and I went to put Minia on it and failed. So Arya has a different configuration, a whole different setup than Bob does. So they have very different open call setups and they're they're spawning sessions in different ways and they have all kinds of different like they're just very different. And so when Miniax went to go in there, it failed. And the worst is Arya said, "Hey, I can't do that. Like there's a bug in open claw and I can't actually do it." And I'm like, "Okay, fix the bug." And so what it actually did is it went out first. First it created this bug report and after it created the bug report I'm like no no no no fix it and then it actually went in and fixed it you know so but yes it it actually went in wrote the code that it needed and believe responds now. So, and now I have now I have that running on Arya. So, it actually runs. So, Miniax is now running on Arya and it wouldn't, which is pretty cool. Like it I configured it all. I thought the config was right. And by the way, Opus was actually one the one who did all this. You wonder like what what like I didn't I didn't write this PR, right? Opus wrote this PR. Opus went in and fixed you can see like this is this is fix add minimax to the provider compatibility and set set the supports developer rule to false. You know I mean that's pretty incredible that it was actually able to go over and do that. But yes, that so that's what it actually ended up doing. It created that and also created the root cause analysis and actually fixed it for me. Like it literally has a local patch. Now the next time I update it will break miniax compatibility assuming this is not fixed. So it warned me of that already. But that is pretty amazing that you can actually fix the core software with Opus. I mean that's the power. I mean to give you an idea like I don't think any of the other models would have been able to create that patch properly and execute it and have working Minia. So there's still times whenever I'm going to be switching to it for sure and realize that these influencers that are hyping using you know cheaper models are using Opus forces like that that is what they're actually using. So that's my two cents on this. I do think this is an amazing technology. I think everyone should check it out. I'll be back on Thursday at 6:30. I'm going to get another live stream going. I'll probably go for 30 minutes. If anyone has any questions, hop on it and you can ask your questions there. My game plan is to try to do these every Tuesday and Thursday. and I will come with a whole bunch of updates on what I'm doing, how things are going, how the training's going, the fitness reports are are looking, and how much progress I've made on the dashboard, on the conbon board, the apps that I'm building with it. Obviously, the BJJ1 needs a lot of work. I think this is a pretty amazing piece of technology and I'm really really excited to be working with it. That's the conclusion for today streaming.