Fable Is Back: Claude Code, AI Agents & Fitness Tech Livestream

Hello and welcome back to Tobion Fitness Tech. So this time the audio should be quite a bit better. So I I I have my lapel mic on this time. So we shouldn't be getting system audio, so it should sound a lot clearer. Um been doing a lot of updates to everything uh that I've been building and I figured I would go through those today. I also wanted to talk a little bit about the Fable launch. So, there's been a massive set of changes in the AI world since the last time I came on here and discussed anything AI related. Um, some discussion points around things that I've uh been talking to people online about and uh like LLMs versus tools and some people not even knowing the difference between the things. uh the different the different like what is an LLM versus what is a tool that is called and not understanding that difference I think is uh you know dangerous like in terms of uh your capabilities and also my approach to how I learn things I think that might be meaningful and impactful if you're newer to starting in development which this person clearly was and then instead of whenever I pointed out that they didn't know the difference between a a large language model and a tool that is called by the large language model instead of saying hey I I should learn these things. Oh no no no they just said you're playing semantics and you know basically calling me names and stuff like that which uh I was shocked to be honest with you. I mean I guess I'm not that shocked that it was like an online conversation uh specifically in one of the Hermes groups where I'm trying to learn a little bit more about Hermes. And this person responded with, "Here's some large language models that edit videos." And it's like, "They don't edit videos. That's not what these things do." Uh, you could make the case that some of them have the capability to do some of what we're talking about by stitching things together, by stitching uh clips together, right? Some of them can do that. WAN can do that. Some of them, and that's what I thought they were referring to. No, they were referring to the direct editing which is being done by FFmpeg and other tool calls. So, I wanted to go into tool calls versus LLMs and I wanted to go over in general where everything's at and start with Fable. So, let's start with the the the elephant in the room, right? So, we had Fable for a few days, we lost Fable, and now Fable is back. And I have not noticed any degradation in Fable whereas a lot of other people online are reporting massive degradation. So I am using uh and this might be the reason by the way. Uh one thing that there's other content creators that I listen to and this one I can't remember the channel name but he he puts out some pretty good content. He's of East Asian descent. And um I know that doesn't sound like it would matter, but I think it matters in this case because he's reporting that his fable is basically becoming a dumbed down opus and dumber than even the the latest generation opus when he sends it in prompts. And we know that the US government stopped all fable uh fables release and specifically wanted to lock it behind a US citizen gate. and not just US citizen but US citizen of non-foreign descent, right? Which I don't know the specifics of this individual, whether he's a US citizen, anything like that, whether he's even in the US. Um, I know that he's he he specifically can can uh read read and write Mandarin, so like he he he he knows, you know what I mean? Like so like he can speak obviously English perfectly plus, you know, plus Chinese, right? Um, so I'm almost wondering, and he kind of hinted at it in the video, like is he being specifically targeted, and I could see them doing that, believe it or not, like based upon the credit card that you've inserted and the name on the credit card, they can trace the last name on the credit card to possible common Chinese last names, uh, and then, you know, give you degradated results. I'm not saying that's what's happening with him in Fable, but there is also another smoking gun. So in my so inside Fable I am always using Fable from inside of the terminal, right? So I'm always using Claude Code uh command line. And what's by the way, what's this is another thing that I got into a little bit of an argument online about what's confusing about Claude Code is that all of these things have a distinct name. So this is the Claude Code command line agent. Then they have the Claude code desktop agent and they have all of these things named Claude, right? So all of their all of their models are also Claude Opus this Claude, but the company name is Enthropic, right? And they don't name any of their models or any of their developer studios anthropic anything which is just kind of bizarre that the company names all of their stuff like all of their stuff with this Claude nomenclature but Claude isn't really a it's a brand and not really a a individual piece of software or denotes whether it's an LLM or you know like for example this thing is you know the command line app and I've still found the command line app to be really really good. Uh they were trying to use Fable inside of cursor. So it could either be hey they're Asian descent and it's being dumbed down because of that or it could be something to do with the cursor routing of their requests and what they're trying to do because their experience has not been mine. Uh post the return of Fable I can tell you I have had uh the same kind of success that I had with Fable in the first place. So, Fable for sure of any model I've ever seen, this is another argument that I got into that I just thought was asinine. People are claiming that these open- source models are anywhere close to Fable in terms of their capabilities. And I can tell you it's night and day. Like I loved Miniax M3, but if Fable was sticking around in Clawude Code Command line and it was going to and you could use 100% of your usage for Fable, I would be paying the $200 and dropping all other subscriptions and ditch project Arya that I was working on, which has routing to multiple different models. Like all of that would go away and it would just be fable because it's it's been that good in my experience. So this model is gamechangingly good in my opinion. It's the best thing I have ever seen in terms of you prompt it once and then it delivers a full-blown feature set based upon that. Tested with quality, very few mistakes. Um it it's honestly been the most impressive thing I've ever seen. And and before it went away and I actually got a little bit I I didn't make any videos in its absence. um on AI because honestly it was a huge letdown [laughter] you know to have to have a model that powerful and and in my opinion by the way the only person that can be blamed for this as anthropic themselves. A lot of people don't want to point the blame at them. They were saying that Fable and Mythos were so dangerous that no one could have them and then they could only release them to certain people. And this company caused this problem like they definitively caused people to not be able to use their models because they were and and then of course they had you know vulnerability issues in terms of you know prompting it to do things that it wasn't supposed to be able to do and um and what's funny is other models had that same problem and they were able to showcase that to the government but to save face they put even more restrictions on this model which is why I think that person who is of Chinese descent or of Asian descent um is getting [clears throat] targeted with, you know, with with degradation of the model because that is a thing that Anthropic openly admits that they do that they will dumb the model down without even telling you in certain circumstances because they don't want distillation attacks. They don't want uh bad actor attacks. And this person was like, "Hey, I'm using it for like website development, which is something that I do, right? My whole website, Toby Fitness Tack, was built through AI. Obviously, it's a whole bunch of a QA process of me going back and forth, but it was built using my AI system. And uh that individual was, you know, was was um w was, you know, saying, "Hey, I'm just using it for website work, but what if your website uh work was actually just a front to try to distill the model?" Right? So, like there's going to be a whole lot of fuzzy matching on this. And I'm not terribly shocked that that is that is a thing that happened to this user. And my speculation would be it's because of of their their targeting around that. But yes, Fable, in my opinion, is the best model. Uh GBT 5.6 should be coming out soon, and I'm pretty excited to test that versus Fable. That's where I pay for a larger subscription. And you can see already since Fable's return, I have I've gotten to 52% of the usage that I have available for that, which is 50% of your total model usage. If you didn't know, um the way the way they're doing it with the Fable return is we only have I believe actually until next Wednesday. We don't have much time with Fable before it goes only to API pricing. And and I'm on I'm on the pro plan, right? I paid yearly for it. And that's that's a decent value. I would have switched it had Fable been around and there wasn't GPT56 on the corner, which sounds like it's also Mythos class. Um, so it doesn't make sense for me to especially when they're going to remove this. We know that when GPT56 rules out to everybody, they're probably not going to gate it the way that Anthropic gates things. Uh, that so my personal opinion is yes, this is the best model. I'm pretty excited to see GBT56. I'm finally excited about AI again. For a while there, I wasn't very excited because there just wasn't new models coming out and the moon new models. Now, by the way, the future preference, the future precedent on this is bad for our government to be meddling in making decisions around who gets what model. Not in decisions of whether the model is safe to be released. that I can understand. But to make decisions around which companies or individuals can use the model after its release, I think is absolutely a bad precedent um it should be they should not be doing that. They they're they're picking winners, right? This is how we've gotten government monopolies in the past is the government picking winners. that's not the business they should be in in this case uh with specifying who the model can go to. It should in my opinion actually be the way Trump originally said it where it's open to all US citizens. Um you know and if he wants to specify not of foreign descent, I don't think that makes any sense. I think it should be all US citizens, right? If you're a US citizen, you should have access to the latest models. I don't think you should be putting that kind of specifier on it. But, okay, let's just say you decide to put that specifier on it. That's fine. That's, you know, a thing that you can do. It's weird and I don't like it. But these models should at least be open to US citizens and we shouldn't be picking winning companies. That's my personal opinion on this that it's a really bad resident to do that. So, yeah, we have so we have this back. I think it's the best model out there. My personal opinion is everyone should go out and try Fable. I see online where people say local LLMs are just as good as Fable. That's wrong. They're not. Like Minia, like this person was saying Miniax M3 is just as good as Fable. That's not true. Like I'm a huge Miniax M3 fan. You saw in my last video where I was talking about Miniax M3 and how incredible it was and what it's capable of. I still have the Miniax. So the Miniax application, which is actually pretty good. surprisingly good. So, Miniax Code, their application is actually surprisingly good. Uh, if I pull it up here. So, um, you can see I haven't used it as much in a while because I haven't been building quite as much uh, with that system, but yes, this Miniax uh, this Miniax code is actually a pretty good. And then Miniax is also inside my Hermes agents. I was using it but I switched over to GPT for both of my Hermes agents. I'll talk about them in a little bit but to say miniaaxes and they're using the thing is they were also comparing apples to oranges. They were using miniax inside Hermes um and then they were using fable inside cursor one fable inside cursor makes no sense. like that's not the logical choice there. Especially because I think you're losing its recursive loop capabilities that it has inside of cloud code. So the thing you notice inside cloud code is you put a prompt into it and then it will sit there and like it seems like what it's doing is it's actually gathering all of the information it can locally and running a whole bunch of tool calls locally. Then it's going out with all of the information it surfaced back to anthropic using a ton of tokens coming back with with a result and then checking the result again which I think inside cursor the fable model is not going to do as well as inside their own harness. So, I think it was a a very unfair comparison to to go inside of Hermes with Miniax M3 where uh Hermes has a a recursive loop of learning. So, as he's been making requests, Hermes is learning and learning and learning on top of what the LLM itself can do. And so, Hermes is so his Hermes is actually farther along in its development. And cursor doesn't have any I don't believe cursor has any of these capabilities. It was built as a it was built as a coding tool, right? As a as a basic coding tool. So you're looking at two, you know, it's an AB comparison. And the person is also, you know, of Asian descent, which I think may have been and he was also speculating that it may actually be giving him worse results than it should be. But in terms of in terms of capabilities, this Fable, first of all, running Miniax M3 on I don't even think the weights are available yet to run it locally. But even when they do become available, you're going to have to have some serious hardware that even I don't have. And speaking of hardware, like I do have decent hardware. So I have an M3 Ultra with 96 gigabytes. I have uh the GX10 which is actually the the ASOS G GX10 which is actually um which is actually a uh that's a DJX Spark and this is the exact model that that is the DGX Spark. So that's and I finally have that loaded into my Exo configuration. So all three of these uh together are running qu3 coder and it's actually incredibly fast to communicate back and forth with quen coder next inside of this exo configuration. I've been really impressed with it so far, but realize this is not the primary daily driver of anything. It's uh it's, you know, GPT uh 5.5 and and Fable, right? Those are the primary drivers. Um, so that that's where I'm at with that. So, let's actually go over LLMs versus tools, right? So, that was the next thing that I wanted to talk about is what is an LLM's role and what is a tools role because of this argument online that I just thought was absolutely crazy. So a user was saying WAN and um LTX these are these are LLMs that are designed for creating video and images I think mostly video from images but these are specifically LLMs right and they were saying that these things could edit videos no that is a tool call to something called ffmpeg typically I mean there's many different tools that you could use but ffmpeg is the common open source tool. One thing is the the base foundation. We have to start at the base foundation. the base foundation of uh Hermes agent, openclaw, all of these powerful uh you know even codeax, anti-gravity, miniax code, all of these things you know claude code the base foundation of these are actually a bunch of open source tools that have been built throughout the years um alongside proprietary tools right let's not exclude those but open source and proprietary tools that people have built throughout the years to do a bunch of functionality So, the LLM is not editing the video. It's not chopping it up into segments. It's not splicing things into the video, right? It's doing the thought process of where to put the text on the screen. It's doing all of these thought processes. So, the thing that the human would normally do with these tools in the past, the LLM is actually powering those pieces, right? So for example, a good thing to think about is if you go into codecs, right? And you ask it something where it has to go to a website, it will actually go to the website and take pictures of screens and then analyze those pictures to try to figure out where to move the mouse cursor to click, right? And you can see those um visual representations of what it's actually doing behind the scenes. So, what it's doing is it's screen capturing, reading the information off the screen capture, then figuring out where to move the mouse, and then doing that. This is all a whole bunch of tool calls, very complicated tool calls on your local system. So, your local system is hugely important for these AI systems, right? And this is why these harnesses have become so important and so powerful. And also, what's on your system is very important. This person didn't realize that his harness is downloading all of this stuff to his machine to make these videos happen. Right? If you're making a video and you're doing it strictly using Fable or strictly using uh you know codeex or any of these, it is pulling down a whole bunch of things to make that happen. And that was kind of the aha moment with me with Open Claw very early on pointing it way back in the day at one of the earlier opuses. Right. Right. So I had well actually before that sorry when I very first started I started with a Google subscription. So I started with Google pointing it at a because I already had you know everyone has there's even a free tier of Google where you can point that and they've kind of nerfed it to where it's almost unusable. But you could take that free tier and point it at OpenClaw and back then the it wasn't extremely restrictive. you get quite a bit of usage and you could use the Gemini model and it was unbelievable to see it pull down and install all of the things needed to get I forget what I oh I can tell you what I was asking it was the very first version of my fitness tracking system. So, the very first version of my fitness tracking system, it pulled down all of the things it needed. GitHub repos, um, uh, GitHub repos, Garmin's Garmin Connects API, uh, it pulled down all of these things and and, and actually, let's go through them one by one because I think it's important like this user was online talking about how it's all LLMs and it's like, no, no, no, no, no. There are so many other pieces of the puzzle. So what you have in inside of any of these agentic workflows nowadays is you have the combination of the LLM which people think of it as the brain right I do not think of it as a brain right I think of it actually as the heart right because without it blood doesn't flow anywhere so people think brain I think that's wrong right I think that's absolutely I think the LLM is the part of the operation, right? It's the blood flow. It's the I if you have if you have a bad heart, you know what I mean? Then everything is going to fall apart. Um but if you have a functional heart, then everything will go very well. And in fact, the heart analogy is even more important than people realize because think about it this way. If you have a heart that can pump a lot of blood, right? So think about um Fable is a great example of it. Uh chatgbt55, right? These are things that especially on the cloud with high inference, right? With high connectivity, high high speed transfers, right? Uh these things can pump a lot of blood through very quickly and you can get a lot of information and and you can do a lot, right? But if you're saddled with a local model that isn't running like like my local model is where it's distributed across three machines and it's using the GX10 for the prefill and then it's going out to these, right? Like like my local model is going to run very very quickly with this type of setup. It just is right. I mean, this is a this is a very um this is a much smaller model than what I would be capable of running with this setup. However, it's a model that can give me very very fast responses. But if you load this thing up with um let me pick one of the models that I can't even do, right? So, if I tried to load up GLM4.7 or Miniax 25, right? One of these these higherend models. If I tried to load one of these bad boys up inside that, you know, one, they're they're going to fail. It's going to crash, right? I'm going to have to unload a lot of memory to to make that happen. But two, it's going to be very very slow, right? And that's your heart not being able to pump enough through it. So you have this, you have the LLM, which is the heart of the operation, and then you have what I consider to be the brain of the operation. So the brain is actually the harness. So people are wrong. So my opinion is the harness is actually the brain, right? And what and and and also like the the father like I was listening to the father of LLMs in his argument and and this is I think actually the most salient point why LLMs cannot ever be the brain and in fact these companies who are trying to do this are going to fail. So realize they will fail if they believe that the LLM itself is a brain of the operations. And I don't believe anthropic or open AI actually believe that. I think it's just us in our simplicity in our in our not understanding how this technology works as individuals. Us oversimplifying the technology has allowed us to believe that the LLM is a brain and it's just not. and especially because it is the most humanlike uh piece of this, right? The LLM is actually the most humanlike in that it's trained on human data, right? It's uh but it actually has the least amount of animalistic and humanideed humanoid capacities. And that was actually the one of the the founding fathers of LLMs was talking about this directly. And what he was saying specifically is an LLM can never uh can never produce AGI. And his argument is because LLMs do not have goals and LLMs do not have uh what's the other thing? So sorry, goals was the main one. And uh let's just start with goals, right? Because LLM cannot have goals and that is actually fundamentally true. An LLM on its own cannot have next token prediction does not have a goal in mind. Right? So a true humanistic uh long form goal, right? However, a harness can have a goal, can be given a goal, can retain that goal, can have a memory structure. The harness is the brain. So realize that in this world of LLM, the harness is actually the piece of this uh technology that gives it the brain-like function, right? So you have this thing that actually is the brain, right? And I think that that's very very important that we have the LLM, we have the harness, and then we have these tools, right? And what do I mean by tools? Well, the tools are actually like ffmpeg um you know all of these different pieces of software that run on your machine. Those are all tools, right? So, this is software on the machine and these are being called in conjunction by the harness who has access to them and the LLM who is actually deciding what to call, right? and in some cases deciding what to install on the machine to be able to call which which is mind-boggling that it it has this capability right so we have uh tools so those tools are are fundamental and you can think of them as um well they're tools right so you know we have a heart we have a brain right and then the human itself has has tools Right? We have hammers, we have knives, we have matches, right? We have all of these, you know, we have fire, right? That we've created uh through through the use of tools, right? So this this is this is a powerful component and understanding that these are components. We also have these things called APIs or MCPs, right? So we have these different interfaces. So an API is just an interface to talk with data, right? And I I wanted to bring up APIs in particular because another argument that I got into is someone wants an API for their Speediance data. So Speediance is going to launch later this year an API for their data and this person is waiting to build anything on top of their technology on top of you know on top of their systems until this API comes out. APIs in the modern era are dumb. Like I'm just going to say that APIs are dead. Like APIs uh right now we have MCPs. Those are also probably going to die. Like none of this stuff is going to be necessary in the future. All we're going to need to be able to do is use tools to get to endpoints to grab data. Um as long as we have access to the endpoints, we will have access to the data. That's always been my principle. Like APIs are great, right? in terms of like giving you an understanding of how to call data but functionally at the end of the day it's always about calling data and getting data right I don't care whether I have an interface to tell me how to get the data or not as long as I can get the data right so that's where uh with speed in particular a user had reverse engineered um basically the the mobile apps endpoints and have a way of talking to the mobile apps endpoints to be able to get all the data that we needed, right? That's how this is working. And my fear is Speedian is going to shut that down when they build this API and give us less access to our own data. And I'm going to flip out when it happens. I mean, because it will happen. And this user is waiting for this. And he listed off all of these things that you can do with the API. It's about a tenth of what I do today. That's what's mindboggling. So his entire list of what they could possibly deliver with this API, by the way, it might not be that comprehensive. We don't know what they're going to deliver. So he's speculating on what they could deliver and that's very much not the case. And we see that with, you know, I I'm I'm going to be a little bit frustrated with Whoop in particular. Whoop has an API. That API is not comprehensive. So there is a list of of features inside the Woo inside the Woo platform that you cannot access via the API. Now I think I now have ways of accessing that information if I would want to. Um I'll go into detail on those once I actually do it if I decide to do it. But one of the things you can't access is the uh steps data. So it's just not there in the API. And there's no way to access your steps data on your device because it's not in their API and I don't have the ability to call the endpoints directly like I do with speed speeds. Um there is I think there is another workaround that I could use. I'll talk about it briefly, but um they now have a Whoop web app that runs through their Whoop um artificial intelligence system that allows you to upload like it's it's meant for uploading, but you can also use it to ask questions. And I could actually have it so that um my Hermes or OpenClaw or Codeex, right, goes in and asks the question every day of giving me my steps data and and reads the steps data and inserts it into into Arya's system. So realize that like there's all kinds of workarounds to getting at the data that you want. And that's kind of what I wanted to highlight that in a world of AI, getting our data is pro well, it's always been the most important thing. Getting getting access to your data has always been the most important thing for me. Um when when there's specific things that I want to build, but in this case, I I also wear a Garmin and the Garmin API has the steps data. Um so I just, you know what I mean? I I just use the Garmin steps data. The only problem is if I forget to wear the Garmin uh every now and then I'll take it off for some reason like to charge it or um or to lift. So both of those I take the Garmin off and I won't have the steps data and all of a sudden it'll show no steps for the day or uh it doesn't autosync at night sometimes uh like the Whoop does. The Whoop's API constantly has the latest data uh because it's constantly syncing back and forth to the server. it doesn't have the kind of internal memory that the Garmin has. And so, uh, that's a both a pro and a con, right? The pro is the Whoop API always has the latest data. The con is the Garmin is a lot smarter of a device in and of itself. So, if you lose network connectivity for days, you'll still have data. You'll still be getting daily suggested workouts, all that kind of stuff will still flow through because it's all one device. So these are two different types of devices and that's that's another big thing that I'm actually glad I'm talking about because the other big thing is devices right so devices as in the local computer devices as in the the IoT internet of things devices that you have devices are going to be a very very uh you know powerful set of this equation right so you have your LLM you have your harness you have tools that sit kind of alongside the harness like these go these go in here, right? This goes in here and then you have these devices, right? So these devices are uh anything that you're connecting to, right? So devices are kind of like the limbs, right? So if we were making this analogous to a h human um tools are actually tools, [laughter] right? APIs are just uh descriptions of things that to make to make it easier for the brain or easier for the heart to to use, right? So easier for the brain to understand and easier for the heart to actually use. So these APIs would be like, you know what I mean? Be like uh you know things in your body that like like a muscle something like that. But a device is, you know, devices are like your limbs, right? They are they are the they are the uh they are the thing that is going to allow you to to to make this all matter, right, in the end of the day. And that could be your your local physical hardware, right, that it's running on. That could be uh the devices that you're connecting into it to to read data off of. you know, it could be any sort of device, right, that you're using. So, I wanted to go over that very, very briefly. Uh, not that briefly, but I wanted to go over this uh just as my concept around LLMs versus tools. Now, let's talk about specific harnesses, right? So, we have harnesses, right? And I wanted to talk specifically about Google Anti-gravity because of the harnesses, I actually think it is the I'll just be honest with you. I think it's the worst of these harnesses in terms of overall capabilities, but also it's kind of amazing that this comes with decent amounts of usage with their with their pro plan. And I honestly think this pro plan is unbelievable, you know. So if I go where is it? Where did they move this to? So it's going to be models. Yeah. So they put it in models. [laughter] Sorry, trying to find where they put this thing. They made the look of this. These things are constantly changing like look and feel. But this Google AI Pro plan um you can see they say upgrade to to receive higher limits. I found when I was on the Google AI Ultra plan, which I did try at one point in time, the limits were downright terrible. Like, they were not impressive for what it was. So, I was really disappointed of all of these ultra plans, these $200 plans, because I've now had all of them, right? I had the Codeex $200 plan at one point in time. I had uh Google's $200 plan, and of course, I had Anthropics $200 plan. of all of them, their ultra plan was terrible. Um, but I really was sad whenever I got rid of it because I lost the adree YouTube capability and upgrading to Google AI Pro gets you back the adree YouTube capability. So, even outside of you can see I barely used uh my model quota for for uh anti-gravity. I'm I'm not using it in the way that I use like you can see my claw my claude usage or let's go into the codeex usage right so if I go that's a wrong one back settings usage remaining so I'm at 98 but I'm at 84 of the week so I'm I'm I'm not using AI nearly as much as I was in the past in terms of the sheer usage because most of the things that I wanted to build I have built and I'm focusing on my running and my jiu-jitsu training I've I've upped the intensity of my jiu-jitsu training and the amount of jiu-jitsu training. I actually put all this information on the Tobon fitness tech website. So, I forgot that it was all there for a while there, but because Arya is doing all this automatically and keeping track of all this, it's up there. So, you can see my BJJ sessions. It's four BJJ sessions for the week, which is actually five, but one was not tracked, right? So, I'm doing BJJ, you know, nearly all week and it was two last week. I think I actually made it three times, but again, one was not tracked. Um, which meant that I didn't train hard enough to have it track the session. But you can see my my running. I'm finally back up to the amount of running that I would like to do. Um, obviously I'm back lifting again. And that lifting I'll talk about you know briefly in a little bit but backlifting uh using a whole new protocol uh that's built out using AI. So uh now my training is uh being done by the Arya system. So and you can see the miles obviously I'm I'm running a lot more. So uh very pleased with the progress week over week. Right. I'm I'm still under 10 million pounds lifted, but you can see, you know, this is all since I started tracking things. Just uh you know, I I like having this information here. It's probably a little bit more information than anyone would need to know, but it's you know, it's pretty cool to have. But yes, in terms of in terms of uh Google's anti-gravity, in terms of harness itself, it is probably not the best one. But man oh man, is this Google AI Pro uh especially the week before last I I was having some health issues. Uh that's why the numbers were so low for the week before. So I was having a few health issues that I had to take care of. And it looks like I'm back I mean based upon those numbers, I'm back to full health. uh in a way that I haven't felt honestly I haven't felt this way since I was hospitalized. So I made some changes um some medication changes and stuff like that actually taking more heart specific medications and uh all of a sudden I went from feeling the worst I felt post hospitalization to the best and uh I wish I had done it sooner or thought to do it sooner but uh it was something that when I went to my cardiologist and I I'll just give you the story. Um, so I went to my cardiologist. They put me on blood thinners. Obviously, I do jiu-jitsu. Uh, that is very dangerous and honestly I questioned the whole medical system after that because I also had a deviated septum. Um, so I had I had a deviated septum and they put me on blood thinners and I do jiu-jitsu, which meant uh yeah, I was getting nosebleleeds for hours that wouldn't stop and I was uh and to the point where I dialed back their dose of blood thinners that they gave me to half, which basically makes them useless. Uh, go back to the cardiologist and I'm like, "Hey, I dialed this back in half because I couldn't stop these nose bleeds." And they're like, "Well, now it's useless. you might as well not take them. And it's like, you probably should have had the deviated se. By the way, the deviated septum is fixed. So, I finally went in and had my nose fixed. And oh my gosh, if you if you get chronic nose bleeds, go and have your nose fixed. Just do it. Like, it it took maybe an hour. It I mean, it less than an hour, you know, in and out. They they did the whole thing. Um it was like they numb the area. It's basically painless. You can smell the burning. I mean, that's a little weird. You know what I mean? But man, oh man, like like the ability to breathe again perfectly. The ability to not have to worry about my nose bleeding every five seconds when I'm doing jiu-jitsu. Like I should have had this done years ago. Years ago. But um that story aside, yeah, the cardiologist didn't recommend any heart I I specifically asked them, should I be on any heart medications? And their answer was no. Uh but when my health started really like I started actually going back into Aphib at night. Um and I didn't because I had a lot of stress going on. That's why no new, you know, AI videos. Also, that's why the AI quotas are still low. I'm getting back into the training, but not quite back into the AI stuff yet. Uh, not as much of it other than playing around with Fable to build some new stuff. Um, other than, you know, I I spent a lot of time with Google AI Pro watching ad for YouTube videos. I mean, that ad for YouTube is amazing to think that they like with this subscription that's like 10 bucks. Right now I'm paying 10 bucks a month and I'm just gonna switch to the yearly plan which I forget what the numbers are but it's not terrible. I think it it's less than $20 a month um with their yearly plan. I mean this is in my opinion in AI if you can only buy one thing I mean if you can only buy one thing uh even though the harness is the worst of all these harnesses like and you watch YouTube right like it's got to be and you watch YouTube but if you watch YouTube content uh you know having ad for YouTube on top of having a harness that is actually quite good. Like anti-gravity is not terrible. Uh I honestly think this is the value for someone who watches YouTube. If you don't watch YouTube, then Miniax M3 is is is still But after having that week where it was basically downtime and I wasn't doing hardly any AI stuff at all, having that ad for YouTube at this cost plus having anti-gravity, I can honestly remove a subscription at some point in time. And by the way, the one that would get removed, but I paid I I prepaid for a year would be Enthropic subscription. Um, I'm tempted to downgrade my Kodak subscription. Right now, that's still at the $100 a month. I do think I could drop that down to the $20 a month subscription, but I'm hesitant to do it just because both of my Hermes agents are are running on that right now. But I'm I'm tempted to do that also. And I'm I'm still going back and forth debating about doing that for next month. Uh especially because we'll be in full training season. So I'll be doing the most training I ever do. And it it looks like I have pretty much everything down as far as all the additions that I wanted to make to my training system. So that's where we're at with this. And then Hermes agent. I've switched everything over to my Hermes agents off of open claw. So almost I mean the open claw uh instance is still there on this machine but there's a Hermes agent on the M3 which is the machine I'm on right now and there is a Hermes agent on the DJX Spark uh that handles the website stuff and various other things. It handles the website stuff. It handles what used to be project Arya's uh fitness tracking and um my agent stack daily podcast that I do is sitting here on the M3 uh that I by the way the reason why that podcast even exists is because I wanted to get the daily news digested in a way that I I you know what I mean I enjoy and so that is actually something that I listen to in the morning and then publish it out. So, um, and every now and then it does a terrible job making an episode, and that's why there's days where it misses. That's why it's not necessarily every day. Um, because if I don't like the episode, I don't release it. But, uh, yeah, I listen to it and then kind of go over the stories and um, that that's where that comes from. But mo everything's being done by the Hermes agents. It's just the telegram from what's funny is the telegram for uh for my open claw is actually still heavily used uh for these things. So if I go through these telegrams. So if I go through these telegrams, you can see the um the Hermes agent, right? So the Hermes agent was so it it published this one. So the Hermes agent on this machine just published my agent agent uh you know agent stack daily. That's what it was working on. uh Arya. Uh this is actually coming from the DGX Spark. So it's coming from this Hermes agent, but I still have it publish all of my uh information to So this is like the morning report. So if we listen to this >> morning Toby recovery is 76%, HRV 25 milliseconds, resting heart rate 75 after 8 hours 6 minutes at 92% efficiency. Whoop strain today is 2.4 and load ratio is 0.9. Today's step target is 12,600. Yesterday you hit 14,49 steps against a 14,700 target. Trend context recovery 76% versus 71% 7-day average 62% 28-day average 34 points above this day last year. HRV 25 milliseconds two above 28-day baseline. Resting heart rate 75, two below baseline. Sleep 8.1 hours, 1.9 above your 7-day average. This 7-day block, 22,898 lb lifted, 10.2 mi run, 22.3 BJJ strain. Year-over-year, average strain is 9.1, down 11% versus this week last year. Last year, that same window had 30.7 BJJ strain. Recommended training moderate volume cardio guidance green light for hard cardio if you feel ready. Recovery and load ratio both support it. Today's a training day. Your body is ready. Make it count. Bottom line. Train to the recovery signal. Keep the main session focused. Fuel it and protect sleep tonight. >> So after I get done here, I'm going to do this training. So the first thing is my new workout that Arya now builds for me. So this workout is actually built by Arya using my system. You can see this 20 RM. So using the exact cadence that I did before Arya. Um but it's choosing the workout. It's choosing the exercises now. And not only is it choosing uh exercises, it's it's also giving me prep and cool down stuff, stuff that I never did in the past. And oh my gosh, uh some of these cool down exercises are absolutely brutal for someone who has very limited flexibility. Uh I I was I was doing them downstairs yesterday and I had my wife try to do the arm circles workout that it had me do. You're basically just sitting uh Indian style and you're doing arm circles and I can't actually do those because my shoulders are so locked up. So, it's pretty uh it was pretty funny to watch and she was able to do them pretty easily and it's like uh this hurts, you know, but I think it's useful to do these things. I just would never force myself to do them on my own, which is why I gave I handed everything over to Arya. Whenever I came back out of, you know, whenever uh when the health issues ended, I handed back over to Arya all my all my exercises. And I'm using all the distilled knowledge that I had from having a personal trainer in the past to have Arya generate workouts that would be similar to what my personal trainer would have given me. Uh, and it's still a work in progress, but it is definitely way better than what I was doing myself and way more dynamic. Right? These are very dynamic based upon uh how how much capacity I have for training, you know, based on everything that goes into it, right? So, it's basing it ba based on my training yesterday. It's giving me more 20, you know, RM20 exercises based upon, you know, how well recovered I am. Uh, and how hard I pushed yesterday. So, I'm very very impressed with what what's happening here so far. Uh, and then you can just see kind of where I'm at with this thing. Uh, but that's, you know, but yeah, almost everything's moved over to these Hermes agents. And I don't know, by the way, if I'm going to, the reason why I left Open Claw running is because I'm finding the same problems with the Hermes agents, uh, updates breaking things, uh, Open AI not working sometimes inside the Hermes agent, uh, in a way that I didn't have problems with Open Claw other than one update. So, every now and then it just seems to break its connection. um in a way that I wasn't expecting. So, I don't know if I'll stick with Hermes. Uh I was impressed with some of its abilities. Like, I'm very much impressed with its ability to update the website, with its ability to uh handle these adaptive workouts because that was all done on the Hermes agent with a little bit of initial prep work with Fable. So, Fable did the initial prep work and then the Hermes agents been doing all the updates to it. So that gives you an idea of kind of my workflow for anything that's big, massive, and comprehensive, it's going to Fable now. And anything that is incremental or adaptive, those go to the Hermes agents on each machine. So if I want a big massive uh agent stack daily uh update that ends up being a Fable thing, you can see Fable was actually making some changes today. uh I thought it was going to be a big massive change and it turns out it was just a bug that it wasn't showing all the stories that it was actually uh using for me. So it it turns out that it was just straight up a bug in my code uh where I had 14 stories and it was showing me eight or nine of them and uh and so it wasn't something that would have had to go to Fable. It could have been done by the Hermes agent on this machine, but you know that that's something that I didn't know until I I prompted Fable with it. And then I just let Fable finish it and take care of it. But yes, uh, you know, the project Arya project is still alive, but it's split across, you know, two different Hermes agents and, uh, two different Hermes agents using Fable, using Codex, right? All of them are kind of in in concert doing this work together. And uh and in terms of my own personal fitness, uh I think finally, you can't see it anymore, but I had a a a massive black eye also from a a headbutt. [laughter] So, an accidental headbutt in jiu-jitsu. I think it's finally going away, but it it was there for two weeks. So, I got clipped right above the eye and the whole thing turned black and blue. Um but you know, that kind of stuff happens in jiu-jitsu. And uh yeah, so so in terms of my own personal fitness, I'm feeling good. Um you can hear where last year I was a little bit fitter than I am this year at this point in time, but with everything I went through, you know, um being able to to to finally like post hospitalization, I was telling the cardiologist, I don't feel the same. And I told my primary care physician, I don't feel the same as and I can't push as hard as I used to. Like what'll happen is my heart rate will get to 155 and then I'll just quit during runs and every now and then it'll happen during jiu-jitsu too where I will cardio tap and uh it's it's noticeably when my heart rate hits 155. Wearing all these trackers, I can see that. But my my max heart rate is 180. So why am I quitting at a 155? And they all said it was mental. Well, I started taking heart medication specifically and uh all of a sudden I could hit my m max heart rate again. My heart rate is the limiter now on these runs and my whole body would feel like it was shutting down at 155 and that is not the case now. Uh could be placebo. I mean the doctors could be right. Um you know taking and by the way you're wondering well if your general practitioner didn't give it to you, if your cardiologist didn't give it to you, who gave it to you, right? Um, and honestly, I went to Algo RX. Um, not sponsored in any way. Um, but I listed the symptoms and, uh, they sent me the medications, right? So, uh, and and there's other sites that do this, but basically, uh, they have, uh, they I don't want to say they rubber stamp them, but they, uh, as long as there isn't any major contra indications, they will, I mean, if we look at their site, they will they will they have prescription medications. Um, I'll pull it up for you guys just so that you can see it. So again, not sponsored in any way. There's other sites that do this. Uh these these guys might not even be the cheapest one, but um you know, for what I needed. I didn't really care about cost and I didn't really feel like trying to find it on the underground or anything like that, anything crazy. I just wanted to order the thing, have it compounded and shipped. And so uh here's here's the the medications and in particular the the medication I'm talking about is a zettoide and neol. So those are the the medications that I'm talking about in case you're wondering what what I was talking about heart medications. Those are the two medications. And again these are prescription medications. These are not something to take lightly. These are something to talk to your doctor with. I just happen to have two doctors who said it's all in your head. and I haven't felt the same since being hospitalized and uh I reached out to these guys. They approved the medications and I'm kind of trial running them. Um and then I will talk to I will talk to like a physi not my primary care that I have now but I'll probably talk to um I'll probably talk to my uh endocrinologist about it next year. I know that's a long time, but basically I got 150 days worth of medication anyways. So, I will talk to them uh when I next talk to them just because uh they're the only ones that So, when I said it, two of the doctors said, "It's all in your head." Um the endocrinologist ran some additional tests and said, "No, something's wrong, but I have no idea what and it's not my area of expertise. I'll send you to a cardiologist." I'm like, "Been there. They said it was in my head." It's like, "Oh, well then I don't know where to send you, you know. I don't I don't know where to send you." So, uh, this is this is a a massive problem in in our healthcare in in the US, right? Uh, because something like that, it could just be in my head. It could be placebo that these medications are working. But I can tell you, I feel like me again. Like this this, you know, five runs, 10 miles versus 1.3 miles the week before. Um, my my runs are two minutes per mile faster than they've been in the past year. So, and the only time I had cardio like this, I was I I had given up basically doing BJJ altogether to be able to do the running. Like, usually one of these things has to slide. And you can see this week I was able to do all of them, right? Have BJJ sessions, have running sessions, and have I mean, not as much lifting as I'd like. I'd like that to be 50,000 pounds plus. I mean, I can hit way more lifting than I have been doing, but I'm easing into it and trying to make sure that like these Arya lifting workouts are restorative. So, a lot of like even if you look at what it gave me for today, if we go back into the Telegram and you know what it gave me today, right, it it's a lot of 20RM stuff, right? So, a lot of 20 RM lifts, which are kind of recovery lifts, but then it has a couple key lifts that it puts in there. So, I'm really liking these ARA workouts. Uh, we're keeping it minimal for right now so that I can focus on jiu-jitsu and focus on running, but that's what I'll be doing after this. And just wanted to thank everyone for their time. I really appreciate it. uh you know the these these uh and I want to thank all the subscribers. I mean I have more subscribers than I ever expected to have to be honest with you. And I just really appreciate everyone who listens and watches and uh gives me feedback, you know, of things that they'd like to talk like me to talk about. Every now and then I'll get a feedback item and think, "Oh, well that's actually a pretty good topic to talk about." So, if you have anything in particular that you want me to talk about, anything on the speeds, anything on AI, uh any any of my views that you would care about, you know, there's there's a lot of things that I have viewpoints and opinions on. You can tell uh especially around around AI and fitness, right? Those are those tend to be my lane. Um even on the jiu-jitsu front, right? I'm a jiu-jitsu brown belt, but I don't I don't have I have opinions on some things in in the jiu-jitsu space, but I'm not, you know, I am I I've relegated myself to the lane of hobbyist jiu-jitsu practitioner. And I always thought about like why do I do jiu-jitsu? Because, you know, it inevitably comes up because you have these injuries that, you know, people are like, why are you doing this? You know, a good example is uh two of the people that were in the line, you know, I I pick up my kid from school and uh two of the kids two of the people in line were like, "Yeah, I feel the best I've ever felt, you know, like at their ages and they're both in their 40s and I I'm 39 and I'm like, I feel like I'm destroyed and I can barely walk today." It was after a hard jiu-jitsu training session and just everything hurt and and and and also at the time I had a broken finger so that was fun. I had and I was still training. I just tape up the finger and continue to train with it. And they're like, "Why would you do that?" You know what I mean? And uh I don't have a great answer for it to be honest with you, but I can tell you there was a point in time that I was thinking about quitting doing jiu-jitsu altogether. And uh one of my buddies at jiu-jitsu kept reaching out to me over and over again being like, "Why are you not here?" Right? Because I I am one I at the time I was still one of the upper belts, right? I'm a brown belt at the end of my brown belt now. And at the time I was I was just becoming a brown belt and I didn't feel like I deserved the brown belt which was part of the problem. And I was trying to work on getting the leanest and fittest I've ever been. So you know at that point in time I was like uh you know I was not making it a priority. I was trying to that was during that whole weight loss uh journey of going down from 250 the whole way down to 188 pounds, right? That I document on my website and I you know and I kind of document it everywhere. Let me see if I can find the So here here was 230. This is so I realized I was 250 pounds at one point in time. I wouldn't take the picture anytime I went above 230 lbs. Like like this was disgusting. But realize I look more disgusting than this. So like realize that's what I look like at one point in time. And you can actually see um like you can see my weight, you know, my weight over time. It fluctuates greatly. I'm going to have to have this actually updated to to show my weight now because guess what? It's going back down this direction. So I'll have that I I'll actually have that updated just so that people can see it. Um but in these progress pictures, right, you can see Right? You can see kind of where I ended up like February 2026. What's wild is look, this is 230 pounds. And that's what I wanted to show is like these are the same weights, right? But realize this is three years of strength training consistently, right? To go from that to that at the same weight. Um they're just wildly different people, right? And the goal is to get to looking right now. I'm, you know, 217 and I look closer to somewhere between these two pictures, right? A little bit better than this picture. H about like this picture, but with you can see the sides going out. But like if we were critiquing my my current weight, um it's just this, you know, this transformation was massive. And at the time I I really thought about quitting jiu-jitsu because every time I would go into jiu-jitsu, I would get banged up and I would get moved not as far in this progression chain. So, it was really detrimental to my ability to bodybuild at the time. And it's it's really bad in terms of in terms of and you can see this has me at over a million pounds. And it's because this guy or or one Yeah. 1 million pounds. This adds in a bunch of lifts that were done off both devices. So that's why this number is so much different. Um might want to upgrade this so that just so that the numbers match on both places. Um but that's something that's something that I can have have it like have it actually look at and do. But um you know my favorite is my wife took a look over this and I put 6 months no excuses and she's like well you might not have had any excuses but you had a lot of complaints [laughter] like realize to do this like it it was it was hard on the family for me to do this uh for me to do this transformation. This was not easy on the family and was not [laughter] and I was not a nice person while I was trying to lose that much weight that quickly. It was it was it was definitely uh stressful. But I knew I needed to get the weight off so that I could slowly put it back on with more muscle and uh but yeah, I I thought about quitting and my buddy who was there, you know, kept coaxing me back and coaxing me back. Well, I had a tree fall on my property and he was over today uh helping me cut up the tree so that it would be off the sidewalk. Uh so that at least people could, you know, at least people could, you know, come and and it's just the the rel what I'm getting at is the relationships that you make in jiu-jitsu. Um it's one of the few places where you can form relationships with people that are meaningful and impactful as an adult. Uh, of course you could do something like uh, you know, you can you can do that at church. Um, you can do that at a bunch of different places, right? There are, but like um, but it's just one of the few activities you can do like where it is cardiovascular in its nature. Um, and it's not a revolving door of individuals, right? One of the biggest problems is uh you know anytime you've gone into any kind of like other other thing you know a lot of those people come and go very very quickly whereas jiu-jitsu people tend to stay for a long long time right I've been at the same gym for seven years or so now something like that and it's not like people hop jiu-jitsu gyms uh with no meaning or no like like there's a real reason if someone switches which is jiu-jitsu gyms. Whereas anything else in life, you know, if you're doing like yoga or anything like that, you could go to a new yoga facil facility any time, it's a little bit more structured around who your coach is. So, I guess what I'm getting at is it's, you know, for me it's the people, right? It's actually the people of jiu-jitsu that makes me continue to do it. And I had to have a deep thought around it. It's not the training. It's not uh it's not, you know, I enjoy it. I think it it's mentally it's it's a combination of physically stimulating and mentally stimulating. Jiu-Jitsu is uh so it has both aspects of mental and physical. Um and and and the combination of the people, right? I actually enjoy the people that that I meet and they're all focused around becoming stronger, faster, better. Like they're f the one thing at church is you're not going to have like you'll have people focused obviously on becoming more spiritual, right? Which is meaningful and impactful for a lot of people, but you won't get people who are trying to better themselves athletically, right? Th those will be rarer and smaller subsets of that population. whereas everyone who's at jiu-jitsu is trying to better themselves physically. Anyways, that was all I had for today. Uh, little bit of AI, a little bit of jiu-jitsu, a little bit of kind of everything that I cover. Uh, thank you. I really appreciate it. Um, let me know if there's anything else you'd like to hear about. Yeah.