So, I wanted to make this video for a while now. I just call it a genetic AI trading for beginners because I get a lot of questions, emails, and stuff about how to kind of get into starting to learn about playing around with this. And I think using a genetic AI and crossing it over with trading like for example on the stock market, crypto exchanges, prediction markets, and stuff like that is a really good way to learn both about a genetic AI and some finance if you're interested in that as I am personally. So, today I thought I'd just go through some of the questions I get, some of the most common questions I said like that gets asked to me. So, basically what I want to cover is how to get started. I had done this before in previous videos, so I'm not going to go specifically into that. But the two platforms I use the most for now is like Hyperliquid and Polymarket. The reason is because using this on fiat currency is a bit more Yeah, it's it's not that easy to get going. But I've seen like Robinhood, I guess I can pull it up here, just launched this. So, basically today when I wanted to make this video or something. So, they are also getting into this. So, I think it's perfect timing for me and it's good timing for you to get into learning more about this. I'm not going to cover their setup here, but there is an option if you don't want to go into kind of cryptocurrency and stuff like that and stable coins. But I think that is very flexible if you want to get started because you don't really need that yeah, KYI really hard setup if you just want to try it. So, I would recommend like Hyperliquid or Polymarket. I will leave some links down in the description if you want to support the channel. So, if you want to sign up, you can follow those, but you don't have to do that. You can just skip that if you want to. So, I have also done other videos on the channel where you kind of set up your account. Uh so, I'm not going to cover that in this video, but if you want to see that, you can go back in my previous videos to see how you kind of set up your account on both Polymarket and on Hyperliquid, I think. So, go do that. So, the next topic I want to cover is um data. This is why it's really important because we need our AI agent, our agentic AI, to have data. Because without data, we don't can't really make any models. We want to follow, right? So, that is really important. So, I'm going to look at kind of how you can collect data yourself using APIs. Some places, like I think Hyperliquid, can fetch some data. Uh so, you can backtrack some to look at different models you want to run. And of course, you can also look into more traditional data sets, right? Uh if you want to do that uh sets, I guess. Uh like on Kaggle or something, or maybe look at uh some stock um APIs and stuff like that. Think you can find some data on there. But, I will mostly cover the the platforms I'm using. For research, another very important point, of course, because if you want to get into agentic trading, you need to have some kind of model. I think we're going to do more of like a hybrid version. So, we're going to use, let's say, Codex 5.5, Claude Opus 4.8 to look at the data we have collected, and we're going to create a model based on a goal we have, right? So, let's say we have a goal of making like yeah, I don't know, $10, something like that. Uh that goal or that model has to be different if we want to make, let's say, $100 uh based on our balance, of course. We want the model to help us create the best financial model, trading, agentic model to reach our goal, right? Because it depends if you want to make a $100 and you only have like 100, then you need to take massive risk. If you have a $1,000, you want to take make $10, that's your goal, you don't have to take that much risk, right? So, the models is going to be different. So, what I was thinking we can do is some kind of hybrid version. We're going to use Codex to create our financial model. Hopefully, it's a profitable model. And we're going to use Agentic to monitor that model and it can change it on the fly. That's I think we're going to do it in this video. And that is the next point here. That is the AI agent that is running this. We do want some kind of harness here. So, this is the tip the part that's going to execute and monitor our setup, our agentic AI trading setup. So, for example, that could be something like Codex that I think we're going to use today. This is from OpenAI. But we also have Claude Code, right? Claude Code from um from Anthropic. And we also have something like Open Code. That is also very good. OC for Open Code. If you want to try maybe local models, that's also an opportunity. So, this was a really long intro, but I just wanted to explain kind of what we're going through in this video. I'm not going to drag it out too much. Like it's not going to be like a 1-hour long video, but basically I want to go through these points and I think we're kind of going to skip the first point here since I already covered this on my previous videos, right? So, you can check those out if you want to go back and see that, okay? So, I think we're just going to get into it, look at my account, how I set this up, and we're going to get started looking at how we can get some data uh to get started trading with the AI. So, like I said in the intro, uh for this video, I will be using something called HyperLiquid. This is a platform where we can do cryptocurrency trading, and we can do some perps with stocks. We can trade different products here. But, I think for the video today, because of the volatility in cryptocurrency lately, I think it's a good place to start just by experimenting a little bit with like cryptocurrency and stuff like that, because it's a really easy way to get data. And, yeah, it's very easy just to get started on that if you want to, yeah, just get into it. But, basically, you can see what I have done now. I have my account on HyperLiquid. I deposited some money. You can see I have $955 in this account, okay? And, now I'm basically ready to get going. So, what I have done is I created this beginner.md file. I'm going to leave a link to this in the description below if you if you want to check it out. So, this should help you get started setting up this if you feed it to like a Codex or Cloud Code Open Code or something like that. I think if you have this file and you feed this as context, we should get up and running pretty fast. And, just to show you, this is the environment variables I have. And, I'm just going to go into my traders directory here, and I'm going to launch Codex. I'm going to do the Yolo, so we can do this autonomously. Okay. So, you can see now I am in this directory, and here we can do read at beginner, because I uploaded this. So, this should read our setup now, and we're going to start going and see if we can get going as fast as possible just to see that everything works on hyperliquid and we can start setting up our research, getting some data, and start doing this. So, let's say I just follow up with just going to say create a framework based on this our ENB and so we can make our first trade on cryptocurrency on hyperliquid. So, hopefully now Codex kind of understands all of this setup that we gave it here, the instructions, and it's going to help us get going. So, we're just going to write all the code we need from the hyperliquid API to just get this set up. But, if you want to read more, you could of course just go to if you go to more and probably just documentation here and you can read more about the documentation here around the API. I think that's pretty smart. Like, if you kind of want to read more a bit more into it and just following what I say. If you want to dive a bit deeper, okay? So, that's could be good to do and you can read all about the fees here because there are different fees, how much you trade. So, don't it's not free, but it's a very low fee at least the way volume I do. So, here you can learn a lot about or you can just point your agent or your framework to this and ask questions. That's also a good idea I really recommend. If you head back to Codex now, you can see it's started building out our framework okay? So, I'm just going to let this run. Hopefully, that beginner.md file is enough to get us up and running and we can start looking at data and start planning out our financial model we want to try because that's the most interesting part, right? So, let's see here now. You can see built the first trade framework in this repo. Perfect. That looks pretty good. I verified implementation against the SDK. Yes, and docs. So, I think we just going to do like a test trade here now. So, so I'm just going to say as a test place a bit uh Bitcoin long for $10, okay? So, we should if this works now, if we can get connected, we should be able to see our position here down on Hyperliquid. So, just to know that everything is up and running before we started collecting um the data to building out our model, I just want to see that we are up and running here. So, we can now see that our agent is running our script, right? And hopefully, that will place a position here on uh Hyperliquid now. So, remember we said $10 on Bitcoin long. Uh I don't think we have any margin or leverage on this. So, let's just see if this position kind of pops up here now. And here we have it, right? So, you can see $10 on Bitcoin long. You can see we have all the uh the P&L here. So, yeah, I can see the entry price. Damn, Bitcoin is low now. So, the volatility is quite good at the moment. So, I think there are opportunities here on the cryptocurrency today. So, I think I said in the intro, that's what we're going to focus on today. So, I'm just going to say good uh exit trade. So, now again, we can just ask Claude Code, Code X, open code, our agent to just uh execute that code for us, and you should see that trade uh yeah, will be removed in the background here. So, you can see how fast it was, too. That was really low latency. Okay? So, that's pretty good. So, now we are up and running. Now, we are ready to start gathering some data to create our financial model. We want to uh just try out today, I guess. Let's say Let's talk about data. So, on Hyperliquid, like I said, there are historical data that we can get from this. So, just make it easy. If you know exactly what data you need, you could be a more bit more specific, but I'm just going to copy this page here. And I'm going to head over to my cursor. Just going to create a folder called docs. And I'm going to call it just going to paste in this information here. Okay? All right. So, that's good. So, we have that. And now I'm just going to say, read this documentation file we just got out. So, we want to trade cryptocurrency on Hyperliquid today with the goal of making $10 today. It can be anything, right? But like I said, we will come back to different strategies and risk-taking. To build and research a good model, we need some data for backtesting. Based on your knowledge on finance, math, trading, collect the relevant data. We need to look for opportunity. This is a bit simple, right? But basically, the idea today is just to give you some inspiration to get going and start exploring this because I want to mention right quite early because I do this mostly for fun. I don't have any like big aspirations of quitting my job to start doing this. This is like a side hustle for me. I like diving into AI, as you probably know from this channel. And I'm also very interested in like finance and stuff like that. So, this is kind of the perfect crossover for me. And I really enjoy just experimenting, learning, and getting better each day. And of course, for the future, I think this could be very interesting. So, you can see now Codex comes back with monthly S3 archive for deep historical book data. But it's not timely enough for today. So, for a same trading same-day trading model, I'm going to collect API data we can use immediately, recent candles, order books, funding, market context, and account risk state. So, it's quite good to kind of outsource this to to the Codex, to the models, if you are a beginner, because then you really don't know so much, and it's nice to just get like a good expert, because it has read everything about finance data and stuff like that, just to help you get started. And when you get more experience, you can be more specific in your data collection. Today, I only want to collect data on Bitcoin, maybe on oil, maybe I want to get into some sentiment analysis on Reddit, maybe I want to see their Wall Street Bets. So, there's a lot of stuff we can learn, uh, going forward. But let's just see now. You can see, uh, Codex is setting up, uh, to pull some data now, so we're just going to wait for that. And hopefully, when we have all the data, uh, we can start testing. So, you can see, I will write a time set file under Hyperliquid, so the back tests are reproducible. This is really good. So, let's just wait for this, and hopefully, we have something soon we can start testing out. So, while we wait for the data to finish up, we can take a look now. So, if we head to data, you can see, we can go into, uh, let's just do raw, and we can do, let's say, books, and we look at Bitcoin. Here, you can see all the data we have collected. We can check out, was it candles? Let's say, 1 hour. Okay. Yeah, we have collected some data. That's good. Funding. A lot of data here. Okay. So, this is a good start, right? So, if we head back here now, it's going to keep continue a bit more, and let's just wait for that. Uh, let's start asking it to build some, uh, some models for us. So, now that we have all the data here, you can see, collected data here. Yeah, we had a look at that. And it says everything it collected here. Uh more useful files. Uh initial findings. Recent data is bearish. The simple 5-minute trend model performed better on shorts than long after 8 bps round trip uh cost assumptions. And it did some initial back testing here. This not This does not prove probability or guarantee the $10 goal today, but it tells us not to force long-only trades right now. The next practical model should be a short or flat intraday strategy with strict spread dip and filter hard stop loss. Okay, so now we going to just going to get into it. I'm going to do a prompt here, and let's see what uh our model comes up with here. So now I'm just going to say, based on the data we collected, create three hypotheses of a model that looks promising to meet our goal today. That was like the $10, right? Uh be analytical, use math and data, look for that edge, and be creative. Multiple trades is allowed. Think hard and deep. Find the three best options. You have the full uh budget on the account to your dis- posal. Yeah, whatever. Okay, so I'm just going to launch that. Again, >> [snorts] >> this is just for the video's sake, right? Uh but let's see what it comes up with. We definitely going to run the one that we think is best or that the one that the AI thinks is the best. That is the idea here. So the reason I do three uh options is because I just found that a bit more interesting. You have the option to learn a bit more about different strategies, and we can also get the model to come back and reevaluate its ideas, and pick the one it thinks looks the strongest based on the data we have. So I'm just going to wait for that. It's going to probably do some back testing, build out a few different models to see what looks best uh both uh on the data side, right? Okay, so you can see now we have three different options. We have the major coin short trend basket. Short those three when 5-minute EMA 20 is less than this. One return is negative and you can see it kind of did some calculations on the win rate here. This is for today. Okay, but we have had a lot of volatility today, so of course yeah, but again, this is for this video and the second one is hype mean reversion short we have salt trend as a single aggressive bet. Okay, maybe not the most interesting strategies, but let's see now. So now I'm going to see from 1 to 10. But you can see always says a ranking here. So I'm just going to see yeah. That's fine. Maybe not the most interesting setup, but uh We're just going to go with it today. And see how it turns out. But I would even if I was not doing the video, I would explore this a bit more, right? Get some other perspectives and stuff like that maybe look a bit further back. But for the video's sake, we just going to try this out. You always learn something anyway, right? So now we just going to get Codex to help us write the the code we need for this. If you go back and look, it's not that complicated. It's basically this. So yeah, that's not so hard to set up, I think. So that should go pretty fast. Yeah, and you can see kind of corrected me here. So it says capping margin usage by default full account available should not mean collateral capacity not should mean collateral capacity not unlimited loss. So yeah, that's pretty fine. That's fine, I guess. So here we can see the model is writing this this model in Python. So, this is just going to run. It's based on the parameters it set. So, what I want to do now in the next phase, that is kind of the agentic setup, is I want the model to observe the behavior of this model. So, this is where we kind of get into the new fields because we can always write models in Python before and make them run autonomously. But, what I really like about AI and agentic trading is that we can have, let's say Codex in this case, monitor a trade and make live adjustments over time. So, let's say I set this to run for 24 hours. Over time, it can adjust the strategy based on the new data. That is what's super interesting and that's what I want to try to do now. Okay, so it looks like the test went pretty good. You can see we set everything up. And yeah, we didn't enter. So, what I want to use now is this is specifically related to Codex that we're using today and that is using something called a goal. So, this is good for long running running running work, okay? So, that could be pretty interesting for this setup as I talked a bit about in my previous video. Okay, so what I did now is instead of kind of trying to write the {slash} goal prompt myself, I just got Codex to write this from me from everything we have talked about so far that we want to do, right? And I said, "Explain this very simple what the goal will do. It will run this trading script. That's what we're going to do, right?" And the goal is check on this every 2 minutes, check the logs and state, check the signal, update status.md, adjust parameters if needed. And we're going to check if the how the goal is doing, the profit. And yeah, basically, I think we're just going to do it and hopefully this starts. So, I'm just going to say start the model script then I will launch the gold prompt myself. Okay. This is a bit specific to Codex now, but I just wanted to show off this because of this I think this is kind of the the big benefits of a AI agentic trading that we can launch these agents that will monitor our trade, okay? So, let's just get this up and running and then I'm going to launch the gold prompt that's going to monitor our trade, uh, yeah, agentic, I guess. So, you can see now we have something called one background terminal task running and this is our model that is looking at the market now, looking for those parameters we set to enter a short position on those, uh, Solana, Bitcoin, Ethereum, and what we set. So, the runner works. Uh, okay, so there was a small issue here, so it's going to try to fix that before we are ready to launch our, yeah, gold monitor agent. Okay, so now we can see we have our runner up and running. We are monitoring the signal state of, uh, our trade. So, you can see not ready, not ready, not ready, and, yeah, we updated our gold prompt. So, basically, hopefully, all we have to do now {slash} goal, paste in our prompt, and hit enter. And hopefully now everything should be agentic, autonomous. You can see it down in the uh right corner here, pursuing goal. So, before I run to the store, get me some food, I'm going to leave this running in the background. I'm going to record the screen to if anything happens. But, uh, I just want to see that our, uh, agent is up and running, coming in at 2 minutes, checking everything, and going back to sleep. So, let's just wait a bit and see. Okay, so now we have it. Now Now can see our agent is checking. The runner is healthy and still only observation mode. Uh signal has moved further away from the short setup, okay? Uh red trailer and RSI are higher than the last checkpoint. So, I'm doing a direct account check balance before updating the status. You can see it runs some code here. Yeah, the balance is unchanged. We haven't made any trades. And now it's going to update the status position again, right? And hopefully the idea now is it's going to wait now 2 minutes before it checks back in again. Let's see if we set that up as we want to. So, you can see I did a small adjustment to the goal here. I just said wait 60 seconds before the next checkpoint. So, that means that we are up and running and everything is uh running. We have our monitor agent checking up on our trade. It's going to do that continuously every second minute. Uh no positions yet because we kind of have moved away from our initial uh setup, but that's good because then we can see if the model's trying to do any changes on the fly here. Like I said, I'm going to leave this running. I'm going to head to the store, grab some food and stuff while this is trading in the background for me. >> [music] >> So, I want to take you back on this because this was a very interesting point, right? So, you can see I also acted on the $10 today priority since the short ball is repeating a block by the bullish EMA structure. So, it's not bearish now. I found uh the current market fits pullback mean reversion long better than short basket. Yes, this is exactly what I wanted to see because now we are kind of reactive or reacting to the market. We are kind of updating our initial plan. Let's say we had this short bearish signal. We just going to run it overnight, but the market kind of changed a bit. The structure changed on the market. And we didn't get to update our model, but now with this agentic setup, we can actually change strategy on the fly okay without me intervening. So, that's pretty interesting. Sorry about the fan in the background, but uh this means that we now have a bit of a more new strategy. So, let's see if this goes live and then we can see if we come into any trades soon. Okay, so that was perfect timing. Sorry about the fan, but now you can see we are entering some positions here. So, the agent found some kind of new strategy. It wants to do Let's take a look at them. Sorry about this. So, you can see we are going in on Ether, BTC, and Solana. Uh we do the margins here and basically we're going to do a 3x leverage. And you can see our goal is $10. Okay, that's almost $10. So, let's see what happens. Okay, so that's 12. So, we should uh exit now, but I don't know. Let's see what happens. I'm not going to intervene because I want to learn something too. So, maybe we should have had like a take profit there now. But, let's just see what happens. I'm just going to let it run and I'll come back to you when we see something happens. Okay, so you can see we are now exiting our positions. Wow, that was easy. But, I don't know what happened like how did we actually do? It was a bit slow but I want to ask how we did and that should mean basically that our goal was completed. But, let's just wait and see a bit. Okay, so you can see direct account verification is now flat. USDC is 962. So, the profit target was hit and positions were closed. Yeah, not really. It was a bit lagging there, but basically we at least did $7 in profit. So, it calculated it to 6633. But, I guess that's fine. I think we've proven that it's this works. It run autonomously. It can change its strategy on the fly. And this is ex- exactly what I wanted to see. I'm happy. Let's call it a day because that kind of proved everything I wanted to see from this. So, we ran for how long? 56 minutes. We did like $6 profit. I think if we just let that run a bit, we could have gotten more, but that was not the point of this. The point was that we switched our strategy during the during the trade or like the agentic AI trading. We switched our strategy on the fly, and we turned that would have been like a bad strategy for this market into a small profit. So, that was pretty cool. And that's that's exactly what I hoped to get out of this video. So, you can kind of see that this is possible to set up now in mid-2026. So, yeah, I'm super happy how this turned out, and I really like this framework. So, I'm think we're just going to wrap this up, talk a bit more, and I'm going to share the beginner MD on a GitHub or something if people are interested. So, just to confirm, you can see our P&L 662 in the last 24 hours. Okay, not bad, not bad. But, just to wrap things a bit up, I think we actually we didn't spend too much time on getting started because we had that covered in other video. We did some data collection. I think that went very well. We used the API from Hyperliquid in our setup example, and we got the data we needed for today. So, that was really good. We did some research. We started out with a goal of making $10, but we started with a short strategy based on the latest data. But, those data kind of faded and we switched into like a long position at the end. And that got us to our $10 gold right. So, that was pretty good. I was really happy how we turned that around. And we can actually monitor with the slash gold set up. That was really good. That worked perfect. We even added that 60-second sleep. So, we got into that 2-minute cycle of always looking at our trade. That was the agentic part. It worked out really good. And I think the AI agent executed pretty good. I'm really happy with Coin Rex on this setup. But Cloud Code, Open Code, I'm sure a lot of other setups would work great as well. So, yeah. That was basically my agentic AI trading for beginners. Just like an introduction if you kind of want to dive into this. And I also want to mention we have a Discord. If you want to talk to people that kind of likes these things too. Just come to the link in the description. I just play around with this basically mostly to learn. And to have some money to play around with. Maybe like a small side hustle. Let's say you have a goal of like $100 a month. $1,000 a month is a bit much. Maybe like 100 a month or something. Or 100 a week or something like that. That's pretty fun. I think that's pretty interesting. But yeah. That's what I wanted to share today. And like I said, if you want to go to see how to get started, check out the different videos we have here on my previous videos. So, yeah. Thank you for tuning in. Hope this gave you some inspiration to get going if you want to. And hopefully I'll see you again soon.