Key Players Pivot to Robots, AI Agents Proliferate | Trading the Markets With AI

21:47And then I'm also looking into how to leverage AI for businesses. Right. And even looking into questions like what is the best business strategy? Sorry. There's a strategy to even target within the agentic economy. That's a very interesting question. So there's a lot of problems I've been working on in AI. And I'm happy to show you the workflow and everything like that. And then hopefully I can help some folks out and then show you the Rebel Terminal product if you would like. I'm still doing local development, so it's not on the website yet. So whatever's on the website now is like a bad demo.

22:18Kris Bullock:No worries at all. All right. Well, Isaiah, thanks for giving us that breakdown. Everyone, come on and join us on the Real Vision platform to watch Isaiah kind of walk through what he's been working on. We've got some questions for him, and we will see you over on the platform.

22:40Kris Bullock:All right, everyone, here we are on the Real Vision platform. We're back with Chris Bullock and Isaiah Morales. Isaiah, thanks again for joining us. So I'm going to just stop talking. I'm just going to let you dive into it. I'll let you, I guess, start where you think is best to kind of, you know, show us. And then Chris and I may interrupt you with some questions. Yeah. Yeah. So I think we talked about like the glory over the hill. So this is like the report I've been working on. Sorry. This is a report I published last year, kind of looking at the probabilistic forecast and see if I can just kind of scroll down real quick to show you guys that.

23:17Sorry, a little bit, a little bit of a long report. It's always funny when you write these reports, you're like, it's only going to be 10 pages. And then 30 pages later, you're like, oh, my God, what did I do to myself? So, yeah, so this is like one of the things I've been working on. But a lot so a lot of modeling and just trying to understand, like, you know, am I going to win with this model with this probabilistic forecast or not? And if so, if I do win, how much am I going to win? What's what's the probability? Just trying to quantify the probability uncertainty to get a forecast out and try to quantify that as best as I can.

23:46Because I'm a Bayesian by definition. So Bayesians believe nothing with certainty, just different degrees of belief. And you do Bayesian updating. So in other words, you're constantly, you're never sure of anything. And you're often fine being wrong. Matter of fact, we were wrong on this forecast. It worked out for the one month. Three months did not work out so well for us because of the geopolitical risk, right? Hard to factor that into models because models only know what they know. And geopolitical risk definitely is not one of them most of the time in the data. So as a probabilistic forecaster, It's fun to kind of work on these problems and attack them.

24:18So this was like one of the problems I've been attacking with AI. And I am making like a consumer product. Like this is kind of like, like what you guys saw here is kind of, was kind of like a simple demo. I do have the system in a much more advanced place now. So we can kind of hop into some of the charts if you would like real quick, or we can kind of talk further about the report or like what, what the importance is of like publicistic forecasting versus like, you know, just a number go up or, you know, whatever people do these days. I don't know. Hard to say really.

24:46Bijan Maleki:I think real quick before we get too deep into this, let's talk a little bit about your background. I know you're a developer by trade, right? You've been working, you're a machine language programmer, if I'm not mistaken. Is that correct? Yeah, I'm like an AI guy. I do a lot of ops work now. Okay. Just because that's what our company needs. And a lot of other things too. We're a lot of guys. We're a quick. But that's kind of like the power of AI, I would say. is like it's becoming, you know, what's so funny is like you hear a lot of coders, right? And some coders used to be really gifted. I mean, like developers that make like half a million dollars at Amazon, right?

25:20These guys are like all stars, man. And it's like all that knowledge they have is like, doesn't matter anymore. You know, they even had been at themselves now. So it's like the question is like everyone's asking themselves like on the developer side is like, well, how do I leverage this to expand my capability, right? Kind of like a bicycle for the mind, I guess, is like the way Steve Jobs used to call the Mac. But now it's like, how do we do that for the AI and get a multiple on like what we can do for our business, our workflows, or, you know, even if we're just doing our normal jobs, you know, our salary jobs or whatever we have to do during the daytime or whatever.

25:52Bijan Maleki:I'm always fascinated by this because I always love to see what people like you, like career coders, how they use AI. Like, you know, like you said, just to kind of expand on your capability versus somebody like myself who is not a developer at all. I'm doing much more rudimentary stuff. And so it always fascinates me to see the level of depth that just goes way beyond even anything I can comprehend. So I'm excited to see how you're leveraging AI to improve your own outcomes and things like that. So basically, I'm writing it at the core level. Like most people, they just hop into a quad like here and they start just typing stuff out.

26:26I don't know, like high. They just write their query out and they get an answer back right um i'm actually building on the back end of this so you can look at um like if you're familiar with like lang chain at all um this is actually how the back end this is actually how you design like the ai systems uh the state system itself so i'm actually writing in the native ai framework itself and that's actually what rebel terminal is based off of um so there's machine learning and there's ai they're very very different like friends um so that's that's a little bit of a confusing point for a lot of folks uh but i guess like the core difference is like i'm still doing what y 'all are doing, but I'm doing it in a different way.

27:01Like I'm actually developing the AI tools I need. So I'm developing skills. I'm developing MCP servers. Um, I'm developing agents. Um, developing, uh, my own AI framework, all this stuff. Um, and then just try to figure out how I can actually, um, I guess monetize that or I don't, I don't know what kind of, like you said, like, how do we actually do something with this thing that's actually translates, um, into something actionable.

27:23Bijan Maleki:So what kind of models are you using for this? um that's a good question so the system is an ai system right so the it's funny um i'm gonna show you this on rebel terminal so i do have like a let's see if i can pull this up real quick yeah so this might be a little hopefully this is okay to show uh but yeah so the idea is that i wanted to design a system so this is what i actually wrote in the lane chain right i wanted to design a flow kind of like what we have in chat dbt you know with codex we can just ask any question you know what's the federal what's the probability of like a fed rate cut in six months right what's the probability of btc going up in 12 months um because the reason is like back in the day like this is like when i first started my career you have to design these machine learning algorithms to answer these questions but it takes a lot of time like probabilistic forecasting is not as simple as people think of like oh i'm just gonna pull the data down and then just just do uh you know a train test split or something and then suddenly get a prediction out that makes sense it's like no man you're not even close to that's not a good back test at all um so i think a lot of the traditional like way of thinking in traditional finance is just plain wrong um so a lot of this is like trying to quantify the correct way to do things and then designing an ai system to control machine learning system to actually answer very difficult questions uh even though they may not seem difficult um you know like questions like you know what's the probability of btc being up in 12 months or six months or one month which is like what glory over the hills was trying to answer so you could ask those three questions here so what happens is this system takes that, passes it to a link chain, breaks it up, breaks it up.

28:49It has a router and then it starts designing a pipeline. It actually writes a machine learning pipeline and then optimizes it on a self-improvement loop. Um, and then gets you the probabilistic forecast you need basically has assumptions checked. So kind of like what we're talking about with Microsoft, right? Microsoft is very smart. They're a little, they're a little late to the party, in my opinion. There's another product I have there to show you guys. But hey, you know, better late than never on Microsoft's part. Don't hate them too much, though. They're a Fortune 500 company. It's kind of like a big cruise ship they're in, man.

29:21And I'm in a little speedboat. So I have a big advantage on them. But anyway, so this product is taking advantage of that. I'm able to move fast. It's questioning its own assumptions, right? And then if it doesn't, if the assumptions don't hold by these LLM judges, then it goes back in and then forces the re-optimization of your answer. In other words, the answer you're going to get, we're going to be able to understand whether not just the probability, but like how certain we are that the probability is even right sort of thing. So that's really what I'm working on here. And that's actually what's going to end up powering Rebel Terminal.

29:53Just to kind of step back, Isaiah.

29:56Kris Bullock:So you're not necessarily using like what we like, you know, a vibe coder would use like Anthropic API or ChatGPT. You basically like built your own. Oh, no, I totally vibe code. Actually, here, I can show you right now. This is my VibeCode session going on here. However, I'm designing my own tools. So for example, this is my own tool. You know that product you showed Microsoft had? Yep. That's what this is. Okay, got it. Because like it's like the big problem you notice and it's not a very, it's very obvious if you work with these models is like one model has a pretty good idea, but then it like kind of goes down that idea even if it's on the wrong track a lot of times.

30:35So you can kind of balance that out with those three. So Microsoft had a good idea there and I'm doing that and exploiting the subscription. I don't know if exploiting is the right word, but using the subscriptions I have to get the correct answer. So for rebel terminal, for example, I actually do have an objective function that decides whether something like what's something like, how do we tune the system to even answer those questions? Right. I don't want to be doing that as a person, like looking at all the inputs and outputs. So the system being like, that's good. That's bad. That's good.

31:02I could, but I just don't have enough time to evaluate like a million models or something um so one of the things that's a very hard problem but uh i could use the ai hive mind between chat gbt54 opulus 46 um and then gemini 31 pro preview whatever it was i forget the exact the latest model from them but anyway i can use those three and have it cross communicate and then form a consensus between the three to solve different problems so this is like how my workflow is a little bit different from like a mainstream uh user is one i'm using like uh codex i'm using products like Codex, Claudia Code, the Gemini CLI as well.

31:41But on top of that, I'm just building tools that automatically leverage them. So like, in other words, I can just type out like my normal prompts like this. This is like a prompt I just typed out, matter of fact. Actually, let me see if I can find an example of one for you real fast. It would be something like this. Like implement this, but let's say this was a really hard problem. The system would automatically know to query those three backend systems to find the correct answer for like very hard problems. So that's the difference between me and like some of the other people is like, I'm not just using CloudyCode as it's built.

32:15I'm building on top of CloudyCode to solve hard, hard, very hard problems. To put it lightly. And just kind of building up my own AI toolkit, basically, because there's a lot of good things to that are out there right now but there's not a lot of cool things like I need per se so I'm kind of just building my own tooling as I need it um and how I kind of see the the field and the discipline too because it's like it's kind of like art in a way it's like it's kind of up to you decide like how to attack the problem and it's very open-ended how to do that so that's a little bit of the difference here that and I'm blowing through my quad subscription which is why I just

32:51Kris Bullock:had an API error yeah and probably won't let you as of Saturday quad doesn't let you use your subscription anymore either so that's just great um uh not for open claw yeah so if you're using it i don't use i do you're right right correct yeah so i do so that's the other thing like it's that's that's another part of the workflow of the funnel i'm working on is like well how do you actually manage the business right because i'm just one person um you know like running this whole like website or whatever this whole business this is one of the products or businesses i'm going to work on right but what if i want to do like four or five businesses like in the traditional economy that would never happen not without you having to hire a ton of people right but the world we're into like that don't matter anymore you can start like answering these really hard questions and designing like really cool businesses around them if it's solving a core problem um and you can do that on a very limited budget really you know you just got to have a quad subscription and then if you have something like open claw you know let's say you don't even have to answer support tickets right maybe like you do but like like claws and answer like sorry open claws and answer like probably like 90 of them and then just escalate to the top 10 to you so there's like strategies you can do it's like you're basically designing a semi-autonomous company a lot of times it feels like now it's kind of like what the medic it's like what's the metagame it's like the question like all right you know i'm a gamer i'm like a nerd right it's like so what's the you know i'm old starcraft two player back of it it's like what's the metagame that i have to run is it six row roach rush or what you know it's like starcraft firms right right out here it's like what what is it what's the strategy we're exactly doing um and so anyway this is like a personal passion project of mine, which is like the macro side.

34:23But some of the other business ideas I do have are actually very, very different. So they attack very different parts of the funnel. Like this one in particular is the main competitor probably to what Microsoft would be offering. And it just automatically like leverages your subscriptions to attack those hard problems, like I mentioned. So this is the protocol I developed to do that, but this is a different business. Like I said, like we're entering into, it's not just like one business you might execute on. It might just be a few of them that really kind of interest you. And that's what you're really working on like over time if that makes sense yeah no absolutely so you've actually you've

34:56Kris Bullock:referenced rebel your rebel terminal quite a bit uh can you kind of just let's take a step back tell us what the rebel terminal is and then let's start diving into like what you built out in it and all these different yeah happy to yeah it looks awesome yeah yeah um it's so this is kind of like a preview and stuff like that um it's so this is kind of like the the web like the landing page i mean this is something i haven't really optimized on uh but what i am doing is i'm trying to make both a combination of like different types of dashboards and charts kind of like, but also a much more powerful version of TradingView because I really hate PineScript.

35:30Like I really, really, really hate PineScript. I cannot emphasize how much I hate PineScript. So I refuse to like use TradingView basically other than for charting and it's a bad toy. I mean, it's great for charting. Don't get me, it's a powerful toy, but for like the actual quant type work like i don't want to touch it with a six foot pole so so this product is really meant to address that and give you some of the churning capabilities and so do a lot of the cycle analysis type stuff so like here's a price of btc right and let's let's kind of fit a sign curve to that um but there's kind of like a problem with this right this this is not like detrended at all right so that's like one of the things you have to just know is like a quant or whatever we need to detrim this stuff well no problem we can just do that here click this button uh let me get rid of this first.

36:16And this series has been detrended. So you can kind of see like this price, we're up here, right? We went down, we've crashed and we basically crashed about as low as we're going to go. We could go lower here in 2023, right? According to this detrind method and set up parameters, right? This is very sensitive to what the parameters are. So I should say that that's the problem with models, right? All models lie, but some models are useful, right? So do keep that in mind whenever we're talking about these things. This is all probability. And it was very sensitive to like how you tune your models, I'd say.

36:47So, yeah, so this is why I built this product is I want to be able to look at the data, decompose it very quickly and then just fit different things I want to fit on them, like different types of sine waves, fits to understand what part of the cycle we're on. You can see here, it looks like we're on the downward part of the expansion, even on the detrinted basis. So maybe the correction we have, maybe we just have entry to the bear market, who knows? And this is a cycle fit. So it's really just interesting to kind of look at these sort of things. So this is kind of like the tooling I'm working on.

37:14The big one though, is I've been working on liquidity. So liquidity is a big one for me because that's kind of what leads everything. It's kind of like the life force of the market is like the way I like to think about it. So that's like one of them I've been working on is like, how do we actually get the liquidity? Like what is the liquidity formula, right? So I had like different custom definitions of that here. And that's what I've been working on. So you can kind of like click that and then something like this. And then you have idea like where we are. very very custom because it depends on how you define liquidity right it's like kind of like the wind you can't see it but you kind of feel it you feel the force of it so that's what i'm working on is like uh trying to visualize ways of decomposing these things because a lot of people think visually right we don't think in terms of like like algorithms or like tables of data even i even myself somebody that's a programmer and you know a bayesian person or whatever um still likes to look at charts a lot because charts really tell you the story and just kind to help you digest things, simplify concepts down.

38:14So this is part of what I'm working on here. There's a lot of other cool stuff I'm working on. Like I said, the big one is just the liquidity stuff. And the main formula I do have does show a main contraction in liquidity. This is actually the downward part of the liquidity cycle, matter of fact. And that was due to the same. I had it on the other, sorry, I didn't pull up this exact formulation, but this is one of the formulations you could pull up on the chart. And this is my favorite. So that's why I put it here as a signal card. And you can kind of see like what ended up happening when liquidity, yeah, this is kind of like a preview.

38:46I need to actually go back and verify this in the back end of the system. But anyway, so don't take this data too seriously. But yeah, this is correct. This has definitely been in contraction for quite a bit. If you look at the sign fit, we are in the downward part of the cycle. So I just want to understand like where we are in the cycle, basically, as an investor, right? As part of, you know, I'm like a quant, right? Writing these reports out. It's good to have probability, but it's also going to understand what part of the cycle we're in in terms of a sign like uh you know is it the up part of the wave or the crest or whatever uh because that kind of tells you like how you need to risk manage this out essentially and so that's what i'm working on but like i said this is not my model i designed a lot in the future these are all going to be uh entirely created by that um agentic ai system i'm designing essentially in the back and there's not going to be any like personal opinion about it uh i might publish some of my own research still kind of like this, maybe kind of interpreting it or maybe even making my own models if I get really bored.

39:42But for the most part, I will just kind of use that, what I'm building and kind of eat my own dog food, more or less to use that terminology if you're familiar with that. Still working on the business side of it, but this is kind of what I put together. The really powerful future is, in addition to that, I do have a workspace enabled. So basically like a Visual Studio editor so you can actually run real python code uh a matter of fact uh do same things you would do in pine script the idea is that you'd be able to write like real python code and then actually have that implemented on the chart essentially here uh because the terminal actually what you don't see is the terminal is actually i can't do this because of oh my gosh uh guys don't know how to move this very this is really annoying on zoom my bad um yeah so there actually is a terminal on the bottle of every window so this thing does follow you around so that's actually what i'm not showing here um so the idea i'm a terminal guy very very very terminal uh love terminal uh but in addition to that um i want to set up something to where you can actually have a like a very the way pine script sort of been essentially in my opinion without it being a very bad toy um so that's what i'm working on is allowing writing my own libraries to allow people to do like different things like pine script offers like zscore so it's still easy to use um but you can also use different frameworks you want like pandas or whatever you need to do to do um do your own type of analysis and then apply that over an overlay on say on top of a chart uh like btc and stuff so i'm designing some preset offerings so for power sorry for normal users and then also designing a product for power users um that um that will run and everyone can actually get access to this uh if you're on like the free tier you're probably just going to run on serverless and then And I'm thinking about having a dedicated tier with like provision VMs and stuff like that.

41:29But I'm not too sure if I'm going to do that route or just go to serverless route. So anyway, just like some technical questions. But the idea is that for this front facing product anyway, I do want to have an ability to just basically chart things out and start asking like really hard problems to like probably not to the AI system because that's more of like an enterprise product. But definitely leveraging the outputs of my most advanced models. The users can kind of leverage that and understand where we are in the cycle. um understand how to risk manage uh their well i guess they have to make that personal decision themselves but they should have like insight of like is it risky to buy at this price or not and they can make a judgment of what they want to do from there so not financial advice you know it's you know this is just math and some nerds talking here uh so consider it just entertainment but so this is what i'm working on um and this is but this is like only the front end like i said the real power is in the back end like in terms of like what this would be worth uh it's probably a one out of 10 and the back end system is probably a nine out of 10 because that's where the real power is is actually in the back end of the system because the thing you're you know you have to learn you have to understand about ai is it's only as good as the data you have and that's true of machine learning too the old model used to be garbage in garbage out it's very true for ai too you know even though i designed this like fancy consensus protocol and all that right this model mux thing that's doesn't matter you can have like the most powerful ai model in the world but But if you have bad data and you ask the LLMs like a question of your data, it's like you're going to run into a lot of problems, man.

42:55And even worse is they will self-reinforce your beliefs that are wrong.

43:01Kris Bullock:Yeah, the sycophantic AI. Absolutely. That's it. It's a big problem. So so anyway, so the back end system and the data curation and all that is like some of the stuff I'm kind of like looking at optimizing and I'm doing it all like from an AI driven perspective. So that's kind of like the reason for that is because it's so time intensive to design these pipelines. Like the so the pipelines you saw, you go back to that research article real quick. Yeah. So like the pipelines you see here take a very long time to write by hand. Like they really do, because you have to like you're not just designing one pipeline.

43:34It's a misnomer. If the methodology I did, you're designing. Let's say you have 100 samples you're designing and you have a window size of 10. That means you're designing 90 models. It's 100 minus 10. So that's 90 models. Each time you're doing a you're moving. It's a window. It's anyway, this gets into a lot of math. But basically, it's not as easy as you think. It's not just one model. It's really like 90 models that you're kind of trying to decompose, invalidate your approach with. And then the last model is the one you kind of use. The last couple models are the ones used to understand where we are going into the future.

44:04Kind of like a weather report. But you have a model for like, what's the weather on day one, day two, day three, four, five, six, seven, eight, nine, ten, whatever it is, the 10 day forecast. Um, it's not as simple as people. It's not just one model in my approach. It's like 10 different models, but it's, it's kind of the correct way to do it in terms of time series, uh, forecasting, um, and like backwards validation. Um, so I have a very different approach just because of the discipline. I am a programmer, but I also come from like, uh, the neuroscience background. I was a comp neuro guy. So I did all the math models for how the brain operates and things like that when I worked in a research lab.

44:36And so those same things we used to do on like calcium imaging data, understanding chaotic day medical systems i'm applying over to financial markets but instead of having to spend a lot of time on this stuff tuning it by hand the ai is going to do that for me um and not just one problem because this is one problem type we're solving this is like a problem type known as a classification um or like we're just we're just predicting like a label zero or a one and that's it that's all these models do but it's like 90 of them sounds simple but it's not really simple. But anyway, these things do take a lot of time.

45:09And I just want to design a system to optimize that and then also attack other problems in the same way with the same rigor that I would this type of problem. Because you can ask other problems, like what's give me the, not just what the probability of it being higher is, but what's the actual forecast of the price chart? What's the price target forecast in 12 months? So that's a very different data science problem um a machine learning problem that you would need to answer um and um so like again this is like this is a paper i think that's kind of important because and i bring it up because it's like a lot of people like well obviously use chat tpt for that it's like no man that's not good enough and this paper shows you exactly why excuse me getting over a little bit of cold here from the easter weekend but this paper basically just kind of shows you that like even if you're thinking in terms of probability the problem is that the models are not they don't understand um it's not that they don't necessarily understand probability enough the problem is the user can like cause like delusional spiraling basically as as is in the paper's title being it's kind of like a drunk guy trying to find his keys um but like in a dark space with just a flashlight right like the the lm's uh gonna tell you like look over here but then you have no idea if it's even right and even worse is you're gonna keep looking in that right area for your keys even if it's not over there there so that's the problem is like what using chat gpt is like yeah you can try to find your keys in the dark when you're drunk or whatever but you're probably not going to find them and even worse even even worse is if you do find them because it may not actually be your keys you might only think you're your keys but you found your neighbor's keys or something so that's that's the problem so that's why i'm designing this back in the system because i want to actually find the key uh my keys to my you know my house or whatever when i stumble and drunk at night or whatever to use this terrible metaphor we used to use in graduate school um so that's that's why i'm designing this system is because you can spend a lot of time talking with the ai um even even with these like the three model that microsoft and i were we're wanting very different i have my own framework microsoft has theirs but we have a very similar idea in mind we want to have convergence between the ai hide mind basically even if you do that this is still going to be a problem.

47:20Even if you're trying to reason about it in Bayesian terms, in pure math. So that's, that's already way past what most users even talk to the AI in, right? Nobody's talking to the AI in math really, except people, weird people like me or somebody. So that's, that's problematic. So that's why I'm really designing this backend system. And then Rebel Terminal is really like the, the front end facing product for that, for the consumer. But the backend would be mostly for enterprise hedge funds, family offices, that type of thing. Even better is we can tailor that to their data. So this engine that I developed can be designed on their data.

47:59So this assumption checker, we can design everything for them. That's powerful. It's extremely powerful.

48:06Kris Bullock:That's what the enterprise is. They want customization. Yeah. So if you ask, I don't know, somebody like Raw, I don't know Raw. Your day's my new number, man. You know, my days as a font analyst, too, might be numbered. I think our time might be coming up here. So, Isaiah, I have a little bit of a selfish question for you. As an open claw vibe coder myself, when can I use Model Mux as my Marty's brain? Marty's my open claw. That's a very good question. As soon as I can publish it. So, first off, I'm debugging this still. um so that's the thing as i'm slowly bugging it but it's going to work with open router so if you're familiar with that you're going to have asked any models even better though is if you have subscriptions so for me i have like the cloth subscription i'm paying two hundred dollars a month for the max i have the opening i have the twenty dollar plan and then i have the gemini i think the twenty thirty dollar whatever it is the thirty whatever plan it is um but i noticed like really fast like what chris was saying is like well i copied it here copied it there and it's like it's a pain in the ass man think about it think how many times you're having to copy back and forth Just look at all these files.

49:14Actually, all these little files you see here on the side here in Visual Studio, these were all from just one session where the AIs were talking back and forth. Now, imagine having to get this, copy, paste it, and then wait. It's like you're wasting so much time on that. It's like, what are you doing? So this is why I designed this protocol out. So the way you would want to share these out. So this is actually the AI economy too, right? This is actually the other business I'm working on. This is how we attack the funnel. This is a question. My opinion is like, I don't want to market to humans very much.

49:46Like Rebel Terminal is one thing for like the backend system. That's cool. That's one way to do it is to attack it like a SaaS business. So that's what Rebel Terminal is. It's an attack the funnel, but it's really marketing to people still. I think we're way past that point now. That's not the world we're entering into, my friends. We're entering into the world where it's like, we're entering into the Asian economy, not the human economy, the Asian economy. So that's the thing is you got to start building out these tools, these MCP servers, these skills, you know, the soul, soul.skill, you know,.md, whatever these things are that the AIs use, the actual agents themselves, the fine tuning, that sort of thing.

50:27That's the kind of world we're entering into. But I'm really building out a genetic, I'm starting to think about building out a genetic business. But this is going to be available. I think I will make this just available to the Rebel Terminal members just as like a complimentary thing. They like buy like the pro tier or whatever for like 20, 30 bucks. But I will have it for friends to test out first because I do need to debug this thing. So, yeah, just reach out to me and then we can kind of set something up because I do need to get this tested and debug it. But for a broad rollout, this will probably take a little bit of a minute because I do need a test to make sure it's working how I think it's working.

50:58Right. As my old advisor used to say, this is true of attacking data science and AI and just general things. Never forget you're the easiest person to fool. So, you know, in other words, just make sure you do your homework and, you know, the thing's doing what you think it's doing.

51:09Kris Bullock:Yeah. And you're not fooling yourself of doing that, which is really easy to do. But yeah, so anyway, so this is why I designed this protocol. I got tired of doing this stuff manually to attack these hard problems and leverage the AI and that sort of thing. And even better is it uses, I forgot to mention it, it uses the local inferencing too. So Ollama is the way I run. And Gemma, actually, this is a really cool callout. You see if I can just pull it up real fast. This was actually a really cool callout by somebody in the Discord. so if you're not in the real vision discord you better hop in there uh go hang out with us in the dgen channel absolutely have you there man because this is the kind of stuff we're talking about um and this has been i think i showed this it might have been chris yeah thanks for going to get out man because so this is actually controlling the back end of my optimization for rebel terminal right now because i got tired of paying haku 4.5 and sonnet 4.5 because i'm paying for api pricing that like it's so expensive for that api pricing because the the subsidy on the subscription is like two hundred dollars gives you five thousand dollars in credits think about that uh so if you're paying on the api pricing that's a lot of money and so to kind of get around that i'm inferencing gemma 4 on my m4 mac mini pro um so this is this is open source right this is all open source i installed on olama and that's actually what's running uh empowering the backed into my system now um in addition to opulus 4.6 so i kind of have like a local i kind of have like a hybrid strategy of like cloud and um uh sorry and then local everything as well um so there's a lot of really cool stuff going on in the ecosystem right now there's a lot of really cool tooling but um anyway long story short is uh you should probably just be hanging out with us in the discord or something and there's a lot of other people working on this thing too you know they're answering that question like well how do i do this for my business or whatever and we're kind of really talking about that um there's a cool guy called burger you know so it's just funny names on the discord uh like this guy called burger i'm really cool friends with really cool guy he's using kids like stock uh stock reports and stuff like that um and he has something like a thousand strategies he identified and he's like well how do i like quantify the best strategy and i was like don't worry burger i got something for you matt i got something for it burke so that way mean you don't end up flipping burgers together when this goes sideways the hive mind you are building a hive mind and you also show us how a hive mind works in action so that's really cool yeah so um anyway long story short uh if you're interested just kind of reach out you can reach out to religion they'll know me hang out you can uh message on discord look up some of my articles or just email me um ai for pure at rebel terminal.com um but yeah it'd be cool to get in contact with you if you're interested in getting some of the stuff because i'm still building it out now and prototyping it, but it's always nice to talk to people ahead of time if they're interested.

53:57Especially if you're like a family office or like a hedge fund, because like the data side is something I'd be interested in prototyping the system out for. Because I'm mostly using just open source data right now, which is a lot of Fred data to answer some of these questions. But it'd be very cool to work with like a family office or something like that to fine tune, I don't know about fine tune this thing further, but definitely kind of make the system more robust for the problems they need. Because I can only like, that's the funny thing is like i'm benchmarking it on like 47 different types of questions i can think of in the i can right but who knows if that's what it's going to be people need to actually use the system for sort of thing so anyway it's really cool to always have collaborators that's that's really the power of the discord too is like we're really collaborating on all this stuff like all the members are we're just kind of talking about it how we're applying it to our business or trading strategies and things like that um so it's just been a really cool community man like that's that's kind of the thing is we're just kind of all figuring this out together like uh you know it's like we're vibe coding, we're figuring out like what works, what doesn't like an open, like here's my open claw.

54:52I vibe code. I don't, I don't code who even codes anymore. Like nobody code. There's not like, there's no such thing as vibe coding. It's just coding. It's just coding now. That's what people need on the own. It's like, there is no vibe coding anymore. Like the best developers in the world. That's, that's not my opinion. I'm telling you that like they're, they're not wasting their time. You think the best developers that you, you know, five years ago are still doing the same thing. It's like, hell no. Unless they're trying to like, you know, don't get me wrong. There are people doing this that are like ostriches that put their head in the sand like i ain't doing anything you know using ai screw that man i want to keep my job and then you know yeah there's all trust me i see that all the time and it's like you're just like whatever like your problem not mine and you just easily beat them you know really easily beat them and make them look foolish and they don't hey they want to do it hey i never said i was here to save everybody right this is

55:44Kris Bullock:business after all so it's not personal business yeah that's right thank you so much for uh for for showcasing this i know there's probably a lot more we could have dived into but guys you can you can check them out in the discord he's in the dgen channel uh you know tag him in discord if you have any questions or if you want me to put you in touch just find me and i I will get you in touch with Isaiah as well, or Chris too. So again, thank you everyone for tuning in. This has been another episode of Trading the Markets. We'll be back next Wednesday for another showcase of what AI is doing. Who knows where it'll be next week?

56:26Kris Bullock:So we'll see you guys then. Don't fuck this stuff. Have a great week, everyone. You obviously enjoyed the episode because you're here with me at the end. But listen, don't forget to go to realvision.com forward slash join and grab a free membership. It's an incredible community packed with alpha, great investment ideas, and the research that you need to help you unfuck your future. So get started now. Go to realvision.com forward slash join.

56:53Bijan Maleki:Ever wanted to explore the world of online trading, but haven't dared try? The futures market is more active now than ever, and Plus 500 Futures is the perfect place to start. Plus 500 gives you access to a wide range of instruments, S &P 500, Nasdaq, Bitcoin, gas, and much more. Explore equity indices, energy, metals, forex, crypto, and beyond. With a simple and intuitive platform, you can trade from anywhere, right from your phone. Deposit with a minimum of$100 and experience the fast, accessible futures trading you've been waiting for. See a trading opportunity? You'll be able to trade in just two clicks once your account is open.

57:34Bijan Maleki:Not sure if If you're ready, not a problem. Plus 500 gives you an unlimited risk-free demo account with charts and analytic tools for you to practice on. With over 20 years of experience, Plus 500 is your gateway to the markets. Visit us at plus500.com to learn more. Trading in futures involves the risk of loss and is not suitable for everyone. Not all applicants will qualify. Plus 500. It's trading with a plus.